Soot blower driving control method based on visual monitoring

Abstract

The invention relates to the technical field of image processing, in particular to a soot blower driving control method based on visual monitoring, which comprises the following steps: firstly, acquiring a visible light image and a thermal infrared image of a target soot blowing area and blowing position data of a soot blower in the target soot blowing area, and analyzing the flying dust existence probability of each pixel position in the images; and then taking the flying dust existence probability as the credibility of the image features, obtaining a dust accumulation state according to the image features, and finally obtaining a control strategy of the soot blower based on the dust accumulation state and carrying out driving control on the soot blower. Compared with the prior art, the method has the advantages that the flying dust existence probability is accurately calculated by fusing the multi-modal information of the visible light image and the thermal infrared image and combining the blowing position data, the image analysis process is dynamically optimized by taking the flying dust existence probability as the credibility of the image features, the reliability of dust retention state evaluation is ensured, and the evaluation accuracy is improved. The problem that in the prior art, the control precision and efficiency of a soot blower can be affected by flying dust is solved.

Description

Technical Field

[0001] This invention relates to the field of image processing technology, and in particular to a soot blower drive control method based on visual monitoring. Background Technology

[0002] A sootblower is a device used to remove ash buildup inside equipment such as boilers and heat exchangers. It achieves efficient ash removal through shock waves, sound waves, and steam. Currently, sootblowing control methods based on visual monitoring have become an important technological direction for improving sootblowing quality due to their non-contact and real-time characteristics. Through image acquisition and analysis, the distribution of ash accumulation and the ash removal effect can be monitored in real time, thereby optimizing the sootblower's propulsion path and injection parameters.

[0003] However, in actual operation, the sootblower generates a large amount of dust during blowing, which significantly reduces visibility inside the equipment, directly affecting the quality assessment and control of sootblowing based on visual monitoring. If the system is stopped and the dust is allowed to dissipate before image acquisition, the single sootblowing cycle will be prolonged, reducing the overall cleaning efficiency and making it difficult to meet the demand for efficient and continuous cleaning in industrial production.

[0004] Therefore, there is a need for a dust blower control method that can overcome the impact of dust on image acquisition and achieve precise control in dusty environments. Summary of the Invention

[0005] Therefore, the present invention provides a visual monitoring-based sootblower drive control method to solve the problem that the control accuracy and efficiency of sootblowers in the prior art are affected by dust.

[0006] This invention provides a visual monitoring-based sootblower drive control method, comprising: Acquire visible light and thermal infrared images of the target soot blowing area, as well as the blowing position data of the soot blower in the target soot blowing area; Based on the image features of visible light and thermal infrared images, combined with the blowing location data, the probability of dust presence at each pixel location in the image is analyzed. Using the probability of dust presence as the reliability of image features, the dust accumulation status at each pixel location in the image is obtained based on the image features of visible light images and thermal infrared images. Based on the dust accumulation status, a control strategy for the sootblower is obtained and driven accordingly.

[0007] In a preferred implementation: based on the image features of the visible light image and the thermal infrared image, combined with the blowing location data, the probability of dust presence at each pixel location in the image is analyzed, including: Based on the degree of blur in the corresponding region of the visible light image according to the jetting position data, a blur score is obtained for each pixel position in the image; Based on the temperature information in the thermal infrared image, and combined with the corresponding position of the jetting location data in the thermal infrared image, a temperature anomaly score is obtained for each pixel in the image. Based on the real-time operating conditions of the target dust blowing area, the fuzz score and temperature anomaly score are fused to obtain the probability of dust presence at each pixel location in the image.

[0008] In a preferred implementation: based on the blurriness of the corresponding region in the visible light image according to the jetting position data, a blur score is obtained for each pixel position in the image, including: Acquire a visible light image sequence that includes a visible light image of the target; Based on the jetting position data, the motion trajectory corresponding to the jetting position data in the visible light image sequence is obtained; In each visible light image in the visible light image sequence, the neighborhood of the motion trajectory is selected as the ROI region; The degree of blurring of the target pixel location in the ROI region of each visible light image in the visible light image sequence is quantized to obtain multiple blurring quantization values ​​of the target pixel location; The average value of the quantized blur level is taken as the blur score of the target pixel position in the visible light.

[0009] In a preferred implementation: based on the temperature information in the thermal infrared image, and combined with the corresponding position of the jetting location data in the thermal infrared image, a temperature anomaly score is obtained for each pixel location in the image, including: Based on the temperature information from the thermal infrared image, the initial temperature gradient at the target pixel location is obtained; Based on the position of the jetting location data in the thermal infrared image, the initial temperature gradient is corrected to obtain the temperature gradient score of the target pixel position. Based on the visible light image, obtain the grayscale value of the target pixel location; By combining the temperature gradient score and the grayscale value, the temperature anomaly score of the target pixel location is obtained.

[0010] In a preferred implementation: the probability of dust presence is used as the reliability of image features. Based on the image features of the visible light image and the thermal infrared image, the dust accumulation state at each pixel location in the image is obtained, including: Image features are extracted from visible light and thermal infrared images and mapped to obtain an initial type score for each pixel location in the image. The magnitude of the type score represents the severity of dust accumulation. Using the probability of dust presence as the confidence level of image features, the initial type score is corrected to obtain the actual type score for each pixel location in the image; Based on the actual type score, the dust type at each pixel location in the image is obtained as a type of data in the dust state.

[0011] In a preferred implementation: image features are extracted from the visible light image and the thermal infrared image, and these image features are mapped to obtain an initial type score for each pixel location in the image, including: Based on a preset image recognition model, semantic segmentation is performed on visible light images to obtain a gray accumulation probability map. Threshold segmentation is performed on the thermal infrared image based on a preset threshold to obtain a coking probability map; Based on the preset infrared thermal resistance model, the ash accumulation thickness distribution map is obtained from the thermal infrared image; The initial type score for each pixel in the image is obtained by summing the values ​​at the same pixel location in the ash accumulation probability map, the coking probability map, and the ash accumulation thickness distribution map.

[0012] In a preferred implementation: based on the dust accumulation state, a control strategy for the sootblower is obtained and driven accordingly, including: Based on the dust accumulation status, the value score of each location in the target soot blowing area is obtained, where the value score is used to characterize the degree of impact of soot blowing at that location on boiler efficiency; Based on the probability of dust presence, the interference score is obtained for each location in the target dust blowing area. The interference score is used to represent the degree of risk that dust will cause a decrease in image quality at that location. Based on the value score and the disturbance score, the control strategy for the soot blower is obtained and driven.

[0013] The present invention also provides a visual monitoring-based sootblower drive control system, comprising: The data acquisition module is used to acquire visible light and thermal infrared images of the target soot blowing area, as well as the blowing position data of the soot blower in the target soot blowing area; The dust analysis module is used to analyze the probability of dust presence at each pixel location in the image based on the image features of visible light and thermal infrared images, combined with the jetting location data. The image analysis module is used to determine the reliability of image features based on the probability of dust presence. It obtains the dust accumulation status at each pixel location in the image based on the image features of visible light and thermal infrared images. The drive control module is used to obtain the control strategy of the sootblower based on the dust accumulation status and drive it.

[0014] The present invention also provides an electronic device, comprising: Memory and processor; The memory is used to store the program, and the processor is used to execute the steps in any of the above-described visual monitoring-based sootblower drive control methods when the program is executed.

[0015] The present invention also provides a computer-readable storage medium for storing a computer-readable program or instruction, which, when executed by a processor, can implement the steps in any of the above-described visual monitoring-based sootblower drive control methods.

[0016] The beneficial effects of adopting the above scheme are: This invention provides a visual monitoring-based sootblower drive control method. First, it acquires visible light and thermal infrared images of the target sootblowing area, as well as the sootblower's blowing position data within that area. Then, based on the image features of the visible light and thermal infrared images, combined with the blowing position data, it analyzes the probability of dust presence at each pixel location in the image. Next, using the dust presence probability as the reliability of the image features, it obtains the dust accumulation state at each pixel location based on the image features of the visible light and thermal infrared images. Finally, based on the dust accumulation state, it obtains the control strategy for the sootblower and drives it accordingly. Compared to existing technologies, this invention achieves accurate calculation of the dust presence probability by fusing multimodal information from visible light and thermal infrared images and combining it with real-time blowing position data of the sootblower, thereby effectively distinguishing between actual dust accumulation and dust interference. Meanwhile, by using the probability of dust presence as the credibility weight of image features, the image analysis process is dynamically optimized, avoiding the misjudgment problem caused by dust interference in traditional methods, ensuring the reliability of dust accumulation status assessment, and finally adaptively generating the optimal sootblower control strategy based on the high-precision dust accumulation status analysis results, achieving precise blowing and efficient dust removal, and solving the problem that the control accuracy and efficiency of sootblowers in the existing technology are affected by dust. Attached Figure Description

[0017] Figure 1 A flowchart illustrating the visual monitoring-based sootblower drive control method provided by this invention; Figure 2 for Figure 1 A detailed step diagram of step S102 is shown below; Figure 3 for Figure 1 A detailed step diagram of step S103 is shown below; Figure 4 The system architecture diagram of the soot blower drive control system based on visual monitoring provided by the present invention is shown. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] Soot blowers, also known as soot blowing equipment, are key devices in industrial boilers, heating furnaces, waste heat boilers, and other thermal equipment used to remove ash, slag, and scale from heated surfaces. Their core function is to efficiently clean heat exchange surfaces through physical methods (such as mechanical impact, high-speed airflow, or sonic vibration) to maintain equipment thermal efficiency, prevent corrosion and clogging, and extend service life. Specific types of soot blowers include shock wave soot blowers, steam soot blowers, and sonic soot blowers. These devices are typically used in combination based on boiler structure, fuel characteristics, and ash type to form targeted soot blowing solutions, ensuring long-term stable and efficient operation of the thermal system. It is understood that the soot blower mentioned in this invention is not limited to any single type of soot blowing equipment.

[0020] Taking a rake-type soot blower as an example, its main components include: Shock tank: As a power source, a large-capacity (e.g., 200L) shock tank can ensure a sufficient supply of gas.

[0021] Motor: Used to drive the soot blower's forward and backward movements.

[0022] Soot blowing control box: Used to control the operation of the soot blower, including forward and backward operations.

[0023] Rake rod main pipe: As the main structure of the soot blower, it connects the shock tank and the rake rod nozzle. It is mainly used to extend into the inside of equipment that requires soot blowing, such as boilers or heat exchangers, and it extends and retracts inside the equipment under the drive of a motor.

[0024] Rake-bar nozzle: The nozzle is equipped with nozzles for blowing shockwave airflow to remove accumulated ash. The nozzles are generally tapered at the front and coarser at the back, with a duckbill-shaped outlet. This design allows for a more concentrated airflow and greater blowing force. When the heated surface tubes are arranged in a straight line, the lower part of the nozzle is positioned directly opposite the tube gap; when the heated surface tubes are arranged in a staggered manner, the nozzles are staggered according to the angle of the tube gaps to ensure optimal blowing effect.

[0025] Support: Used to fix the soot blower rake rod and ensure its stable operation.

[0026] The working steps of existing soot blowers generally include: Step 1: Inflate the shock tank (inflation time can be adjusted according to actual needs).

[0027] Step 2: The shock tank releases for a second, and the dust is instantly blown out through the nozzle of the soot blower, using high-pressure airflow to remove the accumulated dust.

[0028] Step 3: Start the motor and the soot blower moves forward (the length of the advance can be determined according to the experimental running time).

[0029] Each time the soot blower sprays air, it moves forward one step. After reaching the forward limit switch, the soot blower moves backward one step. After reaching the backward limit switch, it stops running, completing one operating cycle.

[0030] It can be seen that existing sootblower technologies rely on fixed-strategy control, which cannot be flexibly adjusted according to actual conditions. For example, when coking occurs on a certain part of the heat exchange tube, the fixed-strategy blowing method may not be effective in removing it. Combining visual monitoring with sootblower control is a feasible approach, but it's conceivable that the sootblower generates a large amount of dust during blowing, significantly reducing visibility inside the equipment and directly affecting the visual monitoring-based sootblowing quality assessment and control. If image acquisition is performed after stopping the machine and waiting for the dust to dissipate, it will prolong the single sootblowing cycle, reduce overall cleaning efficiency, and fail to meet the demands of efficient and continuous cleaning in industrial production. Therefore, this invention provides a visual monitoring-based sootblower drive control method to solve the above problems.

[0031] Combination Figure 1 As shown, a specific embodiment of the present invention discloses a visual monitoring-based sootblower drive control method, comprising: S101. Acquire visible light and thermal infrared images of the target soot blowing area, as well as the blowing position data of the soot blower in the target soot blowing area; S102. Based on the image features of the visible light image and the thermal infrared image, and combined with the blowing location data, analyze the probability of dust presence at each pixel location in the image; S103. Using the probability of dust presence as the credibility of image features, the dust accumulation status of each pixel in the image is obtained based on the image features of the visible light image and the thermal infrared image. S104. Based on the dust accumulation status, obtain the control strategy for the sootblower and drive it.

[0032] In the above, the blowing position data can be indirectly obtained from the position of the nozzle in the soot blower, and the nozzle position can be indirectly obtained from the specifications and dimensions of the equipment and the operating parameters of the motor, etc. Visible light images and thermal infrared images can also be obtained through existing methods such as installing a high-temperature resistant camera inside the equipment. It should be noted that the image content corresponding to visible light images and thermal infrared images is the same. Therefore, in this embodiment, "each pixel position in the image" can refer to a pixel in the visible light image or a pixel in the thermal infrared image. They are essentially the same and both refer to a position in the target soot blowing area.

[0033] Image features refer to features that can reflect information about the target blowing area, obtained from visible light images and thermal infrared images using any existing method. These features can be flexibly set according to actual needs.

[0034] Combination Figure 2 As shown, in a preferred embodiment, step S102, based on the image features of the visible light image and the thermal infrared image, and combined with the blowing location data, analyzes the probability of dust presence at each pixel location in the image, specifically including: S201. Based on the blurring degree of the corresponding region in the visible light image according to the blowing position data, obtain the blur score of each pixel position in the image; S202. Based on the temperature information in the thermal infrared image, and combined with the corresponding position of the blowing position data in the thermal infrared image, obtain the temperature anomaly score for each pixel position in the image. S203. Based on the real-time operating conditions of the target dust blowing area, fuse the fuzzy score and the temperature anomaly score to obtain the probability of dust presence at each pixel location in the image.

[0035] In this embodiment, by utilizing the jetting location data and calculating a fuzziness score based on the degree of blurring in the visible light image, dynamic dust interference caused by jetting is identified from a visual perspective. Simultaneously, by combining temperature information from the thermal infrared image and the jetting location data, a temperature anomaly score is calculated based on the characteristic that the temperature change in the dust area lags behind the heated metal surface due to the poor thermal conductivity of particles, thus distinguishing between real dust accumulation and instantaneous high-temperature dust from a thermodynamic perspective. Finally, the fuzziness score and temperature anomaly score are further fused, and the weights are dynamically adjusted considering real-time operating conditions to obtain a precise dust presence probability distribution. This embodiment overcomes the limitations of single sensors being susceptible to interference and bypasses the analysis of complex gas movements inside the equipment, enabling reasoning about dust states under complex operating conditions. This provides a more reliable data foundation for subsequent dust accumulation state assessment and soot blowing control strategy optimization.

[0036] Specifically, in a preferred embodiment, step S201, which obtains a blur score for each pixel position in the image based on the blur degree of the corresponding region in the visible light image according to the blowing position data, specifically includes: Acquire a visible light image sequence that includes a visible light image of the target; Based on the jetting position data, the motion trajectory corresponding to the jetting position data in the visible light image sequence is obtained; In each visible light image in the visible light image sequence, the neighborhood of the motion trajectory is selected as the ROI region; The degree of blurring of the target pixel location in the ROI region of each visible light image in the visible light image sequence is quantized to obtain multiple blurring quantization values ​​of the target pixel location; The average value of the quantized blur level is taken as the blur score of the target pixel position in the visible light.

[0037] It is understood that the aforementioned target visible light image is the visible light image to be analyzed currently, and the visible light image sequence refers to multiple visible light images within a time window before the current acquisition of the target visible light image. Furthermore, the blur quantization value refers to any existing index capable of quantifying the degree of blur in pixels within an image. For example, the blur degree can be quantified using simple methods such as calculating the average pixel value, or it can be quantified using more complex existing image recognition algorithms (such as using Gaussian convolution kernels).

[0038] This embodiment accurately locates the dynamic region of interest (ROI) affected by dust by tracking the motion trajectory of the jetting position. Utilizing jetting position data avoids misjudgments that might occur with other algorithms. Simultaneously, it employs an averaging method of quantified blur values ​​from multiple image frames, effectively smoothing out noise interference that may exist in a single frame image, significantly improving the stability and accuracy of the fuzzy score. This embodiment can adapt to the dust diffusion characteristics under different jetting intensities and operating conditions, dynamically adjusting the ROI range and fuzzy evaluation parameters to achieve accurate quantification of dust interference. This temporal fuzzy evaluation method based on motion trajectory analysis overcomes the limitations of static image analysis in dynamic dust scenarios, significantly enhancing the accuracy of dust identification.

[0039] A more specific embodiment of step S201 above is as follows: Nozzle motion trajectory extraction: Obtain the spray position data, connect the coordinates of the position data to obtain the trajectory, and map the trajectory onto the image.

[0040] Define the region of interest (ROI) around the trajectory: a rectangular area 50 pixels wide, centered on the trajectory (covering the main dust diffusion range, which can be adjusted according to the actual situation).

[0041] Blur quantization: Using optical flow velocity as the quantization value for blur level, for each pixel within the ROI, calculate its average optical flow velocity within the time window (the window of the acquired image sequence) to obtain the blur score. The specific formula is as follows: ; in, Represents a pixel position. pixel position Fuzzy score, This represents the optical flow velocity at that pixel location within a visible light image. This represents the total number of images in the visible light image sequence.

[0042] In this embodiment, the optical flow velocity can be obtained using any existing algorithm, such as the Lucas-Kanade algorithm. The optical flow velocity directly reflects the intensity of pixel motion in the image and has a clear physical correlation with the dynamic blurring effect caused by dust. It can capture the instantaneous dust characteristics during the blowing process more accurately than traditional blurring algorithms (such as the gradient variance method). In addition, optical flow calculation can simultaneously obtain information in two dimensions: motion direction and velocity magnitude. By quantizing the velocity components (especially the components perpendicular to the imaging plane), it is possible to more accurately distinguish between dust disturbances and normal imaging blur.

[0043] Furthermore, in a preferred embodiment, step S202, which involves obtaining a temperature anomaly score for each pixel in the image based on the temperature information in the thermal infrared image and the corresponding position of the jetting location data in the thermal infrared image, specifically includes: Based on the temperature information from the thermal infrared image, the initial temperature gradient at the target pixel location is obtained; Based on the position of the jetting location data in the thermal infrared image, the initial temperature gradient is corrected to obtain the temperature gradient score of the target pixel position. Based on the visible light image, obtain the grayscale value of the target pixel location; By combining the temperature gradient score and the grayscale value, the temperature anomaly score of the target pixel location is obtained.

[0044] This embodiment uses temperature gradient instead of absolute temperature as the basic feature, effectively eliminating interference from inherent temperature differences in different areas of the boiler, enabling the system to more sensitively capture instantaneous temperature anomalies caused by soot blowing. Furthermore, by dynamically correcting the initial temperature gradient using soot blowing location data, a temperature change model directly related to the soot blowing action is established, significantly reducing the probability of false positives. Simultaneously, this embodiment innovatively combines temperature gradient scores with visible light grayscale values, utilizing the temperature sensitivity of thermal infrared radiation while compensating for the limitations of thermal imaging in terms of material reflectivity differences through visible light information, forming a complementary judgment mechanism. This method is particularly suitable for scenarios with complex heat conduction characteristics, such as boiler tube banks, and can identify instantaneous temperature rises caused by soot blowing while eliminating false positives caused by environmental radiation or the thermal inertia of the tubes themselves, significantly improving the accuracy and reliability of temperature anomaly detection.

[0045] A more specific embodiment of step S202 above is as follows: The temperature gradient score at the target pixel location is calculated using the following formula: ; in, Indicates the target pixel position Temperature gradient score, This indicates the temperature obtained from the thermal infrared image. and Representing the target pixel positions respectively The coordinates. The first term in the formula represents the initial temperature gradient calculated through differentiation, and the second term... This means that the correction amount is obtained based on the spray position data and the coordinates of the target pixel position (the specific calculation method can be flexibly set according to actual needs). It can be understood that the closer the target pixel position is to the spray position, the greater the probability that it will be affected by the blowing at the current moment. At this time, the temperature reflected in the thermal infrared image at this position may be lower than the actual temperature.

[0046] After obtaining the temperature gradient score, anomaly assessment can be performed by combining it with the grayscale values ​​of the visible light image. ; in, Indicates the target pixel position Temperature anomaly score, It is a natural exponential function. It serves as the reference temperature gradient for the heated metal surface, used to measure the difference between the current pixel temperature gradient and the metal substrate. The temperature gradient distribution of the metal surface under normal operating conditions can be determined through calibration experiments or historical data. The location of the target pixel in the visible light image grayscale value, It serves as the reference grayscale value for the heated metal surface, used to distinguish the metal substrate from the dust accumulation area. Similarly, the typical grayscale range of the metal surface can be determined through image segmentation or calibration experiments. and These are adjustable normalization parameters, which can also be obtained from experiments or experience.

[0047] After obtaining the fuzzy score and temperature anomaly score, step S203 can be performed: based on the real-time operating conditions of the target dust blowing area, the fuzzy score and temperature anomaly score are fused to obtain the probability of dust presence at each pixel location in the image. A specific embodiment of this step is as follows: First, dynamically allocate weights based on real-time operating conditions: In high-load conditions (such as steam flow rate > 80%), the airflow disturbance is strong and motion fuzziness dominates. The weight ratio of fuzzy score and temperature anomaly score can be set to 7:3.

[0048] If it is a low-load operating condition (such as steam flow rate <30%), the dust movement is weakened and the thermal characteristics are more significant. In this case, the weight ratio of fuzzy score and temperature anomaly score can be set to 3:7.

[0049] Then, by weighting and summing the fuzzy score and the temperature anomaly score based on the above weights, the probability of dust pollution can be obtained.

[0050] Furthermore, it's conceivable that if the probability of dust presence is high at a certain location, the accuracy of the dust accumulation status represented by image features will be poor. Therefore, this invention uses the probability of dust presence as the reliability of image features for further processing, allowing the system to optimize control strategies only for locations with high reliability (i.e., locations where dust is not present). For locations with poor accuracy, image acquisition and dust blowing can be performed only when the sootblower moves to other locations (changing the dust-affected area), thereby improving overall soot blowing efficiency without stopping the system and reducing the frequency of blowing on areas that do not require soot blowing, thus avoiding waste of air source energy. Specifically, in combination Figure 3 As shown, in a preferred embodiment, step S103, using the probability of dust presence as the reliability of image features, and obtaining the dust accumulation state at each pixel location in the image based on the image features of the visible light image and the thermal infrared image, specifically includes: S301. Extract image features from visible light and thermal infrared images and map the image features to obtain an initial type score for each pixel location in the image, where the magnitude of the type score represents the severity of dust accumulation. S302. Using the probability of dust presence as the credibility of image features, the initial type score is corrected to obtain the actual type score of each pixel position in the image. S303. Based on the actual type score, obtain the dust type at each pixel location in the image as a type of data in the dust state.

[0051] This embodiment employs a dynamic reliability weighting method, which adaptively corrects the initial type score based on the dust probability. This effectively suppresses misjudgment interference in high-dust areas, enabling the system to accurately identify the actual dust accumulation area (actual type score). Simultaneously, in conjunction with the aforementioned "spatiotemporal misalignment optimization" strategy, control decisions for high-dust areas are temporarily suspended. Data collection and analysis are then performed after the dust dissipates following the displacement of the sootblower. This significantly improves the accuracy of dust accumulation status assessment, avoiding efficiency losses caused by downtime and enabling continuous operation, thereby greatly improving sootblowing efficiency.

[0052] Most importantly, in the application scenario of this invention, there is a positive correlation between the degree of dust accumulation and the severity represented by different dust accumulation types. Simultaneously, the severity of different dust accumulation types is also positively correlated with their cleaning requirements, and the accuracy of image recognition of pixel locations (affected by dust) is also positively correlated with their cleaning requirements. Therefore, this embodiment cleverly utilizes this point, mapping image features and, through a hierarchical processing mechanism (initial score - credibility correction - final state determination), implicitly incorporating the probability of dust presence into the information representation of dust accumulation types. Pixel locations that cannot be accurately determined are converted into type scores that do not require cleaning. While ensuring judgment accuracy, it intelligently reduces unnecessary cleaning frequency (only cleaning areas that clearly require cleaning), effectively saving air source energy. This method, which combines image credibility assessment, dynamic decision optimization, and resource management, breaks through the traditional "one-size-fits-all" control mode of soot blowing systems, maximizing energy efficiency while improving cleaning quality, and possesses significant technological advancement and engineering practical value. Furthermore, it's conceivable that other parameters can be combined for comprehensive judgment in practice. For example, for superheaters, changes in steam temperature need to be monitored; for economizers, changes in water temperature are key; and for air preheaters, changes in air temperature are crucial. In practice, visual analysis can be combined with other existing technologies to more accurately identify the shape, adhesion, and other physical properties of ash.

[0053] Specifically, after obtaining image features, any existing method such as neural networks can be used to map the image features, while this invention provides a more convenient and preferred method. In a preferred embodiment, step S301 above, extracting image features from visible light images and thermal infrared images and mapping the image features to obtain an initial type score for each pixel position in the image, specifically includes: Based on a preset image recognition model, semantic segmentation is performed on visible light images to obtain a gray accumulation probability map. Threshold segmentation is performed on the thermal infrared image based on a preset threshold to obtain a coking probability map; Based on the preset infrared thermal resistance model, the ash accumulation thickness distribution map is obtained from the thermal infrared image; The initial type score for each pixel in the image is obtained by summing the values ​​at the same pixel location in the ash accumulation probability map, the coking probability map, and the ash accumulation thickness distribution map.

[0054] This embodiment utilizes a ternary fusion of a preset model (semantic segmentation), a physical model (infrared thermal resistance), and threshold segmentation. It uses these three image features to construct a multi-dimensional initial score that includes the probability of dust accumulation, the probability of coking, and the thickness distribution, so that the severity of different dust accumulation types can be accurately characterized. On this basis, the dust accumulation type can be cleverly transformed into a type score adjustment mechanism, so that areas that cannot be accurately judged will automatically receive a low score (i.e., low blowing priority), avoiding unnecessary blowing caused by misjudgment in traditional methods.

[0055] A more specific embodiment of the above steps S301-S303 is as follows: In visible light images, the U-Net model is used as the preset image recognition model to perform semantic segmentation and output a graying probability map (the pixel values ​​in the image can be represented by 0-1 to indicate the degree of graying; in practice, any other existing neural network model can also be used as the preset image recognition model). In infrared images, the temperature of the focal region is higher than the surrounding temperature, so a focalization probability map can be obtained through threshold segmentation (pixel positions exceeding the threshold are determined to be focal).

[0056] The thickness can be calculated using the following preset infrared thermal resistance model: ; in, For the target pixel position The thickness of the dust accumulation, The preset thermal resistance coefficient depends on factors such as the thermal conductivity of the ash accumulation material and the geometry of the heated surface. It can be calibrated through experiments or by fitting field data. The gas temperature can be obtained by installing a temperature sensor at the flue gas emission outlet. The temperature of the heated surface can also be obtained using embedded thermocouples or by detecting the temperature of the medium discharged from the heat exchange tubes. For, the location of the target pixel in the thermal infrared image The temperature.

[0057] After obtaining the above indicators, the initial type score for each pixel in the image can be obtained by summing the values ​​at the same pixel locations. Then, the probability of dust presence is used as the confidence level for correction. The formula for this process is as follows: ; in, For the target pixel position The actual type score, The target pixel position in the dust accumulation probability map The value, The target pixel position in the focus probability map The value, The probability of dust pollution exists. , and These are preset coefficients used for weight allocation, unit adjustment, etc. The result of the weighted sum is the initial type score.

[0058] Then, based on the actual type score mentioned above, the dust accumulation type can be determined, for example: A score greater than 4 indicates coking, 3-4 indicates sticky ash, 2-3 indicates loose ash, and less than 2 indicates no ash accumulation or possible dust accumulation. The image recognition result is unreliable.

[0059] It is understood that the dust accumulation status, in addition to the dust accumulation types mentioned above, can also include any other relevant data such as dust accumulation thickness and temperature. Therefore, further, in a preferred embodiment, step S104 above, obtaining the control strategy for the sootblower based on the dust accumulation status and driving it, specifically includes: Based on the dust accumulation status, the value score of each location in the target soot blowing area is obtained, where the value score is used to characterize the degree of impact of soot blowing at that location on boiler efficiency; Based on the probability of dust presence, the interference score is obtained for each location in the target dust blowing area. The interference score is used to represent the degree of risk that dust will cause a decrease in image quality at that location. Based on the value score and the disturbance score, the control strategy for the soot blower is obtained and driven.

[0060] In this embodiment, the value score directly correlates dust accumulation status with boiler efficiency improvement. By quantifying the impact of soot removal at each location on overall thermal efficiency, it ensures that soot blowing resources are prioritized for high-value areas, significantly improving the energy efficiency ratio of soot removal operations. The innovative design of the interference score transforms the probability of dust presence into a quantifiable risk indicator, enabling the system to intelligently avoid inefficient blowing in high-dust areas, ensuring both image acquisition quality and energy waste. For example, if dust persists at a certain location for an extended period (with a persistently high probability of dust presence), the strategy can be adjusted to reduce the number of nearby blowing operations, ensuring image acquisition quality and allowing the system to accurately acquire furnace information and make decisions. This embodiment, through a value-interference dynamic balance algorithm, achieves intelligent decision-making of "prioritizing high-value, low-interference operations and avoiding low-value, high-interference operations," significantly reducing ineffective blowing operations while ensuring boiler efficiency and lowering gas consumption.

[0061] A more specific embodiment of step S104 above is as follows: Value scores are calculated based on the following criteria: Dust accumulation type: coking (most difficult to remove) > sticky dust > loose dust > no dust accumulation.

[0062] Dust accumulation thickness: The greater the thickness, the higher the value score (e.g., the score increases significantly when the thickness is >2mm).

[0063] Thermal resistance contribution: The greater the thermal resistance caused by dust accumulation (calculated using a preset infrared thermal resistance model), the higher the score.

[0064] The calculation formula is as follows: ; in, For position The value score, , and These are preset weights. The meaning is that the more difficult the dust accumulation is to remove, the greater its impact on heat transfer, and the higher the score.

[0065] Value scores are calculated based on the following criteria: Dust presence probability: The higher the dust presence probability, the stronger the dust interference, and the higher the interference score.

[0066] Historical dust frequency: The more times the probability of dust occurrence in a certain area is greater than the set threshold in the past N periods, the higher the interference score.

[0067] Jet sensitivity: Jet blowing is more likely to generate dust in densely structured areas, resulting in a higher interference score.

[0068] The calculation formula is as follows: ; in, For position The value score, For position Historical dust frequency For position The spray sensitivity can be preset manually. , and These are also preset weights.

[0069] After obtaining the above scores, the strategy can be adjusted: Objective: To dynamically adjust the blowing time and position based on dust conditions and confidence level, while reducing the blowing frequency in high-dust areas to improve image quality.

[0070] The specific process for adjusting the blowing time includes: For areas with high-value scores: increase blowing time to ensure removal of stubborn dust buildup.

[0071] For areas with high interference scores: reduce the blowing time to avoid excessive dust generation.

[0072] Other areas: Continue spraying according to the existing strategy.

[0073] The specific process for adjusting the spray position includes: Prioritize coverage of high-value score areas: Generate nozzle movement paths using path planning algorithms (such as the A* algorithm) to ensure that high-value score areas are fully covered.

[0074] Avoid high interference score areas: Set "no-go zones" in path planning (e.g., 50mm around high interference score areas), or reduce the nozzle's dwell frequency in these areas.

[0075] The above process can also be implemented through optimization algorithms. For example, by encoding the spraying time and spraying position into a two-dimensional vector group (e.g., one-hot encoding of multiple positions), and then combining the value score and interference score to establish a function to evaluate the merits of the two-dimensional vector group, optimization algorithms such as genetic algorithms can be used to achieve optimization.

[0076] Combination Figure 4 As shown, the present invention also provides a visual monitoring-based sootblower drive control system, comprising: The data acquisition module 410 is used to acquire visible light images and thermal infrared images of the target soot blowing area, as well as the blowing position data of the soot blower in the target soot blowing area; The dust analysis module 420 is used to analyze the probability of dust presence at each pixel location in the image based on the image features of the visible light image and the thermal infrared image, combined with the jetting location data. The image analysis module 430 is used to determine the credibility of image features based on the probability of dust presence, and obtains the dust accumulation status at each pixel location in the image based on the image features of the visible light image and the thermal infrared image. The drive control module 440 is used to obtain the control strategy of the sootblower based on the dust accumulation state and drive it.

[0077] It should be noted that the corresponding systems provided in the above embodiments are computer program products that can implement the technical solutions described in the above method embodiments. The specific implementation principles of the above modules or units can be found in the corresponding content in the above method embodiments, and will not be repeated here.

[0078] The present invention also provides an electronic device, comprising: Memory and processor; The memory is used to store the program, and the processor is used to execute the steps in any of the above-described visual monitoring-based sootblower drive control methods when the program is executed.

[0079] The present invention also provides a computer-readable storage medium for storing a computer-readable program or instruction, which, when executed by a processor, can implement the steps in any of the above-described visual monitoring-based sootblower drive control methods.

[0080] This invention provides a visual monitoring-based sootblower drive control method. First, it acquires visible light and thermal infrared images of the target sootblowing area, as well as the sootblower's blowing position data within that area. Then, based on the image features of the visible light and thermal infrared images, combined with the blowing position data, it analyzes the probability of dust presence at each pixel location in the image. Next, using the dust presence probability as the reliability of the image features, it obtains the dust accumulation state at each pixel location based on the image features of the visible light and thermal infrared images. Finally, based on the dust accumulation state, it obtains the control strategy for the sootblower and drives it accordingly. Compared to existing technologies, this invention achieves accurate calculation of the dust presence probability by fusing multimodal information from visible light and thermal infrared images and combining it with real-time blowing position data of the sootblower, thereby effectively distinguishing between actual dust accumulation and dust interference. Meanwhile, by using the probability of dust presence as the credibility weight of image features, the image analysis process is dynamically optimized, avoiding the misjudgment problem caused by dust interference in traditional methods, ensuring the reliability of dust accumulation status assessment, and finally adaptively generating the optimal sootblower control strategy based on the high-precision dust accumulation status analysis results, achieving precise blowing and efficient dust removal, and solving the problem that the control accuracy and efficiency of sootblowers in the existing technology are affected by dust.

[0081] It should be noted that the various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. The same or similar parts between the various embodiments can be referred to each other.

[0082] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A sootblower drive control method based on visual monitoring, characterized in that, include: Acquire visible light and thermal infrared images of the target soot blowing area, as well as the blowing position data of the soot blower in the target soot blowing area; Based on the image features of visible light and thermal infrared images, combined with the blowing location data, the probability of dust presence at each pixel location in the image is analyzed. Using the probability of dust presence as the reliability of image features, the dust accumulation status at each pixel location in the image is obtained based on the image features of visible light images and thermal infrared images. Based on the dust accumulation status, a control strategy for the sootblower is obtained and driven accordingly.

2. The soot blower drive control method based on visual monitoring according to claim 1, characterized in that, Based on the image features of visible light and thermal infrared images, combined with the jetting location data, the probability of dust presence at each pixel location in the image is analyzed, including: Based on the degree of blur in the corresponding region of the visible light image according to the jetting position data, a blur score is obtained for each pixel position in the image; Based on the temperature information in the thermal infrared image, and combined with the corresponding position of the jetting location data in the thermal infrared image, a temperature anomaly score is obtained for each pixel in the image. Based on the real-time operating conditions of the target dust blowing area, the fuzz score and temperature anomaly score are fused to obtain the probability of dust presence at each pixel location in the image.

3. The soot blower drive control method based on visual monitoring according to claim 2, characterized in that, Based on the blur level of the corresponding region in the visible light image according to the jetting position data, a blur score is obtained for each pixel position in the image, including: Acquire a visible light image sequence that includes a visible light image of the target; Based on the jetting position data, the motion trajectory corresponding to the jetting position data in the visible light image sequence is obtained; In each visible light image in the visible light image sequence, the neighborhood of the motion trajectory is selected as the ROI region; The degree of blurring of the target pixel location in the ROI region of each visible light image in the visible light image sequence is quantized to obtain multiple blurring quantization values ​​of the target pixel location; The average value of the quantized blur level is taken as the blur score of the target pixel position in the visible light.

4. The soot blower drive control method based on visual monitoring according to claim 2, characterized in that, Based on the temperature information in the thermal infrared image, and combined with the corresponding location of the jetting position data in the thermal infrared image, a temperature anomaly score is obtained for each pixel location in the image, including: Based on the temperature information from the thermal infrared image, the initial temperature gradient at the target pixel location is obtained; Based on the position of the jetting location data in the thermal infrared image, the initial temperature gradient is corrected to obtain the temperature gradient score of the target pixel position. Based on the visible light image, obtain the grayscale value of the target pixel location; By combining the temperature gradient score and the grayscale value, the temperature anomaly score of the target pixel location is obtained.

5. The soot blower drive control method based on visual monitoring according to claim 1, characterized in that, Using the probability of dust presence as the reliability of image features, the dust accumulation status at each pixel location in the image is obtained based on the image features of visible light and thermal infrared images, including: Image features are extracted from visible light and thermal infrared images and mapped to obtain an initial type score for each pixel location in the image. The magnitude of the type score represents the severity of dust accumulation. Using the probability of dust presence as the confidence level of image features, the initial type score is corrected to obtain the actual type score for each pixel location in the image; Based on the actual type score, the dust type at each pixel location in the image is obtained as a type of data in the dust state.

6. The soot blower drive control method based on visual monitoring according to claim 5, characterized in that, Image features are extracted from visible light and thermal infrared images and mapped to obtain an initial type score for each pixel location in the image, including: Based on a preset image recognition model, semantic segmentation is performed on visible light images to obtain a gray accumulation probability map. Threshold segmentation is performed on the thermal infrared image based on a preset threshold to obtain a coking probability map; Based on the preset infrared thermal resistance model, the ash accumulation thickness distribution map is obtained from the thermal infrared image; The initial type score for each pixel in the image is obtained by summing the values ​​at the same pixel location in the ash accumulation probability map, the coking probability map, and the ash accumulation thickness distribution map.

7. The soot blower drive control method based on visual monitoring according to claim 1, characterized in that, Based on the dust accumulation status, a control strategy for the sootblower is obtained and its drive is controlled, including: Based on the dust accumulation status, the value score of each location in the target soot blowing area is obtained, where the value score is used to characterize the degree of impact of soot blowing at that location on boiler efficiency; Based on the probability of dust presence, the interference score is obtained for each location in the target dust blowing area. The interference score is used to represent the degree of risk that dust will cause a decrease in image quality at that location. Based on the value score and the disturbance score, the control strategy for the soot blower is obtained and driven.

8. A soot blower drive control system based on vision monitoring, characterized in that, include: The data acquisition module is used to acquire visible light and thermal infrared images of the target soot blowing area, as well as the blowing position data of the soot blower in the target soot blowing area; The dust analysis module is used to analyze the probability of dust presence at each pixel location in the image based on the image features of visible light and thermal infrared images, combined with the jetting location data. The image analysis module is used to determine the reliability of image features based on the probability of dust presence. It obtains the dust accumulation status at each pixel location in the image based on the image features of visible light and thermal infrared images. The drive control module is used to obtain the control strategy of the sootblower based on the dust accumulation status and drive it.

9. An electronic device, characterized in that, include: Memory and processor; The memory is used to store the program, and the processor is used to execute the steps of any one of the vision monitoring-based sootblower drive control methods of claims 1-7 when executing the program.

10. A computer-readable storage medium, characterized in that, Used to store computer-readable programs or instructions, which, when executed by a processor, are capable of implementing the steps in any one of the visual monitoring-based sootblower drive control methods of claims 1-7.

Images