Accepted Manuscript

Flame propagation and flashback characteristics in a kerosene fueled supersonic combustor

equipped with strut/wall combined fuel injectors

Junlong Zhang, Juntao Chang, Zi’ao Wang, Lin Gao, Wen Bao

PII: S1270-9638(19)31752-3
DOI: https://doi.org/10.1016/j.ast.2019.105303
Article Number: 105303
Reference: AESCTE 105303

To appear in: Aerospace Science and Technology

Received date: 28 June 2019

Revised date: 12 July 2019
Accepted date: 15 July 2019

Please cite this article in press as: J. Zhang et al., Flame propagation and flashback characteristics in a kerosene fueled supersonic
combustor equipped with strut/wall combined fuel injectors, Aerosp. Sci. Technol. (2019), 0, 105303,
https://doi.org/10.1016/j.ast.2019.105303

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.
1 Flame propagation and flashback characteristics in a kerosene fueled

2 supersonic combustor equipped with strut/wall combined fuel injectors

3 Junlong Zhang1, Juntao Chang2, Zi’ao Wang3, Lin Gao4, Wen Bao5

4 Harbin Institute of Technology, 150001 Heilongjiang, People’s Republic of China

5 Abstract: Flame propagation characteristics in a liquid kerosene fueled supersonic combustor were

6 observed in this paper. The strut/wall combined fuel injection scheme was used to achieve the multi-

7 staged fuel injection, and the thin strut was also adopted as the flame holder. A series of experiments

8 and numerical simulations were carried out in the condition of Ma=2.8. In order to represent the

9 flame characteristics during the flame propagation process, high-speed photography and pressure

10 measurement were taken to record the data. Based on the basic date, the variation of flame area and

11 Mach number distributions during the experimental process were analyzed, and the flowing field

12 characteristics were also discussed according to the numerical results. Results indicated that the core

13 flame would diffuse to the primary flow with the increasing of equivalence ratio, enlarging the flame

14 width. The flame flashback phenomenon of both core flame and wall flame, induced by the flame

15 propagation, was detected and analyzed. Then, the interaction between combustion and flowing in

16 the supersonic combustor was investigated, based on which, the mechanism of flame flashback

1. PhD Candidate, School of Energy Science and Engineering, junlongzhang1990@163.com

2 Professor, Academy of Fundamental and Interdisciplinary Sciences, changjuntao@hit.edu.cn.

˄Corresponding author˅

3 PhD Candidate, School of Energy Science and Engineering, 1360211958@qq.com.

4 Master Candidate, School of Energy Science and Engineering, gaolin_0527@163.com.

5. Professor, School of Energy Science and Engineering, baowen@hit.edu.cn.

1
1 phenomenon was also revealed. Numerical results showed that the generation of recirculation and

2 low-speed region in boundary layer was the inducing mechanism of flame flashback. With the

3 investigations in this paper, a depth understanding of the flame propagation process in supersonic

4 airflow was achieved, based on which, a further optimization of combustion performance could be

5 conducted.

6 Keyword: flame propagation, flame flashback, strut/wall combined fuel injection, strut flameholder

2
1 Nomenclature

A = Cross-sectional area

Cf = Wall skin friction coeƥcient

D = Hydraulic diameter

dw = Wetted perimeter

ER = Equivalence ratio

ERwall = Equivalence ratio of strut injected fuel

ERwall = Equivalence ratio of wall injected fuel

Ma = Mach number

pw = Wall static pressure, MPa

Pt = Total pressure, MPa

Tt = Total temperature, K

t = Time during the experiment, s

W = Flame width

¢ = Cone angle of the core flame, degree

2 1. Introduction

3 Combustor is the key component of a scramjet [1]-[3], and some issues about fuel injection and

4 flame holding need to be solved urgently to improve the combustor performance [4]-[7]. Combustion

5 process is organized within a supersonic flow, and rapid fuel/air mixing [9]-[10], reliable ignition [11]-

[13]
6 and stabilized combustion [14]-[16] must be realized with a short residence time of an order of

7 milliseconds, which makes it more difficult to achieve flame stabilization in scramjet combustor. In
4
1 order to improve combustor performance further, more attention should be paid on investigations

2 of flame characteristics.

3 Because of the high speed and the complex flow characteristics in supersonic combustion, the

4 combustion is turbulent burning. The complex turbulent flow will generate a large-scale vortex, that

5 assists the flame in propagating to the core flow to enhance the combustion [17]. At the same time, a

6 thermal runaway will also appear and be exacerbated in the turbulent flow, and could possibly

7 extinguish the flame. Therefore, the flame propagation characteristics in turbulent flow are the

8 critical factors for achieving efficient and stable combustion in the supersonic combustor. Mitani [18]

9 numerically investigated the flame establishment and flame propagation process in a hydrogen-

10 fueled combustion under a Mach 6 flight condition. Small flames were generated around the

11 individual fuel jets in the combustor, and a large-scale diffusion flame was produced downstream

12 in the engine. In the supersonic combustor, the flame propagated upstream to reach the backward-

13 facing steps at a speed of approximately 500 m/s. The flame propagation characteristics under

14 different fuel injection schemes were analyzed and presented by Rasmussenu [19] sing LIF of OH

15 and CH2O measurements. The flame propagation characteristics under different fuel injection

16 schemes were analyzed and presented. The structure and location of the flames were significantly

17 changed as the fuel flowrate was decreased toward lean fuel conditions. Yuan [20] considered that

18 there was an essential distinction between supersonic and subsonic combustion. The spreading of

19 the disturbances in the combustor could be faster than the speed of sound, so that the propagation

20 mechanism in supersonic flow was different from that in the subsonic flow. Flame propagation rules

21 could be directly affected by the equivalence ratio, the initial temperature and the initial pressure

[21]
22 . different types of propellants, including a hydrogen/air mixture, a methane/air mixture, and an
5
 1 ethylene/air mixture, were tested, based on the flame propagation velocity. The results indicated that

2 the flame propagation velocity increases with increasing initial temperature of the propellant, and

3 decreases with increasing pressure in the combustor. Two equivalence ratios were tested in a dual-

4 mode combustor, and combustion modes both with and without thermal choking were obtained [22].

5 The flame transition process was divided into three parts, namely the initial stage, the propagation

6 stage and the stable stage. In the propagation stage, the phenomenon of flame up-propagation was

7 captured by the high-speed camera. In the study, a new method of calculating the flame propagation

8 speed was proposed using flame image gray values, and the average propagation speed of the global

9 flame was approximately 136 m/s during the flame establishment progress under a thermal choking

10 condition.

11 In the fuel/air premixed region, because of the violent combustion, the flame propagates

12 upstream at high speed, which is defined as flame flashback. The flame flash back will induce the

13 combustion instabilities in the combustor, and a review of active control approaches in stabilizing

14 combustion systems in the aerospace industry was conducted by Zhao [23]. Flame flashback will

[24]-[26]
15 destroy the combustion stability  leading to flame blowout and inlet unstart, and will

16 significantly reduce the scramjet performance. In the cavity-based combustor, the flame propagates

17 upstream along the combustor side walls, that are low-speed zones, instead of in the core region [27].

18 O’Byrne [28] indicated that the flame flashback in the separation region was induced by the high

19 backpressure because of the heat release in the supersonic combustors. Studies indicated that the

20 flame kernel originates near the ignitor, and expands to form a flame in the cavity [11],[29]. The initial

21 flame kernel then propagates along the shear layer of the cavity, and ignites the fuel distributed

22 downstream. With the propagation of the flame, the pressure downstream also increases, leading to
6
 1 pre-combustion shock trains and the flame region moving against the stream. A global flame is

2 produced and stabilized in the combustor because of the propagation of the flame. In addition, the

3 structure and configuration of the flame holder has an effect on the fuel diffusion and flame

4 propagation process.

5 Flame propagation characteristics are mainly investigated in scramjet combustor using cavity

6 as flame holder, while there is a significant difference of flame characteristics between cavity-flame-

7 holding mode and strut-flame-holding mode. Huang did the conduction on the ram-scram transition

[30]
8 mechanism in a strut-based scramjet combustor , and the influences of the inlet boundary

9 conditions at the entrance of the isolator, the injection strategy and the jet-to-crossflow pressure

10 ratio on the mode transition have been investigated. Flame characteristics have also been widely

11 investigated in a thin strut equipped supersonic combustor [31]-[34]. In this paper, a thin strut is

12 equipped in a liquid kerosene fueled dual-mode combustor, and flame propagation characteristics

13 are experimentally studied and discussed based on pressure distributions and variation of flame area.

14 With the investigations in this paper, a depth understanding on flame oscillation in dual mode

15 combustor is achieved.

16 2. Experimental Setup

17 2.1 Test facility

18 Experiments were conducted on the directly connected scramjet combustor test rig in Harbin

19 Institute of Technology, as Fig. 1 shows. An air heater was used to simulate the supersonic incoming

20 air flow, inside which, ethanol and oxygen were burned to produce the high enthalpy inflow. Then,

7
1 the high enthalpy air flow was accelerated to Ma = 2.8 through a Laval nozzle to achieve the

2 generation of incoming flow.

1 3 4 5 6

7
3
4 Fig. 1 Photograph of the experimental system:
5 1, Heater and Laval nozzle; 2, Visualization section; 3, Reflector; 4, Pressure sensors,
6 5, Support structure of camera; 6, Model combustor; 7, Strut

7 The model combustor is placed at the duct exit of Laval nozzle. The main dimensions of the

8 experiment combustor model are shown in Fig. 2. The total length of the experimental model is

9 1480mm, with a total expansion ratio of 2.5. The first part acts as an isolator with a length of 300mm.

10 An expansion part is connected to the exit of the isolator. The third part with a constant cross section

11 area is installed at the outlet of the expansion part. An expansion part with a length of 100mm is the

12 fourth part. The fifth section is a 420 mm long duct with the constant cross section area, and the

13 sixth one is a nozzle with the length of 100mm.

8
 Laval nozzle Isolator Combustor

Ma 2.0/2.8
4.3 deg

strut plasma ignitor

A B x/mm

1 0 300 560 860 980 13801480

2 Fig. 2 Schematic of scramjet combustor and the camera field of view

3 A strut is equipped in the center of airflow, as is shown in Fig. 2. The trailing edge of the strut

4 is aligned with the exit of isolator. In order to take flame pictures at the strut back, a quartz window

5 is equipped in upper wall of the combustor. A high-speed camera is used to record flame images.

6 The thickness of the strut is 6mm with the front blockage of 6%. There are two rows of injectors in

7 the strut, named with A and B. A-injectors are fuel injector. Direction of the fuel injection is normal

8 to the leading edge of strut. Oxygen injectors are set at the tailing edge of strut. With the help of the

9 oxygen, the flame is stablished and enhanced. Fuel is ignited by a plasma ignitor mounted at the

10 strut back. An amount of nitrogen is fed into the cavity of the ignitor, ionized to produce high-

11 enthalpy plasma to achieve fuel ignition.

12 2.2 Time sequence and post-processing method

Heated air flow

PJ ignitor

Kerosene injection

Time
13 t 0 t 1 t 2 t 3 t 4 t 5

14 Fig. 3 Time sequence in the experiments

9
 1 Experiment time sequence is shown in Fig. 3. Heated airflow begins at t0, and becomes stable

2 at t1. Plasma ignitor is switched on as soon as the steady flow is established. Half a second later,

3 when the plasma torch ignitor is stable, liquid kerosene is injected into combustor. One second later,

4 at t3, plasma torch ignitor is switched off, and combustion can be sustained itself until the end of

5 experiment. To measure the static pressure, a total of 28 pressure-tap ports are distributed on the

6 center line of one side of the combustor wall. There are 28 pressure transducers mounted on the side

7 wall of the combustor, which measure the pressure within the range of 0~1MPa with the maximum

8 error of ±0.25% at the frequency of 0.5 kHz. The transducers are calibrated using least squares linear

9 regression fits. All pressure values use to generate the calibration lines within 0.5% of the

10 corresponding linear fit values.

11 To further investigate the combustor performance in experiments, solving the typical quasi-

12 one-dimensional mass, momentum, and energy conservation equations based on a 1-D analysis

13 approach is employed by reference paper [35]. In this paper, the 1-D model of combustor is based on

14 the following equations (1) for Mach number Ma. A is the cross-sectional area of combustor. The

15 wall skin friction coeƥcient Cf was estimated by the Van Driest method [35]. The hydraulic diameter

16 D is de¿ned to be 4A/dw, where dw is the wetted perimeter of duct.

†‫ܽܯ‬ ͳ ݇െͳ †‫ܽܯ݇ ܣ‬ଶ ൅ ͳ †ܶ୲ ‫ܽܯ‬ଶ †‫ݔ‬

ൌ ଶ
൬ͳ ൅ ‫ܽܯ‬ଶ ൰ ቈ െ െ Ͷ‫ܥ‬୤ ቉ (1)
‫ܽܯ‬ ‫ ܽܯ‬െ ͳ ʹ ‫ܣ‬ ʹ ܶ୲ ʹ ‫ܦ‬

17 2.3 Computational fluid dynamics method

18 In order to get the flow field characteristics, some numerical simulations were carried out in

19 the region nearby the strut.

20 In the experimental model, the strut is mounted in the center of combustor, the whole
10
 1 combustor model being shown in Fig. 4a. In order to reduce the numbers of the mesh nodes, the

2 numerical model only takes a part of the real model, as is marked with the green zone, and the whole

3 computation domain is 180 mm×40mm×3.6mm. For meshing convenience, the circular orifices A

4 and orifices B are all surrogated by square orifices. The size of the square orifices is 0.45 mm h

5 0.45mm, ensure the same area as the circular orifices in the real model. Data collection is processed

6 in a 2D plane, which is just right across the kerosene injectors and the oxygen injectors. In the

7 current study, the numerical methods for calculating flow fields refer to that of Feng [36]-[39], and the

8 calculating results of the combustor static pressure remains almost coincidence with that of the

9 experimental results.

38mm

3.6mm

(a) combustor model

0.45mm
0.45mm

Oxygen injector

Kerosene injector

10 (b) numerical simulation zone

11 Fig. 4 The combustor and the numerical simulation zone

12 The boundary of combustor inlet was set as the pressure inlet, with the stagnation state of

13 Tt=1680K, pt=1.87MPa, Ma=2.8. In the current study, a Navier-Stokes solver was provided to

11
 1 calculate the steady flow field surrounding the combustor model. The governing RANS equations

2 of continuity, momentum and energy were coupled together using the density based solver. The

3 renormalization group k–İ turbulence model was employed in this paper. A no-slip and adiabatic

4 boundary condition was imposed along the solid walls by setting the velocity components to zero

5 and nullifying the energy contributions of the wall faces to the dissipative fluxes. The viscosity and

6 thermal conductivity were evaluated using a mass-weighted mixing law.

7 3. Result and Discussion

8 Fuel is fed into the combustor both from the strut injectors and the wall injectors. The strut-

9 fuel is ignited to form the core flame, and wall flame is produced by the combustion of wall-fuel.

10 The propagation processes of both the core flame and the wall flame are analyzed and discussed in

11 this paper.

12 3.1 Core flame diffusion characteristics

13 In this section, the diffusion characteristics of core flame is discussed, and the fuel is with a

14 linearly increasing equivalence ratio from 0.25 to 0.60. As the ER rise, the core flame is enhanced

15 and diffuse to the primary flow. Fig. 5 presents the flame images with different equivalence ratio

16 captured during the whole experimental process.

12
 a.ER=0.25 b.ER=0.25 c.ER=0.30

d.ER=0.35 e.ER=0.40 f.ER=0.45

g.ER=0.50 h.ER=0.55 i.ER=0.60

2 Fig. 5 Flame images in different ER conditions

3 As is shown in the figure, the flame mainly gathers at the downstream of strut, and the leading

4 edge of flame is attached to the trailing edge of strut. In Fig. 5a, the brightest zone is the position of

5 the plasma ignitor, and the flame is still stabilized in the combustor when the ignitor is switched off,

6 as shown in Fig. 5b. In Fig. 5b ~ i, the equivalence ratio linearly increases from 0.25 to 0.60. For

7 the reason that there is a little amount pure oxygen injected into the trailing edge of strut, a stable

8 pilot flame forms in the shear layer region, and the highlight region in the flame images is the

9 position of pilot flame. With the assistance of the pilot flame, the other part of the fuel is ignited to

10 form the global flame. From Fig. 5b to Fig. 5f, along with the increasing of ER, both the flame area

11 and the flame width are enlarged, while the flame always remains in a V-shape. In the condition of

12 ER=0.50, the global flame diffuses to the side wall direction, attaching to the side wall of the

13 combustor at the rightest side of the view. As soon as attaching to the side wall of the combustor,

14 the global flame will diffuse to the upstream of the main airflow along the side wall since there is a

15 boundary layer with a lower velocity in this region. From Fig. 5g to Fig. 5i, a significant flame up-

16 propagation is observed by the flame image.

17 During the experimental process of ER increasing, the core flame diffuses to the direction of

13
1 combustor sidewall and the upstream of the main airflow, resulting in the variation of flame shape.

2 In order to get the flame diffusion characteristics in detail, two variables, namely the flame width

3 and the flame cone angle, are defined, among which the flame width is the width of the flame

4 perpendicular to the direction of the main airflow, denoted by the symbol W, and the flame cone

5 angle is the angle between the boundary of the flame, denoted by the symbol ¢. To further study

6 the flame characteristics, some post processing of the flame images has been done. The color image

7 is transformed into gray pictures, which is expressed as a two-dimensional matrix with gray value.

8 Fig. 6 presents the flame image and the post processing result of the image in the experimental

9 condition of ER=0.60, among which, Fig. 6a shows the raw flame image and the flame boundary.

10 The flame image is divided into two parts, namely the upper part and the lower part, and the

11 boundary between these two parts is the axis of the strut. The flame boundaries in these two parts

12 are not fully symmetrical for the axis. The pilot flame area of the upper part is larger than that of the

13 lower part. In the upper part of the flame, the boundary of global flame near the position of x=300

14 diffuses to the main airflow since the effect of a stronger pilot flame, that proves that the strength of

15 pilot flame will deeply influence the diffusion process of global flame.

14
 150
Upper part of flame
100
strut
50 Lower part of flame
x=250 x=300 x=350x=400 x=450

y/pixel
0 100 200 300 400 500
150
pilot flame
100 Upper part of flame
strut
50 Lower part of flame
global flame
0 100 200 300 400 500
x/pixel
a. flame boundary and flame image
300 Lower part of flame Upper part of flame
250
x=250
Gray value

200 x=300
150 x=300
x=400
100 x=450
50 W
0 20 40 60 80 100 120 140
y/pixel
1 b. gray value distributions in different x-position

2 Fig. 6 Flame characteristics in the condition of ER=0.60

3 In order to analyze the flame diffusion process along the direction of the main airflow, five

4 typical positions (x=250, 300, 350, 400, 450), as marked in the flame image in Fig. 6a, are chosen

5 to reveal the flame width distributions in different locations. Fig. 6b is the gray value distribution in

6 different x position. The gray value of flame in the pilot flame region is the largest, and decreases

7 accompanied with the flame diffusing to side wall of combustor. Delimiting that the pixel belongs

8 to the flame area if the its gray value exceeds 20, so the boundary of the pixel in the flame area

9 could express the flame width. Along the flowing direction, the peak of gray value gradually

10 decreases. In the lower part of flame, the flame width also increases accompanied with x increasing.

11 While, in the upper part of flame, the variations of gray value in x=300~450 make an insignificant

12 change due to the global flame diffusion near the region of x=300. In summary, the flame diffuses

13 to the direction perpendicular to the main flowing with a non-symmetry characteristic, resulting in
15
 1 the enlargement of flame width.

2 Fig. 7 shows the variation of flame width with ER increasing in different x position. With the

3 increasing of ER, the flame width also gradually grows, while the growth trends are varied in

4 different x position. In the position of x=250, both the oscillation amplitude and the value of the

5 flame width are the least among all the four positions. The oscillation amplitude in this position

6 nearly keeps in a constant value during the whole experimental process. In the position of x=300,

7 the flame width is enlarged. In the initial stage of the experiment (0.25<ER<0.48), the flame width

8 keeps in an oscillation status with a small amplitude. When ER increases to 0.49, a large amplitude

9 oscillation appears in this position, and the oscillation amplitude is nearly two times of that in the

10 initial stage. The flame instability results in the strong oscillation in these ER conditions. In the

11 position of x=350, the flame width and the oscillation amplitude of it increases further. The flame

12 width begins to transform from large amplitude status to the small amplitude status in the condition

13 of ER=0.36. Experimental results show that the flame oscillation status is associated with the flame

14 instability. In the downstream of the main flowing, the flame oscillation is distinctly stronger than

15 that in the upstream of the main flowing, that proves that the flame instability forms in the

16 downstream firstly, then propagates to the upstream.

140

120 x=250
x=300
x=350
100 x=450
W/pixels

80

60

40

20
0.25 0.30 0.35 0.40 0.45 0.50 0.55 0.60
17 ER

16
 1 Fig. 7 Variations of flame width with ER increasing in different x position

30
e.ER=0.40
25

20 ¢

¢/degree
15

10

0
0.25 0.30 0.35 0.40 0.45 0.50 0.55 0.60
2 ER

3 Fig. 8 Variations of flame cone angle with ER increasing

4 Fig. 8 shows the variation of flame cone angle in different ER conditions. In the initial stage

5 (ER=0.25), the average value of the core angle is about 7.5degrees, and it rises with the ER

6 increasing. In the condition of ER=0.55~0.60, the average value of the core angle keeps in the value

7 of about 19degree, with an insignificant change along with the ER increasing. In this condition

8 (ER=0.55~0.60), the core flame has attached to the sidewall of the combustor as shown in Fig. 5, so

9 that the core flame changes little in the later experimental process.

10 The flame is stabilized in the combustor under the effect of strut and the injection of pure

11 oxygen, then the diffusion of core flame results in the flame oscillation in the combustor. With the

12 increasing of equivalence ratio, flame diffuses to the primary flowing, and the flame width and the

13 oscillation amplitude of it grows gradually.

14 3.2 Flame flashback characteristics

15 3.1.1. Core flame flashback characteristics

16 With the enhancement of combustion, the core flame not only diffuses to the side wall of

17
 1 combustor, but also propagates to the upstream of the primary flowing. When ER increases to a

2 certain extent, the unsteady flame propagation against the incoming airflow will occur in the main

3 combustion region, resulting in the phenomenon of flame flashback. Fig. 9 shows the typical period

4 of flame images, which demonstrates the details of the core flame flashback process.

5 In this experimental condition, the equivalence ratio maintains in a constant value of 0.7, and

6 in the flame establishment and flame stabilization process, the flame is in the unstable stage. In Fig.

7 9, flame images in three typical periods are presented. In the initial stage of period-a, the flame

8 gathers at the strut back, then the boundary of the core flame diffuses to the combustor sidewall,

9 enlarging the flame width. Then, in the trailing edge of the strut, the flame begins to propagate to

10 the upstream along the strut sidewall. As the experiment proceeds, the flame gradually propagates

11 upstream. While, in this period, the core flame could not maintain in the flashback status, and it will

12 be blown off from the sidewall of the strut. In period-b, the core flame also suffers the process of

13 flame up-propagation, and the strength of core is enhanced. At the ending of this period, the leading

14 edge of the core flame nearly reaches the position of fuel injectors. With the development of the

15 experimental process, the strength of core flame gradually enhanced. In period-c, the brightest

16 region of the core flame gathers in the recirculation region at the strut back, and the thermal release

17 is concentrated in this region. The thermal blockage near the strut installed region is larger in this

18 period. The flame propagates to the upstream along the strut sidewall firstly, and the core flame

19 could be maintained in this low velocity region. Then, core flame diffuses to the direction of

20 combustor sidewall, distributing in the main airflow. Under the influence of the flame diffusion, the

21 core flame could propagate upstream further in the combustor. By the analysis of the core flame

22 characteristics, we could find that the core flame flashback could be divided into two stages. In the
18
1 first stage, the core flame propagates upstream in the low velocity region formed at the sidewall of

2 the strut, and the leading edge of the core flame is still in the downstream of the fuel injectors. In

3 the second stage, the core flame propagates upstream further, and diffuses to the main airflow in the

4 combustor. The core flame in this stage also forms in the upstream of the fuel injectors.

t increasing
452ms 619ms 788ms
t increasing

453ms 620ms 789ms

5 period- a period-b period-c

6 Fig. 9 The phenomenon of core flame flashback

7 3.1.2. Wall flame flashback characteristics

8 The fuel injection scheme could also make a deep influence in the flame flashback

9 characteristics [24], and Sun found that when the fuel injection upstream of the cavity flameholder

10 produced a premixed region with sufficiently high global equivalence ratio, a rapid flame flashback

11 occurred against the incoming supersonic flow. In the strut/wall combined fuel injection scheme, a

12 part of fuel is also injected into the combustor from the wall injectors, and the flame flashback

13 characteristics in this condition is also different from that in the strut-only fuel injection condition.

14 The core flame propagates to the upstream of the incoming airflow along the side wall of strut. The

19
 1 propagation characteristics of wall flame are quite different from that of core flame, due to the

2 difference of flame stabilization mechanism between these two kinds of flame. An experiment was

3 conducted with the strut/wall combined fuel injection scheme. A part of fuel was fed into the

4 combustor from the strut injectors, with ERstrut=0.47, and the rest part of fuel was fed into the

5 combustor from the wall injectors arranged at the same axial position with strut injectors, with

6 ERwall=0.25. The flame images during the experimental process were captured by the high-speed

7 camera, and Fig. 10 shows the typical periods of flame images, which demonstrates the details of

8 the wall flame flashback process.

t increasing
232ms 305ms 361ms
t increasing

233ms 306ms 362ms

9 period- a period-b period-c

10 Fig. 10 The phenomenon of wall flame flashback

11 In this experimental condition, core flame generates at the strut back firstly, then diffuses to

12 the direction of the combustor sidewall. For the reason that there is not the ignitor and flame holder

13 near the side wall of the combustor, the wall fuel cannot achieve the ignition and flame stabilization.

14 The core flame plays the role of ignitor to ignite the wall-fuel to form the wall flame. During the

15 experimental process, the wall fuel begins to be injected into the combustor after a stable core flame

16 is formed. In period-a, the core flame reaches the side wall in the rightest side of the view, with the
20
1 effect of which, the wall-fuel is successfully ignited in this region. The wall fuel is not ignited as

2 soon as injected into the combustor, and the wall flame forms in the downstream of the combustor

3 firstly since the width of the core flame in this region is the largest. The flowing velocity near the

4 combustor sidewall is significantly lower than that in the main flow, so that the wall flame could

5 easily propagate to the upstream. The separation zone also forms near the combustor wall under the

6 effect of the thermal release of wall flame. In the separation zone, the flowing velocity is further

7 decreased, with is beneficial for the up-propagation of wall flame. In period-a, the wall flame begins

8 to propagate upstream, but does not reach the trailing edge of strut. In period-b, the wall flame

9 propagates upstream further, and the flame gradually reaches the trailing edge of the strut. In this

10 period, the main combustion region is located in the region near the strut, and the wall flame begins

11 to propagate to the upstream of the strut, forming the wall flame flashback phenomenon. In period-

12 c, an obvious flame flashback phenomenon generates in the combustor. The wall-fuel is ignited as

13 soon as injected into the combustor, forming the wall flame near the combustor sidewall. The wall

14 flame diffuses to the primary flow to enhance the core flame. In addition, the wall flame also

15 propagates to the upstream of the wall fuel injectors, and the wall flame attaches to the leftist side

16 of the view. The wall flame could not maintain in the upstream, and it will be blown off. Then, the

17 wall flame will propagate upstream again, forming a flame oscillation in this stage.

18 During the experimental process, the flame shape is in an unsteady state, although the

19 equivalence ratio keeps in a constant value. The thermal release resulting from the combustion

20 process is nonlinearity, so that the back pressure in the combustion region is also in an unstable

21 status. The unstable back pressure leads to a variation of flowing conditions, further has an impact

22 on the flame shapes. In addition, the thermal release will produce the high-temperature gas in the
21
 1 main combustion region, and the high-temperature gas will be entrained to the upstream in the low-

2 speed boundary layer near the combustor sidewall, resulting in the flame flashback phenomenon. In

3 the supersonic combustor, the flame characteristics are not only affected by the combustion process

4 but also the flowing characteristics. The interaction between combustion and flowing should be paid

5 attention on.

6 3.3 Interaction between combustion and flowing

7 In supersonic combustor, thermal release resulting from combustion will affect the flowing

8 characteristics, and flowing state of the supersonic airflow will have a deep influence on combustion

9 process conversely. The interaction between combustion process and flowing state is a complex

10 issue.

11 In order to study the interaction between combustion process and flowing state, an experiment

12 was conducted with strut/wall combined fuel injection scheme. In the initial stage of the experiment,

13 fuel was all fed into combustor only from the strut injectors, with ERstrut=0.25, and the process lasted

14 one second. One second later, the fuel with variable ER was fed into combustor, with a linearly

15 increasing ERwall from 0.25 to 0.60 within 1.5s. At the same time, wall fuel was also fed into the

16 combustor from the injectors arranged at the same axial position with the strut, with ERwall=0.42.

17 The combustion process will lead to a thermal blockage, which may induce the pressure rising in

18 the main combustion region. If the strength of the combustion is large enough, the wall pressure

19 rising will propagate to the upstream of the main airflow, resulting in the pressure rising of the 1st

20 pressure sensor. It is to say that the back pressure has influenced the parameter in the entrance of

21 combustor, which is called unstable combustion mode in this paper and is an abnormal operating

22
 1 status of combustor.

0.4
ER increasing a. ER=0.25
b. ER=0.50
0.3 c. ER=0.72
d. ER=1.02

pwall/MPa
0.2

0.1

fuel off
0
500 x/mm 1000 1500
2

3 Fig. 11 Pressure distribution variation during the experimental process

3.0
a. ER=0.25
2.5 b. ER=0.50
c. ER=0.72
2.0
Ma

1.5

1.0
Ma=1
0.5
0 500 x/mm 1000 1500
4

5 Fig. 12 Mach number distributions in different ER conditions

6 Fig. 11 shows the wall pressure variation during the experimental process with different

7 equivalence ratio distribution schemes, and Fig. 12 presents the Mach number distributions in

8 different ER conditions. In the initial stage of the experimental process, the fuel is fed into the

9 combustor through the strut injectors, with an equivalence ratio of 0.25. The main combustion

10 region is located in 375mm<x<975mm, and the maximum value of the wall pressure is 0.15MPa.

11 The strut is placed in the position of x=246~300mm, and the main combustion region is in the

12 downstream of the strut in this experimental condition. As is shown in Fig. 12, the Mach number

13 decreases to a certain extent in the main combustion region, but the airflow always keeps in the
23
 1 supersonic status. With the increasing of ER, the wall pressure gradually rises up, and the main

2 combustion region is also enlarged. In the experimental condition of ER=0.50, the pressure rising

3 propagates to the upstream of the main flowing, and the pressure in the strut equipment region also

4 begins to raise up. Under the effect a larger thermal release, the Mach number in this condition

5 further reduces, and is in the subsonic states in the region between x=515m and x=800mm. In this

6 condition, the flame up-propagation results in the thermal blockage in the combustor duct, causing

7 the deceleration of the main flowing. In the condition of ER=0.72, the pressure rising has crossed

8 the strut region, and an obvious pressure rising forms in the isolator (50mm<x<300mm). The Mach

9 number distributions in the main combustion region are all in the subsonic status. Although there is

10 not the fuel injection in the isolator, the result of interaction between combustion and flowing creates

11 the pressure rising in this region. The condition of ER=0.72 is a critical state, in which condition,

12 the back pressure resulting from the combustion has propagated upstream to make an influence on

13 the parameter in the entrance of combustor. In the later experimental process, the pressure detected

14 by the 1st pressure sensor continues to grow along with the increasing of ER, and the combustor

15 turns into unstable combustion mode.

16 The flame images during the experimental process are also captured by the high-speed camera

17 with the camera parameter of 5000 f/s, as is shown in Fig. 13. The whole experimental process is

18 divided into three stages according to the flame status and the pressure distribution characteristics,

19 namely the core flame stage, the interaction stage between combustion and flowing, and the unstable

20 combustion stage. In the core flame stage, the fuel is all injected into the combustor through strut

21 injectors with the ER=0.25, corresponding to the status in Fig. 11a. In this stage, the core flame

22 generates downstream of the strut, gradually diffusing to the main airflow. While, the flame does
24
 1 not propagate to the upstream of the airflow. In the interaction stage between combustion and

2 flowing, the core flame begins to propagate to the upstream of the strut with the increasing of ER.

3 In Fig. 13b, the flame propagates to the upstream of the main flowing along the sidewall of the strut.

4 With the further increasing of ER in Fig. 13c, the flame gradually propagates to the upstream of the

5 injectors, and an obvious flame flash phenomenon forms in this condition, corresponding to the

6 pressure up-propagation process in Fig. 11c. The strut-equipped region is filled with flame, and the

7 main combustion region is also enlarged. In Fig. 13d, the flame has already reached the leftmost

8 side of the view, affecting the airflow characteristics in the entrance of the combustor, and the

9 combustion is transformed into an unstable combustion stage. In this stage, the Mach number

10 distributions are all less than one in the whole combustor duct, accompanying with a large

11 combustion oscillation.

a.

Core flame stage

b.

c.

Interaction stage between combustion and flowing

d.

Unstable combustion stage

12

13 Fig. 13 Coupling effect of combustion and flow in the process of combustor operation

14 By the analysis of the results in Fig. 11~Fig. 13, the characteristics of interaction between

15 combustion and flowing will be discussed and revealed. In Case c (ER=0.72), the Mach number in

16 the main combustion region is less than 1, as shown in Fig. 12, that proves that the mode transition

17 phenomenon has occurred in the experimental process with a linearly increasing ER. The mode
25
 1 transition process is corresponded with the flame oscillation and flame flashback presenting in Fig.

2 13. In the experimental process of ER linearly increasing, the back pressure resulting from the

3 combustion gradually grows in the main combustion region. With the thermal blockage of back

4 pressure on the incoming airflow, the shock train generates in the isolator to compress the incoming

5 flowing, under the effect of which, the velocity of the airflow is slowed down. During the mode

6 transition process, the airflow turns from supersonic status into subsonic status, and the combustion

7 characteristic also varies accompanied with the mode transition process. In the subsonic airflow, the

8 combustion is enhanced since the slowing down of velocity and the growth of the static temperature

9 in the main combustion region. Under the effect of the enhanced combustion, the Mach number in

10 the combustor decreases further. The interaction between combustion and flowing causes the

11 unstable variation of the flame image shown in Fig. 13b~c. In Fig. 13d, the combustion is in the

12 unstable stage, and the Mach number distributions in the isolator are all in the subsonic status. In

13 this condition with a lower flowing velocity and higher static temperature, both the chemical

14 reaction rates and the mixing efficiency of the fuel are improved. The flame flashback is induced by

15 the decreasing of flowing velocity, which proves that the flowing characteristics will have a deep

16 influence on the flame propagation process.

17 The strut/wall combined fuel injection scheme is also adopted to study the interaction between

18 combustion and flowing, and numerical simulation is used to get the flowing field characteristics in

19 different strut/wall fuel injection scheme. Fig. 14 and Fig. 15 show the numerical results of the Mach

20 number distributions and the density gradient distributions, respectively. In the numerical results,

21 Case a is in the numerical condition without fuel injection, Case b being the numerical condition of

22 ERstrut/ERwall = 0.3/0.15, Case c being the numerical condition of ERstrut/ERwall = 0.3/0.3. Fig. 14a
26
 1 presents the numerical results of the flowing field characteristics without fuel injection. In the

2 supersonic combustor, there is a local subsonic region, generating in the recirculation region at the

3 strut back under the effect of strut. In Fig. 15a, an oblique shock wave forms in the leading edge of

4 the strut, and an expansion wave emerges at the trailing edge of the strut. The oblique shock wave

5 reflects to the sidewall of the combustor, causing the separation of the boundary layer to form a

6 local subsonic region at the position of x=0.155m. A low-speed recirculation region also generates

7 at the strut back, which is in the position of the subsonic region shown in Fig. 14a. Corresponding

8 to the experimental results of flame image, the initial flame generates in the subsonic region firstly.

9 Fig. 14b and Fig. 15b are the numerical results of Case b, and in this condition, both the core flame

10 and wall flame are established in the combustor. In the main combustion region, the airflow is

11 transformed from supersonic status to subsonic status. The wall-fuel will also make a disturbance

12 to the main flowing, forming a shock wave at the position of wall fuel injectors, as is shown in Fig.

13 15b. Under the effect of the shock waves and the combustion impacts, a large scale of boundary

14 layer separation generates in the location of x=0.145m, in which position, the airflow velocity is

15 lower than that in the primary flow. In the lower-velocity subsonic region, the wall flame flashback

16 is easily induced. Meanwhile, core flame also propagates to the upstream along the strut sidewall.

17 In this numerical condition, the flame flashback is triggered by the flowing characteristics

18 influenced by the shock waves and boundary layer separation. The ERwall is further increased to 0.30,

19 with an enhancement of the wall flame in Case c. In Fig. 14c, the area of subsonic region is made a

20 further increment since the strength of the wall flame is increased in this numerical condition. The

21 subsonic region diffuses to the upstream of the main flowing, reaching the position of wall fuel

22 injectors, that proves that the wall-fuel is successfully ignited as soon as injected into the combustor.
27
1 The core flame also propagates upstream further, although the ERstrut in Case c is the same with that

2 in Case b. In this numerical condition, the flame propagation and combustion enhancement make a

3 deep influence on the flowing process, enlarging the subsonic region and inducing the boundary

4 layer separation, which will all enhance the combustion process further.

Case a

Case b

Subsonic region

Case c

5 Supersonic region Subsonic region

6 Fig. 14 Mach number distributions in different numerical conditions

Case a

Case b

Wall injectors

Case c

Wall injectors

7
8 Fig. 15 Density gradient distributions in different numerical conditions

28
 1 By the analysis of the experimental and numerical results, the interaction between combustion

2 and flowing in supersonic combustor is obtained preliminary. The local subsonic region in the

3 combustor is essential for the establishment of initial flame. The diffusion and propagation of flame

4 will affect the flowing field characteristic, inducing the boundary layer separation and enlarging the

5 subsonic region, that will enhance the flame propagation process further. In brief, the flowing

6 characteristics and combustion process in supersonic combustor are coupled and interacted with

7 each other.

8 3.4 Mechanism of flame flashback

9 The flame flashback phenomenon of both core flame and wall flame were detected and

10 analyzed in this paper, and results proved that the phenomenon was associated with the interaction

11 between combustion and flowing. In this section, the mechanism of flame flashback will be

12 discussed and revealed.

Ma=2.8 Recirculation region Wall fuel injectors

13

14 Fig. 16 Streamline distributions under combustion heat release and wall fuel injection

15 Fig. 16 is the streamline distributions under the combustion hear release near the region of fuel

16 injectors. The combustion heat release causes a back pressure and thermal blockage in the main

17 combustion region. For the thermal blockage of the combustion, the shock trains generate in the

18 duct of combustor. Under the influence of the back pressure and thermal blockage, the separation

29
1 zone forms in the boundary layer near the combustor wall, and the back pressure propagates to the

2 upstream of the main flowing to form the recirculation region in the separation zone. In the position

3 of fuel injectors, there are also two small-scaled recirculation regions generating near the combustor

4 wall. The flowing velocity near the combustor is lower than that in the primary flowing, and the

5 flame and the pressure rising could propagate to the upstream along the combustor sidewall. The

6 wall-fuel is also entrained into the recirculation region. In the recirculation region, a part of fuel

7 diffuses to the upstream of the wall injectors, and this part of fuel is ignited to form the wall flame.

8 Under the effect of the back pressure and the recirculation region, wall flame appears in the upstream

9 of the main airflow and wall fuel injectors, forming the phenomenon of wall flame flashback.

10 The core flame flashback was divided into two stages. In the initial stage, the core flame

11 propagates upstream along the sidewall of strut. In the later stage, the core flame propagates to the

12 upstream of the leading edge of the strut and even diffuses into the primary flow. Fig. 17 shows the

13 flowing field characteristics in the initial flame flashback stage. The combustion product

14 distributions could present the core flame characteristics, and Fig. 17a is the basic shape of the core

15 flame. The core flame generates at the strut back, and begins to propagate upstream in this condition.

16 As is shown in Fig. 17b, the Mach number at the strut back and near the strut sidewall is obvious

17 lower than that in the primary flow. In the lower velocity region, the propagation velocity of flame

18 is larger than the flowing velocity of airflow, so that the core flame could propagate to the upstream

19 in this region. The core flame maintains in the position that the flowing velocity matches well with

20 the flame propagation velocity, and the generation of lower velocity region induces the core flame

21 flashback in the initial stage.

30
 mH2O
0 0.35 0.70

a. Combustion product distributions

Ma
0 1.5 3.0

1 b. Mach number distributions

2 Fig. 17 Flowing field characteristics in the initial flame flashback stage

3 In the later stage, the core flame has propagated to the upstream of the fuel injectors, and the

4 mechanism of core flame flashback is also quite different from that in the initial stage. Fig. 18 shows

5 the fuel distributions in different experimental conditions, among which, Fig. 18a presents the fuel

6 distributions without combustion in the combustor. The fuel is blown downstream along the strut

7 sidewall under the effect of the high-speed airflow. As is shown in Fig. 18a, the penetration depth

8 of the fuel is not large enough to make the fuel diffuse to fill the combustor duct, and the diffusion

9 boundary of the fuel is only about 15mm in the direction perpendicular to the primary flow. By

10 analyzing the side view of the fuel distribution, we could also find that the fuel is mainly arranged

11 at the downstream of the fuel injectors, and there is no fuel diffusing to the upstream of the strut.

12 Fig. 18b shows the flame images of the core flame flashback process, and during the process, the

13 core flame has already propagated to the upstream of the fuel injectors. The flame area is also larger

14 than the area of fuel distribution. When the core flame is established in the combustor, the thermal

15 release and pressure rising generate in the main combustion region, under the effect of which, the

16 fuel is forced to diffuse to the primary flowing, ignited to form the flame in this region. In addition,

31
 1 an adverse pressure gradient also forms near the fuel injectors. The fuel is entrained to the upstream

2 of the fuel injectors due to the adverse pressure gradient, so that the core flame could be established

3 in the upstream of the strut, causing the core flame flashback phenomenon.

Top view

Side view without fuel

Strut Fuel boundary

Side view with fuel

Strut

4 a. fuel boundary without combustion b. flame images and fuel boundary

5 Fig. 18 Fuel entrainment process under the effect of combustion

6 4. Conclusions

7 Flame propagation process could make a deep influence in the flame stabilization and

8 combustion performance. Flame diffusion and propagation characteristics in a strut/wall combined

9 fuel injection combustor were experimentally and numerically investigated in this paper, and the

10 methods of high-speed photography and pressure measurement were used to record the basic data.

11 The main conclusions are as follows:

12 1. The flame diffusion and propagation phenomenon are discussed. With the increasing of

13 equivalence ratio, the core flame gradually diffuses to the direction of the combustor sidewall,

14 resulting in the enlargement of flame width. Both the core flame and the wall flame could propagate

15 to the upstream of the fuel injectors to cause the flame flashback phenomenon.

16 2. The interaction between combustion and flowing is experimentally and numerically analyzed.

32
 1 The diffusion and the propagation of flame will affect the flowing field characteristic, inducing the

2 boundary layer separation and enlarging the subsonic region, that will enhance the flame

3 propagation process further. The flowing characteristics and combustion process in supersonic

4 combustor are coupled and interacted with each other.

5 3. The flashback mechanism of both core flame and wall flame is revealed. By the analysis of the

6 numerical results in flowing field characteristics, we could find that the boundary layer separation

7 and the recirculation region generation near the combustor wall was critical important, under the

8 effect of which, the phenomenon of wall flame flashback is induced. The core flame flashback is

9 caused by the generation of the low-speed region near the strut wall and the fuel entrainment

10 resulting from the thermal release near the fuel injectors.

11 Acknowledgments

12 This research work is supported by the National Natural Science Foundation of China (Grants

13 No. 91741123 and No.51676204).

14 References

15 [1] Wang Z, Wang H, Sun M, Review of cavity-stabilized combustion for scramjet applications,

16 Proceedings of the Institution of Mechanical Engineers, Part G: Journal of Aerospace

17 Engineering 228(14) (2014) 2718-2735.

18 [2] Vanyai T, Bricalli M, Brieschenk S, Boyce R, Scramjet performance for ideal combustion

19 processes, Aerospace Science and Technology 75 (2018) 215-226.

20 [3] Chang J, Zhang J, Bao W, Yu D, Research progress on strut-equipped supersonic combustors

33
1 for scramjet application, Progress in Aerospace Sciences 103 (2018) 1-30.

2 [4] Wang H, Wang Z, Sun M, Wu H, Combustion modes of hydrogen jet combustion in a cavity-

3 based supersonic combustor, International Journal of Hydrogen Energy 38(27) (2013) 12078-

4 12089.

5 [5] Zhu S, Xu X, Yang Q, Jin Y, Intermittent back-flash phenomenon of supersonic combustion

6 in the staged-strut scramjet engine, Aerospace Science and Technology 79 (2018) 70-74.

7 [6] Tian Y, Yang S, Le J, Study on flame stabilization of a hydrogen and kerosene fueled

8 combustor, Aerospace Science and Technology 59 (2016) 183-188.

9 [7] Sun M, Geng H, Liang J, Wang Z, Flame characteristics in supersonic combustor with

10 hydrogen injection upstream of cavity flameholder, Journal of Propulsion and Power 24(4)

11 (2008) 688-696.

12 [8] Chen S, Zhao D, RANS investigation of the effect of pulsed fuel injection on scramjet HyShot

13 II engine, Aerospace Science and Technology 84 (2019) 182-192.

14 [9] Kummitha O, Pandey K, Gupta R, Optimization of scramjet performance with different fuel

15 injection techniques and flame holder cavities, Acta Astronautica 152 (2018) 908-919.

16 [10] Huang W, Liu W, Li S, Xia Z, Liu J, Wang Z, Influences of the turbulence model and the slot

17 width on the transverse slot injection flow field in supersonic flows, Acta Astronautica 73

18 (2012) 1-9.

19 [11] Sun M, Gong C, Zhang S, Liang J, Liu W, Wang Z, Spark ignition process in a scramjet

20 combustor fueled by hydrogen and equipped with multi-cavities at Mach 4 flight condition,

21 Experimental Thermal and Fluid Science 43 (2012) 90-96.

22 [12] Li F, Yu X, Tong Y, Shen Y, Chen J, Chen L H, Chang X Y, Plasma-assisted ignition for a

34
1 kerosene fueled scramjet at Mach 1.8, Aerospace Science and Technology 28(1) (2013) 72-78.

2 [13] Wang Y, Wang Z, Sun M, Wang H, Effects of auto-ignition on combustion characteristics in

3 a hydrogen-fueled dual-mode scramjet combustor, Acta Astronautica 153 (2018) 154-158.

4 [14] Tian Y, Xiao B, Zhang S, Xing J, Experimental and computational study on combustion

5 performance of a kerosene fueled dual-mode scramjet engine, Aerospace Science and

6 Technology 46 (2015) 451–458.

7 [15] Xiao B, He C, Xing J, Zhang S, Experimental and numerical investigations of freejet and

8 direct-connect dual-mode scramjet, Aerospace Science and Technology 75 (2018) 297-303.

9 [16] Shin J, Sung H, Combustion characteristics of hydrogen and cracked kerosene in a DLR

10 scramjet combustor using hybrid RANS/LES method, Aerospace Science and Technology 80

11 (2018) 433-444.

12 [17] Schetz J, Injection and mixing in turbulent flow, In Vol. 68 of Progress in Astronautics and

13 Aeronautics, AIAA 1980.

14 [18] Mitani T, Kouchi T, Flame structures and combustion efficiency computed for a mach 6

15 scramjet engine, Combustion and Flame 142(3) (2005) 187-196.

16 [19] Rasmussen C, Dhanuka S, Driscoll J, Visualization of flameholding mechanisms in a

17 supersonic combustor using PLIF, Proceedings of the Combustion Institute 31(2) (2007) 2505-

18 2512.

19 [20] Yuan S, The discussion on supersonic combustion, Science in China (Series A) 28(8) (1998)

20 735-741. In Chinese.

21 [21] Fan Z, Liu W, Sun M, The research on flame speed at high temperature, Journal of Projectiles,

22 Rockets, Missiles and Guidance 31(3) (2011) 105-107. In Chinese.

35
1 [22] Hu J, Bao W, Chang J, Flame transition in dual-mode scramjet combustor with oxygen piloted

2 ignition, Journal of Propulsion and Power 30(4) (2014) 1103-1107.

3 [23] Zhao D, Lu Z, Zhao H, Li X, Wang B, Liu P, A review of active control approaches in

4 stabilizing combustion systems in aerospace industry, Progress in Aerospace Sciences 97

5 (2018) 35-60.

6 [24] Sun M, Cui X, Wang H, Bychkov V, Flame flashback in a supersonic combustor fueled by

7 ethylene with cavity flameholder, Journal of Propulsion and Power 31(3) (2015) 976-981.

8 [25] Wang Z, Sun M, Wang H, Yu, Liang J, Zhuang F, Mixing-related low frequency oscillation of

9 combustion in an ethylene-fueled supersonic combustor, Proceedings of the Combustion

10 Institute 35(2) (2015) 2137-2144.

11 [26] Wang H, Sun M, Qin N, Wu H, Wang Z, Characteristics of oscillations in supersonic open

12 cavity flows, Flow, turbulence and combustion 90(1) (2013) 121-142.

13 [27] Mathur T, Gruber M, Jackson K, Donbar J, Donaldson W, Jackson T, Billig F, Supersonic

14 combustion experiments with a cavity-based fuel injector, Journal of Propulsion and Power

15 17(6) (2001) 1305-1312.

16 [28] O'Byrne S, Stotz I, Neely A, Boyce R, Mudford N, Houwing F, OH PLIF imaging of

17 supersonic combustion using cavity injection, AIAA Paper 2005-3357.

18 [29] Wang H, Wang Z, Sun M, Qin N, Large eddy simulation of a hydrogen-fueled scramjet

19 combustor with dual cavity, Acta Astronautica 108 (2015) 119-128.

20 [30] Huang W, Yan L, Numerical investigation on the ram–scram transition mechanism in a strut-

21 based dual-mode scramjet combustor, International Journal of Hydrogen Energy 41(8) (2016)

22 4799-4807.
36
1 [31] Zhang J, Chang J, Ma J, Zhang C, Bao W, Investigation of flame establishment and

2 stabilization mechanism in a kerosene fueled supersonic combustor equipped with a thin strut,

3 Aerospace Science and Technology 70 (2017) 152-160.

4 [32] Zhang J, Chang J, Ma J, Wang Y, Bao W, Investigations on flame liftoff characteristics in

5 liquid-kerosene fueled supersonic combustor equipped with thin strut, Aerospace Science and

6 Technology 84 (2019) 686-697.

7 [33] Zhang J, Chang J, Ma J, Zhang Y, Bao W, Local and global flame characteristics in a liquid

8 kerosene fueled supersonic combustor equipped with a thin strut, Aerospace Science and

9 Technology 76 (2018) 49-57.

10 [34] Zhang J, Chang J, Qiu H, Kong C, Bao W, Flame oscillation characteristics in a kerosene

11 fueled dual mode combustor equipped with thin strut flameholder, Acta Astronautica 161

12 (2019) 222-233.

13 [35] ER Van Driest, Turbulent boundary layer in compressible fluids, Journal of the Aeronautical

14 Sciences 18(3) (1951) 145-160.

15 [36] Feng S, Chang J, Zhang J, Zhang C, Bao W, Numerical and experimental investigation of

16 improving combustion performance of variable geometry dual-mode combustor, Aerospace

17 Science and Technology 64 (2017) 213-222.

18 [37] Feng S, Chang J, Zhang C, Wang Y, Ma J, Bao W, Experimental and numerical investigation

19 on hysteresis characteristics and formation mechanism for a variable geometry dual-mode

20 combustor, Aerospace Science and Technology 67 (2017) 96-104.

21 [38] Feng S, Chang J, Zhang Y, Zhang C, Wang Y, Bao W, Numerical studies for performance

22 improvement of a variable geometry dual mode combustor by optimizing deflection angle,

37
1 Aerospace Science and Technology, 68 (2017) 320–330.

2 [39] Zhang C, Chang J, Feng S, Ma J, Zhang J, Bao W, Pressure rising slope variation

3 accompanying with combustion mode transition in a dual-mode combustor, Aerospace Science

4 and Technology 68 (2017) 370–379.

38