Boiler leakage prevention online monitoring method and monitoring system

Abstract

The invention discloses a boiler leakage prevention on-line monitoring method and system, and relates to the technical field of boiler leakage monitoring, and the method comprises the following steps: S100, collecting an infrared image and a visible light image of a monitoring area in real time through an image collection module, and transmitting the data to a preprocessing module; s200, a preprocessing module carries out denoising and edge strengthening processing, conventional edge features are reserved, and tiny temperature difference features are focused; s300, the control module rejects interference factors based on the similarity of the infrared image and the visible light image to obtain a suspected leakage image; s400, carrying out data acquisition on the suspected leakage region by using an acoustic imaging module, analyzing and recording the region data by a deep learning module based on historical detection data, and constructing a time sequence model; and S500, the control module analyzes and judges the overlapping state of the pipeline at the leakage point according to the visible light image and the information collected by the sound wave imaging module, and locates a leakage source.

Description

Technical Field

[0001] This invention relates to the field of boiler leakage monitoring technology, and in particular to an online monitoring method and system for preventing boiler leakage. Background Technology

[0002] Boilers are core equipment in industrial sectors such as thermal power generation and chemical production, and their operational safety directly affects production efficiency and personnel safety. However, boiler tube leakage is a long-standing technical problem, accounting for more than 60% of all power plant accidents. Leaks are difficult to detect in their early stages, but if not addressed promptly, the eruption of high-pressure media can trigger a chain reaction of tube ruptures, equipment damage, and even unplanned unit shutdowns, resulting in significant economic losses.

[0003] Chinese Patent Application No. 2023108349753 discloses a method and device for detecting gas leaks in steam boilers based on image processing. The method includes: acquiring an image of the steam boiler to be detected, the image comprising a set of visible light images and a set of infrared images; determining a first abnormal region in each infrared image in the infrared image set, and determining a second abnormal region in each visible light image in the visible light image set; calculating the similarity between the two abnormal regions, and determining a human interference region within the first abnormal region based on the similarity; removing the human interference region from the first abnormal region to obtain a third abnormal region; and then determining whether the steam boiler is leaking. This invention eliminates human interference regions based on the similarity of abnormal regions in the visible light and infrared images of the steam boiler, reducing interference from humans and the steam boiler itself in the detection of steam boiler gas leaks, and improving detection speed and accuracy.

[0004] Similar to the existing technologies mentioned above, although dual-modal image monitoring can eliminate human interference, the dynamic thermal radiation generated by high-temperature components of the boiler body (such as pipes and valves) or steam emissions, along with the leaking gas, will show high-temperature anomalies in infrared images. Especially when local temperature anomalies are caused by pipe corrosion or scaling, infrared grayscale differences alone cannot effectively identify whether a leak has occurred. Furthermore, due to the high temperature inside the boiler room, steam is easily generated, leading to steam diffusion or insufficient light, which blurs the edge contours of the visible light images and reduces the accuracy of the normalized contour histogram, thus weakening its ability to eliminate human interference and increasing the misjudgment rate.

[0005] Secondly, in order to improve the utilization rate of boilers, the pipes laid on the boilers are usually laid in multi-layer pipe arrays. In the existing technology, dual-modal image monitoring relies on contour histogram comparison. When a leak is detected in a certain area, the overlapping pipes will form a composite hot spot. Visible light is also difficult to penetrate the multi-layer pipes, making it impossible to accurately locate the leak.

[0006] Finally, bilateral filtering is often used for image denoising. Compared to traditional Gaussian filtering, bilateral filtering is based on the geometric distance between pixels, with pixels closer to the center point having higher weights, thus preserving edge structure. It also relies on the gray-level similarity of pixel values, with pixels having lower gray-level differences having higher weights. This suppresses noise while avoiding blurring edges, thereby distinguishing between edges and noise. However, this method is problematic when dealing with tiny leak areas. Since the infrared characteristics of tiny leaks are low-contrast, weak-gradient edges, similar to noise, bilateral filtering, which relies on gradient information to preserve edges, may not be able to effectively distinguish between the two. The pixel gray-level differences may be judged as noise by the value range weights and smoothed out, thus affecting the monitoring results and leading to misjudgments.

[0007] Therefore, it is necessary to invent an online monitoring method and system for boiler leakage prevention to solve the above problems. Summary of the Invention

[0008] The purpose of this invention is to provide an online monitoring method and system for boiler leakage prevention, so as to solve the problems mentioned in the background art.

[0009] To achieve the above objectives, the present invention provides the following technical solution: an online monitoring method for boiler leakage prevention, comprising the following steps: S100: The image acquisition module acquires infrared and visible light images of the monitored area in real time and transmits the data to the preprocessing module. S200, The preprocessing module performs noise reduction and edge enhancement on the data, retaining conventional edge features while focusing on minute temperature difference features, and transmits the processed data to the control module. S300, The control module eliminates interference factors based on the similarity between infrared images and visible light images to obtain suspected leak images; S400: The acoustic imaging module collects data on the suspected leak area to determine whether a leak exists, traces the leak point, and transmits the information synchronously to the control module and the deep learning module. The deep learning module analyzes and records the data in the area based on historical detection data and constructs a time series model. S500: The control module analyzes and judges the overlapping state of the pipeline at the leak point based on the information collected by the visible light image and the acoustic imaging module, and adjusts the position of the acoustic imaging module according to the judgment result to collect data from multiple angles at the leak point and locate the leak source.

[0010] Preferably, step S200 includes the following steps: S210, dual-modal data processing; normalized histogram reconstruction of infrared images, analysis of concentrated temperature distribution ranges, and location of low-contrast regions corresponding to minute leaks, marking them as target regions to be enhanced; bilateral filtering is used on visible light images to preserve structural details and improve the signal-to-noise ratio of minute leaks in infrared images; S220: Convolutional cloud feature extraction, constructing a high-order convolutional model, and inputting infrared and visible light images. By embedding multiplicative interactive convolutional kernels, multi-scale feature maps are output, attention weight maps of infrared and visible light features are calculated, spatial positions are aligned, pseudo-color mapping is performed on the fused feature map, low-concentration gas regions are magnified, and data is transmitted to the control module.

[0011] Preferably, in S300, gas leakage manifests as temperature anomalies in infrared images, but there is no corresponding entity in visible light images; human activity will form abnormal regions with similar outlines in both types of images. By comparing the similarity of abnormal regions in the two types of images, the real leakage and interference can be distinguished.

[0012] Preferably, S400 includes the following steps: S410. Use the acoustic imaging module to perform acoustic detection on the suspected leak area in step S300 and determine whether there is a leak. If yes, it means that there is a leak in this area; proceed to S500; if no, it means that the abnormality in this area is thermal radiation interference. Mark this area and transmit the data to the deep learning module, and proceed to S420. S420: The image acquisition module and the acoustic imaging module collect data on the area at different times and transmit the data to the deep learning module. The deep learning module analyzes whether any abnormality has occurred in the pipeline in the area based on historical data and real-time data and then proceeds to S430. S430. The deep learning module constructs a time series model based on the temperature change of the abnormal area and determines whether the temperature change of the area matches the historical data. If yes, it indicates that the pipeline in the area is in a normal state, the mark is removed, and the process proceeds to S100. If no, it indicates that corrosion or scaling has occurred in the pipeline in the area, causing a local temperature anomaly, and the data is fed back to the control module to proceed to S440. S440. Based on the data fed back by the deep learning module, the control module issues maintenance and alarm commands to prompt the operator to perform maintenance on this area.

[0013] Preferably, in S500, the visible light image can be used to determine whether the pipes in the area are arranged in an array. If they are arranged in an array, the drive module is used to move the acoustic imaging module to monitor the pipes in the area one by one. If it is a single pipe, the leak point is located directly through the acoustic imaging module.

[0014] Preferably, in S500, the control module identifies the structural features of the pipeline where the leak point is located based on visible light images, and constructs a three-dimensional sound field model by combining multi-angle acoustic signals collected by the acoustic imaging module. Through sound wave propagation path analysis and sound energy attenuation law, the specific location and leakage intensity of the leak source are inverted, thereby achieving accurate location of the leak point and assessment of the leakage amount.

[0015] The present invention also provides an online monitoring system for boiler leakage prevention, used to implement the above-mentioned monitoring method. The system includes an image acquisition module, a preprocessing module, an acoustic imaging module, a control module, and a data transmission module. The data transmission module is used to wirelessly transmit the data from the above modules.

[0016] Preferably, the image acquisition module includes an infrared image acquisition unit and a visible light acquisition unit, which are distributed in various areas of the detection scene to achieve global monitoring of the scene.

[0017] Preferably, the preprocessing module has a built-in convolutional neural unit and a similarity analysis unit. The similarity calculation unit uses the Bach distance of the gradient magnitude histogram to calculate the similarity between the infrared image and the visible light image. The convolutional neural unit is used to construct a high-order convolutional model, whose model architecture includes an input layer, a backbone branch, an interaction layer, a fusion layer, and an output layer. The input layer is connected to infrared light and visible light data. The backbone branch extracts data from the infrared image and the visible light image, respectively. The interaction layer has a built-in multiplication interaction module to realize dynamic feature weighting. The fusion layer can aggregate features across scales. The output layer can output multi-scale feature maps.

[0018] Preferably, the acoustic imaging module uses a multi-channel microphone array to synchronously capture reflected acoustic signals. The location of the sound source is determined by calculating the time difference and phase difference of the sound waves arriving at each microphone. The acoustic imaging module is mounted on the drive module and can follow the movement of the drive module to change its position.

[0019] The technical effects and advantages of this invention are as follows: 1. This invention integrates dual-modal image acquisition, intelligent preprocessing, acoustic imaging localization, and deep learning time series analysis to achieve high-precision, interference-resistant online monitoring of boiler leaks under high-temperature environments and accurate location of leak sources, significantly improving the reliability, sensitivity, and intelligence level of leak detection. Attached Figure Description

[0020] Figure 1 This is a schematic diagram of the monitoring method of the present invention.

[0021] Figure 2 This is a flowchart illustrating the sub-steps of step S200 of the present invention.

[0022] Figure 3 This is a flowchart illustrating the sub-steps of step S400 of the present invention.

[0023] Figure 4 This is a schematic diagram of the monitoring logic flow of the present invention. Detailed Implementation

[0024] 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.

[0025] To address the three core problems faced by traditional dual-modal image monitoring in boiler leak detection: misjudgment due to high temperature interference, the impact of multi-layered pipe structures on positioning accuracy, and the difficulty of existing denoising algorithms in preserving weak leak features.

[0026] like Figures 1 to 4 As shown, the first embodiment of the present invention provides a boiler leak prevention online monitoring system to solve the above-mentioned problems. The system includes an image acquisition module, a preprocessing module, an acoustic imaging module, a control module, and a data transmission module. The data transmission module is used to wirelessly transmit the data in the above modules.

[0027] In this embodiment, the image acquisition module includes an infrared image acquisition unit and a visible light acquisition unit. The infrared acquisition unit and the visible light acquisition unit are distributed in various areas of the detection scene, enabling global monitoring of the scene.

[0028] In this embodiment, the preprocessing module has a built-in convolutional neural unit and a similarity analysis unit. The similarity calculation unit uses the Bach distance of the gradient magnitude histogram to calculate the similarity between the infrared image and the visible light image. The convolutional neural unit is used to construct a high-order convolutional model. Its model architecture includes an input layer, a backbone branch, an interaction layer, a fusion layer, and an output layer. The input layer is connected to infrared light and visible light data. The backbone branch extracts infrared images and visible light images respectively. The interaction layer has a built-in multiplication interaction module to realize dynamic feature weighting. The fusion layer can aggregate features across scales. The output layer can output multi-scale feature maps.

[0029] In this embodiment, the acoustic imaging module uses a multi-channel microphone array to synchronously capture reflected acoustic signals. The location of the sound source is determined by calculating the time difference and phase difference of the sound waves arriving at each microphone. The acoustic imaging module is mounted on the driving module and can follow the movement of the driving module to change its position.

[0030] In another embodiment of the present invention, a detection method based on the monitoring system in the above embodiments is also provided, the detection method comprising the following steps: S100: The image acquisition module acquires infrared and visible light images of the monitoring area in real time and transmits the data to the preprocessing module.

[0031] The S200 preprocessing module performs noise reduction and edge enhancement on the data, retaining conventional edge features while focusing on minute temperature difference features, and then transmits the processed data to the control module.

[0032] In this implementation, S200 includes the following steps: S210, dual-modal data processing; normalized histogram reconstruction of infrared images, analysis of concentrated temperature distribution ranges, and location of low-contrast ranges corresponding to minute leaks, marking them as target regions to be enhanced; bilateral filtering is used on visible light images to preserve structural details and improve the signal-to-noise ratio of minute leaks in infrared images.

[0033] S220: Convolutional cloud feature extraction, constructing a high-order convolutional model, and inputting infrared and visible light images. By embedding multiplicative interactive convolutional kernels, multi-scale feature maps are output, attention weight maps of infrared and visible light features are calculated, spatial positions are aligned, pseudo-color mapping is performed on the fused feature map, low-concentration gas regions are magnified, and data is transmitted to the control module.

[0034] It should be noted that when plotting the gradient magnitude histogram, the Sobel operator is used to calculate the horizontal and vertical gradients for infrared and visible light images respectively: , Gradient increment formula: .

[0035] Where I is the pixel grayscale value matrix of the infrared or visible light image.

[0036] When constructing the histogram, the gradient magnitude range is divided into N equal-width intervals, and the number of pixels in each interval is counted to form a normalized probability distribution histogram.

[0037] The S300 control module eliminates interference factors based on the similarity between infrared and visible light images to obtain suspected leak images.

[0038] In this implementation, in S300, gas leakage manifests as temperature anomalies in infrared images, but there is no corresponding entity in visible light images; human activity will form anomaly regions with similar outlines in both types of images. By comparing the similarity of the anomaly regions in the two types of images, the real leakage and interference can be distinguished.

[0039] S400 uses the acoustic imaging module to collect data on suspected leak areas, determine whether a leak exists, trace the leak point, and transmit the information synchronously to the control module and deep learning module. The deep learning module analyzes and records the data in the area based on historical detection data and builds a time series model.

[0040] In this embodiment, S400 includes the following steps: S410. Use the acoustic imaging module to perform acoustic detection on the suspected leak area in step S300 and determine whether there is a leak. If yes, it means that there is a leak in this area; proceed to S500; if no, it means that the abnormality in this area is thermal radiation interference. Mark this area, transmit the data to the deep learning module, and proceed to S420.

[0041] S420: The image acquisition module and the acoustic imaging module collect data on the area at different times and transmit the data to the deep learning module. The deep learning module analyzes whether any abnormalities have occurred in the pipeline within the area based on historical data and real-time data and then proceeds to S430.

[0042] S430: The deep learning module constructs a time series model based on the temperature changes in the abnormal area and determines whether the temperature changes in the area match historical data. If so, it indicates that the pipeline in the area is in a normal state, the marking is removed, and the process proceeds to S100. If not, it indicates that corrosion or scaling has occurred in the pipeline in the area, causing local temperature anomalies, and the data is fed back to the control module, proceeding to S440.

[0043] It should be noted that a dual-channel LSTM was used to construct the time-series model, processing real-time temperature data and historical leakage records separately to capture long-term dependencies. An attention mechanism was added to focus on periods of abnormal temperature abrupt changes. The output layer was connected to a TimeDistributed Dense layer to achieve multi-step prediction. The training data for this model came from a historical database, i.e., boiler operation data accumulated for ≥6 months (including normal, leakage, corrosion, scaling, etc.). If the real-time monitoring data showed a slow, step-like increase in temperature, it indicated corrosion inside the pipeline; if the temperature suddenly rose and then plateaued, it indicated scaling inside the pipeline.

[0044] S440: Based on the data fed back by the deep learning module, the control module issues maintenance and alarm commands to prompt operators to perform maintenance on this area.

[0045] The S500 control module analyzes and judges the overlapping state of the pipeline at the leak point based on the information collected by the visible light image and the acoustic imaging module, and adjusts the position of the acoustic imaging module according to the judgment result to collect data from multiple angles at the leak point and locate the leak source.

[0046] In this implementation, in S500, the visible light image can be used to determine whether the pipes in the area are arranged in an array. If they are arranged in an array, the drive module is used to move the acoustic imaging module to monitor the pipes in the area one by one. If it is a single pipe, the leak point is located directly through the acoustic imaging module.

[0047] In this implementation, in S500, the control module identifies the structural features of the pipeline where the leak point is located based on visible light images, and constructs a three-dimensional sound field model by combining multi-angle acoustic signals collected by the acoustic imaging module. Through sound wave propagation path analysis and sound energy attenuation law, the specific location and leakage intensity of the leak source are inverted, thereby achieving accurate location of the leak point and assessment of the leakage amount.

[0048] It should be noted that when judging the overlapping state of pipelines, the pipeline outline is identified by visible light images, and stereo vision algorithms such as binocular imaging or structured light measurement are used to calculate the pipeline spacing, intersection angle, and depth information of the overlapping area. If the pipeline spacing is less than 1.5 times the pipe diameter and they are arranged in parallel, it is judged as array overlap; if the pipeline intersection angle is greater than 30°, it is marked as cross overlap. The overlapping area will form local temperature anomalies due to differences in heat conduction, such as the temperature gradient at the intersection point being significantly higher than that in the non-overlapping area. The temperature distribution pattern of infrared images is used to verify the overlapping state. When the drive module moves the acoustic imaging module, it moves along the pipeline axis and scans segment by segment at fixed intervals. A "cross grid scan" is used. First, the horizontal movement is used to locate the leakage plane, and then the vertical movement is used to determine the depth.

[0049] It should be noted that this invention integrates dual-modal image acquisition, intelligent preprocessing, acoustic imaging localization, and deep learning time series analysis to achieve high-precision, interference-resistant online monitoring of boiler leaks under high-temperature environments and accurate location of leak sources, significantly improving the reliability, sensitivity, and intelligence level of leak detection.

[0050] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for online monitoring of boiler leakage prevention, characterized in that, Includes the following steps: S100: The image acquisition module acquires infrared and visible light images of the monitored area in real time and transmits the data to the preprocessing module. S200, The preprocessing module performs noise reduction and edge enhancement on the data, retaining conventional edge features while focusing on minute temperature difference features, and transmits the processed data to the control module. S300, The control module eliminates interference factors based on the similarity between infrared images and visible light images to obtain suspected leak images; S400: The acoustic imaging module collects data on the suspected leak area to determine whether a leak exists, traces the leak point, and transmits the information synchronously to the control module and the deep learning module. The deep learning module analyzes and records the data in the area based on historical detection data and constructs a time series model. S500: The control module analyzes and judges the overlapping state of the pipeline at the leak point based on the information collected by the visible light image and the acoustic imaging module, and adjusts the position of the acoustic imaging module according to the judgment result to collect data from multiple angles at the leak point and locate the leak source.

2. The online monitoring method according to claim 1, characterized in that, S200 includes the following steps: S210, dual-modal data processing; normalized histogram reconstruction of infrared images, analysis of concentrated temperature distribution ranges, and location of low-contrast regions corresponding to minute leaks, marking them as target regions to be enhanced; bilateral filtering is used on visible light images to preserve structural details and improve the signal-to-noise ratio of minute leaks in infrared images; S220: Convolutional cloud feature extraction, constructing a high-order convolutional model, and inputting infrared and visible light images. By embedding multiplicative interactive convolutional kernels, multi-scale feature maps are output, attention weight maps of infrared and visible light features are calculated, spatial positions are aligned, pseudo-color mapping is performed on the fused feature map, low-concentration gas regions are magnified, and data is transmitted to the control module.

3. The online monitoring method according to claim 1, characterized in that, In S300, gas leaks appear as temperature anomalies in infrared images, but there is no corresponding entity in visible light images; human activity will form anomaly regions with similar outlines in both types of images. By comparing the similarity of the anomaly regions in the two types of images, we can distinguish between real leaks and interference.

4. The online monitoring method according to claim 1, characterized in that, The S400 includes the following steps: S410. Use the acoustic imaging module to perform acoustic detection on the suspected leak area in step S300 and determine whether there is a leak. If yes, it means that there is a leak in this area; proceed to S500; if no, it means that the abnormality in this area is thermal radiation interference. Mark this area and transmit the data to the deep learning module, and proceed to S420. S420: The image acquisition module and the acoustic imaging module collect data on the area at different times and transmit the data to the deep learning module. The deep learning module analyzes whether any abnormality has occurred in the pipeline in the area based on historical data and real-time data and then proceeds to S430. S430. The deep learning module constructs a time series model based on the temperature change of the abnormal area and determines whether the temperature change of the area matches the historical data. If yes, it indicates that the pipeline in the area is in a normal state, the mark is removed, and the process proceeds to S100. If no, it indicates that corrosion or scaling has occurred in the pipeline in the area, causing a local temperature anomaly, and the data is fed back to the control module to proceed to S440. S440. Based on the data fed back by the deep learning module, the control module issues maintenance and alarm commands to prompt the operator to perform maintenance on this area.

5. The online monitoring method according to claim 1, characterized in that, In the S500, the visible light image can be used to determine whether the pipes in the area are arranged in an array. If they are arranged in an array, the drive module is used to move the acoustic imaging module to monitor the pipes in the area one by one. If it is a single pipe, the leak point is located directly through the acoustic imaging module.

6. The online monitoring method according to claim 1, characterized in that, In S500, the control module identifies the structural features of the pipeline where the leak point is located based on visible light images, and constructs a three-dimensional sound field model by combining multi-angle acoustic signals collected by the acoustic imaging module. Through sound wave propagation path analysis and sound energy attenuation law, the specific location and leakage intensity of the leak source are inverted, thereby achieving accurate location of the leak point and leakage amount assessment.

7. A boiler leak prevention online monitoring system, used to implement the method as described in claim 1, characterized in that, The system includes an image acquisition module, a preprocessing module, an acoustic imaging module, a control module, and a data transmission module. The data transmission module is used to wirelessly transmit the data from the above modules.

8. The online monitoring system according to claim 7, characterized in that, The image acquisition module includes an infrared image acquisition unit and a visible light acquisition unit, which are distributed in various areas of the detection scene, enabling global monitoring of the scene.

9. The online monitoring system according to claim 7, characterized in that, The preprocessing module incorporates a convolutional neural unit and a similarity analysis unit. The similarity calculation unit uses the Bach distance of the gradient magnitude histogram to calculate the similarity between the infrared image and the visible light image. The convolutional neural unit is used to construct a high-order convolutional model, whose model architecture includes an input layer, a backbone branch, an interaction layer, a fusion layer, and an output layer. The input layer is connected to infrared and visible light data. The backbone branch extracts data from the infrared and visible light images, respectively. The interaction layer incorporates a multiplication interaction module to achieve dynamic feature weighting. The fusion layer can aggregate features across scales. The output layer can output multi-scale feature maps.

10. The online monitoring system according to claim 7, characterized in that, The acoustic imaging module uses a multi-channel microphone array to synchronously capture reflected acoustic signals. The location of the sound source is determined by calculating the time difference and phase difference of the sound waves arriving at each microphone. The acoustic imaging module is mounted on the drive module and can follow the movement of the drive module to change its position.

Images