A method and system for monitoring the combustion state of a furnace flame

Abstract

The application discloses a method and system for monitoring the flame combustion state of a heating furnace, comprising: acquiring images of each nozzle of the heating furnace under the conditions of no flame and stable flame combustion, then identifying the edge distribution features in the flame area and flame profile of each stable combustion image, and obtaining a first threshold value and a second threshold value for evaluating whether the flame is in the stable combustion state according to the flame edge of the flame area and the flame profile respectively; acquiring real-time flame images of each nozzle in the actual combustion process, and identifying the edge distribution features of the flame area and the flame profile, and calculating a first parameter for describing the current flame area edge features and a second parameter for describing the current flame profile edge features; comparing the first parameter with the first threshold value and the second parameter with the second threshold value respectively to obtain the real-time flame combustion state of each nozzle. The application realizes effective monitoring of the flame combustion state of the heating furnace.

Description

Technical Field

[0001] This invention belongs to the field of clean and efficient production and explosion safety control in petrochemical industry, and in particular relates to a method and system for monitoring the combustion state of a heating furnace flame. Background Technology

[0002] Currently, the Ministry of Emergency Management's "Catalogue of Outdated Hazardous Chemical Safety Production Processes, Technologies and Equipment (First Batch)" and the Refining Division's "Guiding Opinions on the Setting of Interlock Protection Systems for Tubular Heaters in Refining Units" have clearly outlined the safety requirements for heater flame identification. Monitoring the combustion status of heater flames and assessing their combustion efficiency is an inevitable trend towards clean, safe, and efficient production.

[0003] In developing this invention, the inventors discovered that existing gas turbine combustion status monitoring methods utilize flame detectors to detect combustion status. If a flame is detected, the gas turbine is considered to be burning normally; if no flame is detected, an alarm is triggered and the next step is initiated. Then, the gas turbine speed and turbine exhaust temperature are monitored. If the gas turbine speed is greater than a first threshold 'a' and the turbine exhaust temperature is less than a second threshold 'b', the flame is considered extinguished; otherwise, the gas turbine is considered to be burning normally. This method monitors the ignition status of the gas turbine by combining changes in gas turbine speed and exhaust temperature under different conditions with the presence or absence of flame detection by the flame detector. This solves the problem of misjudgment caused by hardware damage or blocked detection optical paths when using a simple flame detector to monitor the gas turbine combustion status, and provides a simple and rapid combustion status monitoring method. However, this solution only monitors the ignition status and uses it as the basis for determining normal combustion status. It does not distinguish between abnormal combustion states under ignition status and normal combustion states; that is, it does not achieve true combustion status monitoring in the real sense.

[0004] Secondly, existing technology also discloses an intelligent analysis system and method for boiler combustion efficiency in thermal power plants. This system and method combine open-loop and closed-loop control, including a real-time boiler measurement and calculation module, a target control module, a strategy control module, and a boundary judgment module. The analysis method is based on the O2 / CO coupling relationship, with boiler thermal efficiency as the primary control objective, while also considering NO. x CO concentration; through combustion optimization experiments, the relationship between O2 / CO concentration and ash carbon content and NO concentration was determined. xThe correlation between concentrations determined the operating parameters of the pulverizing system and the optimal synergistic operating parameters of the boiler O2 / CO under different loads. Based on the experimental results, an optimal synergistic operating scheme for O2 / CO was developed, and the relevant control logic was modified on the DCS. Ultimately, an intelligent combustion efficiency analysis system was formed. This system can intelligently indicate the currently applicable optimal operating scheme and guide operators to make combustion adjustments, resulting in an improvement in boiler efficiency of over 1%. However, this scheme did not monitor the combustion status.

[0005] In addition, existing technologies also disclose a boiler combustion status monitoring system, including: a video signal acquisition module for acquiring video signals of the furnace flame within a predetermined time period under full load in the combustion area to be monitored; a time-series image module connected to the video signal acquisition module for obtaining a time-series image of the flame center based on the video signal; a pixel calculation module connected to the time-series image module; a scale reference module connected to the pixel calculation module for comparing the number of pixels in all time-series images to determine the image scale reference image; a furnace combustion image acquisition module for acquiring furnace combustion images to be identified in real time; and a furnace combustion image classification module connected to the furnace combustion image acquisition module. This monitoring system can avoid the drawback of boiler operation status monitoring being greatly affected by subjective human factors, thus improving the accuracy of boiler monitoring.

[0006] Therefore, the existing technology has the following shortcomings: (1) Monitoring the combustion state of the heating furnace requires measuring the flue gas temperature and calculating the corresponding characteristic parameters that characterize the flame combustion state. Most existing technologies combine flue gas temperature measurement with image recognition, or only use image recognition to determine the presence or absence of flames. They cannot monitor and evaluate the combustion state, and therefore cannot quantify and evaluate the flame combustion state of the heating furnace, a complex production process. (2) Existing image recognition methods are not suitable for the complex environment in the actual application scenario of the heating furnace, nor can they meet the requirements of multi-nozzle monitoring and measurement. Flame monitoring based on image recognition has the risk of false alarms, and cannot quantify and evaluate the combustion state of the heating furnace flame. (3) Existing technologies have relatively simple methods for flame recognition, and currently have not coupled an optimized and accurate monitoring and evaluation scheme. Summary of the Invention

[0007] To address the aforementioned problems, this invention provides a method for monitoring the combustion state of a heating furnace flame, comprising: acquiring images of each nozzle on the heating furnace under both flameless and stable combustion states; identifying the flame region and flame outline in each stable combustion image; identifying edge distribution features in the flame region and flame outline; and obtaining a first threshold representing the evaluation of a stable combustion state based on the flame edge of the flame region and a second threshold representing the evaluation of a stable combustion state based on the flame edge of the flame outline, respectively; in the step of obtaining the first threshold and the second threshold, selecting stable combustion images of any nozzle within a continuous time period from the stable combustion images of each nozzle on the heating furnace to be monitored; subsequently integrating the stable combustion images within each time period to generate a flame-defined image for each time period. The system generates a first image of the region features and a second image with flame contour features for each time period. The first image of consecutive time periods is used to train a first preset model to obtain a first model for recognizing the first threshold. The second image of consecutive time periods is used to train a second preset model to obtain a second model for recognizing the second threshold. The first and second models are then used to identify the first and second thresholds for each nozzle. Real-time flame images of each nozzle during actual combustion are acquired, and the edge distribution features of the flame region and flame contour are identified. A first parameter describing the edge features of the current flame region and a second parameter describing the edge features of the current flame contour are calculated. The first parameter is compared with the first threshold, and the second parameter is compared with the second threshold to obtain the real-time flame combustion state of each nozzle.

[0008] Preferably, the step of identifying the flame region in each stable combustion image includes: identifying the gray value of each pixel in each stable combustion image, and based on this, using the maximum inter-class variance method to obtain a critical gray value for distinguishing the flame region from the non-flame region, thereby determining the region in the stable combustion image of each nozzle with a gray value greater than the critical gray value as the flame region.

[0009] Preferably, the step of identifying the flame outline in each stable burning image includes: identifying the brightness information of each pixel in each stable burning image, analyzing and comparing the brightness between each pixel and its neighboring pixels to obtain a brightness variable distribution characterizing the brightness difference between each pixel and its neighboring pixels, and then determining the shape of the flame outline based on the brightness variable distribution of each image; and obtaining the average color channel value of the flame outline of each image based on the shape of the flame outline and in combination with the format attributes of each stable burning image to construct the second threshold.

[0010] Preferably, the training process for the first preset model and the second preset model includes: based on the first preset model, using the flame region of a first image over a continuous time period as input information for the first preset model, and using a first threshold as output information for the first preset model, thereby obtaining the first model through training the first preset model, wherein the first threshold includes a first flame height threshold and a first flame deviation distance threshold; based on the second preset model, using the flame outline of a second image over a continuous time period as input information for the second preset model, and using a second threshold as output information for the second preset model, thereby obtaining the second model through training the second preset model, wherein the second threshold includes a second flame height threshold, a second flame deviation distance threshold, and an average color channel threshold for the flame outline.

[0011] Preferably, the first parameter includes a first flame height parameter representing the vertical distance between the lowest point and the highest point of the flame profile, and a first flame deviation distance parameter representing the horizontal distance between the highest point of the flame profile and the flame centerline. The step of calculating the first parameter includes: for each nozzle, superimposing several real-time images generated within a preset unit time, and determining the flame profile of each nozzle based on the probability of the occurrence of pixels in several sub-regions constituting the flame region after superposition, thereby calculating the first flame height parameter and the first flame deviation distance parameter.

[0012] Preferably, the second parameter includes a second flame height parameter representing the vertical distance between the lowest point and the highest point of the flame profile, a second flame deviation distance parameter representing the horizontal distance between the highest point of the flame profile and the flame centerline, and an average color channel parameter of the flame profile. The step of calculating the second parameter includes: for each nozzle, identifying the flame profile features of each real-time image generated within a preset unit time, and calculating the second flame height parameter, the second flame deviation distance parameter, and the average color channel parameter of the flame profile for each real-time image; and using the average value of the second flame height parameter, the average value of the second flame deviation distance parameter, and the average color channel parameter of the flame profile of all real-time images generated by each nozzle within a unit time as the second flame height parameter, the second flame deviation distance parameter, and the average color channel parameter of the flame profile for each nozzle.

[0013] Preferably, the method further includes: obtaining a first evaluation result by calculating the ratio of the first parameter of each nozzle to the first threshold; obtaining a second evaluation result by calculating the ratio of the second parameter of each nozzle to the second threshold; and using the integrated result of the first evaluation result and the second evaluation result as the evaluation result of the real-time flame combustion state of the corresponding nozzle.

[0014] Preferably, in the process of obtaining the first threshold, the method further includes: performing error analysis on the first threshold, and correcting the first threshold according to the analysis results to eliminate the influence of flame pulsation on the flame combustion state, thereby comparing the corrected first threshold with the first parameter.

[0015] Preferably, in the process of obtaining the second threshold, the method further includes: establishing a normal distribution function of the second threshold based on the edge distribution characteristics in the flame profile, obtaining a standard threshold for describing the edge characteristics of the flame profile, and using the standard threshold as the second threshold, thereby comparing the current second threshold with the second parameter.

[0016] Preferably, the step of obtaining the standard threshold includes: performing curve fitting based on a normal distribution function on the second flame height threshold, the second flame deviation distance threshold, and the average color channel threshold of the flame contour in the second threshold according to the edge distribution characteristics in the flame contour, so as to obtain the average value and standard deviation of each threshold in the second threshold, and thus obtain the standard threshold.

[0017] Preferably, the method further includes: using a two-out-of-one voting alarm method to issue an alarm for the abnormal flame combustion state in the first evaluation result and / or the abnormal flame combustion state in the second evaluation result.

[0018] On the other hand, the present invention also provides a system for monitoring the combustion state of a heating furnace flame. The system includes the following modules: a flame feature acquisition module, which acquires images of each nozzle on the heating furnace under flameless and stable combustion states, and identifies the flame region and flame outline in each stable combustion image; a threshold generation module, which identifies the edge distribution features in the flame region and the flame outline, and obtains a first threshold for evaluating the flame as being in a stable combustion state based on the flame edge of the flame region and a second threshold for evaluating the flame as being in a stable combustion state based on the flame edge of the flame outline; a parameter calculation module, which acquires real-time flame images of each nozzle during actual combustion, identifies the edge distribution features of the flame region and the flame outline, and calculates a first parameter for describing the edge features of the current flame region and a second parameter for describing the edge features of the current flame outline; and a combustion state generation module, which compares the first parameter with the first threshold and the second parameter with the second threshold to obtain the real-time flame combustion state of each nozzle.

[0019] Compared with the prior art, one or more embodiments of the above solutions may have the following advantages or beneficial effects: This invention proposes a method and system for monitoring the combustion state of a heating furnace flame. The method involves pre-capturing images of each nozzle on the furnace under both flameless and stable combustion conditions. Image recognition technology is used to identify the flame region and flame outline in each stable combustion image, further identifying the edge distribution features of the flame region and flame outline. Then, based on machine learning, a first threshold representing the flame's stable combustion state based on the flame edge of the flame region and a second threshold representing the flame's stable combustion state based on the flame edge of the flame outline are obtained. Next, real-time flame images of each nozzle during actual combustion are acquired, and again, image recognition technology is used to identify the edge distribution features of the flame region and flame outline. A first parameter describing the current flame region's edge features and a second parameter describing the current flame outline's edge features are calculated. Finally, by comparing the first parameter with the first threshold and the second parameter with the second threshold, the real-time flame combustion state of each nozzle is obtained. This invention achieves effective monitoring of the combustion state of a heating furnace flame, providing technical support for clean, safe, and efficient production.

[0020] Other features and advantages of the invention will be set forth in the description which follows, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures particularly pointed out in the description, claims, and drawings. Attached Figure Description

[0021] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with the embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings: Figure 1 This is a step diagram of a method for monitoring the combustion state of a heating furnace flame according to an embodiment of this application.

[0022] Figure 2 This is a block diagram of a system for monitoring the combustion state of a heating furnace flame according to an embodiment of this application. Detailed Implementation

[0023] The embodiments of the present invention will be described in detail below with reference to the accompanying drawings and examples, so that the process of how the present invention uses technical means to solve technical problems and achieve technical effects can be fully understood and implemented accordingly. It should be noted that, as long as there is no conflict, the various embodiments and features in the various embodiments of the present invention can be combined with each other, and the resulting technical solutions are all within the protection scope of the present invention.

[0024] Furthermore, the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases the steps shown or described may be executed in a different order than that shown here.

[0025] Currently, the Ministry of Emergency Management's "Catalogue of Outdated Hazardous Chemical Safety Production Processes, Technologies and Equipment (First Batch)" and the Refining Division's "Guiding Opinions on the Setting of Interlock Protection Systems for Tubular Heaters in Refining Units" have clearly outlined the safety requirements for heater flame identification. Monitoring the combustion status of heater flames and assessing their combustion efficiency is an inevitable trend towards clean, safe, and efficient production.

[0026] In developing this invention, the inventors discovered that existing gas turbine combustion status monitoring methods utilize flame detectors to detect combustion status. If a flame is detected, the gas turbine is considered to be burning normally; if no flame is detected, an alarm is triggered and the next step is initiated. Then, the gas turbine speed and turbine exhaust temperature are monitored. If the gas turbine speed is greater than a first threshold 'a' and the turbine exhaust temperature is less than a second threshold 'b', the flame is considered extinguished; otherwise, the gas turbine is considered to be burning normally. This method monitors the ignition status of the gas turbine by combining changes in gas turbine speed and exhaust temperature under different conditions with the presence or absence of flame detection by the flame detector. This solves the problem of misjudgment caused by hardware damage or blocked detection optical paths when using a simple flame detector to monitor the gas turbine combustion status, and provides a simple and rapid combustion status monitoring method. However, this solution only monitors the ignition status and uses it as the basis for determining normal combustion status. It does not distinguish between abnormal combustion states under ignition status and normal combustion states; that is, it does not achieve true combustion status monitoring in the real sense.

[0027] Secondly, existing technology also discloses an intelligent analysis system and method for boiler combustion efficiency in thermal power plants. This system and method combine open-loop and closed-loop control, including a real-time boiler measurement and calculation module, a target control module, a strategy control module, and a boundary judgment module. The analysis method is based on the O2 / CO coupling relationship, with boiler thermal efficiency as the primary control objective, while also considering NO. x CO concentration; through combustion optimization experiments, the relationship between O2 / CO concentration and ash carbon content and NO concentration was determined. xThe correlation between concentrations determined the operating parameters of the pulverizing system and the optimal synergistic operating parameters of the boiler O2 / CO under different loads. Based on the experimental results, an optimal synergistic operating scheme for O2 / CO was developed, and the relevant control logic was modified on the DCS. Ultimately, an intelligent combustion efficiency analysis system was formed. This system can intelligently indicate the currently applicable optimal operating scheme and guide operators to make combustion adjustments, resulting in an improvement in boiler efficiency of over 1%. However, this scheme did not monitor the combustion status.

[0028] In addition, existing technologies also disclose a boiler combustion status monitoring system, including: a video signal acquisition module for acquiring video signals of the furnace flame within a predetermined time period under full load in the combustion area to be monitored; a time-series image module connected to the video signal acquisition module for obtaining a time-series image of the flame center based on the video signal; a pixel calculation module connected to the time-series image module; a scale reference module connected to the pixel calculation module for comparing the number of pixels in all time-series images to determine the image scale reference image; a furnace combustion image acquisition module for acquiring furnace combustion images to be identified in real time; and a furnace combustion image classification module connected to the furnace combustion image acquisition module. This monitoring system can avoid the drawback of boiler operation status monitoring being greatly affected by subjective human factors, thus improving the accuracy of boiler monitoring.

[0029] Therefore, the existing technology has the following shortcomings: (1) Monitoring the combustion state of the heating furnace requires measuring the flue gas temperature and calculating the corresponding characteristic parameters that characterize the flame combustion state. Most existing technologies combine flue gas temperature measurement with image recognition, or only use image recognition to determine the presence or absence of flames. They cannot monitor and evaluate the combustion state, and therefore cannot quantify and evaluate the flame combustion state of the heating furnace, a complex production process. (2) Existing image recognition methods are not suitable for the complex environment in the actual application scenario of the heating furnace, nor can they meet the requirements of multi-nozzle monitoring and measurement. Flame monitoring based on image recognition has the risk of false alarms, and cannot quantify and evaluate the combustion state of the heating furnace flame. (3) Existing technologies have relatively simple methods for flame recognition, and currently have not coupled an optimized and accurate monitoring and evaluation scheme.

[0030] Therefore, to address the aforementioned problems, this invention proposes a method and system for monitoring the combustion state of a heating furnace flame. The method involves pre-capturing images of each nozzle on the furnace under both flameless and stable combustion states. Based on image recognition technology, it identifies the flame region and flame outline in each stable combustion image, and further identifies the edge distribution features of the flame region and flame outline. Then, based on machine learning methods, it obtains a first threshold representing the evaluation of a stable combustion state based on the flame edge of the flame region and a second threshold representing the evaluation of a stable combustion state based on the flame edge of the flame outline. Next, again based on image recognition technology, it identifies the edge distribution features of the flame region and flame outline in real-time flame images of each nozzle during actual combustion, and calculates a first parameter describing the edge features of the current flame region and a second parameter describing the edge features of the current flame outline. Finally, by comparing the first parameter with the first threshold and the second parameter with the second threshold, the real-time flame combustion state of each nozzle is obtained. This invention eliminates the reliance on temperature measurement for combustion state monitoring, achieving effective monitoring of the combustion state of the heating furnace flame and providing technical support for clean, safe, and efficient production. Furthermore, the combustion state of the furnace flame is monitored by using a combination of multiple technical means, which meets the requirements of the complex operating conditions of the furnace.

[0031] Example 1 Figure 1 This is a step diagram of a method for monitoring the combustion state of a heating furnace flame according to an embodiment of this application. See below for reference. Figure 1 This will explain each step of the method.

[0032] like Figure 1As shown, in step S110, images of each nozzle on the furnace to be monitored under both flameless and stable combustion states are acquired. Based on this, the flame region and flame outline in each stable combustion image are identified. In this embodiment, firstly, images of each nozzle on the furnace to be monitored under flameless state are acquired to obtain a background image of each nozzle on the furnace to be monitored under flame combustion. Considering the on-site construction conditions, a completely black video image can also be captured as the corresponding background image. Specifically, the background image is acquired by capturing a video of each nozzle under flameless state within a continuous time period (e.g., each time period is 10 seconds), and then decompressing the video into frame images, thereby obtaining an image of each nozzle on the furnace to be monitored under flameless state. The number of decompressed frames is related to the camera configuration (e.g., if the camera captures video at 50fps, then 10 seconds will yield 500 frames). Next, images of each nozzle on the furnace to be monitored under stable combustion state are acquired using a method similar to that used to acquire the background image. After obtaining images of each nozzle on the furnace under stable combustion conditions, the flame region and flame outline in each stable combustion image are identified.

[0033] In one specific embodiment of this application, the camera captures video of each nozzle in a flameless state at 50fps within a continuous time period (each time period lasting 10s), and the decompressed frame count is 500 frames. Then, using a method similar to obtaining the background image, the camera captures video of each nozzle in a stable flame burning state at 50fps within a continuous time period (each time period lasting 30s), and decompresses the corresponding number of frames (1500 frames) of color image.

[0034] In the process of identifying the flame region and flame outline in each stable combustion image, firstly, the color image of each nozzle on the furnace to be monitored under stable combustion state obtained after decompressing the video is converted into the corresponding grayscale image. Next, the grayscale value of each pixel on each stable combustion image is identified. Based on this, the critical grayscale value used to distinguish between the flame region and the non-flame region is obtained by using the Otsu's method. Thus, the region on the stable combustion image of each nozzle with a grayscale value greater than the critical grayscale value is identified as the flame region.

[0035] Specifically, based on the grayscale image of each nozzle on the furnace under monitoring under stable flame combustion conditions, the grayscale value of each pixel in each stable combustion image is obtained. Using the Otsu's method, each stable combustion image corresponding to each nozzle is compared and analyzed with any image of each nozzle on the furnace under monitoring under flameless conditions, thereby obtaining the critical grayscale value on each stable combustion image used to distinguish between flame and non-flame regions. Furthermore, regions with grayscale values ​​greater than the critical grayscale value on the stable combustion image of each nozzle are identified as flame regions, and regions with grayscale values ​​less than the critical grayscale value are identified as non-flame regions.

[0036] Next, the brightness information of each pixel in each stable combustion image is identified, and the brightness of each pixel is analyzed and compared with that of its neighboring pixels to obtain the brightness variable distribution characterizing the brightness difference between each pixel and its neighboring pixels. Then, based on the brightness variable distribution of each image, the shape of the flame outline is determined. According to the nature of combustion, the outermost ring of the flame has the highest brightness; therefore, this embodiment uses Boolean variables to identify the flame outline in each stable combustion image. Specifically, the brightness data of each pixel in each stable combustion image is read, and the brightness data of each pixel is compared and analyzed with the brightness data of its four neighboring pixels to determine whether the brightness of the current pixel is higher than that of its neighboring pixels, thereby obtaining the brightness variable representing the outer ring region of the flame in each stable combustion image. If the brightness variable of a pixel is greater than 0, the Boolean variable at that pixel's location is activated. After obtaining the outer ring of the flame in each stable combustion image, the flame outline in each stable combustion image is further drawn using a function by reading the pixels of the image array, thereby determining the shape of the flame outline.

[0037] Furthermore, in step S120, edge distribution features in the flame region and flame outline are identified. Based on this, a first threshold for evaluating the flame as being in a stable combustion state based on the flame edge of the flame region and a second threshold for evaluating the flame as being in a stable combustion state based on the flame edge of the flame outline are obtained respectively.

[0038] After obtaining the flame region and flame outline in each stable combustion image, the stable combustion image of each nozzle is converted into a binary image based on the flame region and non-flame region determined on the stable combustion image of each nozzle. That is, the value of each pixel is 0 (indicating that the pixel is located in the non-flame region) or 1 (indicating that the pixel is located in the flame region). In other words, if the value of a pixel in each stable combustion image is 1, and the value of the pixels adjacent to that pixel is 0, then the current pixel is determined to be a pixel constituting the flame edge of the flame region, thereby realizing the identification of the edge distribution features of the flame region. In addition, based on the aforementioned identification of the outer ring of the flame in each stable combustion image, the flame edge of the flame outline can be identified. Furthermore, after identifying the edge distribution features in the flame region and the flame outline, a first threshold for evaluating the flame as being in a stable combustion state based on the flame edge of the flame region and a second threshold for evaluating the flame as being in a stable combustion state based on the flame edge of the flame outline are obtained respectively.

[0039] In the steps of obtaining the first threshold and the second threshold, firstly, stable combustion images of any nozzle within a continuous time period are selected from the stable combustion images of each nozzle on the furnace to be monitored; then, the stable combustion images within each time period are integrated to generate a first image with flame region features for each time period and a second image with flame contour features for each time period; finally, a first preset model is trained using the first image within the continuous time period to obtain a first model for identifying the first threshold, and a second preset model is trained using the second image within the continuous time period to obtain a second model for identifying the second threshold, thereby using the first model and the second model to identify the first threshold and the second threshold of each nozzle.

[0040] Specifically, from the stable combustion images of each nozzle on the furnace to be monitored, the continuous flame region of any nozzle is determined by selecting stable combustion images of any nozzle within a continuous time period. Then, the binary images of the stable combustion images within each time period are superimposed, and the resulting image for each time period is used as the first image with flame region features. Next, the probability I of the occurrence of pixels 1 constituting several sub-regions of the flame region is identified based on the first image to determine the flame region on the first image, thus obtaining the flame contour on the first image. Finally, a first threshold is calculated based on the flame contour on the first image to represent the evaluation of a stable combustion state of the flame based on the flame edge of the flame region.

[0041] Next, from the stable combustion images of each nozzle on the furnace to be monitored, stable combustion images of the nozzles used to generate the first image are selected and generated over a continuous time period. A second image is then generated in a similar manner to the generation of the first image, and the flame profile on the second image is determined. Finally, a second threshold is calculated based on the flame profile on the second image to represent the evaluation of a stable combustion state of the flame according to the flame edge of the flame profile.

[0042] This invention obtains the average color channel value of the flame outline in each image based on the flame outline morphology and the format attributes of each stable combustion image, in order to construct a second threshold. After obtaining the flame outline morphology on the stable combustion image of each nozzle on the furnace to be monitored, this embodiment reads the RGB value of each flame outline point according to the image format attributes of each stable combustion image, thereby obtaining the average color channel value of the flame outline in each stable combustion image.

[0043] After obtaining the first threshold and the second threshold, the first preset model is trained using the first image of the selected nozzles in a continuous time period to obtain the first model for identifying the first threshold of each nozzle of the heating furnace to be monitored. At the same time, the second preset model is trained using the second image of the aforementioned nozzles in a continuous time period to obtain the second model for identifying the second threshold of each nozzle of the heating furnace to be monitored.

[0044] During the training of the first preset model, the flame region of the first image in a continuous time period is used as the input information of the first preset model, and the first threshold is used as the output information of the first preset model, thereby obtaining the first model through training. In this embodiment, the flame region of the first image in a continuous time period is used as the input information of the preset machine learning model (the current preset machine learning model is the first preset model), and substituted into the first preset model to start training. After training, the first model is obtained. The first model can identify the first threshold of the first image of different nozzles in each time period and generate the first threshold for different nozzles in each time period. The first threshold includes a first flame height threshold and a first flame deviation distance threshold.

[0045] During the training of the second preset model, the flame contours of the second images over consecutive time periods are used as input information, and the second threshold is used as output information, thereby obtaining the second model through training. In this embodiment, the flame contours of the second images over consecutive time periods are used as input information of a preset machine learning model (currently the second preset model), substituted into the second preset model, and training begins. After training, the second model is obtained. The second model can identify the second thresholds of the second images of different nozzles in each time period, generating second thresholds for different nozzles in each time period. The second thresholds include a second flame height threshold, a second flame deviation distance threshold, and an average color channel threshold for the flame contour.

[0046] Further, in step S130, real-time flame images of each nozzle during the actual combustion process are acquired, and the edge distribution features of the flame region and flame outline are identified. A first parameter describing the edge features of the current flame region and a second parameter describing the edge features of the current flame outline are calculated. Specifically, following a method similar to obtaining stable combustion images of each nozzle on the monitored furnace, a real-time flame combustion video of the monitored furnace is captured to obtain real-time flame images of each nozzle during the actual combustion process. The edge distribution features of the flame region and flame outline on each real-time flame image are obtained to calculate the first parameter describing the edge features of the current flame region. Simultaneously, after obtaining the edge distribution features of the flame outline on the real-time flame images, the second parameter describing the edge features of the current flame outline is calculated for each real-time flame image.

[0047] In this embodiment, the first parameter includes a first flame height parameter representing the vertical distance between the lowest point and the highest point of the flame profile, and a first flame deviation distance parameter representing the horizontal distance between the highest point of the flame profile and the flame centerline. If the lowest point of the flame is difficult to determine, its location is determined based on the average value of the flame bottom lines.

[0048] In the step of calculating the first parameter, for each nozzle, several real-time images generated within a preset unit time are superimposed. Based on the probability of pixel occurrence in several sub-regions constituting the flame region after superposition, the flame contour of each nozzle is determined, thereby calculating the first flame height parameter and the first flame deviation distance parameter. After obtaining the flame region of the real-time flame image for each nozzle, the flame region of the real-time flame image within a preset time period is obtained in a similar manner to generating the first image. First, the real-time flame images generated within a preset unit time (e.g., 1 second) are converted into corresponding binary images and superimposed, thus using the image obtained after superposition within the preset unit time as the first real-time image with real-time flame region features. Next, based on the probability I of pixel occurrence in several sub-regions constituting the flame region in the first real-time image, the flame region on the first real-time image is determined, thereby obtaining the flame contour on the first real-time image. Finally, based on the flame contour on the first real-time image, the first parameters (first flame height parameter and first flame deviation distance parameter) used to represent the evaluation of the real-time combustion state of the flame based on the flame edge of the flame region are calculated.

[0049] Furthermore, the second parameter includes a second flame height parameter representing the vertical distance between the lowest and highest points of the flame outline, a second flame deviation distance parameter representing the horizontal distance of the highest point of the flame outline from the flame centerline, and an average color channel parameter of the flame outline. If the lowest point of the flame is difficult to determine, its location is determined based on the average value of the bottom lines of the flame.

[0050] In the step of calculating the second parameter, for each nozzle, the flame contour features of each real-time image generated within a preset unit time period are identified, and the second flame height parameter, the second flame deviation distance parameter, and the average color channel parameter of the flame contour are calculated for each real-time image. Specifically, the flame contour features in the real-time flame image of each nozzle are identified in a manner similar to that used to identify the flame contour in each stable combustion image. After obtaining the flame contour features in the real-time flame image of each nozzle, the flame contour features of each real-time image generated within a preset unit time period (e.g., 1 second) are determined for each nozzle. Finally, based on the flame contour features of each real-time image within the preset unit time period, the second parameters (second flame height parameter, second flame deviation distance parameter, and average color channel parameter of the flame contour) used to represent the evaluation of the real-time combustion state of the flame based on the flame edge of the flame contour are calculated.

[0051] Next, the average value of the second flame height parameter, the average value of the second flame deviation distance parameter, and the average value of the average color channel parameter of the flame profile of all real-time images generated by each nozzle within a preset unit time period are used as the second flame height parameter, the second flame deviation distance parameter, and the average color channel parameter of the flame profile of each nozzle.

[0052] Further, in step S140, the first parameter is compared with the first threshold, and the second parameter is compared with the second threshold to obtain the real-time flame combustion state of each nozzle. The method of comparing and analyzing data of the same type in the first parameter and the first threshold is used as the first analysis method, and the method of comparing and analyzing data of the same type in the second parameter and the second threshold is used as the second analysis method. By comparing the real-time flame combustion states obtained using the first and second analysis methods, the accurate real-time flame combustion state of each nozzle is obtained.

[0053] Next, a first evaluation result is obtained by calculating the ratio of the first parameter to the first threshold for each nozzle; a second evaluation result is obtained by calculating the ratio of the second parameter to the second threshold for each nozzle. Then, the integrated result of the first and second evaluation results is used as the evaluation result of the real-time flame combustion state of the corresponding nozzle. In this embodiment, for each nozzle, a first percentage corresponding to the ratio of the first parameter to the first threshold (representing a stable flame combustion state) is preset. Then, the real-time percentage corresponding to the ratio of the first parameter to the first threshold (representing a stable flame combustion state) for each nozzle is calculated, and the first percentage is compared with the real-time percentage to evaluate the flame combustion state of each nozzle in the current monitored heating furnace, thus obtaining the first evaluation result of the real-time flame combustion state of the corresponding nozzle. The method for obtaining the second evaluation result is similar to that of the first evaluation result, and therefore will not be repeated here. Next, the first and second evaluation results are analyzed and compared to determine the evaluation result that matches the current real-time flame combustion state of the corresponding nozzle, thus obtaining a comprehensive evaluation result obtained by integrating the first and second evaluation results, thereby comprehensively evaluating the real-time flame combustion state of the corresponding nozzle.

[0054] In obtaining the first threshold, this invention also performs error analysis on the first threshold and corrects it based on the analysis results to eliminate the influence of flame pulsation on the flame combustion state. The corrected first threshold is then compared with the first parameter. Specifically, according to internationally accepted error analysis methods, the error parameter range of the influence of flame pulsation on the first threshold value is calculated, and the first threshold is corrected based on the obtained error parameter range. In this way, the influence of flame pulsation on the flame combustion state is eliminated from the corrected first threshold. Comparing the corrected first threshold with the first parameter yields a more accurate first evaluation result.

[0055] In one specific embodiment of this application, the data of the same type between the real-time first parameter and the corrected first threshold are directly compared, and the flame combustion state when the first parameter exceeds the first threshold is used as an abnormal combustion state for alarm.

[0056] Next, for each nozzle, a second percentage corresponding to the ratio of a first parameter indicating that the flame is in an abnormal combustion state to the same type of data of a first threshold is preset. Then, the real-time percentage corresponding to the ratio of the first parameter to the same type of data of a first threshold for each nozzle is calculated. If the real-time percentage reaches the aforementioned second percentage, it is determined that the corresponding nozzle of the current monitoring furnace is in an abnormal combustion state. At the same time, a third percentage corresponding to the ratio of the second parameter indicating that the flame is in an abnormal combustion state to the same type of data of a second threshold is preset. Then, the real-time percentage corresponding to the ratio of the second parameter to the same type of data of a second threshold for each nozzle is calculated. If the real-time percentage reaches the aforementioned third percentage, it is determined that the corresponding nozzle of the current monitoring furnace is in an abnormal combustion state.

[0057] Furthermore, this invention employs a 1oo2 method (i.e., a two-out-of-two voting alarm method) to trigger alarms for abnormal flame combustion states in the first evaluation result and / or the second evaluation result. Considering the complex combustion conditions of the heating furnace and the high-speed pulsation of the furnace jet flame, this embodiment uses the 1oo2 method to trigger an alarm when either the first or second evaluation result shows an abnormal flame combustion state, and simultaneously prompts for adjustment of the heating furnace air intake. In this embodiment, considering the actual production problem of fuel excess, an adjustment method to increase the combustion-supporting gas volume is suggested.

[0058] Furthermore, in obtaining the second threshold, a normal distribution function for the second threshold is established based on the edge distribution characteristics in the flame contour. A standard threshold for describing the edge characteristics of the flame contour is obtained, and this standard threshold is used as the second threshold. The current second threshold is then compared with the second parameter. Specifically, firstly, the edge distribution characteristics in the flame contour are obtained from each stable combustion image. Then, the second threshold corresponding to each stable combustion image is calculated based on the edge distribution characteristics, thereby establishing normal distribution functions for the second flame height threshold, the second flame deviation distance threshold, and the average color channel threshold of the flame contour, respectively. Based on the normal distribution functions, the average value μ and standard deviation σ of the second flame height threshold, the second flame deviation distance threshold, and the average color channel threshold of the flame contour are obtained, respectively. Finally, the average value and standard deviation are used to describe the edge characteristics of the flame contour; that is, the standard threshold is related to the average value and standard deviation. Therefore, this embodiment uses the average value and standard deviation to represent the standard threshold and compares the current standard threshold as the second threshold with the second parameter, resulting in a more accurate second evaluation result.

[0059] In the step of obtaining the standard threshold, based on the edge distribution characteristics in the flame contour, curve fitting based on the normal distribution function is performed on the second flame height threshold, the second flame deviation distance threshold, and the average color channel threshold of the flame contour in the second threshold, respectively, to obtain the mean and standard deviation of each threshold in the second threshold, thus obtaining the standard threshold. Specifically, the second threshold corresponding to each stable combustion image is calculated based on the edge distribution characteristics. Then, using probabilistic analysis methods, curve fitting based on the normal distribution function is performed on the second flame height threshold, the second flame deviation distance threshold, and the average color channel threshold of the flame contour in the second threshold, respectively, to obtain the mean μ and standard deviation σ of the second flame height threshold, the second flame deviation distance threshold, and the average color channel threshold of the flame contour, respectively. μ-3σ and μ+3σ are used as the standard thresholds for each of the aforementioned threshold types, thereby obtaining the standard threshold of this embodiment.

[0060] In one specific embodiment of this application, the data of the same type between the real-time second parameter and the standard threshold are directly compared, and the flame combustion state when the corresponding type of parameter in the second parameter is lower than μ-3σ and higher than μ+3σ is used as an abnormal combustion state for alarm.

[0061] Example 2 Based on the method for monitoring the combustion state of a heating furnace flame described in Embodiment 1 above, this embodiment of the invention also provides a system for monitoring the combustion state of a heating furnace flame (hereinafter referred to as "flame combustion state monitoring system"). Figure 2 This is a block diagram of a system for monitoring the combustion state of a heating furnace flame according to an embodiment of this application.

[0062] like Figure 2 As shown, the flame combustion state monitoring system in this embodiment of the invention includes: a flame feature acquisition module 21, a threshold generation module 22, a parameter calculation module 23, and a combustion state generation module 24. Specifically, the flame feature acquisition module 21 is implemented according to the method described in step S110 above, configured to acquire images of each nozzle on the furnace to be monitored in both flameless and stable combustion states, and based on this, identify the flame region and flame outline in each stable combustion image; the threshold generation module 22 is implemented according to the method described in step S120 above, configured to identify edge distribution features in the flame region and flame outline identified by the flame feature acquisition module 21 in each stable combustion image, and based on this, obtain a first threshold for representing the flame being in a stable combustion state based on the flame edge of the flame region and a threshold for representing the flame being in a stable combustion state. The second threshold for evaluating whether a flame is in a stable combustion state is based on the flame edge of the flame profile; the parameter calculation module 23 is implemented according to the method described in step S130 above, and is configured to acquire a real-time flame image of each nozzle during the actual combustion process, and identify the edge distribution features of the flame region and the flame profile, and calculate a first parameter for describing the edge features of the current flame region and a second parameter for describing the edge features of the current flame profile respectively; the combustion state generation module 24 is implemented according to the method described in step S140 above, and is configured to compare the first parameter with the first threshold and the second parameter with the second threshold respectively to obtain the real-time flame combustion state of each nozzle.

[0063] This invention proposes a method and system for monitoring the combustion state of a heating furnace flame. The method involves pre-capturing images of each nozzle on the furnace under both flameless and stable combustion conditions. Image recognition technology is used to identify the flame region and flame outline in each stable combustion image, further identifying the edge distribution features of the flame region and flame outline. Then, based on machine learning, a first threshold representing the flame's stable combustion state based on the flame edge of the flame region and a second threshold representing the flame's stable combustion state based on the flame edge of the flame outline are obtained. Next, real-time flame images of each nozzle during actual combustion are acquired, and again, image recognition technology is used to identify the edge distribution features of the flame region and flame outline. A first parameter describing the current flame region's edge features and a second parameter describing the current flame outline's edge features are calculated. Finally, by comparing the first parameter with the first threshold and the second parameter with the second threshold, the real-time flame combustion state of each nozzle is obtained. This invention achieves effective monitoring of the combustion state of a heating furnace flame, providing technical support for clean, safe, and efficient production. Furthermore, the combustion state of the present invention is manifested in heat transfer, which is determined by the flame volume and the flame color. Therefore, it is feasible for the present invention to use image recognition method to evaluate the real-time combustion state of the flame through flame height, deviation distance and RGB.

[0064] The above description is merely a specific implementation example of the present invention, and the scope of protection of the present invention is not limited thereto. Any modifications or substitutions made to the present invention by those skilled in the art within the technical specifications described herein should be within the scope of protection of the present invention.

[0065] Of course, the present invention may have other various embodiments. Without departing from the spirit and essence of the present invention, those skilled in the art can make various corresponding changes and modifications according to the present invention, but these corresponding changes and modifications should all fall within the protection scope of the claims of the present invention.

[0066] Those skilled in the art will understand that the modules or steps of the present invention described above can be implemented using general-purpose computing devices. They can be centralized on a single computing device or distributed across a network of multiple computing devices. Optionally, they can be implemented using computer-executable program code, thereby storing them in a storage device for execution by a computing device, or fabricating them separately as individual integrated circuit modules, or fabricating multiple modules or steps as a single integrated circuit module. Thus, the present invention is not limited to any particular hardware and software combination.

[0067] While the embodiments disclosed in this invention are as described above, the content is merely for the purpose of facilitating understanding of the invention and is not intended to limit the invention. Any person skilled in the art to which this invention pertains may make any modifications and variations in form and detail of the implementation without departing from the spirit and scope disclosed herein; however, the scope of patent protection for this invention shall still be determined by the scope defined in the appended claims.

Claims

1. A method for monitoring the combustion state of a heating furnace flame, characterized in that, include: Images of each nozzle on the furnace under monitoring are acquired in both flameless and stable combustion states. Based on this, the flame region and flame outline in each stable combustion image are identified. The edge distribution features in the flame region and the flame outline are identified. Based on this, a first threshold for evaluating the flame as being in a stable combustion state based on the flame edge of the flame region and a second threshold for evaluating the flame as being in a stable combustion state based on the flame edge of the flame outline are obtained. In the step of obtaining the first threshold and the second threshold, stable combustion images of any nozzle within a continuous time period are selected from the stable combustion images of each nozzle on the furnace to be monitored. Then, the stable combustion images within each time period are integrated to generate a first image with flame region features for each time period and a second image with flame outline features for each time period. Finally, the first image within the continuous time period is used to train a first preset model to obtain a first model for identifying the first threshold. The second image within the continuous time period is used to train a second preset model to obtain a second model for identifying the second threshold. Thus, the first threshold and the second threshold of each nozzle are identified using the first model and the second model. Acquire real-time flame images of each nozzle during actual combustion, identify the edge distribution features of the flame region and flame outline, and calculate the first parameter used to describe the edge features of the current flame region and the second parameter used to describe the edge features of the current flame outline. The real-time flame combustion state of each nozzle is obtained by comparing the first parameter with the first threshold and the second parameter with the second threshold.

2. The method according to claim 1, characterized in that, The step of identifying the flame region in each stable burning image includes: The grayscale value of each pixel in each stable combustion image is identified. Based on this, the critical grayscale value for distinguishing between flame and non-flame regions is obtained using the Otsu's method. Thus, the region in the stable combustion image of each nozzle with a grayscale value greater than the critical grayscale value is identified as the flame region.

3. The method according to claim 1, characterized in that, The step of identifying the flame outline in each stable burning image includes: The brightness information of each pixel in each stable combustion image is identified, and the brightness of each pixel is analyzed and compared with that of its neighboring pixels to obtain a brightness variable distribution characterizing the brightness difference between each pixel and its neighboring pixels. Based on this brightness variable distribution of each image, the shape of the flame outline is determined. Based on the shape of the flame outline and the format attributes of each stable burning image, the average color channel value of the flame outline of each image is obtained to construct the second threshold.

4. The method according to claim 1, characterized in that, The training process for the first and second preset models includes: Based on the first preset model, the flame region of the first image in a continuous time period is used as the input information of the first preset model, and the first threshold is used as the output information of the first preset model, thereby obtaining the first model by training the first preset model, wherein the first threshold includes a first flame height threshold and a first flame deviation distance threshold. Based on the second preset model, the flame contour of the second image in a continuous time period is used as the input information of the second preset model, and the second threshold is used as the output information of the second preset model. Thus, the second model is obtained by training the second preset model. The second threshold includes a second flame height threshold, a second flame deviation distance threshold, and an average color channel threshold of the flame contour.

5. The method according to any one of claims 1 to 4, characterized in that, The first parameter includes a first flame height parameter representing the vertical distance between the lowest point and the highest point of the flame profile, and a first flame deviation distance parameter representing the horizontal distance between the highest point of the flame profile and the flame centerline. The step of calculating the first parameter includes: For each nozzle, several real-time images generated within a preset unit time are superimposed, and the flame profile of each nozzle is determined based on the probability of the occurrence of pixels in several sub-regions that constitute the flame region after superposition, thereby calculating the first flame height parameter and the first flame deviation distance parameter.

6. The method according to any one of claims 1 to 4, characterized in that, The second parameter includes a second flame height parameter representing the vertical distance between the lowest point and the highest point of the flame profile, a second flame deviation distance parameter representing the horizontal distance of the highest point of the flame profile from the flame centerline, and an average color channel parameter of the flame profile. The step of calculating the second parameter includes: For each nozzle, the flame contour features of each real-time image generated within a preset unit time are identified, and the second flame height parameter, the second flame deviation distance parameter, and the average color channel parameter of the flame contour are calculated for each real-time image. The average value of the second flame height parameter, the average value of the second flame deviation distance parameter, and the average value of the average color channel parameter of the flame profile of all real-time images generated by each nozzle within a unit time are used as the second flame height parameter, the second flame deviation distance parameter, and the average color channel parameter of the flame profile of each nozzle.

7. The method according to any one of claims 1 to 4, characterized in that, The method further includes: A first evaluation result is obtained by calculating the ratio of the first parameter to the first threshold for each nozzle; A second evaluation result is obtained by calculating the ratio of the second parameter to the second threshold for each nozzle; The integrated result of the first evaluation result and the second evaluation result is used as the evaluation result of the real-time flame combustion state of the corresponding nozzle.

8. The method according to any one of claims 1 to 4, characterized in that, In obtaining the first threshold, the method further includes: Error analysis is performed on the first threshold, and the first threshold is corrected based on the analysis results to eliminate the influence of flame pulsation on the flame combustion state. The corrected first threshold is then compared with the first parameter.

9. The method according to any one of claims 1 to 4, characterized in that, In obtaining the second threshold, the method further includes: Based on the edge distribution characteristics in the flame profile, a normal distribution function of the second threshold is established to obtain a standard threshold for describing the edge characteristics of the flame profile. The standard threshold is then used as the second threshold, and the current second threshold is compared with the second parameter.

10. The method according to claim 9, characterized in that, The step of obtaining the standard threshold includes: Based on the edge distribution characteristics in the flame profile, curve fitting based on the normal distribution function is performed on the second flame height threshold, the second flame deviation distance threshold, and the average color channel threshold of the flame profile in the second threshold, respectively, so as to obtain the average value and standard deviation of each threshold in the second threshold, and thus obtain the standard threshold.

11. The method according to claim 10, characterized in that, The method further includes: An alarm is triggered by a two-out-of-one voting method for abnormal flame combustion states in the first evaluation result and / or the second evaluation result.

12. A system for monitoring the combustion state of a heating furnace flame, characterized in that, include: The flame feature acquisition module is used to acquire images of each nozzle on the furnace under monitoring in both flameless and stable combustion states. Based on this, the flame region and flame outline in each stable combustion image are identified. A threshold generation module is used to identify edge distribution features in the flame region and the flame outline. Based on this, a first threshold for evaluating the flame as being in a stable combustion state based on the flame edge of the flame region and a second threshold for evaluating the flame as being in a stable combustion state based on the flame edge of the flame outline are obtained. In the step of obtaining the first threshold and the second threshold, stable combustion images of any nozzle within a continuous time period are selected from the stable combustion images of each nozzle on the furnace to be monitored. Then, the stable combustion images within each time period are integrated to generate a first image with flame region features for each time period and a second image with flame outline features for each time period. Finally, the first image within the continuous time period is used to train a first preset model to obtain a first model for identifying the first threshold. The second image within the continuous time period is used to train a second preset model to obtain a second model for identifying the second threshold. Thus, the first threshold and the second threshold of each nozzle are identified using the first model and the second model. The parameter calculation module is used to acquire real-time flame images of each nozzle during actual combustion, identify the edge distribution features of the flame region and flame outline, and calculate the first parameter used to describe the edge features of the current flame region and the second parameter used to describe the edge features of the current flame outline, respectively. The combustion state generation module is used to compare the first parameter with the first threshold and the second parameter with the second threshold respectively to obtain the real-time flame combustion state of each nozzle.

Images