Hearth flame stability detection method based on deep learning

Through deep learning technology, the IRPC-YOLOV9 model is used to monitor the flame of industrial boiler and calculate the flame stability index, which solves the accuracy and automation problems of flame stability monitoring in the existing technology, and improves combustion efficiency and safety.

CN119027660APending Publication Date: 2024-11-26CHANGZHOU UNIV
View PDF 0 Cites 2 Cited by

Patent Information

Application Number
CN202410859785.1
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-06-28
Publication Date
2024-11-26

AI Technical Summary

Technical Problem

When monitoring the flame stability of industrial boilers, the prior art relies on threshold segmentation, edge detection and morphological operations, requiring a lot of manual adjustment and limited generalization capabilities, making it difficult to accurately obtain the geometric, thermodynamic and brightness parameters of the flame, resulting in low combustion efficiency. , pollutant emissions exceed the standard and other problems.

Method used

The furnace flame stability detection method based on deep learning is adopted to monitor the flame in real time through explosion-proof flame surveillance cameras, and image segmentation is performed using the IRPC-YOLOV9 model, flame characteristic parameters are extracted, flame stability index is calculated, and flame stability is judged, including geometric parameters and thermodynamics. Standard deviation processing of parameters and brightness parameters.

Benefits of technology

Accurate monitoring of flame stability is achieved, low combustion efficiency, excessive pollutant emissions and safety hazards are avoided, and the safety and stability of the production process is ensured. Through the automated segmentation and parameter extraction of deep learning algorithms, the need for manual adjustment is reduced.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119027660A_ABST
    Figure CN119027660A_ABST
Patent Text Reader

Abstract

The invention provides a hearth flame stability detection method based on deep learning, and the method comprises the steps: carrying out the real-time monitoring of the flame in a hearth through an anti-explosion flame monitoring camera, and obtaining a monitoring video; a monitoring image is extracted from the monitoring video, image preprocessing is carried out on the monitoring image, and preprocessing comprises the step of removing a vignetting invalid area of the monitoring image through a mask; performing image segmentation processing on the preprocessed monitoring image by using an IRPC-YOLOV9 model to obtain a flame region; extracting flame characteristic parameters of the flame area, wherein the flame characteristic parameters comprise geometric parameters, thermodynamic parameters and brightness parameters; performing calculation processing on the flame characteristic parameters to obtain a flame stability index; and judging the flame stable combustion degree according to the flame stability index. According to the method, a worker can continuously monitor the flame stability index and master the flame stability condition in real time, so that the serious problems of low combustion efficiency, excessive pollutant emission, even explosion and the like are avoided, and the safety and stability of the production process are ensured.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of industrial boiler flames, and in particular to a furnace flame stability detection method based on deep learning. Background Art

[0002] Industrial combustion processes play a vital role in many industrial processes, including power generation, chemical production, and waste management. However, unstable flames can lead to many combustion problems, such as furnace vibration, low combustion efficiency, excessive pollutant emissions, and even flameout. In order to improve the overall operating efficiency of the furnace, ensure the safety of the operating environment, and meet strict energy-saving and emission reduction standards, it is necessary to accurately monitor the stability of the flame in the furnace. Among them, camera-based monitoring technology needs to effectively separate the flame from the background through segmentation, and use the flame area to obtain relevant data on flame stability during the combustion process, such as temperature, brightness, and jitter frequency. Therefore, finding a suitable and accurate flame region segmentation method is not only an important contribution to the control of the combustion process, but also provides the possibility of achieving more efficient and energy-saving combustion process control. Threshold segmentation, edge detection, region growing, and morphological operations proposed in traditional segmentation techniques are effective but rely on a deep understanding of the application, require a lot of manual adjustments, and have limited generalization capabilities. Summary of the invention

[0003] In view of this, the present invention provides a furnace flame stability detection method based on deep learning to solve the above problems.

[0004] The present invention provides a furnace flame stability detection method based on deep learning, comprising: monitoring the flame in the furnace in real time by using an explosion-proof flame monitoring camera to obtain a monitoring video; extracting a monitoring image from the monitoring video, and performing image preprocessing on the monitoring image, wherein the preprocessing comprises removing a dark corner invalid area of ​​the monitoring image by using a mask; performing image segmentation processing on the preprocessed monitoring image by using an IRPC-YOLOV9 model to obtain a flame area; extracting flame characteristic parameters of the flame area, wherein the flame characteristic parameters include geometric parameters, thermodynamic parameters and brightness parameters; calculating and processing the flame characteristic parameters to obtain a flame stability index; and judging the degree of flame stable combustion according to the flame stability index.

[0005] In another implementation of the present invention, the geometric parameters include flame height, which refers to the vertical distance X0(a,b)-X1(0,b) from the flame outlet X0(a,b) to the front end of the flame X1(0,b), and its value is expressed as:

[0006] L=k|a|

[0007] Among them, k is a correction factor used to take into account the characteristics of the lens. It is obtained by placing an object of known size L0 in the view for camera calibration. Then k is calculated as:

[0008]

[0009] Among them, N0 is the length of the calibration object, and L0 is the number of pixels corresponding to the calibration object in the image.

[0010] In another implementation of the present invention, the brightness parameter includes flame brightness, which refers to the average value of the grayscale value of the luminous area. These values ​​are normalized to the maximum grayscale value, reflecting the most basic luminous characteristics of the flame. The calculation formula is expressed as:

[0011]

[0012] Among them, R f is the flame luminous area, A is the total number of pixels in the flame area, and G(i,j) is the grayscale value of point (i,j).

[0013] In another implementation of the present invention, the thermodynamic parameters include average temperature, maximum temperature and flicker frequency; the flame temperature T of each pixel is obtained according to the two-color method, and the average flame temperature T is calculated. av and the maximum temperature T max :

[0014]

[0015] T max =max{T(i,j)}

[0016] Among them, R and G represent the color components of the image, C2 is the second Planck constant, and λ R is the peak response wavelength of the red spectrum, λ G is the peak response wavelength of the green spectrum, G R , G G are the brightness values ​​of the R and G components respectively, S is the device factor, and T(i,j) is the temperature information of each pixel in the flame area in the image;

[0017] The flicker frequency is determined by performing Fourier transform on the average grayscale of each pixel in the flame area and calculating its power spectrum. The flicker frequency is expressed as:

[0018]

[0019] Where F is the flicker frequency, f x is the xth frequency value, p x is the power density of the xth frequency component, and N2 is the total number of frequency components.

[0020] In another implementation of the present invention, the flame characteristic parameters are calculated to obtain the flame stability index, including: calculating the average value M of the standard deviation of the flame characteristic parameters over a period of time. k and standard deviation σ k , determine the flame stability index.

[0021]

[0022] Among them, M k (k=1,2,3,4), representing L, I, T respectively av and T max The average value of N3 data points after normalization, N3 is the number of data points included in calculating the average value of the kth parameter, V ki is the i-th data value of the k-th parameter in N3 data, and σ1, σ2, σ3, and σ4 are the standard deviations of the four parameters respectively.

[0023] Based on the above parameters, the flame stability index is expressed as:

[0024]

[0025] Among them, φ(M i ) is M i The theoretical maximum standard deviation, σ i M i The standard deviation of i M i The more important a parameter is, the higher its weight w i The bigger.

[0026] In another implementation of the present invention, the degree of flame stable combustion is judged according to the flame stability index, including: the flame stability index δ is used to quantify the stability of the flame, with a minimum value of 0 and a maximum value of 1; wherein, "1" indicates that the flame has reached a completely stable state, and when the standard deviation of any flame characteristic parameter approaches or reaches its theoretical maximum value, the index will approach "0", indicating that the flame is in an extremely unstable state.

[0027] The furnace flame stability detection method based on deep learning of the present invention utilizes the improved YOLOV9 algorithm to segment the flame in the industrial environment to obtain the flame area, and combines the relevant data of the flame stability during the combustion process to calculate the flame stability index to judge the stability of the flame, so that the staff can realize the real-time grasp of the flame stability by continuously monitoring the flame stability index, so as to avoid serious problems such as low combustion efficiency, excessive pollutant emissions and even explosions, thereby ensuring the safety and stability of the production process. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the prior art descriptions. By reading the detailed description of the following implementations, the advantages and benefits of the solutions become clear to those skilled in the art. The drawings are only used to illustrate the preferred implementations and are not to be considered as limiting the present invention.

[0029] In the attached picture:

[0030] Figure 1 This is a schematic flow chart of a furnace flame stability detection method based on deep learning according to an embodiment of the present invention.

[0031] Figure 2 The figure is a schematic diagram of a flame image acquisition system according to an embodiment of the present invention.

[0032] Figure 3 The figure is a schematic diagram of a process for removing the dark corner invalid area of ​​a flame image according to an embodiment of the present invention.

[0033] Figure 4 Schematic diagram of the IRPC-YOLOV9 structure of an embodiment of the present invention.

[0034] Figure 5 FIG. 4 is a schematic diagram of the Rep-iRMB structure according to an embodiment of the present invention.

[0035] Figure 6 A schematic diagram of the PC block structure of an embodiment of the present invention.

[0036] Figure 7 FIG. 4 is a schematic diagram of flame segmentation results according to an embodiment of the present invention.

[0037] Figure 8 Schematic diagram of flame and its characteristic parameters according to an embodiment of the present invention. DETAILED DESCRIPTION

[0038] In order to enable those skilled in the art to better understand the technical solutions in the embodiments of the present invention, the technical solutions in the embodiments of the present invention will be described clearly and in detail below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in the field based on the embodiments in the embodiments of the present invention should fall within the scope of protection of the embodiments of the present invention.

[0039] Figure 1 A flow chart of a furnace flame stability detection method based on deep learning provided by an embodiment of the present invention is as follows: Figure 1 As shown, this embodiment mainly includes:

[0040] The flame in the furnace is monitored in real time by an explosion-proof flame monitoring camera to obtain a monitoring video.

[0041] For example, Figure 2 As shown, the explosion-proof flame monitoring camera and the display are connected to NVIDIA's Jetson NX development board through GPIO pins or USB interfaces to monitor the flame in the furnace in real time.

[0042] It should be understood that using NVIDIA's Jetson NX development board as a controller for data collection and transmission, Jetson NX provides up to 21TOPs of accelerated AI computing performance and can run multiple modern neural networks in parallel. While maintaining high performance, Jetson NX is designed with a focus on energy efficiency, which is suitable for applications with strict power requirements and has strong compatibility and scalability, and can provide the necessary computing power and flexibility.

[0043] A monitoring image is extracted from the monitoring video, and the monitoring image is subjected to image preprocessing, wherein the preprocessing includes removing a dark corner invalid area of ​​the monitoring image by masking.

[0044] For example, the combustion flame of the furnace is the object of video monitoring, which directly reflects the combustion status of the boiler. The flame video is monitored through the observation port, but this method leads to invalid areas in the video, and these areas have brightness fluctuations due to light scattering, which may interfere with subsequent analysis. To solve this problem, an elliptical mask with the same size as the actual flame area is defined, and the grayscale value of the invalid area around it is set to 0 to extract the valid area image.

[0045] like Figure 3 As shown in the figure, by setting a 552×552 grayscale mask, separating the flame video color channel and superimposing it with the mask, the invalid area is set to 0, the grayscale value of the valid area is retained, and finally the channels are merged to remove irrelevant areas, thereby eliminating the interference of invalid areas on subsequent image segmentation.

[0046] The preprocessed monitoring image is segmented using the IRPC-YOLOV9 model to obtain the flame area.

[0047] For example, Figure 4 As shown in Figure 1, in order to achieve accurate and reliable flame segmentation, the IRPC framework is improved based on YOLOV9. The backbone of the IRPC framework is responsible for extracting the features of the input image, while the neck further processes the extracted features and fuses feature information at different levels. After the head and post-processing, the category, location information and pixel segmentation results of the detected object are output. The flame segmentation results are shown in Figure 1. Figure 7 shown.

[0048] The backbone of the IRPC framework consists of three modules: PC block, ADown (downsampling), and Rep-iRMB.

[0049] The PC block (Partial Convolution Block) extracts spatial features more efficiently by reducing redundant calculations and memory accesses.

[0050] The ADown module uses a pooling operation to downsample the feature matrix.

[0051] The RepNCSPELAN4 module is improved by introducing the inverse residual moving block iRMB to calculate the Rep-iRMB module, which helps the segmentation task to better aggregate global information in depth and improves the feature extraction capability while ensuring lightweight.

[0052] The neck of IRPC-YOLOV9 consists of SPPELAN, which is a combination of SPPF (Spatial Pyramid Pooling Fast) and ELAN (Efficient Local Aggregation Network).

[0053] SPPF can capture spatial information of different scales and improve the robustness of the model, while ELAN is a lightweight network structure that effectively improves the feature extraction ability of the model through local aggregation and global integration. By combining SPPF and ELAN, the performance of the model is further improved.

[0054] The head network consists of three segment detectors, each of which localizes and segments the target object on feature matrices at different scales, thus improving the detection performance through multi-scale fusion.

[0055] After post-processing of the model output, the prediction results of object positioning and pixel segmentation can be obtained. The improvement process of each module is as follows:

[0056] To meet the requirements of efficient visualization models, the Rep-iRMB module introduces the inverse residual moving block iRMB for the RepNCSPELAN4 module. This attention mechanism combines the lightweight of convolutional neural networks with the dynamic model capabilities of Transformer (consisting of encoders and decoders). Compared with ordinary convolutional neural networks, it can help segmentation tasks better aggregate global information at a deep level, improving feature extraction capabilities while ensuring lightweight.

[0057] F in iRMB is modeled as a cascade of DW-Conv and EW-MHSA convolutions, and the calculation formula is as follows:

[0058] F(·)=(DW-Conv,Skip)(EW-MHSA(·))

[0059] like Figure 5 As shown in the figure, iRMB consists only of 3×3 depthwise separable convolution (DW-Conv) and self-attention mechanism. To optimize the computational efficiency, 1×1 convolution is used to compress and expand the channel, and the depthwise separable convolution is used to capture the feature information of the space. This module can enhance the model's ability to extract features, which is more conducive to segmentation operations.

[0060] PC Block (Partial Convolution Block) is introduced to reduce redundant calculations and memory access. The specific operation principle of partial convolution is to first divide the input X into two parts, X1 and X2, and only perform convolution operations on the input channels of one part (X1), while keeping the number of channels of the other part (X2) unchanged. Finally, the two parts are connected together.

[0061] Given input X, partial convolution Y PC The output can be expressed as:

[0062] X1,X2=f split (X)

[0063] Y PC =f concat (Conv(X1),X2)

[0064] Among them, X1 and X2 are divided in a ratio of 1:3, f concat represents the splicing operation, f split Represents a split operation.

[0065] like Figure 6 As shown in Figure 2, partial convolution uses traditional convolution operations to process a portion of the input channels, while the remaining channels are not processed. Therefore, the FLOPs (floating point operations) of partial convolution are It is 1 / 16 of the normal convolution. PC Block (Partial Convolution Block) introduces two point-by-point convolutions to improve the output of partial convolution, and uses batch normalization and ReLU (Rectified Linear Unit) activation between the two point-by-point convolutions to achieve fast convergence. In order to solve the gradient disappearance problem, skip connections are introduced. Finally, 3×3 convolution is used to adjust the size of the output feature map.

[0066] Given input X, PC block Y PCB The output can be expressed as:

[0067] Y PCB =Conv(PWConv(ReLU(BN(PWConv(Y PC ))))+Y PC )

[0068] Compared with traditional convolution methods, PConv (partial convolution) significantly reduces the amount of floating-point operations, more effectively utilizes the computing resources of terminal devices, and reduces the need for memory access, thereby achieving a lightweight network architecture.

[0069] In addition, the introduction of point-by-point convolution ensures efficient extraction of spatial features. In response to the challenges encountered in furnace flame detection, such as difficulty in detecting small targets, target missing, and high detection costs, point-by-point convolution is combined with partial convolution to form an overall PC Block (Partial Convolution Block), which is integrated into the backbone network of YOLOv9 to replace the original convolution module. This innovation not only improves detection accuracy, but also further reduces the number of model parameters.

[0070] The flame characteristic parameters of the flame area are extracted, wherein the flame characteristic parameters include geometric parameters, thermodynamic parameters and brightness parameters.

[0071] For example, since the flame root is a key area of ​​the combustion reaction, the quantitative analysis is mainly focused on the flame root area.

[0072] The flame characteristic parameters are calculated and processed to obtain a flame stability index.

[0073] The flame stability combustion degree is judged according to the flame stability index.

[0074] The furnace flame stability detection method based on deep learning of the present invention utilizes the improved YOLOV9 algorithm to segment the flame in the industrial environment to obtain the flame area, and combines the relevant data of the flame stability during the combustion process to calculate the flame stability index to judge the stability of the flame, so that the staff can realize the real-time grasp of the flame stability by continuously monitoring the flame stability index, so as to avoid serious problems such as low combustion efficiency, excessive pollutant emissions and even explosions, thereby ensuring the safety and stability of the production process.

[0075] In another implementation of the present invention, the geometric parameters include flame height, such as Figure 8 As shown, it refers to the vertical distance X0(a,b)-X1(0,b) from the flame outlet X0(a,b) to the front end of the flame X1(0,b), and its value is expressed as:

[0076] L=k|a|

[0077] Among them, k is a correction factor used to take into account the characteristics of the lens. It is obtained by placing an object of known size L0 in the view for camera calibration. Then k is calculated as:

[0078]

[0079] Among them, N0 is the length of the calibration object, and L0 is the number of pixels corresponding to the calibration object in the image.

[0080] In another implementation of the present invention, the brightness parameter includes flame brightness, which is a quantitative index, referring to the average value of the grayscale value of the luminous area. These values ​​are normalized to the maximum grayscale value, reflecting the most basic luminous characteristics of the flame. The calculation formula is expressed as:

[0081]

[0082] Among them, R f is the flame luminous area, A is the total number of pixels in the flame area, and G(i,j) is the grayscale value of point (i,j).

[0083] In another implementation of the present invention, the thermodynamic parameters include average temperature, maximum temperature and flicker frequency.

[0084] The change in temperature distribution reflects the process of radiation heat transfer and is closely related to flame stability. Since the captured image has three color components, namely red (R), green (G) and blue (B), two color components are selected in the calculation. In this application, the selection of R and G color components is taken as an example.

[0085] The flame temperature T of each pixel is obtained by the two-color method, and the average flame temperature T is calculated. av and the maximum temperature T max :

[0086]

[0087] T max =max{T(i,j)}

[0088] Among them, R and G represent the color components of the image, and C2 is the second Planck constant, 1.438×10 -2 mk;λ R is the peak response wavelength of the red spectrum, 615nm; G is the peak response wavelength of the green spectrum, 540nm; G R , G G are the brightness values ​​of the R and G components respectively, S is the equipment factor, which can be determined by calibrating the measurement system using a standard blackbody furnace; T(i,j) is the temperature information of each pixel in the flame area in the image.

[0089] The flicker frequency of a flame is an important parameter to measure its stability, which reflects the regularity of the change of the flame intensity signal or image grayscale over time. The flicker frequency is determined by performing Fourier transform on the flame intensity signal (i.e., the average grayscale of each pixel in the flame area) and calculating its power spectrum. The flicker frequency is expressed as:

[0090]

[0091] Where F is the flicker frequency, f x is the xth frequency value, p x is the power density of the xth frequency component, and N2 is the total number of frequency components.

[0092] In another implementation of the present invention, the flame characteristic parameters are calculated to obtain a flame stability index, including: using the extracted flame characteristic parameter values ​​to calculate the flame stability index to judge the flame stability, the method does not need to foresee specific conditions such as fuel characteristics and burner types, and calculates the average value M of the standard deviation of the flame characteristic parameters over a period of time. k and standard deviation σ k , determine the flame stability index.

[0093]

[0094] Among them, M k (k=1,2,3,4), representing L, I, T respectively av and T max The average value of N3 data points after normalization, N3 is the number of data points included in calculating the average value of the kth parameter, V ki is the i-th data value of the k-th parameter in N3 data, and σ1, σ2, σ3, and σ4 are the standard deviations of the four parameters respectively.

[0095] Based on the above parameters, the flame stability index is expressed as:

[0096]

[0097] Among them, φ(M i ) is M i The theoretical maximum standard deviation, σ i M i The standard deviation of i M i The more important a parameter is, the higher its weight w i In this application, the weights of all parameters w i = 2. Since the dynamic ranges of M1, M2, M3, and M4 are uniform [0, 1], the range of δ is [0, 1].

[0098] In another implementation of the present invention, the degree of flame stable combustion is judged according to the flame stability index, including: the flame stability index δ is used to quantify the stability of the flame, with a minimum value of 0 and a maximum value of 1; wherein, "1" indicates that the flame has reached a completely stable state, and when the standard deviation of any flame characteristic parameter (such as geometry, brightness or thermodynamic parameters) approaches or reaches its theoretical maximum value, the index will approach "0", indicating that the flame is in an extremely unstable state.

[0099] It can be seen that in the industrial production process, it is very important to monitor the changes in the flame stability index. Through this means, staff can promptly detect signs of flame instability and prevent potential safety hazards. When the standard deviation of the flame parameters approaches or touches its theoretical maximum value, the flame stability index will quickly approach 0, which clearly warns that the flame is in an extremely unstable state. For industrial furnace flames, this early warning mechanism should take prompt action to avoid serious problems such as low combustion efficiency, excessive pollutant emissions, and even explosions by adjusting combustion conditions or taking other necessary measures. And by continuously monitoring and calculating the flame stability index, staff can grasp the stability of the flame in real time, provide strong support for the monitoring and control of industrial flames, and ensure the safety and stability of the production process.

[0100] Thus far, specific embodiments of the present invention have been described. Other embodiments are within the scope of the appended claims. In some cases, the actions recited in the claims may be performed in a different order and still achieve the desired results. Additionally, the processes depicted in the accompanying drawings do not necessarily require the specific order or sequential order shown to achieve the desired results. In some embodiments, multitasking and parallel processing may be advantageous.

[0101] It should be noted that all directional indications in the embodiments of the present invention (such as up, down, left, right, back, etc.) are only used to explain the relative position relationship, movement status, etc. between the components under a certain specific posture (as shown in the accompanying drawings). If the specific posture changes, the directional indication will also change accordingly.

[0102] In the description of the present invention, the terms "first" and "second" are only used to facilitate the description of different components or names, and cannot be understood as indicating or implying a sequential relationship, relative importance, or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features.

[0103] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art of the present invention. The terms used in the specification of the present invention herein are only for the purpose of describing specific embodiments and are not intended to limit the present invention.

[0104] It should be noted that although the specific embodiments of the present invention are described in detail in conjunction with the accompanying drawings, it should not be understood as limiting the scope of protection of the present invention. Within the scope described in the claims, various modifications and variations that can be made by those skilled in the art without creative work still belong to the scope of protection of the present invention.

[0105] The examples of the embodiments of the present invention are intended to concisely illustrate the technical features of the embodiments of the present invention so that those skilled in the art can intuitively understand the technical features of the embodiments of the present invention, and are not intended to be improper limitations of the embodiments of the present invention.

[0106] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A furnace flame stability detection method based on deep learning, characterized in that: include: The flame in the furnace is monitored in real time by an explosion-proof flame monitoring camera to obtain a monitoring video; Extracting a monitoring image from the monitoring video, and performing image preprocessing on the monitoring image, wherein the preprocessing includes removing a dark corner invalid area of ​​the monitoring image by masking; The preprocessed monitoring image is segmented using the IRPC-YOLOV9 model to obtain the flame area; Extracting flame characteristic parameters of the flame area, wherein the flame characteristic parameters include geometric parameters, thermodynamic parameters and brightness parameters; Calculating and processing the flame characteristic parameters to obtain a flame stability index; The flame stability combustion degree is judged according to the flame stability index.

2. The method according to claim 1, characterized in that The geometric parameters include flame height, which refers to the vertical distance X0(a,b)-X1(0,b) from the flame outlet X0(a,b) to the front end of the flame X1(0,b), and its value is expressed as: L = ka | Among them, k is a correction factor used to take into account the characteristics of the lens. It is obtained by placing an object of known size L0 in the view for camera calibration. Then k is calculated as: Among them, N0 is the length of the calibration object, and L0 is the number of pixels corresponding to the calibration object in the image.

3. The method according to claim 1, characterized in that The brightness parameters include flame brightness, which refers to the average grayscale value of the luminous area. These values ​​are normalized to the maximum grayscale value, reflecting the most basic luminous characteristics of the flame. The calculation formula is expressed as: Among them, R f is the flame luminous area, A is the total number of pixels in the flame area, and G(i,j) is the grayscale value of point (i,j).

4. The method according to claim 1, characterized in that: The thermodynamic parameters include average temperature, maximum temperature and flicker frequency; The flame temperature T of each pixel is obtained by the two-color method, and the average flame temperature T is calculated. av and the maximum temperature T max : T max =max{T(i,j)} Among them, R and G represent the color components of the image, C2 is the second Planck constant, and λ R is the peak response wavelength of the red spectrum, λ G is the peak response wavelength of the green spectrum, G R , G G are the brightness values ​​of the R and G components respectively, S is the device factor, and T(i,j) is the temperature information of each pixel in the flame area in the image; The flicker frequency is determined by performing Fourier transform on the average grayscale of each pixel in the flame area and calculating its power spectrum. The flicker frequency is expressed as: Where F is the flicker frequency, f x is the xth frequency value, p x is the power density of the xth frequency component, and N2 is the total number of frequency components.

5. The method according to claim 1, characterized in that The step of calculating and processing the flame characteristic parameters to obtain a flame stability index includes: By calculating the average value M of the standard deviation of the flame characteristic parameters over a period of time k and standard deviation σ k , determine the flame stability index; Where, Mk (k = 1, 2, 3, 4) represents L, I, Tav and T respectively. max The average value of N3 data points after normalization, N3 is the number of data points included in calculating the average value of the kth parameter, V ki is the i-th data value of the k-th parameter in N3 data, σ1, σ2, σ3, and σ4 are the standard deviations of the four parameters respectively; Based on the above parameters, the flame stability index is expressed as: Among them, φ(M i ) is M i The theoretical maximum standard deviation, σ i M i The standard deviation of i M i The more important a parameter is, the higher its weight w i The bigger.

6. The method according to claim 5, characterized in that The step of judging the degree of flame stability combustion according to the flame stability index comprises: The flame stability index δ is used to quantify the stability of the flame, with a minimum value of 0 and a maximum value of 1; Among them, "1" means that the flame has reached a completely stable state. When the standard deviation of any flame characteristic parameter approaches or reaches its theoretical maximum value, the index will approach "0", indicating that the flame is in an extremely unstable state.

Citation Information

Cited By

  • Method for detecting combustion efficiency of firewood-fired energy-saving furnace based on image processing

    CN120107261A

  • Coal-fired power plant boiler burner flame combustion area segmentation method and device and medium

    CN120107595A