Accepted Manuscript LES of flame acceleration and DDT in small-scale channels Yongyao Zhao, Cheng Wang, Yong Bi PII:
S0950-4230(17)30158-4
DOI:
10.1016/j.jlp.2017.02.011
Reference:
JLPP 3420
To appear in:
Journal of Loss Prevention in the Process Industries
Received Date: 30 October 2016 Revised Date:
15 February 2017
Accepted Date: 15 February 2017
Please cite this article as: Zhao, Y., Wang, C., Bi, Y., LES of flame acceleration and DDT in small-scale channels, Journal of Loss Prevention in the Process Industries (2017), doi: 10.1016/j.jlp.2017.02.011. This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
ACCEPTED MANUSCRIPT
1
LES of flame acceleration and DDT in
2
small-scale channels
RI PT
3 4
Yongyao Zhao1, Cheng Wang1*, Yong Bi1
5
E-mail:
[email protected] 1
State Key Laboratory of Explosion Science and Technology, Beijing Institute of Technology,
SC
6 7
Beijing 100081, China
AC C
EP
TE D
M AN U
8 9
ACCEPTED MANUSCRIPT 10
11
Abstract Large eddy simulation (LES) has been performed to investigate the process of flame
13
acceleration (FA) and deflagration-to-detonation transition (DDT) in a small-scale channel
14
filled with methane-oxygen mixture. A transport model for sub-grid kinetic energy was used
15
to enclose the sub-grid fluxes while the sub-grid scale combustion was modelled by the
16
artificially thickened flame (ATF) model. A fifth-order Weighted Essential Non-oscillatory
17
(WENO) finite difference scheme was used to discretize the advection term of the governing
18
equations. The results showed that the interaction of the flame with the flow and the pressure
19
wave ahead led to the flame acceleration. The overdriven detonation led by the local
20
explosion formed and the flame propagated upstream which then coupled with the precursor
21
shock wave. A retonation wave propagating downstream was further observed. The
22
experimental results substantiate the validation of the calculation and further identified the
23
DDT mechanism.
24
Keywords: large eddy simulation, WENO, flame acceleration, DDT
SC
M AN U
TE D EP
26
AC C
25
RI PT
12
ACCEPTED MANUSCRIPT 27 28
1
Introduction Most gaseous explosions usually start from electrical spark or autoignition, which lead to
30
the transition to detonation. In order to prevent accidental explosions in the production and
31
control the detonation initiation in the propulsion system, it is necessary to investigate the
32
mechanism of DDT. Ciccarellia & Dorofeev,(2008); Liberman et al., (2010); Ivanov et al.,
33
(2011); Kagan & Sivashinsky, (2003); Oran & Gamezo, (2007); Thomas et al., (2010); Valiev
34
et al., (2009); Nie et al. (2014, 2015); Wang et al., (2016) have already conducted various
35
studies on the general initiation of DDT. However, detailed mechanism of FA and DDT
36
remain one of the major unresolved problems in combustion science, hence need for further
37
research
M AN U
SC
RI PT
29
Micro-scale combustions have attracted more attention since several studies on simulations
39
and experiments on the process of DDT in the micro-scale channel have been conducted. Wu
40
et al. (2007, 2011) analyzed the effects of tube diameter and equivalence ratio on the flame
41
propagation of ethylene-oxygen mixtures, and observed several reaction propagation
42
scenarios including DDT/C–J detonation, oscillating flame, steady deflagration, galloping
43
detonation and quenching flame. Liberman et al. (2010) investigated the flame acceleration in
44
channels, and observed that the flame propagation exhibits three distinct stages: (1) the flame
45
accelerates exponentially producing shock waves far ahead from the flame, (2) the flame
46
acceleration decreases and shocks are formed directly on the flame surface, and (3) the final
47
third stage of the actual transition to a detonation. Li et al. (2016) also conducted a DDT study
48
in the macro-scale tube, and noticed that the boundary layer plays more important role in the
49
flame acceleration in the narrow channel and further has an influence on the DDT distance.
50
Manzhalei et al. (1998) studied the near-limiting propagation of detonation waves in a
51
capillary tube filled with acetylene-oxygen mixture. The momentum and the heat losses
52
caused by the wall friction as well as the heat of conduction led to a different propagation
53
modes of the flame. Kuznetsov et al. (2005) studied the effect of initial pressure on the run-up
54
distance to DDT in a smooth tube with a stoichiometric hydrogen-oxygen mixture, found the
55
dependence to be close to the inverse proportionality. The mechanism of DDT through 2D
56
and 3D simulations using detailed chemistry reactions were investigated by Liberman et al.,
57
2010 and Ivanov et al., 2011. They found that, the mechanism of DDT is entirely determined
58
by the features of the flame acceleration in tubes with no-slip walls. Moreover, FA and DDT
AC C
EP
TE D
38
ACCEPTED MANUSCRIPT in micro-scale channels have been studied by (Ivanov et al., 2013) using high-resolution
60
simulations of 3D Navier-Stokes equation with detailed chemical reaction mechanism. They
61
obtained “numerical schlieren” and “numerical shadowgraph” which clarify the meaning of
62
the experimental schlieren and shadow photos. Han et al. (2017) studied numerically the
63
integrated mechanistic behavior of flame acceleration and DDT in micro- and macro-channels
64
by DNS. The pressure wave and flow field play an important role in the combustion
65
dynamics. However, the simulation of DDT and detonation in the combustible mixtures is a
66
formidable task for direct numerical simulation (DNS) (Han et al. 2015; Wang et al., 2013).
67
LES method is a good choice to study the flame acceleration (FA) and DDT. Xiao et al.
68
(2012) undertook both experimental and numerical studies on the formation of a premixed
69
hydrogen/air tulip flame in a closed channel and discussed the evolution of flame structure.
70
They claimed the appearance of adverse pressure gradients resulted in reverse flow in the
71
unburned gas, which caused the onset of the tulip or a distorted tulip flame. Johansen &
72
Ciccarelli (2009) simulated the flame propagation in a closed channel with an obstacle using
73
large eddy simulation with the dynamic Smagorinskye-Lilly subgrid model and the Boger
74
flame surface density combustion model. They found that flame velocity showed oscillations
75
which were as a result of the acceleration and deceleration of the unburned gas flow through
76
obstacles. The flame–vortex interaction was investigated by Sarli et al. (2009) using LES
77
model coupled to the power-law and flame-wrinkling model, and they found that the
78
interaction intensified the rate of flame propagation and the pressure rise. Shin et al. (2008)
79
examined the effect of turbulence on the highly unstable detonation by using Monotone
80
Integrated Large Eddy Simulation (MILES), and compared the results from inviscid Euler
81
equations, Reynolds Averaged Navier-Stokes (RANS) equations, and MILES. They found all
82
approaches are capable of capturing the highly unstable detonation wave, but show distinctive
83
differences in FFT analysis of highly unstable detonation. Yu & Navarroe-Martinez (2014)
84
used a flame thickening approach to capture DDT with relatively coarse meshes, and assessed
85
the capabilities of ATF to capture DDT. Emami et al. (2015) also used LES to study the DDT
86
in an obstructed channel containing premixed stoichiometric hydrogen-air mixture. They
87
studied the effect of grid resolution on the formation of hot spots and the occurrence of DDT,
88
and verified the accuracy of the results in predicting the DDT. They claimed that the flame-
89
vortex interaction with the folding and wrinkling flame were the main mechanism of the
90
flame acceleration in the slow flame regime. Robert et al. (2015) presented a first LES study
91
addressing different scenarios for knock and super-knock observed in a spark ignition (SI)
AC C
EP
TE D
M AN U
SC
RI PT
59
ACCEPTED MANUSCRIPT 92
engine. LES showed that the pressure waves generated by one or a couple of autoignition
93
spots were strong enough to induce locally a strong fresh gases temperature increase leading
94
itself to a substantial decrease of the autoignition delay. From the above discussions on the
95
various articles, it could be observed that only few researches focused on the complete
96
process of DDT by applying LES. Hence this present study seeks to apply the large eddy simulation (LES) methodology to
98
investigate the FA and the methane-oxygen DDT in the channel with no-slip and adiabatic
99
walls. Further validation of the numerical simulation results was compared with the experiments.
101
2
SC
100
RI PT
97
Governing equations
A spatial filter based on local grid size has been applied to the compressible Navier-Stokes
103
equations for mass, momentum, total energy, and mass fraction of a reactant to obtain the
104
governing equations for LES. The resulting LES equations are obtained by applying Favre
105
averaging, which is defined by °f = ρ f / ρ , where the over line indicates volume average. The
106
volume average for flow variable f is given by: f ( x) =
110
111
112
113
114
∂ ρ ∂ ρ u°i + =0 ∂t ∂xi
EP
109
(1)
The governing equations are as following:
∂ ρ u°i ∂ ∂ ∂ + ( ρ u°° (τ ij ) − (τ ijsgs ) i u i + Pδ ij ) = ∂t ∂x j ∂x j ∂x j
AC C
108
∫ f ( x′)F ( x − x′) dx′
TE D
107
M AN U
102
(2)
(3)
° ∂ ∂ qi ∂ρ E ∂ ° + P )u%i ) = ∂ (u%% + (( ρ E − H isgs i τ ij ) − ∂t ∂xi ∂xi ∂xi ∂xi
(4)
∂ ρY° ∂ °% ∂ ∂Y° ∂ sgs ξω& + ρYu i = ( ρ Fξ Di )− ϕi + ∂t ∂xi ∂xi ∂xi ∂xi F
(5)
υ ∂k sgs ∂ u° ∂ ρ k sgs ∂ ∂ + ( ρ u°i k sgs ) = (ρ t ) − τ ijsgs ( i ) − C e ρ ( k ) 3 2 ∆ ∂t ∂xi ∂ xi pr ∂xi ∂x j
P = ρ RT°
(6)
(7)
ACCEPTED MANUSCRIPT 115
where; ρ , u i , P , τ , E , q , Y , and ω& denote density, velocity, pressure, stress, total energy,
116
energy diffusion vector, mass fraction and reaction rate, respectively. Several terms in the
117
LES equations require closure, such as the sub-grid stress τ ijsgs , the unclosed term Hi , the sub-
118
grid diffusion of specie mass ϕi . Here, a closure scheme based on a transport model for the
119
sub-grid kinetic energy k sgs is used to close the momentum, energy and mass fraction sub-
120
grid fluxes, shown as function (6), which is solved along with the LES equations. The three
121
terms at the right hand of function (6) are the sub-grid kinetic energy diffusion, production,
122
and dissipation, respectively. The unresolved momentum fluxes are expressed according to
123
the Boussinesq assumption. The enthalpy fluxes and mass fraction sub-grid fluxes are
124
modeled through a simple gradient assumption. The sub-grid stress, energy flux and specie
125
mass fraction are closed as follows:
sgs
M AN U
SC
RI PT
sgs
1 3
τ ijsgs = − 2 ρυ t ( s° s° ij − ij δ ij ) +
126
H isgs = − ρ
127
ϕisgs = − ρ
TE D
128
υt ∂% h prt ∂xi
υt ∂Y
Sct ∂xi
2 ρ k sgsδ ij 3
(8)
(9)
(10)
Where, υt = Cυ (k sgs )1/2 ∆ , h denote total enthalpy, and h = E + P / ρ . Constants 0.067 and 0.916
130
are employed for model coefficients Cυ and Ce , prt and Sct are valued as 1.0. The chemical
131
source term is closed as
AC C
132
EP
129
ω&= A ρ Y°exp ( −
Ea ) RT
133
The thickened-flame (TF) model (Durand & Polifke, 2007; Colin, 2000) is used in the
134
combustion closed model. In the TF model, the flame front is artificially thickened to be
135
resolved on LES mesh, which is simply achieved by decreasing the pre-exponential factor of
136
the chemical Arrhenius law whereas enhancing the molecular diffusion by a factor F. The
137
response of the thickened-flame to turbulence is considered by incorporating an efficiency
138
function (Charlette et al., 2002) in the governing equations. In this present work, the
139
thickening factor defined as F = N ∆ / δ , N is the grid number in the flame front, ∆ is the cell
140
size and δ is the thickness of laminar flame. The efficiency function ξ is Power-Law Flame-
141
Wrinkling Model (De & Acharya 2008). These parameters of the system are summarized in
ACCEPTED MANUSCRIPT 142
Table 1. They are chosen to model the stoichiometric methane-oxygen mixture. The numerical
143
methods used are the 5thorder WENO scheme for convective terms, the 6thorder central
144
difference for viscosity terms and the 3rd-order TVD Runge-Kutta method for time
145
discretization (Wang et al., 2016; Han et al. 2015, 2017).
146
3
RI PT
Experimental apparatus The experimental setup is shown in Fig. 1 and it is well discussed in our recent article
148
(Wang & Hu, 2016). It consists of a constant volume combustion vessel, a gas mixing system,
149
a high-speed photography system, a high-voltage ignition system, and a data acquisition
150
system. The combustion vessel is a horizontal explosion channel, with a transparent quartz
151
glass on one side. The channel is closed at both ends with a square cross section of 20 mm ×
152
20 mm, and the length is 1.5 m, which was filled with premixed methane–oxygen, shown as
153
Fig. 1. The initial pressure in the channel was 33 kPa, and the initial temperature was also
154
chosen to be 298 K. The ignition position was located at the left end of the channel. The spark
155
igniter, pressure recorder, temperature recorder and high-speed video camera were controlled
156
by the synchronization system.
157
4
TE D
Results and discussions
M AN U
SC
147
In our study, two-dimensional channel was used. The computational domain was 20 mm×
159
1.5 m filled with stoichiometric methane-oxygen premixed gas. The boundary was a solid
160
reflecting and no-slip wall. To ignite the mixture in the simulation, a circular region of hot
161
burned material was placed at the left end of the channel while a small perturbation was
162
further added to the burnt region. Initial velocity, temperature and pressure of 0.0 m/s, 298 K
163
and 33 KPa, respectively were considered in the unreacted mixture. The validity of the
164
numerical method and the tested convergence of the solutions with different mesh sizes (1,
165
0.5, 0.2, 0.1 mm) were verified as shown in Fig. 2. When the grid resolution was 1.0 mm, the
166
solution differs significantly from that obtained with the refine mesh, and the flame
167
propagates faster than that of the experiment. As the mesh become finer, the results tend to
168
converge. The results obtained with the grid resolutions of 0.2 and 0.1 mm were almost the
169
same, and the position of the flame tip change with time, agree with the experimental results
170
before DDT occurs.
AC C
EP
158
171
The flame tip speed obtained from the numerical calculation was also compared with the
172
experimental results, as shown in Fig. 3. The overall trend of the curves was roughly the same.
ACCEPTED MANUSCRIPT They showed several distinct stages which demonstrate the appearance of DDT. In the
174
experimental results, there was a long acceleration period of the flame before an abrupt
175
increase appears with the value of 2500 m/s at x~0.77 m which indicate the occurrence of
176
DDT. The velocity further decays to 2300 m/s. In the numerical simulation, at the first stage,
177
the flame accelerates exponentially with a rapid augmentation of the flame surface area as
178
shown in Fig. 3. This phenomenon has been explained theoretically and confirmed by DNS
179
and experimental studies by Liberman et al. (2011). During the second stage, the rate of flame
180
acceleration decreases. It was noticed that, at a mesh size of 1 mm, the calculated flame
181
velocity was larger than the experimental values before DDT occurs. While, with the finer
182
mesh, the flame velocity tends to be close to the experimental values before DDT. The DDT
183
run-up distance was slightly longer than that of the experimental result. This was because the
184
simulation used 2D, while 3D was used for the experiment. Ivanov et al., (2013) observed that
185
the flame velocity in 3D case is larger than that in 2D case, and the time of DDT is shorter in
186
3D case compared to 2D due to extra dimension. In the stage of DDT, the flame velocity had a
187
sharp increase, reaching the maximum value, and then decays to a steady value.
M AN U
SC
RI PT
173
Fig. 4 shows the process of flame acceleration and DDT for both the experiment and
189
numerical results. Fig. 4a presents a high-speed image of the experiment while Fig. 4b shows
190
the numerical results of temperature contours. After ignition, the initial laminar flame kernel
191
expands, and then the flame gradually spreads to the right end of the channel, showing a
192
fingerlike appearance. In this stage, the brightness of the flame surface was brighter, and then
193
becomes dim along with the flame surface changing. As the flame propagates, a suddenly
194
white light near the flame front appears, which means local explosion occurs. As shown in Fig.
195
3, the flame velocity reaches the maximum value after a while. These explosions initiate the
196
process of DDT, and the detonation front continue to propagate upstream while a retonation
197
wave moves downstream. The white light gradually fill the channel, and the detonation and
198
retonation propagation process are shown by the red lines. Similar to the experimental results,
199
numerical results show the processes of flame acceleration, DDT, and the resulting detonation
200
propagation can be seen from the temperature contours shown in Fig. 4b.
AC C
EP
TE D
188
201
Fig. 5 shows the temperature and pressure profiles in the flame front along the channel at a
202
selected time interval. As the flame propagates forward, the thermal expansion of hot products
203
induces the unburned gas to flow ahead of the flame. In this process, pressure waves are
204
produced constantly. This lead to a small heating of the gas which is less than 500 K in the
205
unburned material ahead of the flame, until the pressure is strong enough for DDT occur. Due
ACCEPTED MANUSCRIPT to the flame interaction of the shock wave formed at the flame front, a feedback mechanisms
207
forms, leading to the development of high flame speed. As time elapsed the shock waves are
208
coalesced, merged, amplified and hence creating a layer of compressed and heated unreacted
209
gas – which is the preheat zone (Liberman et al. 2009). This leads to more unburned mixture
210
with higher density and temperature entering the flame front. Eventually, the flame front
211
coincides with the leading shock, leading to the flame propagating as detonation waves.
RI PT
206
Fig. 6 shows the time-distance trajectory of the flame tip and the leading shock according to
213
the numerical results. As the flame propagates to the right, it acts as a piston producing
214
compression waves in the unreacted gas, which steepens into the shock waves far ahead of the
215
flame front. The leading shock stands in front of the flame front, presenting a typical two
216
waves and three regions structure. The distance between them is about two times of the
217
channel width, which is less than Ivanov’s result (2013), five to seven times of the channel
218
width. This might be as a result of too large calculating area used in the study with regards to
219
the 2D model. The distance between them gradually decreases which tends to zero at about
220
x=0.74 m at t~1.4ms. The flame accelerates constantly and becomes faster at t~1.4ms,
221
corresponding to the velocity increase fast as shown in Fig. 3. Consequently, the flame
222
catches up with and couples the leading shock, which leads to the overdriven detonation.
TE D
M AN U
SC
212
The flame acceleration is not only related to pressure waves, but also to the flow field
224
around flame front. After ignition of the flame kernel, initially, the weak compression waves
225
move away and produces flow ahead of the flame with the velocity, approximately u = (Θ−1)U f ,
226
Θ = ρu ρb ≈ 10 , where ρu and ρb indicate the density ratio of the unburned and burned gas,
227
respectively (Zeldovich et al. 1985). Fig. 7 shows the flow field near the flame front at
228
t=0.727 ms and 0.86 ms. The figure shows the velocity contours and the corresponding
229
longitudinal velocity profiles (right part) in the cross section A and B. Owing to the wall
230
friction, the flow velocity vanishes at the channel walls, and has maximal value at the center
231
of the channel which exceeds 100 m/s. The flame surface is stretched gradually, which results
232
in more fresh fuel being consumed. Higher burning velocity results in an enhancement of the
233
flow velocity ahead of the flame, which in turn gives rise to a larger gradient field and boosts
234
flame stretching. As the flame is stretched in the velocity field ahead of the flame front, a
235
positive feedback coupling is established between the upstream flow and the burning velocity
236
(Ivanov et al., 2013). Whiles it was realized that the flow velocity was dropping to zero at the
237
channel walls, the flow ahead of the flame was uniform in the bulk. This is because, the time
AC C
EP
223
ACCEPTED MANUSCRIPT interval from the flame ignition to the instant of DDT is too short to establish parabolic
239
velocity profiles (i.e. the Poiseuille flow) (Kuznetsov et al., 2010). According to Zeldovich
240
(1985), the estimated time of formation of a parabolic velocity profile in the upstream flow
241
given as t = D 2 / υ , is similar to the magnitude obtained in (Liberman et al., 2009), i is larger
242
than 10 s for the channel with width 20 mm. However, the duration of the whole process of
243
DDT is in the range of 1 ms, which is shorter than the time required for the formation of a
244
parabolic velocity profile in the upstream flow by many orders of magnitude.
RI PT
238
The streamlines near the flame front are drawn at some time instants, as shown in Fig. 8. It
246
has been found that, upto the time of DDT, the flow ahead of the flame remains laminar, and
247
only becomes unparalleled behind the flame surface. At t=1.31 and 1.39 ms, all the
248
streamlines vanish in front of the flame, and this is because pressure waves steepen into shock
249
waves in front of which there is no disturbance.
M AN U
SC
245
The pressure and temperature distributions are presented in Fig. 9 in the process of DDT.
251
The leading shock is located on the right side of the flame, and the preheating zone forms
252
ahead of the flame front at t=1.31 ms (see Fig. 9a). The unburned mixtures were compressed
253
and heated in the preheating zone by the increasing pressure. As a result, a larger amount of
254
mixture enters the reaction zone, which enhance the reaction rate. These effects lead to
255
stronger temperature and pressure gradient as a feedback. Under such interactive circumstance,
256
it causes the formation of a shock wave directly ahead of the reaction zone, and these
257
conditions lead to the coupling of the shock and the reaction zone, which results in the
258
transition to detonation (Ivanov et al., 2011). Fig. 9b shows the change process of pressure
259
contours. At t=1.35 ms, there was a high-pressure zone between the flame front and the wall.
260
Then, a strong circular pressure wave could be seen caused by local explosion, triggering the
261
transition from deflagration to detonation. Finally, the flame coupled with the leading shock,
262
leading to the flame propagate as detonation waves.
AC C
EP
TE D
250
263
Fig. 10 shows the occurrence of local explosion and its process of development. When
264
t=1.383 ms, there appears a local explosion near the wall that causes local pressure rises
265
rapidly. The explosion waves propagate in the channel forming reverse detonation wave and
266
transverse wave at t=1.39 ms. At the same time, this explosion wave propagates forward,
267
catches up with the flame and collides with the precursor shock, triggering another stronger
268
local explosion at t=1.396ms. Finally, a strong overdriven detonation forms, the flame
ACCEPTED MANUSCRIPT 269
coupling with the precursor shock propagates upstream, and another stronger reverse
270
detonation wave propagates downstream, see Fig. 10 at t=1.401 and 1.405 ms.
271
4. Conclusions LES method has been used to investigate the flame acceleration and DDT in small scale
273
channels. The artificially thickened flame approach was adopted since the mesh scale is
274
greater than the thickness of the flame. The comparisons of experimental and numerical
275
results show that, applied method in this study is more effective and can capture the typical
276
characteristic of flame acceleration and DDT.
RI PT
272
The flame velocity initially increase rapidly as a result of the stretching of the flame front
278
within the boundary layer. The interaction between the flame and the flow field around the
279
flame front establishes a positive feedback mechanism, promoting the flame propagation.
280
Though the channel width is 20mm, the Poiseuille flow is not formed in front of the flame up
281
to the time of DDT occurs, and the flow was always laminar.
M AN U
SC
277
The fast acceleration of the flame is due to its coupling with the shock wave formed at the
283
flame front. As the pressure increases, the leading shock is close to the flame front, and the
284
preheating zone appears ahead of the flame, increasing the chemical reaction rate. Finally, the
285
shock-wave complex near the wall becomes severe, which causes local explosion, and then
286
triggers the process of DDT. Then, there appear retention waves propagating backward with
287
detonation wave propagating forward.
288
Acknowledgements
289
This research is supported by the National Natural Science Foundation of China under grants
290
11325209 and 11521062.
291
References
292 293
Bychkov, V., Petchenko, A., Akkerman, V., & Eriksson, L. E. (2005). Theory and modeling of accelerating flames in tubes. Physical Review E, 72(2), 138-148.
294 295 296
Charlette F, Meneveau C, Veynante D. (2002). A power-law flame wrinkling model for LES of premixed turbulent combustion. Part I: Non-dynamic formulation and initial tests. Combust. Flame. 131 (1/2), 159.
297
Ciccarellia G., Dorofeev S., (2008). Flame acceleration and transition to detonation in ducts. Prog
298
AC C
EP
TE D
282
Energ Combust.34:499.
ACCEPTED MANUSCRIPT 299 300 301 302
Colin O. (2000). A thickened flame model for large eddy simulation of turbulent premixed combustion. Phys. Fluids. Inst. 12: 7. De A, Acharya S., (2008). Large Eddy Simulation of Premixed Combustion With a Thickened-Flame Approach. J. Eng. Gas Turb. Power. 131(6):1021-1034. Durand L, Polifke W. (2007). “Implementation of the Thickened Flame Model for Large Eddy
304
Simulation of Turbulent Premixed Combustion in a Commercial Solver,” ASME Paper No.
305
GT2007-28188.
RI PT
303
Emami, S., Mazaheri, K., Shamooni, A., & Mahmoudi, Y. (2015). Les of flame acceleration and ddt in
307
hydrogen–air mixture using artificially thickened flame approach and detailed chemical kinetics.
308
Int. J. Hydrog. Energy, (23), 7395-7408.
SC
306
Han W. H., Gao Y., Law C. K. (2017). Flame Acceleration and Deflagration-to-Detonation Transition
310
in Micro- and Macro-Channels: An Integrated Mechanistic Study. Combust. Flame. 176, 285-298.
311
Han W. H., Gao Y., Wang C., Law C. K. (2015). Coupled pulsating and cellular structure in the
313 314 315 316
propagation of globally planar detonations in free space. Phys. Fluids. 27, 106101. Kagan, L., & Sivashinsky, G. (2003). The transition from deflagration to detonation in thin channels. Combust Flame, 134(4), 389-397.
Kuznetsov, M., Alekseev, V., Matsukov, I., & Dorofeev, S. (2005). DDT in a smooth tube filled with
TE D
312
M AN U
309
a hydrogen–oxygen mixture. Shock Waves, 14(3), 205-215. Ivanov, M. F., Kiverin, A. D., & Liberman, M. A. (2011). Flame acceleration and DDT of hydrogen-
318
oxygen gaseous mixtures in channels with no-slip walls. Int. J. Hydrog. Energy, 36(13), 7714-
319
7727.
EP
317
Ivanov, M. F., Kiverin, A. D., Yakovenko, I. S., & Liberman, M. A. (2013). Hydrogen–oxygen flame
321
acceleration and deflagration-to-detonation transition in three-dimensional rectangular channels
322
with no-slip walls. Int. J. Hydrog. Energy, 38(36), 16427-16440.
323 324 325 326
AC C
320
Jiang, GS, Shu CW. (1996). Efficient implementation of weighted ENO schemes. J. Comput. Phys. 126:202.
Johansen, C. T., & Ciccarelli, G. (2009). Visualization of the unburned gas flow field ahead of an accelerating flame in an obstructed square channel. Combust. Flame, 156(2), 405-416.
327
Khokhlov, A. M., Oran, E. S., & Thomas, G. O. (1999). Numerical simulation of deflagration-to-
328
detonation transition: The role of shock-flame interactions in turbulent flames. Combust. Flame,
329
117(1-2), 323-339.
ACCEPTED MANUSCRIPT 330 331 332 333
Knudsen E., Pitsch H. (2008). A dynamic model for the turbulent burning velocity for large eddy simulation of premixed combustion. Combust. Flame, 154: 740. Li, J., Zhang, P., Yuan, L., Pan, Z., & Zhu, Y. (2016). Flame propagation and detonation initiation distance of ethylene/oxygen in narrow gap. Appl. Therm. Eng., 110, 1274–1282. Liberman, M. A., Ivanov, M. F., Kiverin, A. D., Kuznetsov, M. S., Chukalovsky, A. A., &
335
Rakhimova, T. V. (2010). Deflagration-to-detonation transition in highly reactive combustible
336
mixtures. Acta Astronautica, 67(7-8), 688-701.
RI PT
334
Liberman, M. A., Kuznetsov, M., Ivanov, A., & Matsukov, I. (2009). Formation of the preheated zone
338
ahead of a propagating flame and the mechanism underlying the deflagration-to-detonation
339
transition. Physics Letters A, 373(5), 501-510.
341
Manzhalei V.I., (1998). Detonation regimes of gases in capillaries. Combust., Expl., Shock Waves 34: 662–664.
M AN U
340
SC
337
342
Nie B, He X, Wang C, Lu H, Xue F., (2015). Computational method of the propagation velocity of
343
methane explosion flame based on correlation coefficient of images. Combust. Sci. Technol.,
344
187(8):1157-1166.
346 347 348
Nie B, He X, Zhang C, Li X, Li H. (2014). Temperature measurement of gas explosion flame based on the radiation thermometry. Int. J. Thermal Sci., 78:132-144.
TE D
345
Oran, E. S., & Gamezo, V. N. (2007). Origins of the deflagration-to-detonation transition in gas-phase combustion. Combust. Flame, 148(1–2), 4-47.
Poinsot, T., & Veynante, D. (2005). Theoretical and numerical combustion (2nd ed.). R.T. Edwards.
350
Robert, A., Richard, S., Colin, O., & Poinsot, T. (2015). Les study of deflagration to detonation
351
EP
349
mechanisms in a downsized spark ignition engine. Combust. Flame, 162(7), 2788-2807. Sarli V. Di, Benedetto A. Di, Russo G., Jarvis S., Long E. J., Hargrave G. K., (2009). Large Eddy
353
Simulation and PIV Measurements of Unsteady Premixed Flames Accelerated by Obstacles. Flow
354
Turbul. Combust., 83(2):227-250.
AC C
352
355
Shin J., Cho D., Won S. and Choi J. (2008). Large Eddy Simulation of a Highly Unstable Detonation
356
Wave. 46th AIAA Aerospace Sciences Meeting and Exhibit 7 - 10 January 2008, Reno, Nevada.
357
Thomas, G., Oakley, G., & Bambrey, R. (2010). An experimental study of flame acceleration and
358
deflagration to detonation transition in representative process piping. Process Saf. Environ., 88(2),
359
75-90.
ACCEPTED MANUSCRIPT 360
Valiev, D. M., Bychkov, V., Akkerman, V., & Eriksson, L. (2009). Different stages of flame
361
acceleration from slow burning to Chapman-Jouguet deflagration. Physical Review E,
362
80(0363173).
365 366 367 368 369 370
scale pipeline. Transactions of Beijing Institute of Technology, 8, 784-788. Wang C., Shu CW., Han W., Ning J. (2013). High resolution WENO simulation of 3D detonation
RI PT
364
Wang C., Hu BB, (2016). Experimental study on the explosive flame propagation of CH4-O2 in small
waves. Combust Flame. 160, 447–462.
Wang C., Zhao Y., Han W., Large eddy Simulation of Flame Acceleration and Transition from Deflagration to Detonation. 25th ICDERS August 2 – 7, 2015 Leeds, UK.
SC
363
Wang C., Zhao Y.,Zhang B. (2016). Numerical simulation of flame acceleration and deflagration-todetonation transition of ethylene in channels. J. Loss Prevent. Proc., 43:120-126. Wu, M. H., Burke, M. P., Son, S. F., & Yetter, R. A. (2007). Flame acceleration and the transition to
372
detonation of stoichiometric ethylene/oxygen in microscale tubes. Proc. Combust. Inst., 31(2),
373
2429-2436.
377 378 379 380 381 382 383 384
Xiao H, Shen X, Sun J. (2012). Experimental study and three dimensional simulation of premixed
TE D
376
mixtures. Proc. Combust. Inst. 33:2287-93.
hydrogen/air flame propagation in a closed duct. Int. J. Hydrog. Energy, 37: 11466-73. Yu, S., & Navarro-Martinez, S. (2014). Modelling of deflagration to detonation transition using flame thickening. Proc. Combust.Inst., 35(2), 1955–1961. Zbikowski M., Makarov D., Molkov V. (2008). LES model of large scale hydrogen–air planar
EP
375
Wu M, Wang C. (2011). Reaction propagation modes in millimeter-scale tubes for ethylene/oxygen
detonations: Verification by the ZND theory. Int. J. Hydrog. Energy, 33(18): 4884-4892. Zeldovich YaB, Barenblatt GI, Librovich VB, Makhviladze GM. (1985). Mathematical theory of
AC C
374
M AN U
371
combustion and explosion. New York: Consultants Bureau.
ACCEPTED MANUSCRIPT 385 386
Figure captions
387
Table 1 Physical parameters of the model system.
388
Fig. 1. Sketch of experimental apparatus: (1) pure methane, (2) pure oxygen, (3) gas mixing device, (4)
ignition electrode, (5) horizontal channel, (6) high-speed video camera, (7) pressure and
390
temperature recorder, (8) data recorder, (9) synchronization controller, (10) spark igniter.
391
Fig. 2. Comparison of the flame tip position-time evolution between experimental and numerical
results computed for different mesh, △x= 1, 0.5, 0.2 and 0.1 mm.
SC
392
RI PT
389
Fig. 3. Flame tip velocity against distance along the channel for experimental and numerical results.
394
Fig. 4. Flame acceleration and DDT occurs: (a) high speed video of experiment; (b) time interval of
395 396 397
M AN U
393
temperature contours of numerical results, △=0.5 mm.
Fig. 5. Evolution of temperature and pressure profiles corresponding to leading point along the
channel, time instants are from t=0.727 ms to 1.546 ms.
Fig. 6. Change of distance between the flame tip and the leading shock
399
Fig. 7. Flow field ahead of the flame front at t=0.727 ms (A) and 0.86 ms (B) and corresponding velocity
400
TE D
398
profile along the channel in the cross A and B.
Fig. 8. Flame structure and streamlines around the flame front at different times, t=0.86, 1.21, 1.31, 1.39ms
402
Fig. 9. Temperature and pressure contours at different times for transition to detonation: (a) the
404 405 406 407 408
temperature, and (b) the pressure.
Fig. 10. Pressure contours for local explosion and DDT.
AC C
403
EP
401
ACCEPTED MANUSCRIPT 409
Table 1 Physical parameters of the model system Notations
Values
Initial pressure
P0
33 KPa
Initial temperature
T0
Initial density
ρ0
Polytropic exponent
γ
Molecular weight
M
Preexponential factor
A
Chemical energy release Activation energy 411
AC C
EP
TE D
412
RI PT
Parameter
298 k
1.19 kg/m3 1.15
SC
0.027 kg/mol
1.21×1010 m3/Kg/s
M AN U
410
q
60 RT0/M
Ea
69 RT0
SC
RI PT
ACCEPTED MANUSCRIPT
M AN U
413 414
Fig. 1. Sketch of experimental apparatus: (1) pure methane, (2) pure oxygen, (3) gas mixing device,
415
(4) ignition electrode, (5) horizontal channel, (6) high-speed video camera, (7) pressure and
416
temperature recorder, (8) data recorder, (9) synchronization controller, (10) spark igniter.
EP AC C
418
TE D
417
ACCEPTED MANUSCRIPT 419 420 421
M AN U
SC
RI PT
422
423
Fig. 2. Comparison of the flame tip position-time evolution between experimental and numerical
425
results computed for different mesh, △x= 1, 0.5, 0.2 and 0.1 mm.
EP
427
AC C
426
TE D
424
ACCEPTED MANUSCRIPT 428 429 3000 2500
RI PT
Utip(m/s)
2000 1500
Experimental
1000
△=1 mm
△=0.2 mm
0 0
0.5
1
M AN U
x(m)
430 431
SC
500
1.5
Fig. 3. Flame tip velocity against distance along the channel for experimental and numerical results.
AC C
EP
TE D
432
ACCEPTED MANUSCRIPT 433 434 435
M AN U
SC
Local explosion
RI PT
(a)
(b)
438 439 440
Fig. 4: Flame acceleration and DDT occurs: (a) high speed video of experiment; (b) time interval of temperature contours of numerical results, △=0.5 mm.
AC C
437
EP
TE D
436
ACCEPTED MANUSCRIPT 441 442 5000
30 T p
25
RI PT
4000
T(K)
3000
SC
15
1000
0
447
0.6 x(m)
0.8
5
1
channel, time instants are from t=0.727 ms to 1.546 ms.
TE D
446
0.4
10
Fig. 5. Evolution of temperature and pressure profiles corresponding to leading point along the
EP
445
0.2
AC C
443 444
M AN U
2000
p(bar)
20
ACCEPTED MANUSCRIPT 448 449
1
RI PT
0.8
x(m)
0.6
Leading shock
0.4
Flame tip
0 1
1.1
1.3
1.4
M AN U
Time(ms)
450 451
1.2
SC
0.2
1.5
1.6
Fig. 6. Change of distance between the flame tip and the leading shock.
452
AC C
EP
TE D
453
ACCEPTED MANUSCRIPT 454 455 120
100 0 305
u(m/s)
80
0.05
0.1
x(m)
0.15
60
40
0.2
A
20
0
456
0.005
0.01 y(m)
0.015
0.02
SC
457
RI PT
A
M AN U
150
B
u(m/s)
0 308
0.08
0.12
0.16 x(m)
0.01 y(m)
TE D
corresponding velocity profile along the channel in the cross A and B.
EP
461
0.005
0.015
0.02
Fig. 7 Flow field ahead of the flame front (red line) at t=0.727 ms (A) and 0.86 ms (B) and
AC C
460
B
50
0
458 459
0.2
100
ACCEPTED MANUSCRIPT 462
SC
RI PT
463
M AN U
464
469 470 471
EP
467 468
Fig 8 Flame structure and streamlines around the flame front at different times, t=0.86, 1.21, 1.31, 1.39 ms.
AC C
466
TE D
465
ACCEPTED MANUSCRIPT 472
RI PT
473
477 478 479
Fig. 9: Temperature and pressure contours at different times for transition to detonation: (a) the temperature, and (b) the pressure.
EP
476
AC C
475
TE D
M AN U
SC
474
ACCEPTED MANUSCRIPT 480
M AN U
SC
RI PT
481
482 483
Fig. 10: Pressure contours for local explosion and DDT.
484
AC C
EP
TE D
485
ACCEPTED MANUSCRIPT
Highlights
Yongyao Zhao, Cheng Wang*, Yong Bi E-mail:
[email protected]
RI PT
LES of flame acceleration and DDT in small-scale channels
SC
(1) FA and DDT in small-scale channels are investigated with LES and experimental method.
M AN U
(2) The interaction of the flame with the flow field and the pressure wave is the main mechanism for flame acceleration.
(3) The flow ahead of the flame always keeps laminar up to the time of DDT occurs. (4) Local explosion triggers the DDT and causes the retonation wave propagates
AC C
EP
TE D
downstream.