Industrial furnace combustion diagnosis method based on flame image

Abstract

The invention discloses an industrial furnace combustion diagnosis method based on a flame image. The method comprises the following steps: arrangement of cameras: arranging a camera device on each of a pair of side walls of a boiler flue above a secondary air nozzle of an industrial boiler; two flame cameras are named as a camera I and a camera II respectively, video signals of the camera I and the camera II are accessed to the same computer, and a frame of RGB image is extracted at regular intervals; and then calculating a left-right combustion deviation index Dp1, a front-back combustion deviation index Dp2 and a combustion fluctuation index Db according to the flame area ratio and the gray value, and diagnosing the flame combustion state according to the Dp1, the Dp2 and the Db. The method provided by the invention provides diagnostic information of combustion stability and distribution uniformity of the industrial boiler, has the advantages of low time delay, large monitoring range, good adaptability, high precision and low cost, and provides high-quality diagnostic information for combustion control and automatic operation.

Description

Technical Field

[0001] The invention belongs to the technical field of waste incineration, and in particular relates to an industrial furnace combustion diagnosis method based on flame images. Background Art

[0002] Industrial boilers such as garbage incinerators, biomass boilers, and coal-fired chains generally lack real-time and accurate combustion diagnostic methods during combustion control, resulting in low operating efficiency. The result is that these boilers have large load fluctuations, high excess air coefficients, and low automation, making it difficult to achieve the desired effects of stable combustion, uniform flame distribution, and less manual intervention in the field of combustion control. The difficulty of industrial boiler combustion diagnosis lies in the harsh environment around the boiler, the high latency of conventional diagnostic methods, and the sensitivity to instrument costs. If low-latency diagnosis that can achieve long-term stable operation at a low cost is not used, it will be difficult to promote and apply on a large scale.

[0003] For example, the thermocouple used for furnace temperature monitoring can only measure the local temperature in a very small range, and the time delay for responding to temperature changes after ash accumulation is about 30-60 seconds. The zirconium oxide oxygen meter used to monitor O2 in the waste heat boiler is limited to measuring temperature and corrosive gas environment. It is generally far away from the combustion source, and its response time to oxygen changes is about 40-60 seconds. These conventional detection methods either have a limited detection range or a large time delay, which greatly affects the combustion diagnosis and control of industrial boilers. The industry has an urgent need for non-destructive, large-range, and low-latency diagnostic technology. Summary of the invention

[0004] Technical problems to be solved: In response to the above technical problems, the present invention provides an industrial furnace combustion diagnosis method based on flame images, which can effectively solve the shortcomings of the above detection means, such as limited detection range, large time delay, and great impact on the combustion diagnosis and control of industrial boilers.

[0005] Technical solution: A method for industrial furnace combustion diagnosis based on flame images, comprising the following steps:

[0006] S1. Arrangement of cameras: A camera device is respectively arranged on a pair of side walls of a boiler flue above a secondary air nozzle of an industrial boiler, a pre-buried pipe is arranged in the middle of the side wall of the boiler flue, and a flame camera of the camera device is arranged in the pre-buried pipe. The pre-buried pipe is in an enlarged conical shape, the cone angle of the cone is the same as the viewing angle of the flame camera, and the upper edge is arranged horizontally;

[0007] S2. Image acquisition and processing: The two flame cameras are named camera 1 and camera 2 respectively. The video signals of camera 1 and camera 2 are connected to the same computer. One frame of RGB image is extracted at regular intervals. Then, the left-right combustion deviation index Dp1, the front-back combustion deviation index Dp2 and the combustion fluctuation index Db are calculated according to the flame area ratio and the grayscale value. The flame combustion state is diagnosed according to the size of Dp1, Dp2 and Db. That is, when Dp1 is greater than 0, the combustion area is biased to the side of camera 1, and the larger the value, the greater the degree of bias; when Dp1 is less than 0, the combustion area is biased to the side of camera 2, and the smaller the value, the greater the degree of bias; when Dp2 is greater than 0, the combustion area is biased to the front wall, and the larger the value, the greater the degree of bias; when Dp2 is less than 0, the combustion area is biased to the back wall, and the smaller the value, the greater the degree of bias; the smaller Db is, the smoother the combustion process is; the larger Db is, the greater the fluctuation of the combustion state is.

[0008] Preferably, the specifications of camera 1 and camera 2 are the same, the center lines of the viewing angles of the two cameras converge on the center line of the boiler, the height of the convergence point is greater than the height of the secondary air nozzle, and the height difference H is 0.5 to 2m.

[0009] Preferably, the flame camera is an air-cooled camera or a water-cooled camera, and the front end of the flame camera does not exceed the range of the embedded pipe.

[0010] Furthermore, a compressed air purge pipe is provided between the flame camera and the embedded pipe to prevent dust and coking at the front end of the embedded pipe.

[0011] Preferably, the formula for grayscale conversion of the RGB image in step S2 is as shown in the following formula (1):

[0012] g(x, y)=0.11×R(x, y)+0.59×G(x, y)+0.3×B(x, y) (1)

[0013] Where (x, y) is the pixel coordinate, g(x, y) is the grayscale value of the pixel coordinate (x, y), R(x, y) is the pixel value of the pixel with coordinate (x, y) in the R channel, G(x, y) is the pixel value of the pixel with coordinate (x, y) in the G channel, and B(x, y) is the pixel value of the pixel with coordinate (x, y) in the B channel;

[0014] The formula for the mean flame grayscale of the flame camera at different times is as follows:

[0015]

[0016] Wherein, Z is the judgment function, when (g(x, y)-80)≥0, Z=1; when (g(x, y)-80)<0, Z=0;

[0017] The formula for the average grayscale value gf1 of the flame n seconds before the camera is as follows, where n is 5 to 20:

[0018]

[0019] Use the same method to calculate the average flame gray value gf2 of camera 2 in the first n seconds, as shown below:

[0020]

[0021] The combustion deviation index Dp1 is calculated based on the average of the flame grayscale values ​​of camera 1 and camera 2 in the first n seconds. The formula is as follows:

[0022]

[0023] If Dp1 is positive, the burning area is biased toward the side of camera one. The larger the value, the greater the degree of bias. If Dp1 is negative, the burning area is biased toward the side of camera two. The smaller the value, the greater the degree of bias.

[0024] Furthermore, 50% of the images of camera 1 and camera 2 close to the front wall are extracted and merged horizontally into one image, and the average grayscale value of the flame in the first n seconds gf3 is calculated; 50% of the images of camera 1 and camera 2 close to the back wall are extracted and merged horizontally into one image, and the average grayscale value of the flame in the first n seconds gf4 is calculated. The formula is as follows:

[0025]

[0026] Then the combustion deviation index Dp2 is calculated as follows:

[0027]

[0028] If Dp2 is positive, the combustion area is biased toward the front wall. The larger the value, the greater the degree of bias. If Dp2 is negative, the combustion area is biased toward the rear wall. The smaller the value, the greater the degree of bias.

[0029] Preferably, the calculation process of the flame area ratio in step S2 is as follows:

[0030] Merge the images of camera 1 and camera 2 horizontally, that is, stitch the images of the two flame cameras horizontally and process them as one image. At this time, the number of horizontal pixels of the image is i=i1+i2, and the number of vertical pixels is still j. Calculate the variance of the average area rate of the flame in the first m seconds. m is recommended to be 3-10.

[0031] First, the calculation formula for the average flame area rate Sv of each extracted frame in the first m seconds is as follows:

[0032]

[0033] Wherein, Z is the judgment function, when (g(x, y)-80)≥0, Z=1, when (g(x, y)-80)<0, Z=0;

[0034] Secondly, the calculation formula of the average flame area rate variance Ds in the first m seconds is as follows:

[0035]

[0036] In the formula, S vmean is the average flame area ratio S of the screenshot images at each moment in the previous m seconds v , the mean of;

[0037] Next, the variance of the average grayscale value of the high-temperature flame in the first m seconds of the images of camera 1 and camera 2 is calculated by combining the average grayscale value of the high-temperature flame in each extracted frame in the first m seconds. The judgment basis for high temperature flame is g(x, y)≥120, the formula is as follows:

[0038]

[0039] Wherein, Z′ is the judgment function, when (g(x, y)-120)≥0, Z′=1, when (g(x, y)-120)<0, Z′=0;

[0040] Then calculate the variance Dg of the average gray value of the flame in the first m seconds, the formula is as follows:

[0041]

[0042] In the formula, g high，mean is the average gray value g of the flame in the screenshots at each moment in the previous m seconds high The mean of

[0043] Finally, the combustion fluctuation index Db is calculated using the following formula:

[0044] D B =D s ×D g (13)

[0045] The smaller the combustion fluctuation index Db is, the smoother the combustion process is; the larger the Db is, the greater the fluctuation of the combustion state is.

[0046] Beneficial effects: The method of the present invention can calculate the corresponding indicators according to the flame morphology to reflect the combustion deviation and fluctuation in the industrial furnace, timely diagnose the combustion state, and control the combustion process. The left-right deviation index can be used to intuitively determine whether the flame in the furnace is biased to the left or right side, and the front-back deviation index can be used to determine whether the flame in the furnace is biased to the front and rear walls, so as to adjust the combustion state immediately. This identification index can be used as an input parameter of the automatic combustion control ACC. The combustion fluctuation index can dynamically reflect whether the combustion is stable, indirectly measure the unburned state of gaseous combustibles, and can be used as a basis for combustion adjustment. The method of the present invention has the following advantages:

[0047] The diagnostic method of the present invention has a small delay, and the response time of the camera is generally less than 1 second. The indicator algorithm provided by the present invention can be updated in real time to provide timely combustion diagnosis;

[0048] The detection range is large. Cameras are arranged on the two side walls above the secondary air of the industrial furnace, which is far away from each other, and can completely cover all the visual fields below the monitoring point.

[0049] Good adaptability, combustion diagnosis can be achieved without the help of machine learning, good portability, and the calculation time range of indicators can be flexibly set according to needs and application occasions;

[0050] High precision: the indicator based on the dynamic form of flame provided by the present invention can accurately reflect the biased burning and fluctuation of flame in the furnace, and has the advantage of high precision;

[0051] Low cost, the main equipment is two sets of high-temperature resistant cameras in the furnace, cooling and purge, computers, etc., without high-cost equipment and instruments, with the advantages of low cost and easy installation. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] Figure 1 It is a schematic diagram of the positional relationship between the camera and the industrial boiler in the present invention;

[0053] Figure 2 It is a schematic diagram of the specific position of the camera installation in the present invention;

[0054] Figure 3 It is a schematic diagram of the camera video signal processing flow in the present invention;

[0055] Figure 4 is an operation curve diagram of the Db index and the waste heat boiler O2 in Example 2;

[0056] Figure 5 is an operation curve diagram of Db index and waste heat boiler flue CO in Example 2;

[0057] Figure 6 is an operation curve diagram of the Db index and the main steam volume of the waste heat boiler in Example 2;

[0058] Serial numbers in the figure: 1. industrial boiler, 2. secondary air nozzle, 3. boiler flue, 4. camera device, 41. embedded pipe, 42. flame camera, 43. center line of viewing angle. DETAILED DESCRIPTION

[0059] The present invention is described in detail below in conjunction with the accompanying drawings and specific embodiments:

[0060] Example 1

[0061] A method for industrial furnace combustion diagnosis based on flame images, such as Figure 3 As shown, the following steps are included:

[0062] S1. Camera arrangement: Figure 1-Figure 2 As shown, a camera device 4 is respectively arranged on a pair of side walls of the boiler flue 3 above the secondary air nozzle 2 of the industrial boiler 1, and a pre-buried pipe 41 is arranged in the middle of the side wall of the boiler flue 3. The heights of the pre-buried pipes on both sides are the same, and the pre-buried pipes on both sides are arranged relative to each other; the flame camera 42 of the camera device 4 is arranged in the pre-buried pipe 41, and the pre-buried pipe 41 is in an enlarged conical shape, and the cone angle of the cone is the same as the viewing angle of the flame camera 42. The conical pre-buried pipe is arranged obliquely downward, and the outer contour of the upper end face is at a horizontal angle; the specifications of the two flame cameras are the same, and the viewing angle center lines 43 of the two flame cameras converge on the boiler center line, the height of the convergence point is greater than the height of the secondary air nozzle, and the height difference H is 0.5 to 2m; the flame camera 42 is an air-cooled camera or a water-cooled camera, and the front end of the flame camera 42 does not exceed the range of the pre-buried pipe 41; a compressed air purge pipe is provided between the flame camera 42 and the pre-buried pipe 41 to prevent ash and coking at the front end of the pre-buried pipe 41.

[0063] S2. Image acquisition and processing: The two flame cameras 42 are named as camera 1 and camera 2 respectively. The video signals of camera 1 and camera 2 are connected to the same computer. One frame of RGB image is extracted at regular intervals respectively. Then, the left and right combustion deviation index Dp1, the front and rear combustion deviation index Dp2 and the combustion fluctuation index Db are calculated according to the flame area ratio and the gray value. The flame combustion state is diagnosed according to the size of Dp1, Dp2 and Db, that is, when Dp1 is greater than 0, the combustion area is biased to the side of camera 1, and the larger the value, the greater the degree of bias; when Dp1 is less than 0, the combustion area is biased to the side of camera 2, and the smaller the value, the greater the degree of bias; when Dp2 is greater than 0, the combustion area is biased to the front wall, and the larger the value, the greater the degree of bias; when Dp2 is less than 0, the combustion area is biased to the rear wall, and the smaller the value, the greater the degree of bias; the smaller Db is, the smoother the combustion process is; the larger Db is, the greater the fluctuation of the combustion state is.

[0064] The grayscale value formula of RGB image is as follows:

[0065] g(x, y)=0.11×R(x, y)+0.59×G(x, y)+0.3×B(x, y) (1)

[0066] Where (x, y) is the pixel coordinate, g(x, y) is the grayscale value of the pixel coordinate (x, y), R(x, y) is the pixel value of the pixel with coordinate (x, y) in the R channel, G(x, y) is the pixel value of the pixel with coordinate (x, y) in the G channel, and B(x, y) is the pixel value of the pixel with coordinate (x, y) in the B channel;

[0067] The formula for the mean flame grayscale of the flame camera at different times is as follows:

[0068]

[0069] Wherein, Z is the judgment function, when (g(x, y)-80)≥0, Z=1; when g(x, y)-80)<0, Z=0;

[0070] The formula for the average grayscale value of the flame n seconds before the camera is as follows, where n is 5 to 20:

[0071]

[0072] Use the same method to calculate the average flame gray value gf2 of camera 2 in the first n seconds, as shown below:

[0073]

[0074] The combustion deviation index Dp1 is calculated based on the average of the flame grayscale values ​​of camera 1 and camera 2 in the first n seconds. The formula is as follows:

[0075]

[0076] If Dp1 is positive, the burning area is biased toward the side of camera one. The larger the value, the greater the degree of bias. If Dp1 is negative, the burning area is biased toward the side of camera two. The smaller the value, the greater the degree of bias.

[0077] Extract 50% of the images of camera 1 and camera 2 close to the front wall and merge them horizontally into one image, and calculate the average grayscale value of the flame in the first n seconds gf3; extract 50% of the images of camera 1 and camera 2 close to the back wall and merge them horizontally into one image, and calculate the average grayscale value of the flame in the first n seconds gf4. The formula is as follows:

[0078]

[0079] Then the combustion deviation index Dp2 is calculated as follows:

[0080]

[0081] If Dp2 is positive, the combustion area is biased toward the front wall. The larger the value, the greater the degree of bias. If Dp2 is negative, the combustion area is biased toward the rear wall. The smaller the value, the greater the degree of bias.

[0082] The calculation process of flame area ratio is as follows:

[0083] Merge the images of camera 1 and camera 2 horizontally, that is, stitch the images of the two flame cameras horizontally and process them as one image. At this time, the number of horizontal pixels of the image is i=i1+i2, and the number of vertical pixels is still j. Calculate the variance of the average area rate of the flame in the first m seconds, where m is 3-10.

[0084] First, the calculation formula for the average flame area rate Sv of each extracted frame in the first m seconds is as follows:

[0085]

[0086] Wherein, Z is the judgment function, when (g(x, y)-80)≥0, Z=1, when (g(x, y)-80)<0, Z=0;

[0087] Secondly, the calculation formula of the average flame area rate variance Ds in the first m seconds is as follows:

[0088]

[0089] In the formula, S vmean is the average flame area ratio S of the screenshot images at each moment in the previous m seconds v The mean of

[0090] Next, the variance of the average grayscale value of the high-temperature flame in the first m seconds of the images of camera 1 and camera 2 is calculated by combining the average grayscale value of the high-temperature flame in each extracted frame in the first m seconds. The judgment basis for high temperature flame is g(x, y)≥120, the formula is as follows:

[0091]

[0092] Wherein, Z′ is the judgment function, when (g(x, y)-120)≥0, Z′=1, when (g(x, y)-120)<0, Z′=0;

[0093] Then calculate the variance Dg of the average gray value of the flame in the first m seconds, the formula is as follows:

[0094]

[0095] In the formula, ghigh，mean is the average gray value g of the flame in the screenshots at each moment in the previous m seconds high The mean of

[0096] Finally, the combustion fluctuation index Db is calculated using the following formula:

[0097] D b =D s ×D g (13)

[0098] The smaller the combustion fluctuation index Db is, the smoother the combustion process is; the larger the Db is, the greater the fluctuation of the combustion state is.

[0099] In summary, the most ideal state of flame combustion is: Dp1=0, Dp2=0 and Db=0. The closer the value is to this condition, the better the flame combustion state is.

[0100] The above indicators have guiding significance for combustion adjustment. When Dp1>0, it means that the flame is biased toward the side of camera one. The larger the value, the more serious the bias. It is necessary to adjust the operation to suppress the burning speed of the side with too intense combustion and increase the burning speed of the side with weaker combustion. On the contrary, if Dp1<0, it means that the flame is biased toward the side of camera two. The smaller the value, the more serious the bias. It is necessary to adjust the operation to balance the burning speed on both sides.

[0101] When Dp2>0, it means that the flame is biased towards the front wall. The larger the value, the more serious the bias. When the index exceeds a certain limit, intervention adjustment can be made as needed to prevent the flame from getting too close to the front wall. When Dp2<0, it means that the flame is biased towards the rear wall. The smaller the value, the more serious the bias. When the index exceeds a certain limit, intervention adjustment can be made as needed to prevent the flame from getting too close to the rear wall.

[0102] According to the definition, the indicator Db≥0. The smaller the value, the smaller the flame fluctuation in the first m seconds. When it is equal to 0, there is no flame fluctuation, which is an ideal state. When Db is larger, it means that the flame fluctuates more violently in the first m seconds. The fluctuation includes changes in flame area and flame intensity. Multiple adjustment thresholds can be set according to actual operating conditions. When Db exceeds a certain threshold, corresponding adjustment strategies can be adopted to suppress the fluctuation of combustion in the furnace and make the combustion process smooth, such as reducing the air supply in the combustion section to suppress excessive combustion caused by excessive increase in combustion speed, so as to reduce the probability of excessive unburned CO and drastic changes in load.

[0103] Example 2

[0104] In order to detect and diagnose combustion fluctuations and predict CO, two flame cameras 42 were set on the left and right side walls in a flue of a 300t / d incinerator according to the method of the present invention. The Db index was obtained according to the algorithm of the present invention, and a laser gas analyzer was set at a flue outlet to measure CO. The operating results are shown in FIG. Figure 4-Figure 6 As shown in the figure: Db index has a good early warning effect on the fluctuation of O2 and CO, and also has a strong guiding role for the sharp fluctuation of load; Figure 4 As shown in Figure 2, when O2 fluctuates violently in a short period of time, the Db index will fluctuate greatly; Figure 5 As shown in the figure, when CO fluctuates instantaneously, the Db index will fluctuate accordingly in advance, with a high accuracy. The running time in the figure is 11h; Figure 6 As shown, whenever the operator adjusts the feed and grate combustion to adjust the main steam volume, the Db index tends to fluctuate greatly and has a high accuracy rate. The operating time in the figure is 11h.

[0105] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. An industrial furnace combustion diagnosis method based on flame image, characterized in that: The following steps are involved: S1. Arrangement of cameras: A camera device (4) is respectively arranged on a pair of side walls of a boiler flue (3) above a secondary air nozzle (2) of an industrial boiler (1); a pre-buried pipe (41) is arranged in the middle of the side wall of the boiler flue (3); a flame camera (42) of the camera device (4) is arranged in the pre-buried pipe (41); the pre-buried pipe (41) is in an enlarged conical shape; the cone angle of the cone is the same as the viewing angle of the flame camera (42), and the upper edge is arranged horizontally; S2. Image acquisition and processing: Two flame cameras (42) are named camera 1 and camera 2 respectively. The video signals of camera 1 and camera 2 are connected to the same computer. A frame of RGB image is extracted at regular intervals. Then, the left-right combustion deviation index Dp1, the front-back combustion deviation index Dp2 and the combustion fluctuation index Db are calculated according to the flame area ratio and the gray value. The flame combustion state is diagnosed according to the magnitudes of Dp1, Dp2 and Db. That is, when Dp1 is greater than 0, the combustion area is biased toward the side of camera 1. The larger the value, the greater the degree of bias; when Dp1 is less than 0, the combustion area is biased toward the side of camera 2. The smaller the value, the greater the degree of bias; when Dp2 is greater than 0, the combustion area is biased toward the front wall. The larger the value, the greater the degree of bias; when Dp2 is less than 0, the combustion area is biased toward the rear wall. The smaller the value, the greater the degree of bias; the smaller Db is, the more stable the combustion process is; the larger Db is, the greater the fluctuation of the combustion state is.

2. The method for industrial furnace combustion diagnosis based on flame image according to claim 1, characterized in that: The specifications of the camera 1 and the camera 2 are the same, and the viewing angle center lines (43) of the two cameras converge on the boiler center line. The height of the convergence point is greater than the height of the secondary air nozzle, and the height difference H is 0.5 to 2m.

3. The method for industrial furnace combustion diagnosis based on flame image according to claim 1, characterized in that: The flame camera (42) is an air-cooled camera or a water-cooled camera, and the front end of the flame camera (42) does not exceed the range of the embedded pipe (41).

4. The method for industrial furnace combustion diagnosis based on flame image according to claim 3 is characterized in that: A compressed air purge pipe is provided between the flame camera (42) and the embedded pipe (41) to prevent dust and coking at the front end of the embedded pipe (41).

5. The method for industrial furnace combustion diagnosis based on flame image according to claim 1, characterized in that: The formula for grayscale conversion of RGB image in step S2 is shown in the following formula (1): g(x, y)=0.11×R(x, y)+0.59×G(x, y)+0.3×B(x, y) (1) Where (x, y) is the pixel coordinate, g(x, y) is the grayscale value of the pixel coordinate (x, y), R(x, y) is the pixel value of the pixel with coordinate (x, y) in the R channel, G(x, y) is the pixel value of the pixel with coordinate (x, y) in the G channel, and B(x, y) is the pixel value of the pixel with coordinate (x, y) in the B channel; The formula for the mean flame grayscale of the flame camera at different times is as follows: Wherein, Z is the judgment function, when (g(x, y)-80)≥0, Z=1; when (g(x, y)-80)<0, Z=0; The formula for the average grayscale value gf1 of the flame n seconds before the camera is as follows, where n is 5 to 20: Use the same method to calculate the average flame gray value gf2 of camera 2 in the first n seconds, as shown below: The combustion deviation index Dp1 is calculated based on the average of the flame grayscale values ​​of camera 1 and camera 2 in the first n seconds. The formula is as follows: If Dp1 is positive, the burning area is biased toward the side of camera one. The larger the value, the greater the degree of bias. If Dp1 is negative, the burning area is biased toward the side of camera two. The smaller the value, the greater the degree of bias.

6. The method for industrial furnace combustion diagnosis based on flame image according to claim 5 is characterized in that: Extract 50% of the images of camera 1 and camera 2 close to the front wall and merge them horizontally into one image, and calculate the average grayscale value of the flame in the first n seconds gf3; extract 50% of the images of camera 1 and camera 2 close to the back wall and merge them horizontally into one image, and calculate the average grayscale value of the flame in the first n seconds gf4. The formula is as follows: Then the combustion deviation index Dp2 is calculated as follows: If Dp2 is positive, the combustion area is biased toward the front wall. The larger the value, the greater the degree of bias. If Dp2 is negative, the combustion area is biased toward the rear wall. The smaller the value, the greater the degree of bias.

7. The method for industrial furnace combustion diagnosis based on flame image according to claim 1, characterized in that: The calculation process of the flame area ratio in step S2 is as follows: Merge the images of camera 1 and camera 2 horizontally, that is, stitch the images of the two flame cameras horizontally and process them as one image. At this time, the number of horizontal pixels of the image is i=i1+i2, and the number of vertical pixels is still j. Calculate the variance of the average area rate of the flame in the first m seconds. m is recommended to be 3-10. First, the calculation formula for the average flame area rate Sv of each extracted frame in the first m seconds is as follows: Wherein, Z is the judgment function, when (g(x, y)-80)≥0, Z=1, when (g(x, y)-80)<0, Z=0; Secondly, the calculation formula of the average flame area rate variance Ds in the first m seconds is as follows: In the formula, S vmean is the average flame area ratio S of the screenshot images at each moment in the previous m seconds v The mean of Next, the variance of the average grayscale value of the high-temperature flame in the first m seconds of the images of camera 1 and camera 2 is calculated by combining the average grayscale value of the high-temperature flame in each extracted frame in the first m seconds. The judgment basis for high temperature flame is g(x, y)≥120, the formula is as follows: Wherein, Z′ is the judgment function, when (g(x, y)-120)≥0, Z′=1, when (g(x, y)-120)<0, Z′=0; Then calculate the variance Dg of the average gray value of the flame in the first m seconds, the formula is as follows: In the formula, g high，mean is the average gray value g of the flame in the screenshots at each moment in the previous m seconds high The mean of Finally, the combustion fluctuation index Db is calculated using the following formula: D b =D s ×D g (13) The smaller the combustion fluctuation index Db is, the smoother the combustion process is; the larger the Db is, the greater the fluctuation of the combustion state is.

Images