Cathode material for lithium ion battery and preparation method thereof

Abstract

Disclosed are a jacket underwater crack detection system based on acoustic emission signals and a method thereof, which adopts an artificial intelligence detection algorithm based on deep learning, takes the acoustic emission echo signals of a jacket to be detected by an acoustic emission sensor as input data, and after field transformation of the acoustic emission echo signals in a time sequence under a Cartesian coordinate system to a polar coordinate system, extracts feature distribution information about the jacket focusing on spatial positions and channel contents in the acoustic emission echo signals, and detects and judges whether the jacket has underwater cracks by fusing the feature information of the two. In this way, the jacket can be accurately and intelligently detected for underwater cracks to ensure the safety of offshore wind power jackets.

Description

Technical Field

[0001] This application relates to the field of intelligent detection technology, and more specifically, to a system and method for detecting underwater cracks in a jacket structure based on acoustic emission signals. Background Technology

[0002] Against the backdrop of the nation's vigorous promotion of "carbon peaking" and "carbon neutrality," the wind power, photovoltaic, and pumped storage industries are developing rapidly. Offshore wind power, as the most important component of the wind power industry, is developing at a remarkable pace. With the rapid development of offshore wind power, the trend towards larger offshore wind turbines is inevitable. Jacket foundations, as one of the most stable and reliable foundation types for offshore wind turbines, occupy an indispensable position in the development of offshore wind power. Furthermore, as the transitional structure connecting the upper turbine tower and the lower main body, it plays a crucial role in this process.

[0003] Currently, since jacket platforms are mostly located in marine areas with high temperatures, high salinity, and complex environmental loads, they require regular maintenance and flaw detection to ensure their safety. Traditional inspection technologies include magnetic particle testing (MT), ultrasonic testing (UT), and eddy current testing (ET). These technologies require highly skilled operators, have low cost-effectiveness, and low levels of automation. They also necessitate removing the anti-corrosion layer from the surface of the jacket being tested, leading to high costs, and are only suitable for detecting macroscopic defects. Furthermore, most traditional inspection methods rely on manual operation, resulting in low efficiency and poor real-time performance.

[0004] Therefore, there is a need for an optimized underwater crack detection system for offshore wind farm jackets that can accurately detect underwater cracks in offshore wind farm jackets in real time and intelligently, so as to ensure the safety of offshore wind farm jackets and reduce the maintenance costs for enterprises. Summary of the Invention

[0005] To address the aforementioned technical problems, this application is proposed. Embodiments of this application provide a system and method for detecting underwater cracks in jacket structures based on acoustic emission signals. This system employs a deep learning-based artificial intelligence detection algorithm. It uses the acoustic emission echo signal from an acoustic emission sensor acting on the jacket under test as input data. After performing a field transformation from a Cartesian coordinate system to a polar coordinate system, it extracts feature distribution information regarding the spatial focus and channel content of the jacket structure from the acoustic emission echo signal. The fusion of these two feature information is used to detect and determine whether underwater cracks exist in the jacket structure. This allows for accurate and intelligent detection of underwater cracks in offshore wind farm jacket structures, ensuring the safety of offshore wind farm jacket structures.

[0006] According to one aspect of this application, an underwater crack detection system for jackets based on acoustic emission signals is provided, comprising:

[0007] Acoustic emission echo signal acquisition module is used to acquire the acoustic emission echo signal of the acoustic emission sensor acting on the duct frame under test;

[0008] The domain mapping module is used to perform Gram angle and field transformation on the acoustic emission echo signal to obtain Gram angle and field image;

[0009] The first feature extraction module is used to obtain a first acoustic emission detection feature map by using a first convolutional neural network model with spatial attention to extract the Gram angle and field image;

[0010] The second feature extraction module is used to obtain a second acoustic emission detection feature map by using a second convolutional neural network model with channel attention to extract the Gram angle and field image;

[0011] The feature fusion module is used to fuse the first acoustic emission detection feature map and the second acoustic emission detection feature map to obtain a classification feature map;

[0012] A feature manifold correction module is used to perform high-dimensional feature manifold correction on the classification feature map to obtain a corrected classification feature map; and

[0013] The detection result generation module is used to pass the corrected classification feature map through a classifier to obtain a classification result, which is used to indicate whether the jacket has underwater cracks.

[0014] In the aforementioned underwater crack detection system for jackets based on acoustic emission signals, the first feature extraction module is further configured to: perform depth convolutional encoding on the Gram angle and field images using the convolutional encoding part of the first convolutional neural network model using spatial attention to obtain a first Gram angle and field convolutional feature map; input the first Gram angle and field convolutional feature map into the spatial attention part of the first convolutional neural network model using spatial attention to obtain a spatial attention map; pass the spatial attention map through a Softmax activation function to obtain a spatial attention feature map; and calculate the position-based dot product of the spatial attention feature map and the first Gram angle and field convolutional feature map to obtain the first acoustic emission detection feature map.

[0015] In the aforementioned underwater crack detection system for jackets based on acoustic emission signals, the second feature extraction module is further configured to: input the Gram angle and field images into the multi-layer convolutional layers of the second convolutional neural network model using channel attention to obtain a second Gram angle and field convolutional feature map; calculate the global mean of each feature matrix along the channel dimension of the second Gram angle and field convolutional feature map to obtain a channel feature vector; input the channel feature vector into the Sigmoid activation function to obtain a channel attention weight vector; and weight each feature matrix along the channel dimension of the second Gram angle and field convolutional feature map using the feature values ​​at each position in the channel attention weight vector as weights to obtain a second acoustic emission detection feature map.

[0016] In the aforementioned underwater crack detection system for jackets based on acoustic emission signals, the feature fusion module is further configured to: fuse the first acoustic emission detection feature map and the second acoustic emission detection feature map using the following formula to obtain a classification feature map; wherein, the formula is:

[0017]

[0018] Wherein, F is the classification feature map, F1 is the first acoustic emission detection feature map, and F2 is the second acoustic emission detection feature map. This indicates positional addition, where λ and β are weighting parameters used to control the balance between the first acoustic emission detection feature map and the second acoustic emission detection feature map.

[0019] In the aforementioned underwater crack detection system for jackets based on acoustic emission signals, the feature manifold correction module is further configured to: perform high-dimensional feature manifold correction on the classification feature map using the following formula to obtain the corrected classification feature map; wherein, the formula is:

[0020]

[0021] Where, f i and f j The feature values ​​at each position of the classification feature map are represented, and Let N represent the mean of all feature values ​​of the classification feature map, and let N represent the scale of the classification feature map. exp(·) represents the exponential operation of the numerical value, where the exponential operation represents the value of the natural exponential function raised to the power of the numerical value. i ' represents the feature value at each position in the corrected classification feature map.

[0022] In the aforementioned underwater crack detection system for jackets based on acoustic emission signals, the detection structure generation module includes: a feature map unfolding unit, used to unfold the corrected classification feature map into classification feature vectors according to row vectors or column vectors; a fully connected encoding unit, used to perform fully connected encoding on the classification feature vectors using multiple fully connected layers of the classifier to obtain encoded classification feature vectors; and a classification unit, used to pass the encoded classification feature vectors through the Softmax classification function of the classifier to obtain the classification result.

[0023] According to another aspect of this application, a method for underwater crack detection of jacket structures based on acoustic emission signals is provided, comprising:

[0024] Acoustic emission echo signal acquisition steps: Acquire the acoustic emission echo signal of the acoustic emission sensor acting on the duct frame under test;

[0025] Domain mapping step: Perform Gram angle and field transformation on the acoustic emission echo signal to obtain Gram angle and field image;

[0026] First feature extraction step: The Gram angle and field image are used to obtain the first acoustic emission detection feature map by using a first convolutional neural network model with spatial attention;

[0027] The second feature extraction step: The Gram angle and field image are used to obtain the second acoustic emission detection feature map by using a second convolutional neural network model with channel attention;

[0028] Feature fusion step: Fuse the first acoustic emission detection feature map and the second acoustic emission detection feature map to obtain a classification feature map;

[0029] Feature manifold correction step: Perform high-dimensional feature manifold correction on the classification feature map to obtain the corrected classification feature map; and

[0030] The detection result generation steps are as follows: The corrected classification feature map is passed through a classifier to obtain a classification result, which is used to indicate whether there are underwater cracks in the jacket.

[0031] Compared with existing technologies, the underwater crack detection system and method for jacket structures based on acoustic emission signals provided in this application employs a deep learning-based artificial intelligence detection algorithm. It uses the acoustic emission echo signal from the acoustic emission sensor acting on the jacket structure under test as input data. After performing a field transformation from a Cartesian coordinate system to a polar coordinate system, it extracts feature distribution information regarding the spatial focus and channel content of the jacket structure from the acoustic emission echo signal. The fusion of these two feature information is then used to detect and determine whether underwater cracks exist in the jacket structure. This allows for precise and intelligent detection of underwater cracks in offshore wind farm jacket structures, ensuring the safety of offshore wind farm jacket structures. Attached Figure Description

[0032] The above and other objects, features, and advantages of this application will become more apparent from the more detailed description of the embodiments of this application in conjunction with the accompanying drawings. The drawings are provided to further illustrate the embodiments of this application and form part of the specification. They are used together with the embodiments of this application to explain this application and do not constitute a limitation thereof. In the drawings, the same reference numerals generally represent the same components or steps.

[0033] Figure 1 This is an application scenario diagram of the underwater crack detection system for jackets based on acoustic emission signals according to an embodiment of this application.

[0034] Figure 2 This is a block diagram of an underwater crack detection system for jackets based on acoustic emission signals, according to an embodiment of this application.

[0035] Figure 3 This is a block diagram of the detection structure generation module in the underwater crack detection system for jackets based on acoustic emission signals according to an embodiment of this application.

[0036] Figure 4 This is a flowchart of an underwater crack detection method for jackets based on acoustic emission signals, according to an embodiment of this application.

[0037] Figure 5 This is a schematic diagram of the system architecture of the underwater crack detection method for jackets based on acoustic emission signals according to an embodiment of this application. Detailed Implementation

[0038] Hereinafter, exemplary embodiments according to this application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of this application, and not all embodiments of this application. It should be understood that this application is not limited to the exemplary embodiments described herein.

[0039] Scene Overview

[0040] As mentioned above, traditional inspection technologies require highly skilled operators, have low cost-effectiveness, and low levels of automation. They necessitate removing the anti-corrosion layer from the surface of the jacket structure being inspected, leading to high costs, and are only applicable to detecting macroscopic defects. Furthermore, most traditional inspection methods rely on manual operation, resulting in low efficiency and poor real-time performance. Therefore, an optimized underwater crack detection system for jacket structures is desired, capable of accurately detecting underwater cracks in offshore wind farm jacket structures in real-time and intelligently, thereby ensuring the safety of offshore wind farm jacket structures and reducing subsequent maintenance costs for enterprises.

[0041] Currently, deep learning and neural networks have been widely applied in fields such as computer vision, natural language processing, and speech signal processing. Furthermore, deep learning and neural networks have demonstrated near-human or even surpassed human-level performance in areas such as image classification, object detection, semantic segmentation, and text translation.

[0042] In recent years, the development of deep learning and neural networks has provided new ideas and solutions for underwater crack detection of jacket structures.

[0043] Correspondingly, considering that traditional detection technologies require high operator skills, have low cost-effectiveness, and low levels of intelligence, and that image-processing-based crack detection schemes have low accuracy in detecting cracks in jacket structures due to underwater image acquisition, and are easily confused with rust, scratches, etc. on the jacket structure, traditional detection methods cannot directly reflect the presence of underwater cracks in the jacket structure. However, considering that acoustic emission is a phenomenon where stress concentration in a localized area of ​​a material rapidly releases energy and generates transient elastic waves, and that the acoustic emission signal is a high-frequency stress wave pulse signal released due to lattice distortion of molecules in the material, crack intensification, and plastic deformation, it can effectively reflect the structural strength status information of the jacket structure, thus enabling underwater crack detection.

[0044] Based on this, the technical solution of this application employs a deep learning-based artificial intelligence detection algorithm. The acoustic emission echo signal from the acoustic emission sensor acting on the jacket under test is used as input data. After performing a field transformation from a Cartesian coordinate system to a polar coordinate system, feature distribution information regarding the spatial focus and channel content of the jacket is extracted from the acoustic emission echo signal. The fusion of these two feature information is then used to detect and determine whether underwater cracks exist in the jacket. This enables accurate and intelligent detection of underwater cracks in offshore wind farm jackets, ensuring the safety of offshore wind farm jackets and reducing subsequent maintenance costs for enterprises.

[0045] Specifically, in the technical solution of this application, firstly, the acoustic emission echo signal of the acoustic emission sensor acting on the guide frame under test is acquired. It should be understood that, since the Gramian angular field (GAF) is based on the Gram principle, it can transfer the time series in the classical Cartesian coordinate system to the polar coordinate system for representation. GAF can well preserve the dependence and correlation of the original acoustic emission time series signal, possessing similar time series characteristics to the original acoustic emission signal. Depending on the trigonometric functions used for encoding, GAF can yield the Gramian angular sum field (GASF) and the Gramian angular difference field (GADF). Since the GADF transformation is irreversible, in the technical solution of this application, the GASF transformation method, which allows for inverse transformation, is selected for encoding the acoustic emission echo signal. That is, specifically, the acoustic emission echo signal is subjected to Gramian angular sum field transformation to obtain a Gramian angular sum field image. In a specific example, the encoding steps from the acoustic emission echo signal to the GASF image are as follows: For a time series with C dimensions = {Q1, Q2, ..., QC}, where each dimension contains n sampling points Qi = {qi1, qi2, ..., qin}, the data for each dimension is first normalized. Then, all values ​​in the data are integrated into the range [-1, 1]. After integration, the normalized values ​​are replaced by trigonometric function values ​​(Cos values), and polar coordinates are used instead of Cartesian coordinates, thus preserving the absolute temporal relationship of the sequence.

[0046] Then, a convolutional neural network model, which excels in extracting local implicit correlation features from images, is used to extract features from the Gram angle and field images. Specifically, considering that in detecting underwater cracks in the jacket under test, the focus should be on the spatial location feature distribution information of the jacket and the correlation feature distribution information of content features in the channels, in order to accurately detect and judge underwater cracks in the jacket under test. Given that the attention mechanism can select the focus position, generating more discriminative feature representations, and that the features after adding the attention module will adaptively change as the network deepens.

[0047] Based on this, in the technical solution of this application, in order to improve the extraction effect of implicit features of acoustic emission echo signals in the Gram angle and field images and thus improve the accuracy of underwater crack detection for the jacket under test, a convolutional neural network model with a dual attention mechanism is used to perform feature mining on the Gram angle and field images. Specifically, the dual attention mechanism here is a spatial attention mechanism and a channel attention mechanism. That is, specifically, the Gram angle and field images are used to obtain a first acoustic emission detection feature map by using a first convolutional neural network model with spatial attention, and the Gram angle and field images are used to obtain a second acoustic emission detection feature map by using a second convolutional neural network model with channel attention. It should be understood that the image features extracted by the channel attention reflect the correlation and importance between feature channels, while the image features extracted by the spatial attention reflect the weight of spatial dimensional feature differences, which are used to suppress or enhance features at different spatial locations. Therefore, the channel attention and the spatial attention can focus on the feature content and feature location in the Gram angle and field image, respectively, complementing each other to a certain extent and improving the feature extraction effect of the network. This allows a large amount of effective information about whether the jacket has underwater cracks to be captured, which can effectively enhance the feature discrimination learning ability. During network training, the task processing system focuses more on finding significant useful information in the input image data that is relevant to the current output, thereby improving the quality of the output. Furthermore, the increasing number of attention modules will bring continuous performance improvement.

[0048] Furthermore, the first and second acoustic emission detection feature maps are fused to integrate the implicit feature information about the spatial location of the jacket structure and the content correlation feature information between channels in the acoustic emission echo signal. This information is then used as a classification feature map for classification processing by a classifier to obtain a classification result indicating whether the jacket structure has underwater cracks. This enables intelligent detection of underwater cracks in offshore wind farm jacket structures, ensuring the safety of offshore wind farm jacket structures.

[0049] Specifically, in the technical solution of this application, since the first acoustic emission detection feature map and the second acoustic emission detection feature map are obtained from the Gram angle and field images respectively using a first convolutional neural network model with spatial attention and a second convolutional neural network model with channel attention, the first acoustic emission detection feature map and the second acoustic emission detection feature map can respectively enhance the semantic association features of the Gram angle and field images in the image spatial dimension and the model channel dimension. However, while making full use of features in different dimensions, the feature distribution of the classification feature map obtained by fusing the first acoustic emission detection feature map and the second acoustic emission detection feature map also needs to include feature distributions expressing feature associations in different dimensions. This makes the fitting burden between the classification feature map and the weight matrix of the classifier heavy after the classification feature map is expanded into feature vectors in the classifier, thereby affecting the training speed of the classifier and the accuracy of the classification results.

[0050] Therefore, the applicant of this application flattens the class representation of the classification feature map, specifically as follows:

[0051]

[0052] f i and f j These are the feature values ​​of the classification feature map, and is the mean of all feature values ​​of the classification feature map, and N is the scale of the classification feature map, i.e., width multiplied by height multiplied by the number of channels.

[0053] Here, the class representation flattening of the classification feature map flattens the finite polyhedral manifold used for class representation of feature distribution in high-dimensional feature space, while maintaining the inherent distance between the planes of the manifold and avoiding intersections based on spatial intuition. Essentially, it decomposes the finite polyhedral manifold into a cubic lattice based on intersections of right angles and common vertices, thereby obtaining the flattened "slice" continuity of the class plane, enhancing the fitting performance of the classification feature map to the classifier's weight matrix. This improves the training speed and accuracy of the classification results when the classification feature map is used by the classifier. Consequently, it enables precise intelligent detection of underwater cracks in offshore wind farm jacket structures, ensuring the safety of offshore wind farm jacket structures and reducing subsequent maintenance costs for enterprises.

[0054] Based on this, this application provides an underwater crack detection system for jackets based on acoustic emission signals, comprising: an acoustic emission echo signal acquisition module for acquiring acoustic emission echo signals from an acoustic emission sensor acting on the jacket under test; a domain mapping module for performing Gram angle and field transformations on the acoustic emission echo signals to obtain Gram angle and field images; a first feature extraction module for using a first convolutional neural network model with spatial attention to obtain a first acoustic emission detection feature map; a second feature extraction module for using a second convolutional neural network model with channel attention to obtain a second acoustic emission detection feature map; a feature fusion module for fusing the first acoustic emission detection feature map and the second acoustic emission detection feature map to obtain a classification feature map; a feature manifold correction module for performing high-dimensional feature manifold correction on the classification feature map to obtain a corrected classification feature map; and a detection result generation module for passing the corrected classification feature map through a classifier to obtain a classification result, wherein the classification result indicates whether underwater cracks exist in the jacket.

[0055] Figure 1 This is an application scenario diagram of the underwater crack detection system for jackets based on acoustic emission signals according to an embodiment of this application. Figure 1 As shown, in this application scenario, firstly, the acoustic emission sensor (e.g., Figure 1 The acoustic emission echo signal (e.g., M) acting on the catheter stent under test, as shown in the figure. Figure 1 (as shown in C); then, the acquired acoustic emission echo signal is input to a server deployed with an underwater crack detection algorithm for jackets based on acoustic emission signals (e.g., Figure 1 As shown in S), the server is capable of processing the acoustic emission echo signal using an acoustic emission signal-based underwater crack detection algorithm for the jacket to generate a representation indicating whether an underwater crack exists in the jacket.

[0056] After introducing the basic principles of this application, various non-limiting embodiments of this application will be described in detail below with reference to the accompanying drawings.

[0057] Exemplary System

[0058] Figure 2 This is a block diagram of an underwater crack detection system for jackets based on acoustic emission signals, according to an embodiment of this application. Figure 2As shown, the underwater crack detection system 100 for jackets based on acoustic emission signals according to an embodiment of this application includes: an acoustic emission echo signal acquisition module 110, used to acquire acoustic emission echo signals from an acoustic emission sensor acting on the jacket under test; a domain mapping module 120, used to perform Gram angle and field transformation on the acoustic emission echo signals to obtain Gram angle and field images; a first feature extraction module 130, used to extract the Gram angle and field images by using a first convolutional neural network model with spatial attention to obtain a first acoustic emission detection feature map; and a second feature extraction module 140, used to extract the... The first acoustic emission detection feature map is obtained by using a second convolutional neural network model with channel attention to the Gram angle and field image; the feature fusion module 150 is used to fuse the first acoustic emission detection feature map and the second acoustic emission detection feature map to obtain a classification feature map; the feature manifold correction module 160 is used to perform high-dimensional feature manifold correction on the classification feature map to obtain a corrected classification feature map; and the detection result generation module 170 is used to pass the corrected classification feature map through a classifier to obtain a classification result, which is used to indicate whether there are underwater cracks in the jacket.

[0059] Specifically, in this embodiment of the application, the acoustic emission echo signal acquisition module 110 is used to acquire the acoustic emission echo signal of the acoustic emission sensor acting on the catheter frame under test.

[0060] As mentioned earlier, traditional inspection technologies require highly skilled operators, have low cost-effectiveness, and low levels of automation. They necessitate removing the anti-corrosion layer from the surface of the jacket structure being tested, leading to high costs, and are only applicable to detecting macroscopic defects. Furthermore, most traditional inspection methods rely on manual operation, resulting in low efficiency and poor real-time performance. Therefore, there is a need for an optimized underwater crack detection system for jacket structures that can accurately detect underwater cracks in offshore wind farm jacket structures in real-time and intelligently, thereby ensuring the safety of offshore wind farm jacket structures and reducing subsequent maintenance costs for enterprises.

[0061] Currently, deep learning and neural networks have been widely applied in fields such as computer vision, natural language processing, and speech signal processing. Furthermore, deep learning and neural networks have demonstrated near-human or even surpassed human-level performance in areas such as image classification, object detection, semantic segmentation, and text translation.

[0062] In recent years, the development of deep learning and neural networks has provided new ideas and solutions for underwater crack detection of jacket structures.

[0063] Correspondingly, considering that traditional detection technologies require high operator skills, have low cost-effectiveness, and low levels of intelligence, and that image-processing-based crack detection schemes have low accuracy in detecting cracks in jacket structures due to underwater image acquisition, and are easily confused with rust, scratches, etc. on the jacket structure, traditional detection methods cannot directly reflect the presence of underwater cracks in the jacket structure. However, considering that acoustic emission is a phenomenon where stress concentration in a localized area of ​​a material rapidly releases energy and generates transient elastic waves, and that the acoustic emission signal is a high-frequency stress wave pulse signal released due to lattice distortion of molecules in the material, crack intensification, and plastic deformation, it can effectively reflect the structural strength status information of the jacket structure, thus enabling underwater crack detection.

[0064] Based on this, the technical solution of this application employs a deep learning-based artificial intelligence detection algorithm. The acoustic emission echo signal from the acoustic emission sensor acting on the jacket under test is used as input data. After performing a field transformation from a Cartesian coordinate system to a polar coordinate system, feature distribution information regarding the spatial focus and channel content of the jacket is extracted from the acoustic emission echo signal. The fusion of these two feature information is then used to detect and determine whether underwater cracks exist in the jacket. This enables accurate and intelligent detection of underwater cracks in offshore wind farm jackets, ensuring the safety of offshore wind farm jackets and reducing subsequent maintenance costs for enterprises.

[0065] Specifically, in the technical solution of this application, firstly, the acoustic emission echo signal of the acoustic emission sensor acting on the guide frame under test is acquired.

[0066] Specifically, in this embodiment of the application, the domain mapping module 120 is used to perform Gram angle and field transformation on the acoustic emission echo signal to obtain Gram angle and field image.

[0067] It should be understood that, since the Gramian angular field (GAF) is based on the Gram principle, it can transfer the time series in the classical Cartesian coordinate system to the polar coordinate system for representation. GAF can well preserve the dependencies and correlations of the original acoustic emission time series signal, possessing similar time series characteristics to the original acoustic emission signal. Depending on the trigonometric functions used for encoding, GAF can yield the Gramian angular sum field (GASF) and the Gramian angular difference field (GADF). Since the GADF conversion is irreversible, the technical solution of this application selects the inversely convertible GASF conversion method for encoding the acoustic emission echo signal.

[0068] Specifically, the acoustic emission echo signal is subjected to Gram angle and field transformation to obtain the Gram angle and field image. In a specific example, the encoding steps from the acoustic emission echo signal to the GASF image are as follows: For a time series with C dimensions = {Q1, Q2, ..., QC}, where each dimension contains n sampling points Qi = {qi1, qi2, ..., qin}, the data for each dimension is first normalized. Then, all values ​​in the data are integrated into the range [-1, 1]. After integration, the normalized values ​​are replaced by trigonometric function values ​​(Cos values), and polar coordinates are used instead of Cartesian coordinates, thereby preserving the absolute temporal relationship of the sequence.

[0069] Specifically, in the embodiments of this application, the first feature extraction module 130 and the second feature extraction module 140 are used to obtain a first acoustic emission detection feature map by using a first convolutional neural network model with spatial attention on the Gram angle and field image; and to obtain a second acoustic emission detection feature map by using a second convolutional neural network model with channel attention on the Gram angle and field image.

[0070] Then, a convolutional neural network model, which excels in extracting local implicit correlation features from images, is used to extract features from the Gram angle and field images. Specifically, considering that in detecting underwater cracks in the jacket under test, the focus should be on the spatial location feature distribution information of the jacket and the correlation feature distribution information of content features in the channels, in order to accurately detect and judge underwater cracks in the jacket under test. Given that the attention mechanism can select the focus position, generating more discriminative feature representations, and that the features after adding the attention module will adaptively change as the network deepens.

[0071] Based on this, in the technical solution of this application, in order to improve the extraction effect of implicit features of acoustic emission echo signals in the Gram angle and field images and thus improve the accuracy of underwater crack detection for the jacket under test, a convolutional neural network model with a dual attention mechanism is used to perform feature mining on the Gram angle and field images. Specifically, the dual attention mechanism here is a spatial attention mechanism and a channel attention mechanism. That is, specifically, the Gram angle and field images are used to obtain a first acoustic emission detection feature map by using a first convolutional neural network model with spatial attention, and the Gram angle and field images are used to obtain a second acoustic emission detection feature map by using a second convolutional neural network model with channel attention.

[0072] It should be understood that the image features extracted by the channel attention reflect the correlation and importance between feature channels, while the image features extracted by the spatial attention reflect the weight of spatial dimensional feature differences, which are used to suppress or enhance features at different spatial locations.

[0073] The first feature extraction module is further configured to: perform deep convolutional encoding on the Gram angle and field images using the convolutional encoding part of the first convolutional neural network model using spatial attention to obtain a first Gram angle and field convolutional feature map; input the first Gram angle and field convolutional feature map into the spatial attention part of the first convolutional neural network model using spatial attention to obtain a spatial attention map; pass the spatial attention map through a Softmax activation function to obtain a spatial attention feature map; and calculate the position-wise multiplication of the spatial attention feature map and the first Gram angle and field convolutional feature map to obtain the first acoustic emission detection feature map.

[0074] More specifically, the second feature extraction module is further configured to: input the Gram angle and field images into the multi-layer convolutional layers of the second convolutional neural network model using channel attention to obtain a second Gram angle and field convolutional feature map; calculate the global mean of each feature matrix along the channel dimension of the second Gram angle and field convolutional feature map to obtain a channel feature vector; input the channel feature vector into the Sigmoid activation function to obtain a channel attention weight vector; and weight each feature matrix along the channel dimension of the second Gram angle and field convolutional feature map using the feature values ​​at each position in the channel attention weight vector as weights to obtain a second acoustic emission detection feature map.

[0075] Therefore, the channel attention and the spatial attention can focus on the feature content and feature location in the Gram angle and field image, respectively, complementing each other to a certain extent and improving the feature extraction effect of the network. This allows a large amount of effective information about whether the jacket has underwater cracks to be captured, which can effectively enhance the feature discrimination learning ability. During network training, the task processing system focuses more on finding significant useful information in the input image data that is relevant to the current output, thereby improving the quality of the output. Furthermore, the increasing number of attention modules will bring continuous performance improvement.

[0076] Specifically, in this embodiment of the application, the feature fusion module 150 is used to fuse the first acoustic emission detection feature map and the second acoustic emission detection feature map to obtain a classification feature map.

[0077] Furthermore, the first and second acoustic emission detection feature maps are fused to integrate the implicit feature information about the spatial location of the jacket structure and the content correlation feature information between channels in the acoustic emission echo signal. This information is then used as a classification feature map for classification processing by a classifier to obtain a classification result indicating whether the jacket structure has underwater cracks. This enables intelligent detection of underwater cracks in offshore wind farm jacket structures, ensuring the safety of offshore wind farm jacket structures.

[0078] Further, the first acoustic emission detection feature map and the second acoustic emission detection feature map are fused using the following formula to obtain a classification feature map; wherein, the formula is:

[0079]

[0080] Wherein, F is the classification feature map, F1 is the first acoustic emission detection feature map, and F2 is the second acoustic emission detection feature map. This indicates positional addition, where λ and β are weighting parameters used to control the balance between the first acoustic emission detection feature map and the second acoustic emission detection feature map.

[0081] Specifically, in this embodiment of the application, the feature manifold correction module 160 is used to perform high-dimensional feature manifold correction on the classification feature map to obtain a corrected classification feature map.

[0082] Specifically, in the technical solution of this application, since the first acoustic emission detection feature map and the second acoustic emission detection feature map are obtained from the Gram angle and field images respectively using a first convolutional neural network model with spatial attention and a second convolutional neural network model with channel attention, the first acoustic emission detection feature map and the second acoustic emission detection feature map can respectively enhance the semantic association features of the Gram angle and field images in the image spatial dimension and the model channel dimension. However, while making full use of features in different dimensions, the feature distribution of the classification feature map obtained by fusing the first acoustic emission detection feature map and the second acoustic emission detection feature map also needs to include feature distributions expressing feature associations in different dimensions. This makes the fitting burden between the classification feature map and the weight matrix of the classifier heavy after the classification feature map is expanded into feature vectors in the classifier, thereby affecting the training speed of the classifier and the accuracy of the classification results.

[0083] Therefore, the applicant of this application performs class representation flattening on the classification feature map, that is, performs high-dimensional feature manifold correction on the classification feature map using the following formula to obtain the corrected classification feature map; wherein, the formula is:

[0084]

[0085] Where, f i and f j The feature values ​​at each position of the classification feature map are represented, and Let N represent the mean of all feature values ​​of the classification feature map, and let N represent the scale of the classification feature map. exp(·) represents the exponential operation of the numerical value, where the exponential operation represents the value of the natural exponential function raised to the power of the numerical value. i ' represents the feature value at each position in the corrected classification feature map.

[0086] Here, the class representation flattening of the classification feature map flattens the finite polyhedral manifold used for class representation of feature distribution in high-dimensional feature space, while maintaining the inherent distance between the planes of the manifold and avoiding intersections based on spatial intuition. Essentially, it decomposes the finite polyhedral manifold into a cubic lattice based on intersections of right angles and common vertices, thereby obtaining the flattened "slice" continuity of the class plane, enhancing the fitting performance of the classification feature map to the classifier's weight matrix. This improves the training speed and accuracy of the classification results when the classification feature map is used by the classifier. Consequently, it enables precise intelligent detection of underwater cracks in offshore wind farm jacket structures, ensuring the safety of offshore wind farm jacket structures and reducing subsequent maintenance costs for enterprises.

[0087] Specifically, in this embodiment, the detection result generation module 170 is used to pass the corrected classification feature map through a classifier to obtain a classification result, which is used to indicate whether the jacket structure has underwater cracks. This enables intelligent detection of underwater cracks in offshore wind farm jacket structures, ensuring the safety of offshore wind farm jacket structures.

[0088] In the embodiments of this application, Figure 3 The following is a block diagram of the detection structure generation module in the underwater crack detection system for jackets based on acoustic emission signals according to an embodiment of this application, as shown below. Figure 3 As shown, the detection structure generation module includes: a feature map unrolling unit 210, used to unroll the corrected classification feature map into a classification feature vector according to row vectors or column vectors; a fully connected encoding unit 220, used to fully connect and encode the classification feature vector using multiple fully connected layers of the classifier to obtain an encoded classification feature vector; and a classification unit 230, used to pass the encoded classification feature vector through the Softmax classification function of the classifier to obtain the classification result.

[0089] Furthermore, in a specific example of this application, the corrected classification feature map is passed through the classifier using the following formula to obtain a classification result indicating whether the jacket has underwater cracks; wherein, the formula is:

[0090] O = softmax{(W n B n ): ...:(W1, B1)|Project(F)}, where W1 to W n The weight matrix is ​​B1 to B1. n Let F be the bias vector, and Project(F) be the vector projected onto the corrected classification feature map.

[0091] In summary, the underwater crack detection system 100 for jacket structures based on acoustic emission signals, according to embodiments of this application, is explained. It employs a deep learning-based artificial intelligence detection algorithm, using the acoustic emission echo signal from the acoustic emission sensor acting on the jacket structure under test as input data. After performing a field transformation from a Cartesian coordinate system to a polar coordinate system, it extracts feature distribution information regarding the spatial focus and channel content of the jacket structure from the acoustic emission echo signal. The system then uses the fusion of these two feature information to detect and determine whether underwater cracks exist in the jacket structure. This allows for precise intelligent detection of underwater cracks in offshore wind farm jacket structures, ensuring the safety of offshore wind farm jacket structures.

[0092] Exemplary methods

[0093] Figure 4 This is a flowchart of an underwater crack detection method for jackets based on acoustic emission signals, according to an embodiment of this application. Figure 4As shown, the underwater crack detection method for jacket based on acoustic emission signals according to an embodiment of this application includes: an acoustic emission echo signal acquisition step S110: acquiring the acoustic emission echo signal of the acoustic emission sensor acting on the jacket under test; a domain mapping step S120: performing Gram angle and field transformation on the acoustic emission echo signal to obtain Gram angle and field images; a first feature extraction step S130: using a first convolutional neural network model with spatial attention to obtain a first acoustic emission detection feature map from the Gram angle and field images; and a second feature extraction step S140: processing the Gram angle and field images into a first acoustic emission detection feature map. Gram angle and field image are used to obtain a second acoustic emission detection feature map by using a second convolutional neural network model with channel attention; feature fusion step S150: fuse the first acoustic emission detection feature map and the second acoustic emission detection feature map to obtain a classification feature map; feature manifold correction step S160: perform high-dimensional feature manifold correction on the classification feature map to obtain a corrected classification feature map; and detection result generation step S170: pass the corrected classification feature map through a classifier to obtain a classification result, which is used to indicate whether there is underwater crack in the jacket.

[0094] Figure 5 This is a schematic diagram of the system architecture of the underwater crack detection method for jackets based on acoustic emission signals according to an embodiment of this application. Figure 5 As shown, in the system architecture of the underwater crack detection method for jacket based on acoustic emission signals, firstly, the acoustic emission echo signal of the acoustic emission sensor acting on the jacket under test is acquired; then, the acoustic emission echo signal is subjected to Gram angle and field transformation to obtain Gram angle and field images; next, the Gram angle and field images are used to obtain a first acoustic emission detection feature map by using a first convolutional neural network model with spatial attention; then, the Gram angle and field images are used to obtain a second acoustic emission detection feature map by using a second convolutional neural network model with channel attention; next, the first acoustic emission detection feature map and the second acoustic emission detection feature map are fused to obtain a classification feature map; then, the classification feature map is subjected to high-dimensional feature manifold correction to obtain a corrected classification feature map; and finally, the corrected classification feature map is passed through a classifier to obtain a classification result, which is used to indicate whether there is an underwater crack in the jacket.

[0095] In a specific example, in the above-mentioned underwater crack detection method for jacket based on acoustic emission signals, the first feature extraction step further includes: a deep convolutional encoding sub-step: using the convolutional encoding part of the first convolutional neural network model using spatial attention to perform deep convolutional encoding on the Gram angle and field images to obtain a first Gram angle and field convolutional feature map; a spatial attention map generation sub-step: inputting the first Gram angle and field convolutional feature map into the spatial attention part of the first convolutional neural network model using spatial attention to obtain a spatial attention map; an activation sub-step: passing the spatial attention map through a Softmax activation function to obtain a spatial attention feature map; and a first acoustic emission detection feature map calculation sub-step: calculating the position-based multiplication of the spatial attention feature map and the first Gram angle and field convolutional feature map to obtain the first acoustic emission detection feature map.

[0096] In a specific example, in the above-mentioned underwater crack detection method for jacket based on acoustic emission signals, the second feature extraction step further includes: a convolution sub-step: inputting the Gram angle and field images into the multi-layer convolutional layers of the second convolutional neural network model using channel attention to obtain a second Gram angle and field convolutional feature map; a channel feature vector calculation sub-step: calculating the global mean of each feature matrix along the channel dimension of the second Gram angle and field convolutional feature map to obtain a channel feature vector; an activation sub-step: inputting the channel feature vector into the Sigmoid activation function to obtain a channel attention weight vector; and a second acoustic emission detection feature map generation sub-step: using the feature values ​​at each position in the channel attention weight vector as weights to weight each feature matrix along the channel dimension of the second Gram angle and field convolutional feature map to obtain the second acoustic emission detection feature map.

[0097] Then, a convolutional neural network model, which excels in extracting local implicit correlation features from images, is used to extract features from the Gram angle and field images. Specifically, considering that in detecting underwater cracks in the jacket under test, the focus should be on the spatial location feature distribution information of the jacket and the correlation feature distribution information of content features in the channels, in order to accurately detect and judge underwater cracks in the jacket under test. Given that the attention mechanism can select the focus position, generating more discriminative feature representations, and that the features after adding the attention module will adaptively change as the network deepens.

[0098] Based on this, in the technical solution of this application, in order to improve the extraction effect of implicit features of acoustic emission echo signals in the Gram angle and field images and thus improve the accuracy of underwater crack detection for the jacket under test, a convolutional neural network model with a dual attention mechanism is used to perform feature mining on the Gram angle and field images. Specifically, the dual attention mechanism here is a spatial attention mechanism and a channel attention mechanism. That is, specifically, the Gram angle and field images are used to obtain a first acoustic emission detection feature map by using a first convolutional neural network model with spatial attention, and the Gram angle and field images are used to obtain a second acoustic emission detection feature map by using a second convolutional neural network model with channel attention.

[0099] In a specific example, in the above-mentioned underwater crack detection method for jackets based on acoustic emission signals, the feature fusion step further includes: fusing the first acoustic emission detection feature map and the second acoustic emission detection feature map using the following formula to obtain a classification feature map; wherein, the formula is:

[0100]

[0101] Wherein, F is the classification feature map, F1 is the first acoustic emission detection feature map, and F2 is the second acoustic emission detection feature map. This indicates positional addition, where λ and β are weighting parameters used to control the balance between the first acoustic emission detection feature map and the second acoustic emission detection feature map.

[0102] Furthermore, the first and second acoustic emission detection feature maps are fused to integrate the implicit feature information about the spatial location of the jacket structure and the content correlation feature information between channels in the acoustic emission echo signal. This information is then used as a classification feature map for classification processing by a classifier to obtain a classification result indicating whether the jacket structure has underwater cracks. This enables intelligent detection of underwater cracks in offshore wind farm jacket structures, ensuring the safety of offshore wind farm jacket structures.

[0103] Specifically, in the technical solution of this application, since the first acoustic emission detection feature map and the second acoustic emission detection feature map are obtained from the Gram angle and field images respectively using a first convolutional neural network model with spatial attention and a second convolutional neural network model with channel attention, the first acoustic emission detection feature map and the second acoustic emission detection feature map can respectively enhance the semantic association features of the Gram angle and field images in the image spatial dimension and the model channel dimension. However, while making full use of features in different dimensions, the feature distribution of the classification feature map obtained by fusing the first acoustic emission detection feature map and the second acoustic emission detection feature map also needs to include feature distributions expressing feature associations in different dimensions. This makes the fitting burden between the classification feature map and the weight matrix of the classifier heavy after the classification feature map is expanded into feature vectors in the classifier, thereby affecting the training speed of the classifier and the accuracy of the classification results.

[0104] In a specific example, in the above-mentioned underwater crack detection method for jackets based on acoustic emission signals, the feature manifold correction step further includes: performing high-dimensional feature manifold correction on the classification feature map using the following formula to obtain the corrected classification feature map; wherein, the formula is:

[0105]

[0106] Where, f i and f j The feature values ​​at each position of the classification feature map are represented, and Let N represent the mean of all feature values ​​of the classification feature map, and let N represent the scale of the classification feature map. exp(·) represents the exponential operation of the numerical value, where the exponential operation represents the value of the natural exponential function raised to the power of the numerical value. i ' represents the feature value at each position in the corrected classification feature map.

[0107] Here, the class representation flattening of the classification feature map flattens the finite polyhedral manifold used for class representation of feature distribution in high-dimensional feature space, while maintaining the inherent distance between the planes of the manifold and avoiding intersections based on spatial intuition. Essentially, it decomposes the finite polyhedral manifold into a cubic lattice based on intersections of right angles and common vertices, thereby obtaining the flattened "slice" continuity of the class plane, enhancing the fitting performance of the classification feature map to the classifier's weight matrix. This improves the training speed and accuracy of the classification results when the classification feature map is used by the classifier. Consequently, it enables precise intelligent detection of underwater cracks in offshore wind farm jacket structures, ensuring the safety of offshore wind farm jacket structures and reducing subsequent maintenance costs for enterprises.

[0108] In a specific example, in the above-mentioned underwater crack detection method for jackets based on acoustic emission signals, the step of generating the detection result includes: expanding the corrected classification feature map into a classification feature vector according to row vectors or column vectors; using multiple fully connected layers of the classifier to fully connect and encode the classification feature vector to obtain an encoded classification feature vector; and passing the encoded classification feature vector through the Softmax classification function of the classifier to obtain the classification result.

[0109] Here, those skilled in the art will understand that the specific operations of each step in the above-described method for underwater crack detection of jacket structures based on acoustic emission signals have been referenced above. Figures 1 to 3 The underwater crack detection system for jackets based on acoustic emission signals has been described in detail, and therefore, its repeated description will be omitted.

[0110] The basic principles of this application have been described above with reference to specific embodiments. However, it should be noted that the advantages, benefits, and effects mentioned in this application are merely examples and not limitations, and should not be considered as essential features of each embodiment of this application. Furthermore, the specific details disclosed above are for illustrative and facilitative purposes only, and are not limitations. These details do not limit the application to the necessity of employing the aforementioned specific details for implementation.

[0111] The block diagrams of devices, apparatuses, devices, and systems involved in this application are merely illustrative examples and are not intended to require or imply that they must be connected, arranged, or configured in the manner shown in the block diagrams. As those skilled in the art will recognize, these devices, apparatuses, devices, and systems can be connected, arranged, and configured in any manner. Words such as “comprising,” “including,” “having,” etc., are open-ended terms meaning “including but not limited to,” and are used interchangeably with them. The terms “or” and “and” as used herein refer to the terms “and / or,” and are used interchangeably with them unless the context clearly indicates otherwise. The term “such as” as used herein refers to the phrase “such as but not limited to,” and is used interchangeably with it.

[0112] It should also be noted that in the apparatus, equipment, and methods of this application, the components or steps can be disassembled and / or recombined. These disassemblies and / or recombinations should be considered as equivalent solutions of this application.

[0113] The above description of the disclosed aspects is provided to enable any person skilled in the art to make or use this application. Various modifications to these aspects will be readily apparent to those skilled in the art, and the general principles defined herein can be applied to other aspects without departing from the scope of this application. Therefore, this application is not intended to be limited to the aspects shown herein, but rather to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A system for detecting underwater cracks in a jacket structure based on acoustic emission signals, characterized in that, include: Acoustic emission echo signal acquisition module is used to acquire the acoustic emission echo signal of the acoustic emission sensor acting on the duct frame under test; The domain mapping module is used to perform Gram angle and field transformation on the acoustic emission echo signal to obtain Gram angle and field image; A first feature extraction module is configured to obtain a first acoustic emission detection feature map by applying a first convolutional neural network model with spatial attention to the Gram angle and field images; wherein, the first feature extraction module is further configured to: perform depth convolutional encoding on the Gram angle and field images using the convolutional encoding part of the first convolutional neural network model with spatial attention to obtain a first Gram angle and field convolutional feature map; input the first Gram angle and field convolutional feature map into the spatial attention part of the first convolutional neural network model with spatial attention to obtain a spatial attention map; apply the spatial attention map to the Softmax activation function to obtain a spatial attention feature map; and calculate the positional dot product of the spatial attention feature map and the first Gram angle and field convolutional feature map to obtain the first acoustic emission detection feature map; The second feature extraction module is used to obtain a second acoustic emission detection feature map by using a second convolutional neural network model with channel attention to process the Gram angle and field images; wherein, the second feature extraction module is further used to: input the Gram angle and field images into multiple convolutional layers of the second convolutional neural network model using channel attention to obtain a second Gram angle and field convolutional feature map; calculate the global mean of each feature matrix along the channel dimension of the second Gram angle and field convolutional feature map to obtain a channel feature vector; input the channel feature vector into a Sigmoid activation function to obtain a channel attention weight vector; and use the feature values ​​at each position in the channel attention weight vector as weights to weight each feature matrix along the channel dimension of the second Gram angle and field convolutional feature map to obtain the second acoustic emission detection feature map; The feature fusion module is used to fuse the first acoustic emission detection feature map and the second acoustic emission detection feature map to obtain a classification feature map; The feature manifold correction module is used to perform high-dimensional feature manifold correction on the classification feature map to obtain a corrected classification feature map; the feature manifold correction module is further used to: perform high-dimensional feature manifold correction on the classification feature map using the following formula to obtain the corrected classification feature map; The formula is as follows: ， in, and The feature values ​​at each position of the classification feature map are represented, and This represents the mean of all feature values ​​in the classification feature map, and The scale of the classification feature map is represented. The exponentiation of a numerical value represents the value of a natural exponential function raised to the power of the numerical value. The feature values ​​at each position of the corrected classification feature map are represented by these features. And a detection result generation module, used to pass the corrected classification feature map through a classifier to obtain a classification result, the classification result being used to indicate whether the jacket has underwater cracks.

2. The underwater crack detection system for jackets based on acoustic emission signals according to claim 1, characterized in that, The feature fusion module is further configured to: fuse the first acoustic emission detection feature map and the second acoustic emission detection feature map using the following formula to obtain a classification feature map; The formula is as follows: ， in, For the classification feature map, This is the first acoustic emission detection feature map. This is the second acoustic emission detection feature map. This indicates addition by position. The weighting parameter is used to control the balance between the first acoustic emission detection feature map and the second acoustic emission detection feature map.

3. The underwater crack detection system for jackets based on acoustic emission signals according to claim 2, characterized in that, The detection result generation module includes: The feature map unpacking unit is used to unpack the corrected classification feature map into a classification feature vector according to a row vector or a column vector. A fully connected encoding unit is configured to perform fully connected encoding on the classification feature vector using multiple fully connected layers of the classifier to obtain an encoded classification feature vector; and A classification unit is used to pass the encoded classification feature vector through the Softmax classification function of the classifier to obtain the classification result.

4. A method for underwater crack detection of jacket structures based on acoustic emission signals, characterized in that, include: Acoustic emission echo signal acquisition steps: Acquire the acoustic emission echo signal of the acoustic emission sensor acting on the duct frame under test; Domain mapping step: Perform Gram angle and field transformation on the acoustic emission echo signal to obtain Gram angle and field image; The first feature extraction step involves using a first convolutional neural network model with spatial attention to obtain a first acoustic emission detection feature map from the Gram angle and field images. This first feature extraction step further includes: a deep convolutional encoding sub-step: performing deep convolutional encoding on the Gram angle and field images using the convolutional encoding part of the first convolutional neural network model with spatial attention to obtain a first Gram angle and field convolutional feature map; a spatial attention map generation sub-step: inputting the first Gram angle and field convolutional feature map into the spatial attention part of the first convolutional neural network model with spatial attention to obtain a spatial attention map; an activation sub-step: applying a Softmax activation function to the spatial attention map to obtain a spatial attention feature map; and a first acoustic emission detection feature map calculation sub-step: calculating a positional multiplication of the spatial attention feature map and the first Gram angle and field convolutional feature map to obtain the first acoustic emission detection feature map. The second feature extraction step involves using the Gram angle and field images with a second convolutional neural network model employing channel attention to obtain a second acoustic emission detection feature map. This second feature extraction step further includes: a convolution sub-step: inputting the Gram angle and field images into multiple convolutional layers of the second convolutional neural network model employing channel attention to obtain a second Gram angle and field convolutional feature map; a channel feature vector calculation sub-step: calculating the global mean of each feature matrix along the channel dimension of the second Gram angle and field convolutional feature map to obtain a channel feature vector; an activation sub-step: inputting the channel feature vector into a Sigmoid activation function to obtain a channel attention weight vector; and a second acoustic emission detection feature map generation sub-step: using the feature values ​​at each position in the channel attention weight vector as weights to weight each feature matrix along the channel dimension of the second Gram angle and field convolutional feature map to obtain the second acoustic emission detection feature map. Feature fusion step: Fuse the first acoustic emission detection feature map and the second acoustic emission detection feature map to obtain a classification feature map; Feature manifold correction step: Perform high-dimensional feature manifold correction on the classification feature map to obtain a corrected classification feature map; the feature manifold correction step further includes: performing high-dimensional feature manifold correction on the classification feature map using the following formula to obtain the corrected classification feature map; wherein, the formula is: , in, and The feature values ​​at each position of the classification feature map are represented, and Let N represent the mean of all feature values ​​of the classification feature map, and let N represent the scale of the classification feature map. The exponentiation of a numerical value represents the value of a natural exponential function raised to the power of the numerical value. This represents the feature value at each position in the corrected classification feature map; The detection result generation step involves passing the corrected classification feature map through a classifier to obtain a classification result, which is used to indicate whether the jacket has underwater cracks.

5. The underwater crack detection method for jacket structures based on acoustic emission signals according to claim 4, characterized in that, The feature fusion step further includes: fusing the first acoustic emission detection feature map and the second acoustic emission detection feature map using the following formula to obtain a classification feature map; The formula is as follows: ， in, For the classification feature map, This is the first acoustic emission detection feature map. This is the second acoustic emission detection feature map. This indicates addition by position. The weighting parameter is used to control the balance between the first acoustic emission detection feature map and the second acoustic emission detection feature map.

Images