Steam leakage detection method for waste heat recovery system

Abstract

The invention relates to the field of steam leakage detection, in particular to a steam leakage detection method for a waste heat recovery system, and the method comprises the steps: obtaining a visible light image, an infrared image and a sampling result label at any sampling moment in history; calculating the illumination intensity of the visible light images, clustering the plurality of visible light images according to the illumination intensity to obtain a plurality of clusters, and for the same cluster, training a first model and a second model; and judging the cluster to which the visible light image collected in real time belongs, obtaining a first leakage probability and a second leakage probability, and taking a weighted summation result of the first leakage probability and the second leakage probability as a total leakage probability to complete leakage detection. According to the technical scheme, the accuracy of a steam leakage detection result can be improved, and the energy utilization efficiency of waste heat recovery is improved.

Description

Technical Field

[0001] The present invention relates to the field of steam leakage detection, and more specifically, to a steam leakage detection method for a waste heat recovery system. Background Art

[0002] As a key technology to improve energy efficiency, waste heat recovery system has been fully applied in industrial production. Waste heat recovery system can recover waste heat or excess heat generated in industrial production process and convert it into useful energy, thereby improving energy efficiency, reducing energy consumption, and reducing greenhouse gas emissions. Waste heat recovery can not only effectively save costs, but also reduce environmental pollution, with significant economic and environmental benefits.

[0003] In many high-temperature and high-pressure industrial applications, steam, as an important heat transfer medium, is widely used in waste heat recovery systems. Steam leakage is a technical problem that cannot be ignored in such systems. Steam leakage not only leads to energy waste and reduces waste heat recovery efficiency, but also poses a potential threat to system safety.

[0004] The existing Chinese patent application document with publication number CN109447011A discloses a real-time monitoring method for steam pipe leakage using infrared, including: using the current frame infrared thermal image and temperature parameters to detect whether there is a steam leakage area in the monitoring area; and judging whether there is a steam leakage area at a position related to the steam leakage area detected in the next frame infrared thermal image.

[0005] However, infrared Figure 1 Generally, the detection effect is better at night. During the day, the sunlight is strong and the ambient temperature is high, which may cause the background thermal signal to be confused with the heat source of the steam leak, thus affecting the clarity and detection accuracy of the infrared image. In addition, the temperature difference during the day is small. When the infrared camera captures the steam leakage area, it may face insufficient contrast with the ambient heat source, resulting in inaccurate steam leakage detection results. Summary of the invention

[0006] In order to solve the technical problem of inaccurate steam leakage detection results, the present invention provides a steam leakage detection method for a waste heat recovery system. It includes: obtaining a visible light image, an infrared image and a sampling result label at any sampling time in the history; calculating the illumination intensity of the visible light image, clustering several visible light images according to the illumination intensity to obtain several cluster clusters, for the same cluster cluster, training a first model according to the visible light image and the sampling result label, training a second model according to the infrared image and the sampling result label, traversing to obtain the first model and the second model corresponding to each cluster cluster; judging the cluster cluster to which the real-time collected visible light image belongs, and inputting the real-time collected visible light image into the first model of the cluster cluster to obtain the first leakage probability, inputting the real-time collected infrared image into the second model of the cluster cluster to obtain the second leakage probability, taking the result of the weighted sum of the first leakage probability and the second leakage probability as the total leakage probability, and completing the leakage detection.

[0007] The complementary characteristics of visible light images and infrared images are effectively utilized, and special leak detection models are trained under different lighting conditions to ensure the stability and accuracy of detection results in different environments such as day and night. By combining the leakage probability of visible light and infrared images and obtaining the total leakage probability through weighted summation, more accurate leak detection can be achieved.

[0008] Preferably, calculating the light intensity includes: converting the visible light image into a grayscale image, constructing a grayscale histogram with the grayscale value range as the horizontal axis and the number of pixels as the vertical axis; and taking the grayscale mean of the grayscale value range corresponding to the peak value in the grayscale histogram as the light intensity.

[0009] It can effectively identify the fluctuation of light intensity in the image, especially quantify the light condition by the mean of the peak gray value, and provide accurate light information for subsequent cluster analysis and model training. This light intensity calculation method based on grayscale histogram can classify and optimize images according to different light conditions, and improve the adaptability of the leakage detection system to different environmental conditions in practical applications.

[0010] Preferably, calculating the illumination intensity includes: converting the visible light image into a grayscale image, and taking the grayscale mean of all pixels in the grayscale image as the illumination intensity.

[0011] As a representative of light intensity, the grayscale mean can effectively reflect the brightness distribution of the image and provide a unified standard for measuring the lighting conditions of the image.

[0012] Preferably, the first model is a convolutional neural network, which extracts features from the visible light image to obtain image features, and classifies the image features to output sampling result labels.

[0013] Preferably, the training process of the first model includes: taking all visible light images of the same cluster in the history as input information, and taking the true value of the sampling result label as the network label to obtain a set of training data; inputting the training data into the first model to obtain an output result; based on the output result and the network label, using the cross entropy loss function to calculate the loss value of the first model, back-propagating the error signal according to the loss value, updating the model parameters of the first model to reduce the loss value; iteratively updating the model parameters of the first model, and when the first model reaches the set maximum number of training times or the loss value is less than the set loss value, stopping the update to obtain a trained first model.

[0014] Preferably, the method further includes: for the same cluster, denoising each visible light image in the cluster respectively.

[0015] Preferably, the denoising of each visible light image in the cluster includes: using wavelet transform to obtain a low-frequency image and a detail image for any visible light image, the detail images including a horizontal detail image, a vertical detail image and a diagonal detail image; respectively obtaining a first image with the horizontal detail image removed, a second image with the vertical detail image removed, and a third image with the diagonal detail image removed; using UCIQE to respectively calculate the quality of the first image, the quality of the second image, and the quality of the third image, calculate the degree of noise according to the quality, and calculate the importance of each detail image according to the degree of noise and the amount of steam leakage information; deleting the detail image with the smallest importance within a preset threshold, and obtaining the denoised visible light image through inverse wavelet transform of the remaining detail images and low-frequency images.

[0016] By analyzing the quality and noise level of the detail image, combined with the amount of steam leakage information to evaluate the importance of each detail image, unimportant noise details can be removed in a targeted manner. This method can retain the key information of the image while removing interference, improving the quality and clarity of the image, thereby improving the accuracy and robustness of the subsequent leak detection model.

[0017] Preferably, calculating the noise level includes: for any detail image, calculating the ratio of the quality of the detail image to the maximum quality, and taking the normalized ratio as the noise level of the detail image, the quality of the detail image is positively correlated with the noise level, and the maximum quality is negatively correlated with the noise level.

[0018] The positive correlation between quality and noise level allows noisy details to be effectively identified and processed, while the negative correlation between maximum quality and noise level ensures that noise can be accurately removed while retaining important image details.

[0019] Preferably, the importance satisfies the relationship: , Detailed image The importance of Detailed image The noise level, Removal of detail image The amount of steam leakage information, Represents a constant. The greater the noise level, the less important the detail map is, because higher noise may mean distortion of image details, resulting in reduced contribution in restoring the original information. At the same time, the greater the amount of steam leakage information after removing the detail map, it means that the detail map carries more information in the image, resulting in its relatively higher importance.

[0020] Preferably, the steam leakage information includes: The difference between the first image and the visible light image is taken as the steam leakage information amount of the horizontal detail image, the difference between the second image and the visible light image is taken as the steam leakage information amount of the vertical detail image, and the difference between the third image and the visible light image is taken as the steam leakage information amount of the diagonal detail image.

[0021] Beneficial effects of the present invention: The present invention combines visible light images and infrared images, and utilizes cluster analysis and model training methods to achieve accurate detection of steam leaks in waste heat recovery systems. By calculating the illumination intensity of the visible light image and performing clustering, similar images can be classified into the same category, thereby improving the accuracy of the model. Furthermore, by performing feature extraction on the visible light image and infrared image through a convolutional neural network, and combining the characteristics of the clustering clusters, the probability of steam leakage can be accurately predicted. The denoising process further improves the image quality, reduces interference, and improves the robustness of the model. Ultimately, by combining the detection results of the visible light image and the infrared image, more accurate and stable leak detection can be achieved. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 The present invention is a flow chart of a steam leakage detection method for a waste heat recovery system according to an embodiment of the present invention. DETAILED DESCRIPTION

[0023] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all of the embodiments.

[0024] The specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0025] Reference Figure 1 A steam leakage detection method for a waste heat recovery system includes steps S1 to S3, which are specifically as follows: S1: Obtain the visible light image, infrared image and sampling result label at any sampling time in the history.

[0026] In one embodiment, in a waste heat recovery system, visible light image and infrared image data are collected to monitor and determine the operating status of the system. Each sampling moment is marked with a sampling result label, and the sampling result label includes steam leakage and normal, so as to facilitate fault diagnosis and analysis in the later stage.

[0027] Specifically, at each moment of the system operation, cameras and infrared sensors are installed to obtain real-time visible light images and infrared images respectively. These images are stored in real time and labeled in combination with the operation data at the sampling moment. The sampling result label will be determined based on the image content. For example, if there is an obvious sign of steam leakage in the image, the label will be marked as steam leakage, otherwise it is normal.

[0028] S2: Calculate the illumination intensity of the visible light image, cluster several visible light images according to the illumination intensity to obtain several clusters, for the same cluster, train the first model according to the visible light image and the sampling result label, train the second model according to the infrared image and the sampling result label, and traverse to obtain the first model and the second model corresponding to each cluster.

[0029] It should be noted that the lighting conditions during the day and at night are different, which leads to differences in image quality. Specifically, during the day when the lighting conditions are good, the visible light image can provide a clearer picture, which makes the identification of steam leaks better. However, at night when the light intensity is low, the quality of the visible light image decreases, resulting in poor recognition results. Relatively speaking, infrared images perform better at night because the temperature difference is large at night, there are fewer background heat sources, and the detection of heat sources such as steam leaks is more accurate; but during the day, due to the interference of staff activities and other environmental factors, infrared images may be affected, resulting in poor recognition results.

[0030] In order to cope with these different lighting conditions, by analyzing the light intensity in the image, it is possible to automatically determine whether the image is in a condition of low light intensity, so that different processing methods can be adopted when classifying the image. For example, if the light intensity is low, a detection method that relies more on infrared images may be used; while during the day when the light intensity is high, visible light images are prioritized for analysis.

[0031] In one embodiment, the visible light image is converted into a grayscale image, and a grayscale histogram is constructed with the grayscale value range as the horizontal axis and the number of pixels as the vertical axis; the grayscale mean of the grayscale value range corresponding to the peak value in the grayscale histogram is used as the light intensity.

[0032] Several visible light images are clustered according to the light intensity to obtain several clustering clusters. The clustering method may use Kmeans clustering. After the clustering is completed, several clustering clusters are obtained. Each clustering cluster represents a type of visible light image of light intensity.

[0033] For the same cluster, a first model is trained according to the visible light image and the sampling result label. The first model is a convolutional neural network. The convolutional neural network extracts features of the visible light image to obtain image features, and classifies the image features to output sampling result labels.

[0034] The training process of the first model includes: All visible light images of the same cluster in history are used as input information, and the true value of the sampling result label is used as the network label to obtain a set of training data; the training data is input into the first model to obtain the output result; based on the output result and the network label, the loss value of the first model is calculated using the cross entropy loss function, and the error signal is back-propagated according to the loss value to update the model parameters of the first model to reduce the loss value; the model parameters of the first model are iteratively updated, and when the first model reaches the set maximum number of training times or the loss value is less than the set loss value, the update is stopped to obtain the trained first model.

[0035] Exemplarily, when the number of training times reaches 200 or the loss value is less than 0.0001, the update is stopped.

[0036] The second model is also a convolutional neural network, and the training method and network structure of the second model are the same as those of the first model.

[0037] The difference is that the first model is trained with all visible light images in the same cluster, and after the training is completed, it only accepts the input of real-time visible light images; the second model is trained with the infrared images corresponding to all visible light images in the same cluster, and after the training is completed, it only accepts the input of infrared images collected at the same time as the real-time visible light images.

[0038] It should be noted that one cluster corresponds to one first model and one first second model.

[0039] In one embodiment, calculating the illumination intensity includes: converting the visible light image into a grayscale image, and taking the grayscale mean of all pixels in the grayscale image as the illumination intensity.

[0040] In one embodiment, the method further includes: for the same cluster, denoising each visible light image in the cluster respectively.

[0041] A low-frequency image and a detail image are obtained by using wavelet transform for any visible light image, wherein the detail image includes a horizontal detail image, a vertical detail image and a diagonal detail image; a first image with the horizontal detail image removed, a second image with the vertical detail image removed and a third image with the diagonal detail image removed are obtained respectively.

[0042] It should be noted that the noise in the visible light image is mainly distributed in the high-frequency detail image, so there is no need to analyze the image after removing the low-frequency image.

[0043] Use UCIQE to calculate the quality of the first image, the quality of the second image, and the quality of the third image respectively, and calculate the noise degree according to the quality. The noise degree includes: For any detail image, the ratio of the quality of the detail image to the maximum quality is calculated, and the normalized ratio is used as the noise level of the detail image. The quality of the detail image is positively correlated with the noise level, and the maximum quality is negatively correlated with the noise level.

[0044] The noise level can be expressed using the relation: , Detailed image The noise level, Removal of detail image Quality, represents the maximum mass, Represents the normalization function.

[0045] The difference between the first image and the visible light image is taken as the steam leakage information amount of the horizontal detail image, the difference between the second image and the visible light image is taken as the steam leakage information amount of the vertical detail image, and the difference between the third image and the visible light image is taken as the steam leakage information amount of the diagonal detail image.

[0046] The importance of each detail image is calculated based on the noise level and the amount of steam leakage information. The importance satisfies the relationship: , Detailed image The importance of Detailed image The noise level, Removal of detail image The amount of steam leakage information.

[0047] The detail image with the least importance and the preset threshold is deleted, and the remaining detail image and low-frequency image are transformed by inverse wavelet transform to obtain the denoised visible light image.

[0048] In one embodiment, filtering and denoising are performed on the visible light image and the infrared image respectively.

[0049] S3, determine the cluster to which the visible light image collected in real time belongs, and input the visible light image collected in real time into the first model of the cluster to which it belongs to obtain a first leakage probability, input the infrared image collected in real time into the second model of the cluster to which it belongs to obtain a second leakage probability, and take the result of the weighted sum of the first leakage probability and the second leakage probability as the total leakage probability to complete the leakage detection.

[0050] In one embodiment, the total leakage probability is compared with the leakage threshold. When the real-time total leakage probability is greater than the leakage threshold, it is considered that steam leakage occurs in the waste heat recovery system, and an alarm signal is generated and sent to remind the staff to perform maintenance in time.

[0051] It should be noted that, for ordinary technicians in this field, several modifications and improvements can be made without departing from the concept of this application, which are all within the scope of protection of this application. Therefore, the scope of protection of the patent of this application shall be based on the attached claims.

Claims

1. A steam leakage detection method for a waste heat recovery system, characterized in that: include: Obtain visible light images, infrared images and sampling result labels at any sampling moment in history; Calculate the illumination intensity of the visible light image, cluster several visible light images according to the illumination intensity to obtain several clusters, for the same cluster, train a first model according to the visible light image and the sampling result label, train a second model according to the infrared image and the sampling result label, and traverse to obtain the first model and the second model corresponding to each cluster; Determine the cluster to which the visible light image collected in real time belongs, and input the visible light image collected in real time into the first model of the cluster to which it belongs to obtain a first leakage probability, and input the infrared image collected in real time into the second model of the cluster to which it belongs to obtain a second leakage probability, and take the result of the weighted sum of the first leakage probability and the second leakage probability as the total leakage probability to complete the leakage detection.

2. A steam leakage detection method for a waste heat recovery system according to claim 1, characterized in that: Calculating the light intensity includes: The visible light image is converted into a grayscale image, and a grayscale histogram is constructed with the grayscale value range as the horizontal axis and the number of pixels as the vertical axis; The grayscale mean of the grayscale value range corresponding to the peak value in the grayscale histogram is taken as the light intensity.

3. The steam leakage detection method for a waste heat recovery system according to claim 1, characterized in that: Calculating the light intensity includes: The visible light image is converted into a grayscale image, and the grayscale mean of all pixels in the grayscale image is used as the light intensity.

4. The steam leakage detection method for a waste heat recovery system according to claim 1, characterized in that: The first model is a convolutional neural network, which extracts features from visible light images to obtain image features, and classifies the image features to output sampling result labels.

5. The steam leakage detection method for a waste heat recovery system according to claim 1, characterized in that: The training process of the first model includes: All visible light images of the same cluster in history are used as input information, and the true value of the sampling result label is used as the network label to obtain a set of training data; Input the training data into the first model to obtain the output result; Based on the output results and the network labels, the loss value of the first model is calculated using the cross entropy loss function, the error signal is back-propagated according to the loss value, and the model parameters of the first model are updated to reduce the loss value; Iteratively update the model parameters of the first model. When the first model reaches a set maximum number of training times or the loss value is less than a set loss value, stop updating to obtain a trained first model.

6. The method for detecting steam leakage in a waste heat recovery system according to claim 1, characterized in that: Also includes: For the same cluster, each visible light image in the cluster is denoised separately.

7. A steam leakage detection method for a waste heat recovery system according to claim 6, characterized in that: The denoising of each visible light image in the cluster includes: For any visible light image, wavelet transform is used to obtain a low-frequency image and a detail image, and the detail images include a horizontal detail image, a vertical detail image and a diagonal detail image; a first image with the horizontal detail image removed, a second image with the vertical detail image removed and a third image with the diagonal detail image removed are obtained respectively; UCIQE is used to calculate the quality of the first image, the quality of the second image and the quality of the third image respectively, the degree of noise is calculated according to the quality, and the importance of each detail image is calculated according to the degree of noise and the amount of steam leakage information; the detail image with the smallest importance and a preset threshold is deleted, and the remaining detail images and low-frequency images are subjected to inverse wavelet transform to obtain a denoised visible light image.

8. A steam leakage detection method for a waste heat recovery system according to claim 7, characterized in that: Calculating the noise level includes: For any detail image, the ratio of the quality of the detail image to the maximum quality is calculated, and the normalized ratio is used as the noise level of the detail image. The quality of the detail image is positively correlated with the noise level, and the maximum quality is negatively correlated with the noise level.

9. The method for detecting steam leakage in a waste heat recovery system according to claim 7, characterized in that: The importance satisfies the relationship: , Detailed image The importance of Detailed image The noise level, Removal of detail image The amount of steam leakage information.

10. A steam leakage detection method for a waste heat recovery system according to claim 9, characterized in that: The steam leak information includes: The difference between the first image and the visible light image is taken as the steam leakage information amount of the horizontal detail image, the difference between the second image and the visible light image is taken as the steam leakage information amount of the vertical detail image, and the difference between the third image and the visible light image is taken as the steam leakage information amount of the diagonal detail image.

Images