Experimental Study on Transient Ignition Characteristics of Acoustic Excited Methane Jet Diffusion Flames

Abstract

The ignition process of fuel plays an important role in the flame development and emission characteristics, which has attracted intensive attention in the combustion field. However, the transient ignition process for jet flames under acoustic excitation is rarely reported. In the current study, the effect of external acoustic excitation with different frequencies on the ignition process of methane jet diffusion flames has been studied experimentally using high-speed color and schlieren imaging systems. The fuel nozzle used in the experiment features a concentric ring structure, with fuel in the middle and air around it. The acoustic excitation was added to the air side through the loudspeaker, and the frequency of the acoustic excitation was set as 10 Hz, 30 Hz, 50 Hz and 100 Hz, respectively, while a case without external excitation was used as the control group. It is found that the periodic vortex structure propagates downstream in the flow field after acoustic excitation is added, which leads to an uneven velocity distribution in the flow field and the appearance of a local high-speed zone. The acoustic excitation of 30 Hz and 50 Hz can reduce the probability of successful ignition, which is mainly because the acoustic wave propagates in the flow field and causes drastic velocity changes near the ignition position. For the case of 100 Hz, the acoustic perturbation is confined in a small region near the nozzle exit, while the flow field velocity is slightly higher than the case without acoustic excitation.

1. Introduction

The ignition process of fuel has been one of the most important research topics in the field of fundamental combustion. For example, blowout and re-ignition of internal combustion engines, gas turbines and aeroengines and fire development caused by fuel leakage are closely related to the ignition stage [1,2,3]. The establishment of the initial flame kernel and the following flame development in the ignition process can directly affect the combustion stability, combustion efficiency and emission characteristics. The ignition process is jointly affected by a variety of factors, such as the equivalent ratio at the location of the flame kernel, flow velocity and acoustic pulsation, etc. Therefore, the study of ignition characteristics is essential to deeply understand the mechanism of combustion stability and is also of great significance to the combustion chamber design.

In previous research work, the ignition characteristics have been intensively studied in constant volume bombs and turbulent jet flow, as well as confined combustors. The involved factors include flame kernel formation, the minimum ignition energy (MIE), ignition position, equivalent ratio and fuel compositions, etc. Ju and Chen [4] theoretically described the development process and the internal mechanism of flame from flame kernel to flame ball and then to stably propagating flame in a static flow field. The results demonstrated the influence of flame stretching and radiant heat loss on the ignition process. Beduneau et al. [5] investigated the minimum energy necessary to ignite a laminar premixed methane-air mixture through an experimental study. Ref. [6] studied the ignition problem in a wide Lewis number (Le) range and found that when Le is less than a certain value, the flame ball smaller than the theoretical minimum diameter can be successfully ignited. Shen et al. [7] found that the ignition fuel ratio has considerable influence on combustion efficiency, combustion product and flame stability.

In actual combustion devices, the jet diffusion combustion mode is mostly adopted [8]. The combustion process of a diffusion flame is accompanied by a complex flow phenomenon and a heat and mass transfer process of chemical reaction. Existing research results show that the dynamic characteristics of a jet diffusion flame are closely related to the flow field distribution [9]. Phuoc et al. [10] experimentally investigated the laser spark ignition of a jet diffusion flame. It is reported that the success of ignition depends on whether or not the spark-initiated reacting gas could undergo a transition from hot plasma to a propagating flame. Ahmed et al. [11,12] presented experiments on ignition in turbulent non-premixed methane jets in air. The spark position, energy, duration, electrode diameter/gap, jet velocity and air premixing of the fuel stream are examined to study their effects on the ignition probability defined as a successful flame establishment. Mastorakos [13] has reviewed the initiation of turbulent non-premixed flames through auto ignition and spark ignition. Wang et al. [14] experimentally investigated the ignition characteristics of hydrogen-enriched methane diffusion impinging flames, while the results indicate that the flow field velocity increases with an increasing hydrogen percentage. Ebrahimi et al. [15] investigated the effect of adding hydrogen to methane on the thermal characteristics and ignition delay in a methane-air, oxygen-enriched and oxy-fuel MILD combustion mode, which concluded that the use of hydrogen as an additive to methane is an effective strategy. The influences of a head geometry structure on the ignition process are investigated using experimental and numerical methods [16]. The results have shown that the ignition performance can be improved with the shortening of the sleeve length. Liu et al. [17] investigated the ignition process of a sonic ethylene jet penetrating to heated supersonic crossflows. The simulation results have revealed the fast transient unsteady flame kernel formation and propagation with a cavity-stabilized mechanism, while the chemical reaction and the heat transfer characteristics are also discussed.

Noise disturbance exists in the combustion chambers of aircraft engines, gas turbines and other devices, which may cause serious combustion instability and lead to safety accidents [18,19,20]. Intensive investigations have been conducted to reveal the complex interactions between acoustic perturbation and flame dynamics [21,22]. With acoustic perturbation, the flame may show dramatic changes in both structure and color [23,24]. The large-scale vortex structure in the flow field is a common phenomenon under the acoustic effect, which will enhance the mixing of fuel and air, thus affecting the heat release rate as well as the flame dynamics [25,26,27]. Karimi et al. [28] compared the effects of different acoustic amplitudes on flame transfer function and flame dynamics. The results indicate that an acoustic wave would not only have a linear effect on flame, but also have a nonlinear effect. Williams et al. [29] studied the influence of longitudinal acoustic disturbances of different frequencies and amplitudes applied to the fuel side on a jet methane flame. It was found that under an acoustic disturbance of a certain frequency, an ethylene flame would produce a half-frequency phenomenon; that is, it would oscillate at half the frequency of an external acoustic excitation. Other than half-frequency, more nonlinear interactions in non-premixed jet methane flames have been reported by Wang et al. [30], including sum and difference between the natural oscillating and external acoustic excitation frequencies. Sun et al. [22] studied the dynamic response and blowoff characteristics of a premixed methane-air flame to acoustic disturbances in a longitudinal combustor. The results indicated that the strong velocity fluctuation caused by the acoustic wave is the main reason for blowoff.

As aforementioned, the study of the ignition process has been intensively studied in the static environment and in jet flames. However, the acoustic effect on the transient ignition process in jet flames is still rarely reported. Since the ignition process occurs in a very short time, it requires high frequency diagnostics to reveal the flame kernel formation and propagation process. In the present study, the ignition process under different frequencies of acoustic excitation was visualized and quantitatively analyzed based on high-speed color flame images and schlieren images. Quantitative velocity information has been estimated from schlieren images. Both the successful and the failed ignition processes are presented and compared. The important role that acoustic excitation plays in the ignition process has been revealed by the experimental evidence.

2. Experimental Methods

2.1. Experimental Setup

Figure 1 shows the schematic of the experimental setup and the structure of the nozzle. The fuel passes through the flow meter and the pressure relief valve into the nozzle. The structure of the nozzle is shown in Figure 1b. Methane and air flow go into the nozzle through the inner and outer tubes, respectively. There is a blunt body above the nozzle, which helps to stabilize the flame. On the air side, a ring with evenly distributed holes is used to laminate the air flow. The dimensions of the ring and the holes are shown in Figure 1b. The flow rate of air and methane is set at 41.1 L/min and 3 L/min, with the corresponding Reynolds number at 1110 and 480, respectively. The main focus of the present study is to reveal the influence of acoustic excitation on the transient ignition process. Thus, the flow rates of air and methane are kept as a constant, which correspond to a lean burn condition normally favorable for most applications.

The igniter consists of a lead-acid battery, an ignition coil and a pair of steel electrodes, which can supply a pulsed spark with the energy around 200 mJ. When the igniter starts working it will produce a continuous pulse spark at a frequency of 250 Hz. The ignition probe is located approximately 20 mm above the nozzle. Acoustic excitation is applied on the air side by a loudspeaker. Different acoustic signals are generated and controlled by a signal generator. On the outer tube (the air side), two microphones are mounted to measure the velocity pulsation caused by acoustic excitation, as shown in Figure 1a. The schlieren system (LUFTVIS-250, Luftvis Schlieren Co. Ltd., China) includes a point-light source, two concave mirrors, a knife edge and a high-speed camera (Camera 1, M220, Agile Device Co. Ltd., Shenzhen, China). The schlieren images are taken with a frame rate of 2000 fps, exposure time of 1/6600 s and image size of 1024 × 920 pixels. The flame image was captured by the other high-speed color camera (Camera 2, IDT NXA7S1, Integrated Design Tools Inc., Pasadena, CA, USA), with a frame rate of 2000 fps, exposure time of 1/2000 s and image size of 1024 × 1024 pixels. The two cameras are synchronized to capture the flame and schlieren images simultaneously. The test conditions and specific parameter settings are shown in

Table 1. The test conditions and parameters of experiment setup.

Acoustic excitations	No.	1	2	3	4	5
	Frequency (Hz)	/	10	30	50	100
	Velocity amplitude (m/s)	/	0.39	0.50	0.41	0.18
	Relative amplitude	/	0.54	0.69	0.56	0.25
Flow conditions	Air	Flow rate = 41 slm, average velocity = 0.73 m/s, Re = 1110
	Methane	Flow rate = 3 slm, average velocity = 0.99 m/s, Re = 480
Camera setting	Camera 1	Frame rate = 2000 fps, exposure time = 1/6600 s
	Camera 2	Frame rate = 2000 fps, exposure time = 1/2000 s

. During the experiments, the variation of sound pressure amplitude was recorded by microphones and converted into velocity pulsation amplitude by a two-microphone random-excitation algorithm [31], where the relative amplitude denotes the velocity amplitude over the average velocity.

2.2. Flow-Field Velocimetry Using Schlieren Images

In this study, an optimized optical flow algorithm is used to obtain the velocity field based on schlieren images. Optical flow is considered as the distribution of the apparent velocities of the movement of brightness patterns in an image [32,33]. The basic assumption in optical flow techniques is that the gray levels of objects in subsequent frames do not change over time. One constraint can be established based on the change in image brightness at a point on the image plane due to motion, which can be written as:

$$I_{x}u+I_{y}v+I_t=0$$

(1)

As shown in Figure 2, $I_x=\frac{\partial I}{\partial x}$ represents the brightness change in the horizontal direction; $I_y=\frac{\partial I}{\partial y}$ represents the brightness change in the vertical direction; $I_t=\frac{\partial I}{\partial t}$ represents the brightness change of the same position at different times; I denotes the luminance intensity; t is the time between frames and $u=\frac{dx}{dt}$ and $v=\frac{dy}{dt}$ are velocities. The data conservation term H₁ can be written as:

$$H_1=f_1(I_{x}u+I_{y}v+I_t)$$

(2)

where f₁ is the penalty function. The nonconvex generalized Charbonnier penalty function $f_1(x)={({\sigma }^2+x^2)}^{\epsilon }$ is chosen to improve the robustness, where $\sigma =0.001$ and $\epsilon =0.45$.

In addition, spatial consistency should be considered, which refers to the assumption that points in small neighborhoods have similar velocities. One way to express this constraint is to minimize the square of the magnitude of the optical flow velocity gradient, which can be written as applying the penalty function to the first-order velocity space smoothness [32]:

$$H_2=f_1\left(\frac{\partial u}{\partial x}\right)+f_1\left(\frac{\partial u}{\partial y}\right)+f_1\left(\frac{\partial v}{\partial x}\right)+f_1\left(\frac{\partial v}{\partial y}\right)$$

(3)

Then, the global cost function E (H₁, H₂) can be written as:

$$E(H_1,H_2)=\int H_{1}dxdy+\beta \int H_{2}dxdy$$

(4)

In most cases, these two assumptions cannot be fulfilled simultaneously. Thus, there is competition between the two assumptions, which leads to an optimal compromise to minimize the global cost function. The relative importance of each hypothesis is controlled by a weight parameter β. The density displacement field is estimated by the variational method in which all model assumptions are formulated by the cost-optimization problem. The use of variational methods allows flexibility in combination with modern optimization techniques. In the current study, the optimized optical flow method proposed by Sun et al. [32,33] is adopted. The error of the optical flow algorithm is about 0.319 pixel/frame, estimated by the Middlebury benchmark test database, which corresponds to 0.064 m/s in the current study. For more details of the algorithm, please refer to Ref. [32].

2.3. Image Processing of the Color Images

In the chemical reaction accompanied by combustion, electrons will radiate the electromagnetic spectrum of a specific band when they fall back from the excited state, which is called chemiluminescence. During the combustion of hydrocarbon fuels, the blue and green flames produced by premixed combustion are emitted by chemiluminescence of CH* and C₂* in the visible spectrum. RGB color images obtained by ordinary CCD cameras contain visible light and a small part of near-infrared spectral characteristics. Due to the short exposure time in high-speed imaging, the very weak blue flame is hard to directly observe. Therefore, it is necessary to manually enhance the color flame image. In this paper, the method of image enhancement based on a linear spatial filter is adopted [34]. As shown in Figure 3, for each pixel of the image, the product of its neighborhood pixels and the corresponding elements of the filter matrix is calculated, which are then summed up as the value of the pixel position based on Equation (5) to achieve image filtering. Then, the image contrast is enhanced and image quality and identifiability are improved.

$$R=k_0{s}_0+k_1{s}_1+\cdots +k_8{s}_8$$

(5)

An original and a corresponding color-enhanced image are shown in Figure 4 for comparison. It can be seen that on the original image, there is little information. As shown in Figure 4b, the enhanced image clearly shows the flame structure. Most of the flame is blue and a small part is yellow, which may be caused by the combustion of metal particles generated from the ignition probe.

3. Results and Discussion

3.1. Cold Flow Pattern

The process of fuel injection into the flow field before ignition is defined as cold flow field. In this study, the cold flow field has been visualized by the schlieren images, based on which the quantitative velocity information has been estimated using the optical flow method. It can be observed from the schlieren images that the cold flow field shows a periodic change after the addition of acoustic excitation, which corresponds well to the applied excitation frequency. Schlieren images and the resolved velocity contours with equal time intervals within a period under different excitation frequencies are shown in Figure 5. Images within 100 ms, 33 ms, 20 ms and 10 ms were chosen for the test cases, with external excitations at 10 Hz, 30 Hz, 50 Hz and 100 Hz, respectively. For comparison, images with a time interval of 3.5 ms are displayed for the case without external excitation.

As can be seen from Figure 5a, when there is no external acoustic excitation, the cold flow field has no obvious change in time sequence. The region near the nozzle exit has shown a laminar pattern, while a turbulent structure can be observed in the downstream, as shown in the dashed white box in Figure 5a. The flow in the laminar region develops steadily downstream and the fluid trajectory presents a regular smooth curve, which is caused by the small holes on the ring structure, as shown in Figure 1b. However, the flow in the turbulent region is relatively random, chaotic and has no obvious regularity. The velocity of the main stream area is about 1 m/s, which also shows little change at different time instants.

The acoustic excitation has great influences on the flow field, resulting in an uneven velocity distribution both in time and space. For the case of 10 Hz, the velocity field has shown cyclic variation with time, as shown by Figure 5b. The maximum velocity can reach to 2 m/s at certain time instants, while at 56 ms the velocity is no more than 0.5 m/s. The dramatic change indicates the 10 Hz perturbation has influenced a large-scale area with a distance up to 80–120 mm further to the nozzle exit. For the cases with 30 Hz and 50 Hz perturbations, as shown in Figure 5c,d, the regular formation of a periodical vortex structure can be observed, which accordingly corresponds to the acoustic frequency. The area with strong variation is confined below 80 mm in the downstream, which is smaller than the 10 Hz case. It can be observed clearly from the velocity contours that some high-speed zones are near the nozzle exit, as indicated by the white box, showing velocity around 2 m/s. The discontinuity between the high-speed zone and the surrounding area means the existence of a large velocity gradient in the local area. For the case of 100 Hz, the acoustic perturbation effect is confined to a smaller region, which is not beyond 40 mm downstream. Some local areas also show a high velocity around 2 m/s, while the velocity discontinuity can be observed. However, the perturbation is not able to resolve large-scale cyclic variation, while the region further downstream is similar to the case without acoustic excitation.

3.2. Ignition Process Visualization

Based on the high speed flame/schlieren images, the ignition process under different acoustic excitation frequencies was visualized and quantitatively analyzed. The ignition initiation is defined as the moment when the ignition device generates sparks.

The transient ignition process of a similar methane diffusion flame has been introduced for experimental validation. The selected case is a coflow diffusion methane flame, with the Reynolds number of methane at 55 and the surrounding air 1710, in reference [35]. As shown in Table 1, the Reynolds number of methane and surrounding air in the current study is 480 and 1110, respectively. Figure 6 shows the change of flame height in the ignition process for the selected case in [35] and the case without acoustic perturbation in the present study. It can be seen that the flame heights in both cases are both increasing with time in almost linear trends. The flame propagations in two cases are with similar time and length scales. The flame height of the current study is higher, which is mainly due to the larger Reynolds number of methane. The experiment repeatability is also checked by the flame height for repeating the same case five times, while the flame height variation is within ±5%.

Figure 7 shows the flame/schlieren images and the velocity contours of the five selected cases, with a time interval at 3.5 ms. In the case without acoustic excitation, as can be seen from the schlieren images in Figure 7a, an initial flame kernel is formed after ignition, which then propagates both upward and downward. The flame attaches to the nozzle exit around 10 ms. It can be seen from the color and schlieren images that the flame then mainly propagates downstream and reaches a distance about 100 mm at 28 ms, which indicates that a stable flame is established. With schlieren velocimetry, the velocities of the flame front and the turbulent structure inside the flame are both resolved using schlieren images. The motion of the flame front is caused by the propagation of flame in the unignited gas, while the turbulent hot gas is mainly affected by inertia force and buoyancy force. It can be seen from Figure 7a that the absolute velocity at the front of the flame increases rapidly after ignition, reaching a maximum of about 3 m/s, and decreases slightly during the upward propagation process. The velocity of the turbulent structure inside the flame is lower than the flame front, which is below 2 m/s. During the whole process, the center of the flame front protrudes towards unburned gas, which is called a mushroom flame [36].

As shown in Figure 7b, for the case with 10 Hz acoustic excitation, the flame kernel formation and propagation are similar to the case without external perturbation. A stable flame is established around 28 ms. However, a large-scale vortex structure can be observed to develop from upstream to downstream, as indicated by the white dotted line in the schlieren image. This structure is due to the superposition of the comprehensive buoyancy effect of acoustic excitation. The vortex structure interacts with the flame, which will affect the subsequent flame development. The maximum absolute velocity of the flame front reaches about 3 m/s and decreases gradually during the upward propagation. However, the flame no longer keeps the standard mushroom flame and the center of the flame front protrudes toward the ignition gas, which is called a tulip flame [36].

In the cases of 30 Hz and 50 Hz, similar large-scale vortex structures can be observed in the schlieren images. As can be seen from the flame images in Figure 7c,d, the shape of the flame changes dramatically during the ignition process compared with the case without external acoustic excitation. However, the flame heights at 28 ms for the two cases are lower than the 10 Hz and 0 Hz cases. The flame front still shows high velocities, which is around 2 m/s and smaller than the cases of 10 Hz and 0 Hz. In these two cases, some high speed regions appear inside the flame, as shown in the white dotted box in Figure 7c,d. According to the above analysis of cold flow field, this is related to the propagation of acoustic waves under the conditions with external perturbation.

When the excitation is 100 Hz, the flame and schlieren images in Figure 7e show that the ignition process is similar to the case without excitation. Due to the high frequency of acoustic excitation, the large-scale vortex structure appearing at low frequency can no longer be observed. It can be seen from the color images that the flame propagates to 100 mm downstream at about 17 ms, which was significantly shorter than the case without acoustic perturbation. This indicates that the high frequency acoustic excitation would accelerate the ignition process. Accordingly, the velocity of the whole flow field is slightly higher compared with the case without excitation. It is also noticed that the high-speed zone appeared inside the flame, as shown in the white dotted box, which is similar to the cases of 30 Hz and 50 Hz. The results reveal that the high frequency perturbation is confined in a small region near the nozzle exit, while the downstream flow field is less affected.

3.3. Analysis of Ignition Failure Caused by Acoustic Perturbation

In this experiment, when the igniter starts to work, continuous electric sparks with a frequency of 250 Hz would be generated; the energy of a single pulse spark is about 200 mJ. For the cases without acoustic excitation, multiple attempts show that this energy exceeded the minimum ignition energy of the methane jet flame at the test position. Through the visualization of the color and the schlieren images, it can be found that the first electric spark generated by the igniter can successfully ignite the fuel. However, for the cases of 30 Hz and 50 Hz, the ignition failure was observed at the initial several electric sparks. The flame can be ignited later with a certain spark generation. For each acoustic excitation case, more than twenty attempts were conducted, which excludes the accidental occurrence. As shown in Figure 8, the ignition failure phenomena are visualized by high-speed schlieren images, while the ignition success processes are also presented for comparison. The time interval between images was 0.5 ms, while the onset time was recognized at the moment when a single spark was generated. As can be seen from the figure, in the unignited cases (Figure 8a,c), only a small hot gas kernel was generated, as shown by the red arrow in the white dotted box. The hot gas mass disappeared quickly, lasting less than 2 ms. As can be seen from the figure of successful ignition cases (Figure 8b,d), the electric spark heats the surrounding gas, forms the initial flame core and then develops into flame successfully.

In order to further study the reasons for the failure of single spark ignition caused by acoustic excitation, the velocity contour and vector of a flow field at 0 ms under the above conditions were calculated by the optical flow algorithm, as shown in Figure 9. In Figure 9, the white arrow represents the position of the ignition probe, where the spark is generated. By observing the schlieren images in Figure 8 and the velocity vectors in Figure 9, it can be found that, in the cases of ignition failure, the wave peak excited by acoustic excitation just propagates to the ignition probe position. The velocity at this position increases significantly, which reaches around 2 m/s. The velocity gradient at the boundary increases sharply, resulting in the failure of an electric spark to ignite the fuel/air mixture. In the case of successful ignition, the acoustic wave peak has crossed the ignition probe to reach the downstream. Thus, the velocity of the flow field at the ignition probe does not change drastically, which has enabled the successful ignition. It should be pointed out that the ignition failure is not observed for the cases of 10 Hz and 100 Hz. It can be seen from Figure 5b that the 10 Hz excitation has influenced the large vortex structure, while there is no local velocity peak region near the spark position. For the case of 100 Hz, as shown in Figure 5e, the local velocity peak appears near the nozzle exit, which is confined in a small area. Therefore, the main reason for the failure of single spark ignition caused by acoustic excitation is that the propagation of an acoustic wave leads to the drastic change of velocity at the ignition position in the flow field.

4. Conclusions

In the current study, the transient ignition characteristics of non-premixed methane diffusion flame under acoustic excitations has been experimentally investigated. Two high-speed cameras are synchronized to simultaneously record the color and schlieren images. The enhanced flame images show the flame structure clearly, while the turbulent flow field before and after ignition has been visualized through schlieren images. The quantitative velocity information of the flow field has been estimated using an optimized optical flow algorithm. The results indicate that acoustic excitation has great influences on the cold flow pattern, resulting in an uneven velocity distribution both in time and space. The 10 Hz perturbation has affected the cold flow field further downstream, while the cyclic variation is observed with velocity contours. For higher frequencies, the acoustic excitation effect is mainly confined in the region near the nozzle exit, which causes an uneven velocity distribution with local high-speed regions and discontinuities in the flow field. The acoustic effect also has a significant effect on the transient ignition process. The flame front has the highest velocity in all the cases, which has coupled the combustion and flame propagation velocities. For the cases with 10 Hz, 30 Hz and 50 Hz, the forming of a large vortex can be observed due to acoustic excitation. For 100 Hz excitation, the acoustic-affected region is smaller, which is mainly near the nozzle exit, and the downstream flow field is more similar to the case without acoustic excitation. The acoustic effect has caused higher velocities in the flow field, which has promoted the flame establishment in a shorter time. The ignition failure phenomena have been observed for the cases of 30 Hz and 50 Hz. The velocity at the initial ignition stage has revealed that the failure is mainly caused by the dramatic velocity change at the spark position. The large velocity gradient has induced the extinction of the initial flame kernel. The experimental results reported in the current study have shed light on the complex interactions between acoustic excitation and the fast transient ignition process. The visualization and quantitative measurement results can be served as reliable validation proof for numerical simulations. The technical means are also applicable for combustors with complex turbulent conditions, which are able to supply solid and sound experimental evidence for the design and optimization of combustors. For example, in the ignition process of the combustion chamber with acoustic or mechanical perturbations, a similar low frequency acoustic perturbation in the experiment should be paid attention to as much as possible, which may lead to ignition failure.