Cross-device bearing fault diagnosis method and system, storage medium and electronic device

Abstract

The invention provides a cross-equipment bearing fault diagnosis method and system, a storage medium and electronic equipment. The method comprises the following steps: acquiring source domain data and target domain data; denoising the source domain data and the target domain data; extracting features of the source domain data and the target domain data to obtain an initial feature vector; iteratively updating nodes in the source domain data and the target domain data to obtain a source domain graph and a target domain graph, and calculating kernel values; calculating similarity loss between the two to obtain a graph kernel loss function, calculating a bearing category real distribution probability and a bearing category prediction distribution probability to obtain a cross entropy loss function, and calculating a total loss function; performing gradient calculation on the graph kernel function, the source domain graph, the target domain graph, the nodes and the parameters of the graph kernel model in sequence; supervised learning is carried out on the classification result, and combined optimization is carried out on the graph kernel model in combination with a graph kernel loss function. According to the method, the graphical difference between the source domain graph and the target domain graph is reduced, and the generalization ability of the recognition model is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent fault diagnosis, and in particular to a cross-device bearing fault diagnosis method, system, storage medium and electronic equipment. Background Art

[0002] Axle box bearings are one of the most important rotating components in rail vehicles, and their health status directly affects the performance and life of the equipment. Early diagnosis of bearing faults can effectively improve the accuracy and timeliness of fault detection, reduce operation and maintenance costs, and improve production efficiency.

[0003] However, in the prior art, due to different objects or operating conditions, the data distribution of the measured signals will be different. The deep learning model trained under specific operating conditions may not be suitable for fault diagnosis tasks under other conditions or objects, and the model generalization is poor. Summary of the invention

[0004] In view of the shortcomings of the prior art, the purpose of the present invention is to provide a cross-device bearing fault diagnosis method, aiming to solve the technical problems in the prior art that due to different objects or operating conditions, the data distribution of the measured signal will be different, the deep learning model trained under specific operating conditions may not be suitable for fault diagnosis tasks under other conditions or objects, and the model generalization is poor.

[0005] In order to achieve the above object, the present invention is implemented by the following technical solutions:

[0006] A cross-device bearing fault diagnosis method comprises the following steps:

[0007] Acquire the signal data of the bearing in a known environment and use it as the source domain data, and acquire the signal data of the bearing in other environments and use it as the target domain data;

[0008] Performing denoising processing on the source domain data and the target domain data;

[0009] Using a convolutional neural network to perform feature extraction on the denoised source domain data and the target domain data to obtain an initial feature vector;

[0010] The initial feature vector is regarded as a node, and a graph kernel model is introduced to iteratively update the nodes in the source domain data and the target domain data respectively to obtain a source domain graph and a target domain graph, and a kernel value between the source domain graph and the target domain graph is calculated by a graph kernel function;

[0011] Based on the graph kernel loss function, the similarity loss between the source domain graph and the target domain graph is calculated according to the kernel value, based on the cross entropy loss function, the true distribution probability of the bearing category is obtained according to the source domain data and the target domain data, and the predicted distribution probability of the bearing category is obtained in the process of iteratively updating the nodes in the source domain data and the target domain data through the fully connected layer in the convolutional neural network, so as to calculate the difference between the true distribution probability of the bearing category and the predicted distribution probability of the bearing category, and the total loss function is calculated by combining the graph kernel loss function and the cross entropy loss function;

[0012] Based on the total loss function, sequentially performing gradient calculations on the graph kernel function, the source domain graph, the target domain graph, the nodes, and the parameters of the graph kernel model;

[0013] The cross entropy loss function is used to perform supervised learning on the classification results during the iterative updating of the nodes in the source domain data and the target domain data, and the graph core model is jointly optimized in combination with the graph core loss function.

[0014] According to one aspect of the above technical solution, the denoising process of the source domain data and the target domain data specifically includes:

[0015] Dividing the source domain data into a source domain training set and a source domain test set, and dividing the target domain data into a target domain training set and a target domain test set;

[0016] Dividing the source domain training set, the source domain test set, the target domain training set, and the target domain test set into data samples according to a preset data length, and performing standardization processing on the data samples;

[0017] The abnormal amplitude replacement method is used to correct the abnormal values ​​in the data sample after the standardization process.

[0018] According to one aspect of the above technical solution, the use of a convolutional neural network to extract features from the denoised source domain data and the target domain data to obtain an initial feature vector specifically includes:

[0019] Introducing a convolutional neural network, wherein the convolutional neural network includes multiple convolutional layers, normalization layers, and global average pooling layers;

[0020] Extracting the time domain and frequency domain features of the denoised source domain data and the target domain data layer by layer through multiple convolutional layers;

[0021] Normalizing the denoised source domain data and the target domain data through the normalization layer;

[0022] The global average pooling layer is used to calculate the global average of the internal channels of the input features in the convolutional neural network to obtain an initial feature vector.

[0023] According to one aspect of the above technical solution, the initial feature vector is regarded as a node, a graph kernel model is introduced to iteratively update the nodes in the source domain data and the target domain data respectively to obtain a source domain graph and a target domain graph, and a kernel value between the source domain graph and the target domain graph is calculated by a graph kernel function, specifically including:

[0024] The initial feature vector is regarded as a node, a graph core model is introduced, and label embedding initialization is performed on each node in the graph core model to obtain an initial embedding value;

[0025]

[0026] Among them, E I (v) represents the initial feature vector corresponding to the node v in the source domain data, represents the initial embedding value corresponding to the node v in the source domain data, E I (i) represents the initial feature vector corresponding to node i in the target domain data, represents the initial embedding value corresponding to node i in the target domain data;

[0027] Performing neighbor aggregation calculation on the initial embedding values ​​in the source domain data and the target domain data layer by layer through the graph kernel model;

[0028]

[0029] Among them, k represents the number of neighbor points in the source domain data, N represents the total number of nodes in the source domain data, represents the embedding value corresponding to the node u in the k-1th layer in the source domain data, represents the neighbor aggregation feature of node v in the kth layer in the source domain data, j represents the number of neighbor points in the target domain data, C represents the total number of nodes in the target domain data, represents the embedding value corresponding to the node f at the j-1th layer in the target domain data, Represents the neighbor aggregation features of the target domain data at the jth layer node i;

[0030] Based on the result of neighbor aggregation calculation, updating nodes initially embedded in the source domain data and the target domain data respectively;

[0031]

[0032] Among them, σ represents the activation function, W represents the weight, and b represents the bias. represents the updated node v in the kth layer of the source domain data, represents the updated node i in the jth layer in the target domain data;

[0033] By respectively aggregating updated nodes in the source domain data and the target domain data, an embedding representation of a graph is constructed to obtain a source domain graph and a target domain graph respectively;

[0034]

[0035] in, represents the updated node v, V in the last layer K of the source domain data G represents the total number of updated nodes in the source domain data, h G represents the source domain graph, represents the updated node i in the last layer J in the target domain data, I H represents the total number of updated nodes in the target domain data, h H represents the target domain graph;

[0036] Introducing a graph kernel function to calculate a kernel value between the source domain graph and the target domain graph;

[0037]

[0038] Among them, φ represents the bandwidth of the graph kernel function.

[0039] According to one aspect of the above technical solution, the graph kernel loss function is based on the kernel value, and the similarity loss between the source domain graph and the target domain graph is calculated, and the true distribution probability of the bearing category is obtained according to the source domain data and the target domain data based on the cross entropy loss function, and the predicted distribution probability of the bearing category is obtained in the process of iteratively updating the nodes in the source domain data and the target domain data through the fully connected layer in the convolutional neural network, so as to calculate the difference between the true distribution probability of the bearing category and the predicted distribution probability of the bearing category, which specifically includes:

[0040] Based on a graph kernel loss function, and according to the kernel value, a similarity loss between the source domain graph and the target domain graph is calculated;

[0041] L=1-k θ

[0042] Among them, L represents the graph kernel loss function;

[0043] Customize the fault category items of bearings;

[0044] Based on the definition of the fault category item, obtaining the true distribution probability of the bearing category according to the source domain data and the target domain data;

[0045] Based on the definition of the fault category item, using the fully connected layer in the convolutional neural network, obtaining the bearing category prediction distribution probability in the process of iteratively updating the nodes in the source domain data and the target domain data;

[0046] Based on the cross entropy loss function, according to the true distribution probability of the bearing category and the predicted distribution probability of the bearing category, the difference between the true distribution probability of the bearing category and the predicted distribution probability of the bearing category is calculated;

[0047]

[0048] Among them, P(x a ) indicates that the bearing category in the true distribution is x a The probability of Q(x a ) indicates that the bearing category in the predicted distribution is x a The probability of, F(P, Q) represents the cross entropy loss function.

[0049] According to one aspect of the above technical solution, the total loss function is calculated by combining the graph kernel loss function and the cross entropy loss function, and the total loss function is defined as follows:

[0050] T = L + F (P, Q);

[0051] Among them, T represents the total loss function.

[0052] According to one aspect of the above technical solution, the gradient calculation of the graph kernel function, the source domain graph, the target domain graph, the node, and the parameters of the graph kernel model is performed in sequence based on the total loss function, specifically including:

[0053] Calculating the gradient of the graph kernel function based on the total loss function;

[0054]

[0055] in, Indicates the symbol for partial derivative;

[0056] Based on the gradient of the graph kernel function and by using the chain rule, respectively calculating the gradients of the source domain graph and the target domain graph;

[0057]

[0058] Calculating the gradient of the node based on the gradients of the source domain graph and the target domain graph;

[0059]

[0060] Performing gradient calculation on weight parameters and bias parameters of the graph kernel model;

[0061]

[0062] in, Represents the node v after the last update of neighbor aggregation, represents the node i after the last update of neighbor aggregation, and R represents transpose.

[0063] The present invention also provides a cross-device bearing fault diagnosis system, comprising:

[0064] Acquisition module: used to acquire signal data of the bearing in a known environment and use it as source domain data, and acquire signal data of the bearing in other environments and use it as target domain data;

[0065] A denoising module; used for performing denoising processing on the source domain data and the target domain data;

[0066] The denoising module is specifically used to: divide the source domain data into a source domain training set and a source domain test set, and divide the target domain data into a target domain training set and a target domain test set;

[0067] Dividing the source domain training set, the source domain test set, the target domain training set, and the target domain test set into data samples according to a preset data length, and performing standardization processing on the data samples;

[0068] Using an abnormal amplitude replacement method to correct the abnormal values ​​in the data sample after the standardization process;

[0069] Extraction module: used for performing feature extraction on the denoised source domain data and the target domain data using a convolutional neural network to obtain an initial feature vector;

[0070] The extraction module is specifically used for:

[0071] Introducing a convolutional neural network, wherein the convolutional neural network includes multiple convolutional layers, normalization layers, and global average pooling layers;

[0072] Extracting the time domain and frequency domain features of the denoised source domain data and the target domain data layer by layer through multiple convolutional layers;

[0073] Normalizing the denoised source domain data and the target domain data through the normalization layer;

[0074] Performing global average calculation on the internal channels of the input features in the convolutional neural network through the global average pooling layer to obtain an initial feature vector;

[0075] Graph calculation module: used for treating the initial feature vector as a node, introducing a graph kernel model to iteratively update the nodes in the source domain data and the target domain data respectively to obtain a source domain graph and a target domain graph, and calculating a kernel value between the source domain graph and the target domain graph through a graph kernel function;

[0076] The graph computing module is specifically used for:

[0077] The initial feature vector is regarded as a node, a graph core model is introduced, and label embedding initialization is performed on each node in the graph core model to obtain an initial embedding value;

[0078]

[0079] Among them, E I (v) represents the initial feature vector corresponding to the node v in the source domain data, represents the initial embedding value corresponding to the node v in the source domain data, E I (i) represents the initial feature vector corresponding to node i in the target domain data, represents the initial embedding value corresponding to node i in the target domain data;

[0080] Performing neighbor aggregation calculation on the initial embedding values ​​in the source domain data and the target domain data layer by layer through the graph kernel model;

[0081]

[0082] Among them, k represents the number of neighbor points in the source domain data, N represents the total number of nodes in the source domain data, represents the embedding value corresponding to the node u in the k-1th layer in the source domain data, represents the neighbor aggregation feature of node v in the kth layer in the source domain data, j represents the number of neighbor points in the target domain data, C represents the total number of nodes in the target domain data, represents the embedding value corresponding to the node f at the j-1th layer in the target domain data, Represents the neighbor aggregation features of the target domain data at the jth layer node i;

[0083] Based on the result of neighbor aggregation calculation, updating nodes initially embedded in the source domain data and the target domain data respectively;

[0084]

[0085] Among them, σ represents the activation function, W represents the weight, and b represents the bias. represents the updated node v in the kth layer of the source domain data, represents the updated node i in the jth layer in the target domain data;

[0086] By respectively aggregating updated nodes in the source domain data and the target domain data, an embedding representation of a graph is constructed to obtain a source domain graph and a target domain graph respectively;

[0087]

[0088] in, represents the updated node v, V in the last layer K of the source domain data G represents the total number of updated nodes in the source domain data, h G represents the source domain graph, represents the updated node i in the last layer J in the target domain data, I H represents the total number of updated nodes in the target domain data, h H represents the target domain graph;

[0089] Introducing a graph kernel function to calculate a kernel value between the source domain graph and the target domain graph;

[0090]

[0091] Among them, φ represents the bandwidth of the graph kernel function;

[0092] Loss calculation module: used to calculate the similarity loss between the source domain graph and the target domain graph based on the graph kernel loss function and the kernel value, obtain the true distribution probability of the bearing category based on the source domain data and the target domain data based on the cross entropy loss function, and obtain the predicted distribution probability of the bearing category in the process of iteratively updating the nodes in the source domain data and the target domain data through the fully connected layer in the convolutional neural network, so as to calculate the difference between the true distribution probability of the bearing category and the predicted distribution probability of the bearing category, and calculate the total loss function by combining the graph kernel loss function and the cross entropy loss function;

[0093] The loss calculation module is specifically used for:

[0094] Based on a graph kernel loss function, and according to the kernel value, a similarity loss between the source domain graph and the target domain graph is calculated;

[0095] L=1-k θ ;

[0096] Among them, L represents the graph kernel loss function;

[0097] Customize the fault category items of bearings;

[0098] Based on the definition of the fault category item, obtaining the true distribution probability of the bearing category according to the source domain data and the target domain data;

[0099] Based on the definition of the fault category item, using the fully connected layer in the convolutional neural network, obtaining the predicted distribution probability of the bearing category in the process of iteratively updating the nodes in the source domain data and the target domain data;

[0100] Based on the cross entropy loss function, according to the true distribution probability of the bearing category and the predicted distribution probability of the bearing category, the difference between the true distribution probability of the bearing category and the predicted distribution probability of the bearing category is calculated;

[0101]

[0102] Among them, P(x a ) indicates that the bearing category in the true distribution is x a The probability of Q(x a ) indicates that the bearing category in the predicted distribution is x a The probability of , F(P, Q) represents the cross entropy loss function;

[0103] The total loss function is calculated by combining the graph kernel loss function and the cross entropy loss function. The definition of the total loss function is as follows:

[0104] T = L + F (P, Q);

[0105] Where T represents the total loss function;

[0106] Gradient calculation module: based on the total loss function, sequentially performing gradient calculations on the graph kernel function, the source domain graph, the target domain graph, the nodes, and the parameters of the graph kernel model;

[0107] The gradient calculation module is specifically used for:

[0108] Calculating the gradient of the graph kernel function based on the total loss function;

[0109]

[0110] in, Indicates the symbol for partial derivative;

[0111] Based on the gradient of the graph kernel function and by using the chain rule, respectively calculating the gradients of the source domain graph and the target domain graph;

[0112]

[0113] Calculating the gradient of the node based on the gradients of the source domain graph and the target domain graph;

[0114]

[0115] Performing gradient calculation on weight parameters and bias parameters of the graph kernel model;

[0116]

[0117] in, Represents the node v after the last update of neighbor aggregation, represents the node i after the last update of neighbor aggregation, and R represents transposition;

[0118] Optimization module: used to use the cross entropy loss function to perform supervised learning on the classification results during the iterative update of the nodes in the source domain data and the target domain data, and to jointly optimize the graph core model in combination with the graph core loss function.

[0119] The present invention also provides a storage medium on which a computer program is stored. When the computer program is executed by a processor, the cross-device bearing fault diagnosis method as described above is implemented.

[0120] The present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the cross-device bearing fault diagnosis method as described above when executing the computer program.

[0121] Compared with the prior art, the present invention has the following beneficial effects:

[0122] 1. By combining deep convolutional neural networks with parameterized graph kernel models, local and global features in bearing vibration signals are effectively extracted, significantly improving the accuracy of fault diagnosis;

[0123] 2. By introducing domain-adaptive graph kernel loss function and cross entropy loss function, the distribution difference between source domain and target domain is reduced, making the model's migration effect on different data sets better and improving the generalization ability of the model. The parameterized graph kernel function not only enhances the mathematical interpretability of the model by introducing trainable weights, biases and graph kernel function bandwidth, but also can effectively quantify the differences in graph morphology.

[0124] 3. High degree of automation. The diagnostic process does not need to rely on manual feature engineering, which reduces dependence on expert knowledge and improves diagnostic efficiency and accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0125] Figure 1 Flow chart of a bearing fault diagnosis method across devices in a first embodiment of the present invention;

[0126] Figure 2 This is a schematic diagram of calculating the kernel value between the source domain graph and the target domain graph in the first embodiment of the present invention;

[0127] Figure 3 is a structural block diagram of a convolutional neural network in the first embodiment of the present invention;

[0128] Figure 4 It is a structural block diagram of a cross-device bearing fault diagnosis system in a second embodiment of the present invention;

[0129] Figure 5 is a structural block diagram of an electronic device in a third embodiment of the present invention;

[0130] The following specific implementation manner will further illustrate the present invention in conjunction with the above-mentioned drawings. DETAILED DESCRIPTION

[0131] In order to facilitate the understanding of the present invention, the present invention will be described more fully below with reference to the relevant drawings. Several embodiments of the present invention are given in the drawings. However, the present invention can be implemented in many different forms and is not limited to the embodiments described herein. On the contrary, the purpose of providing these embodiments is to make the disclosure of the present invention more thorough and comprehensive.

[0132] It should be noted that when an element is referred to as being "fixed to" another element, it may be directly on the other element or there may be a central element. When an element is considered to be "connected to" another element, it may be directly connected to the other element or there may be a central element at the same time. The terms "vertical", "horizontal", "left", "right" and similar expressions used herein are for illustrative purposes only.

[0133] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art of the present invention. The terms used herein in the specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. The term "and / or" used herein includes any and all combinations of one or more related listed items.

[0134] See also Figures 1 to 3 , which shows a bearing fault diagnosis method across devices in a first embodiment of the present invention, comprising the following steps:

[0135] S10, acquiring signal data of the bearing in a known environment and using it as source domain data, acquiring signal data of the bearing in other environments and using it as target domain data;

[0136] S20, performing denoising processing on the source domain data and the target domain data;

[0137] S30, using a convolutional neural network to perform feature extraction on the denoised source domain data and the target domain data to obtain an initial feature vector;

[0138] S40, regarding the initial feature vector as a node, introducing a graph kernel model to iteratively update the nodes in the source domain data and the target domain data respectively to obtain a source domain graph and a target domain graph, and calculating a kernel value between the source domain graph and the target domain graph through a graph kernel function;

[0139] S50, based on the graph kernel loss function, and according to the kernel value, calculating the similarity loss between the source domain graph and the target domain graph, based on the cross entropy loss function, obtaining the true distribution probability of the bearing category according to the source domain data and the target domain data, and obtaining the predicted distribution probability of the bearing category in the process of iteratively updating the nodes in the source domain data and the target domain data through the fully connected layer in the convolutional neural network, so as to calculate the difference between the true distribution probability of the bearing category and the predicted distribution probability of the bearing category, and calculating the total loss function in combination with the graph kernel loss function and the cross entropy loss function;

[0140] S60, performing gradient calculations on the graph kernel function, the source domain graph, the target domain graph, the nodes, and the parameters of the graph kernel model in sequence based on the total loss function;

[0141] S70, using the cross entropy loss function, performing supervised learning on the classification results in the process of iteratively updating the nodes in the source domain data and the target domain data, and jointly optimizing the graph core model in combination with the graph core loss function.

[0142] It can be understood that the beneficial effects of the present invention are:

[0143] 1. By combining deep convolutional neural networks with parameterized graph kernel models, local and global features in bearing vibration signals are effectively extracted, significantly improving the accuracy of fault diagnosis;

[0144] 2. By introducing domain-adaptive graph kernel loss function and cross entropy loss function, the distribution difference between source domain and target domain is reduced, making the model's migration effect on different data sets better and improving the generalization ability of the model. The parameterized graph kernel function not only enhances the mathematical interpretability of the model by introducing trainable weights, biases and graph kernel function bandwidth, but also can effectively quantify the differences in graph morphology.

[0145] 3. High degree of automation. The diagnostic process does not need to rely on manual feature engineering, which reduces dependence on expert knowledge and improves diagnostic efficiency and accuracy.

[0146] Specifically, in this embodiment, the source domain data is the bearing installed on the wheel body, and we have already trained the recognition model for this type of image, and the target domain data is the bearing installed on other wheels, and we have not trained the recognition model for this type of image. The purpose of this application is to reduce the morphological difference between the source domain data and the target domain data through image algorithms, and then identify the bearings on different wheels through recognition models, so as to improve the generalization ability of the model.

[0147] Furthermore, the specific steps of step S10 include:

[0148] The source domain data is divided into a source domain training set and a source domain test set, and the target domain data is divided into a target domain training set and a target domain test set; it should be noted that the source domain data and the target domain data are both vibration signal data;

[0149] The source domain training set, the source domain test set, the target domain training set, and the target domain test set are divided into data samples according to a preset data length, and the data samples are standardized; this step is to enhance the robustness of the image recognition model;

[0150] The abnormal amplitude replacement method is used to correct the abnormal values ​​in the data sample after the standardization process.

[0151] It can be understood that the purpose of this step is mainly to reduce the interference of noise on signal data, thereby improving the ability to diagnose bearings in the later stage.

[0152] Furthermore, the specific steps of step S30 include:

[0153] A convolutional neural network (CNN) is introduced, wherein the convolutional neural network includes multiple convolutional layers, normalization layers, and global average pooling layers;

[0154] Extracting the time domain and frequency domain features of the denoised source domain data and the target domain data layer by layer through multiple convolutional layers;

[0155] The denoised source domain data and the target domain data are normalized by the normalization layer; this step can make the distribution of data more standardized, improve the convergence speed of the model and the generalization ability of the model;

[0156] The global average pooling layer is used to calculate the global average of the internal channels of the input features in the convolutional neural network to obtain an initial feature vector; the initial feature vector is a one-dimensional feature vector.

[0157] Furthermore, the specific steps of step S40 include:

[0158] The initial feature vector is regarded as a node, a graph core model is introduced, and label embedding initialization is performed on each node in the graph core model to obtain an initial embedding value;

[0159]

[0160] Among them, E I (v) represents the initial feature vector corresponding to the node v in the source domain data, represents the initial embedding value corresponding to the node v in the source domain data, E I (i) represents the initial feature vector corresponding to node i in the target domain data, represents the initial embedding value corresponding to node i in the target domain data;

[0161] Performing neighbor aggregation calculation on the initial embedding values ​​in the source domain data and the target domain data layer by layer through the graph kernel model;

[0162]

[0163] Among them, k represents the number of neighbor points in the source domain data, N represents the total number of nodes in the source domain data, represents the embedding value corresponding to the node u in the k-1th layer in the source domain data, represents the neighbor aggregation feature of node v in the kth layer in the source domain data, j represents the number of neighbor points in the target domain data, C represents the total number of nodes in the target domain data, represents the embedding value corresponding to the node f at the j-1th layer in the target domain data, Represents the neighbor aggregation features of the target domain data at the jth layer node i;

[0164] Based on the result of neighbor aggregation calculation, updating nodes initially embedded in the source domain data and the target domain data respectively;

[0165]

[0166] Among them, σ represents the activation function, W represents the weight, and b represents the bias. represents the updated node v in the kth layer of the source domain data, represents the updated node i in the jth layer in the target domain data;

[0167] By respectively aggregating updated nodes in the source domain data and the target domain data, an embedding representation of a graph is constructed to obtain a source domain graph and a target domain graph respectively;

[0168]

[0169] in, represents the updated node v, V in the last layer K of the source domain data G represents the total number of updated nodes in the source domain data, h G represents the source domain graph, represents the updated node i in the last layer J in the target domain data, I H represents the total number of updated nodes in the target domain data, h H represents the target domain graph;

[0170] Introducing a graph kernel function to calculate a kernel value between the source domain graph and the target domain graph;

[0171]

[0172] Among them, φ represents the bandwidth of the graph kernel function.

[0173] It can be understood that node label embedding is a technique that maps nodes in a network to a low-dimensional vector space, which can maintain the similarity of nodes in the network so that the node relationship in the embedded space can approximately reflect the structure and properties of the original network. The basic idea of ​​node embedding is to represent each node as a low-dimensional vector that maintains a certain similarity in the network.

[0174] Graph representation (embedded representation of source domain graph and target domain graph) is a data structure that builds relationships between nodes and edges. The graph structure can quickly retrieve specific information in the graph and facilitate the extraction and analysis of graph morphological features.

[0175] By calculating the kernel value between the source domain graph and the target domain graph embedding to measure the similarity between the two, the parameterized graph kernel function can better characterize the differences in graph morphological features between the source domain and the target domain, ensuring a more accurate domain adaptation process.

[0176] This step is performed in two ways, one is to gradually calculate the source domain graph starting from the source domain data, and the other is to gradually calculate the target domain graph starting from the target domain data.

[0177] Further, in the step S50, the similarity loss between the source domain graph and the target domain graph is calculated based on the graph kernel loss function and the kernel value, the true distribution probability of the bearing category is obtained based on the source domain data and the target domain data based on the cross entropy loss function, and the predicted distribution probability of the bearing category is obtained in the process of iteratively updating the nodes in the source domain data and the target domain data through the fully connected layer in the convolutional neural network, so as to calculate the difference between the true distribution probability of the bearing category and the predicted distribution probability of the bearing category, specifically including:

[0178] Based on a graph kernel loss function, and according to the kernel value, a similarity loss between the source domain graph and the target domain graph is calculated;

[0179] L=1-k θ ;

[0180] Among them, L represents the graph kernel loss function;

[0181] Customize the fault category items of bearings;

[0182] Based on the definition of the fault category item, obtaining the true distribution probability of the bearing category according to the source domain data and the target domain data;

[0183] Based on the definition of the fault category item, using the fully connected layer in the convolutional neural network, obtaining the bearing category prediction distribution probability in the process of iteratively updating the nodes in the source domain data and the target domain data;

[0184] Based on the cross entropy loss function, according to the true distribution probability of the bearing category and the predicted distribution probability of the bearing category, the difference between the true distribution probability of the bearing category and the predicted distribution probability of the bearing category is calculated;

[0185]

[0186] Among them, P(x a ) indicates that the bearing category in the true distribution is x a The probability of Q(x a ) indicates that the bearing category in the predicted distribution is x a The probability of, F(P, Q) represents the cross entropy loss function.

[0187] It can be understood that after calculating the source domain graph and the target domain graph, and calculating the kernel value, the inter-domain difference can be minimized through the graph kernel loss term, thereby improving the generalization performance of the model in the target domain. The graph and loss function can calculate the similarity loss between the source domain graph and the target domain graph. Then, the fault category item of the bearing is customized. The true distribution probability of the bearing category can be obtained from the source domain data. The graph embedding is classified using a fully connected network to obtain the fault category of the bearing. The fully connected network contains three layers, each of which uses a ReLU activation function to enhance the nonlinear expression ability. Finally, the Softmax activation function is connected to output the classification result as a probability distribution. In this step, the predicted distribution probability of the bearing category can be obtained in the process of iteratively updating the nodes in the source domain data and the target domain data, and the cross entropy loss function can be constructed. This function can measure the difference between the predicted distribution probability of the bearing category and the true distribution probability of the bearing category for the model.

[0188] Furthermore, in the step S50, the total loss function is calculated by combining the graph kernel loss function and the cross entropy loss function, and the total loss function is defined as follows:

[0189] T = L + F (P, Q);

[0190] Among them, T represents the total loss function.

[0191] It can be understood that the total loss function is defined by combining the cross entropy classification loss and the graph kernel loss function. The total loss function can optimize the graph kernel function to realize the difference in graph morphology between the source domain and the target domain, and minimize this difference.

[0192] Furthermore, the specific steps of step S60 include:

[0193] Calculating the gradient of the graph kernel function based on the total loss function;

[0194]

[0195] in, Indicates the symbol for partial derivative;

[0196] Based on the gradient of the graph kernel function and by using the chain rule, respectively calculating the gradients of the source domain graph and the target domain graph;

[0197]

[0198] Calculating the gradient of the node based on the gradients of the source domain graph and the target domain graph;

[0199]

[0200] Performing gradient calculation on weight parameters and bias parameters of the graph kernel model;

[0201]

[0202] in, Represents the node v after the last update of neighbor aggregation, represents the node i after the last update of neighbor aggregation, and R represents transpose.

[0203] It can be understood that after minimizing the graph differences through the total loss function, the gradient of the network parameters (such as kernel weights, biases, kernel function bandwidth, etc.) can be calculated, and the parameters can be updated to finally achieve cross-device diagnosis of bearing faults.

[0204] The present invention can use Python programming language and mainly relies on PyTorch deep learning framework to build and train the model. The implementation of CNN model includes five convolutional layers, each convolutional layer is followed by batch normalization layer and ReLU activation layer, and finally the global average pooling layer is used to generate features. Graph embedding is mapped through a custom graph embedding layer, and the node embedding update of parameterized WL graph kernel is realized. Domain adaptation is trained by defining an adaptive loss function, combining the kernel similarity between the source domain and the target domain, and using the back propagation process.

[0205] In summary, the cross-device bearing fault diagnosis method in the above embodiments of the present invention, 1. effectively extracts local and global features from the bearing vibration signal by combining a deep convolutional neural network with a parameterized graph kernel model, thereby significantly improving the accuracy of fault diagnosis;

[0206] 2. By introducing domain-adaptive graph kernel loss function and cross entropy loss function, the distribution difference between source domain and target domain is reduced, making the model's migration effect on different data sets better and improving the generalization ability of the model. The parameterized graph kernel function not only enhances the mathematical interpretability of the model by introducing trainable weights, biases and graph kernel function bandwidth, but also can effectively quantify the differences in graph morphology.

[0207] 3. High degree of automation. The diagnostic process does not need to rely on manual feature engineering, which reduces dependence on expert knowledge and improves diagnostic efficiency and accuracy.

[0208] Please refer to Figure 4 , shown is a cross-device bearing fault diagnosis system in a second embodiment of the present invention, comprising:

[0209] Acquisition module 11: used to acquire signal data of the bearing in a known environment and use it as source domain data, and acquire signal data of the bearing in other environments and use it as target domain data;

[0210] Denoising module 12; used for performing denoising processing on the source domain data and the target domain data;

[0211] The denoising module 12 is specifically used to: divide the source domain data into a source domain training set and a source domain test set, and divide the target domain data into a target domain training set and a target domain test set;

[0212] Dividing the source domain training set, the source domain test set, the target domain training set, and the target domain test set into data samples according to a preset data length, and performing standardization processing on the data samples;

[0213] Using an abnormal amplitude replacement method to correct the abnormal values ​​in the data sample after the standardization process;

[0214] Extraction module 13: used to use a convolutional neural network to perform feature extraction on the denoised source domain data and the target domain data to obtain an initial feature vector;

[0215] The extraction module 13 is specifically used for:

[0216] Introducing a convolutional neural network, wherein the convolutional neural network includes multiple convolutional layers, normalization layers, and global average pooling layers;

[0217] Extracting the time domain and frequency domain features of the denoised source domain data and the target domain data layer by layer through multiple convolutional layers;

[0218] Normalizing the denoised source domain data and the target domain data through the normalization layer;

[0219] Performing global average calculation on the internal channels of the input features in the convolutional neural network through the global average pooling layer to obtain an initial feature vector;

[0220] Graph calculation module 14: used for treating the initial feature vector as a node, introducing a graph kernel model to iteratively update the nodes in the source domain data and the target domain data respectively to obtain a source domain graph and a target domain graph, and calculating a kernel value between the source domain graph and the target domain graph through a graph kernel function;

[0221] The graph calculation module 14 is specifically used for:

[0222] The initial feature vector is regarded as a node, a graph core model is introduced, and label embedding initialization is performed on each node in the graph core model to obtain an initial embedding value;

[0223]

[0224] Among them, E I (v) represents the initial feature vector corresponding to the node v in the source domain data, represents the initial embedding value corresponding to the node v in the source domain data, E I (i) represents the initial feature vector corresponding to node i in the target domain data, represents the initial embedding value corresponding to node i in the target domain data;

[0225] Performing neighbor aggregation calculation on the initial embedding values ​​in the source domain data and the target domain data layer by layer through the graph kernel model;

[0226]

[0227] Among them, k represents the number of neighbor points in the source domain data, N represents the total number of nodes in the source domain data, represents the embedding value corresponding to the node u in the k-1th layer in the source domain data, represents the neighbor aggregation feature of node v in the kth layer in the source domain data, j represents the number of neighbor points in the target domain data, C represents the total number of nodes in the target domain data, represents the embedding value corresponding to the node f at the j-1th layer in the target domain data, Represents the neighbor aggregation features of the target domain data at the jth layer node i;

[0228] Based on the result of neighbor aggregation calculation, updating nodes initially embedded in the source domain data and the target domain data respectively;

[0229]

[0230] Among them, σ represents the activation function, W represents the weight, and b represents the bias. represents the updated node v in the kth layer of the source domain data, represents the updated node i in the jth layer in the target domain data;

[0231] By respectively aggregating updated nodes in the source domain data and the target domain data, an embedding representation of a graph is constructed to obtain a source domain graph and a target domain graph respectively;

[0232]

[0233] in, represents the updated node v, V in the last layer K of the source domain data G represents the total number of updated nodes in the source domain data, h G represents the source domain graph, represents the updated node i in the last layer J in the target domain data, I H represents the total number of updated nodes in the target domain data, h H represents the target domain graph;

[0234] Introducing a graph kernel function to calculate a kernel value between the source domain graph and the target domain graph;

[0235]

[0236] Among them, φ represents the bandwidth of the graph kernel function;

[0237] Loss calculation module 15: used to calculate the similarity loss between the source domain graph and the target domain graph based on the graph kernel loss function and according to the kernel value, obtain the true distribution probability of the bearing category according to the source domain data and the target domain data based on the cross entropy loss function, and obtain the predicted distribution probability of the bearing category in the process of iteratively updating the nodes in the source domain data and the target domain data through the fully connected layer in the convolutional neural network, so as to calculate the difference between the true distribution probability of the bearing category and the predicted distribution probability of the bearing category, and calculate the total loss function by combining the graph kernel loss function and the cross entropy loss function;

[0238] The loss calculation module 15 is specifically used for:

[0239] Based on a graph kernel loss function, and according to the kernel value, a similarity loss between the source domain graph and the target domain graph is calculated;

[0240] L=1-k θ ;

[0241] Among them, L represents the graph kernel loss function;

[0242] Customize the fault category items of bearings;

[0243] Based on the definition of the fault category item, obtaining the true distribution probability of the bearing category according to the source domain data and the target domain data;

[0244] Based on the definition of the fault category item, using the fully connected layer in the convolutional neural network, obtaining the bearing category prediction distribution probability in the process of iteratively updating the nodes in the source domain data and the target domain data;

[0245] Based on the cross entropy loss function, according to the true distribution probability of the bearing category and the predicted distribution probability of the bearing category, the difference between the true distribution probability of the bearing category and the predicted distribution probability of the bearing category is calculated;

[0246]

[0247] Among them, P(x a ) indicates that the bearing category in the true distribution is x a The probability of Q(x a ) indicates that the bearing category in the predicted distribution is x a The probability of , F(P, Q) represents the cross entropy loss function;

[0248] The total loss function is calculated by combining the graph kernel loss function and the cross entropy loss function. The definition of the total loss function is as follows:

[0249] T = L + F (P, Q);

[0250] Where T represents the total loss function;

[0251] Gradient calculation module 16: based on the total loss function, sequentially performing gradient calculations on the graph kernel function, the source domain graph, the target domain graph, the nodes, and the parameters of the graph kernel model;

[0252] The gradient calculation module 16 is specifically used for:

[0253] Calculating the gradient of the graph kernel function based on the total loss function;

[0254]

[0255] in, Indicates the symbol for partial derivative;

[0256] Based on the gradient of the graph kernel function and by using the chain rule, respectively calculating the gradients of the source domain graph and the target domain graph;

[0257]

[0258] Calculating the gradient of the node based on the gradients of the source domain graph and the target domain graph;

[0259]

[0260] Performing gradient calculation on weight parameters and bias parameters of the graph kernel model;

[0261]

[0262] in, Represents the node v after the last update of neighbor aggregation, represents the node i after the last update of neighbor aggregation, and R represents transposition;

[0263] Optimization module 17: used to use the cross entropy loss function to perform supervised learning on the classification results in the process of iteratively updating the nodes in the source domain data and the target domain data, and jointly optimize the graph core model in combination with the graph core loss function.

[0264] The present invention also provides an electronic device, see Figure 5 , shown is an electronic device in the third embodiment of the present invention, including a memory 10, a processor 20, and a computer program 30 stored in the memory 10 and executable on the processor 20, and when the processor 20 executes the computer program 30, the above-mentioned cross-device bearing fault diagnosis method is implemented.

[0265] The memory 10 includes at least one type of storage medium, including flash memory, hard disk, multimedia card, card-type memory (e.g., SD or DX memory, etc.), magnetic memory, magnetic disk, optical disk, etc. In some embodiments, the memory 10 may be an internal storage unit of an electronic device, such as a hard disk of the electronic device. In other embodiments, the memory 10 may also be an external storage device, such as a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card, etc. Further, the memory 10 may also include both an internal storage unit of the electronic device and an external storage device. The memory 10 may be used not only to store application software and various types of data installed in the electronic device, but also to temporarily store data that has been output or is to be output.

[0266] Among them, in some embodiments, the processor 20 can be an electronic control unit (Electronic Control Unit, abbreviated as ECU, also known as a vehicle computer), a central processing unit (Central Processing Unit, CPU), a controller, a microcontroller, a microprocessor or other data processing chip, used to run the program code stored in the memory 10 or process data, such as executing access restriction programs, etc.

[0267] It should be pointed out that Figure 5 The structure shown does not constitute a limitation on the electronic device. In other embodiments, the electronic device may include fewer or more components than those shown in the figure, or combine certain components, or arrange the components differently.

[0268] The embodiment of the present invention further provides a readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the cross-device bearing fault diagnosis method as described above is implemented.

[0269] Those skilled in the art will appreciate that the logic and / or steps represented in the flowchart or otherwise described herein, for example, may be considered as an ordered list of executable instructions for implementing logical functions, and may be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or in conjunction with such instruction execution systems, devices or apparatuses. For purposes of this specification, a "computer-readable medium" may be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in conjunction with such instruction execution systems, devices or apparatuses.

[0270] More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk case (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be a paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering or, if necessary, processing in another suitable manner, and then stored in a computer memory.

[0271] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, a plurality of steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or a combination thereof: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0272] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "examples", "specific examples", or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner.

[0273] The above-mentioned embodiments only express several implementation methods of the present invention, and the description thereof is relatively specific and detailed, but it cannot be understood as limiting the scope of the patent of the present invention. It should be pointed out that, for ordinary technicians in this field, several variations and improvements can be made without departing from the concept of the present invention, which all belong to the protection scope of the present invention. Therefore, the protection scope of the patent of the present invention shall be subject to the attached claims.

Claims

1. A cross-device bearing fault diagnosis method, characterized in that: The steps include: Acquire the signal data of the bearing in a known environment and use it as the source domain data, and acquire the signal data of the bearing in other environments and use it as the target domain data; Performing denoising processing on the source domain data and the target domain data; Using a convolutional neural network to perform feature extraction on the denoised source domain data and the target domain data to obtain an initial feature vector; The initial feature vector is regarded as a node, and a graph kernel model is introduced to iteratively update the nodes in the source domain data and the target domain data respectively to obtain a source domain graph and a target domain graph, and a kernel value between the source domain graph and the target domain graph is calculated by a graph kernel function; Based on the graph kernel loss function, the similarity loss between the source domain graph and the target domain graph is calculated according to the kernel value, based on the cross entropy loss function, the true distribution probability of the bearing category is obtained according to the source domain data and the target domain data, and the predicted distribution probability of the bearing category is obtained in the process of iteratively updating the nodes in the source domain data and the target domain data through the fully connected layer in the convolutional neural network, so as to calculate the difference between the true distribution probability of the bearing category and the predicted distribution probability of the bearing category, and the total loss function is calculated by combining the graph kernel loss function and the cross entropy loss function; Based on the total loss function, sequentially performing gradient calculations on the graph kernel function, the source domain graph, the target domain graph, the nodes, and the parameters of the graph kernel model; Using the cross entropy loss function, supervised learning is performed on the classification results during the iterative update of the nodes in the source domain data and the target domain data, and the graph core model is jointly optimized in combination with the graph core loss function.

2. The cross-device bearing fault diagnosis method according to claim 1 is characterized in that: The denoising process of the source domain data and the target domain data specifically includes: Dividing the source domain data into a source domain training set and a source domain test set, and dividing the target domain data into a target domain training set and a target domain test set; Dividing the source domain training set, the source domain test set, the target domain training set, and the target domain test set into data samples according to a preset data length, and performing standardization processing on the data samples; The abnormal amplitude replacement method is used to correct the abnormal values ​​in the data sample after the standardization process.

3. The cross-device bearing fault diagnosis method according to claim 1, characterized in that: The using a convolutional neural network to extract features from the denoised source domain data and the target domain data to obtain an initial feature vector specifically includes: Introducing a convolutional neural network, wherein the convolutional neural network includes multiple convolutional layers, normalization layers, and global average pooling layers; Extracting the time domain and frequency domain features of the denoised source domain data and the target domain data layer by layer through multiple convolutional layers; Normalizing the denoised source domain data and the target domain data through the normalization layer; The global average pooling layer is used to calculate the global average of the internal channels of the input features in the convolutional neural network to obtain an initial feature vector.

4. The cross-device bearing fault diagnosis method according to claim 1, characterized in that: The initial feature vector is regarded as a node, a graph kernel model is introduced to iteratively update the nodes in the source domain data and the target domain data respectively to obtain a source domain graph and a target domain graph, and a kernel value between the source domain graph and the target domain graph is calculated by a graph kernel function, specifically including: The initial feature vector is regarded as a node, a graph core model is introduced, and label embedding initialization is performed on each node in the graph core model to obtain an initial embedding value; Among them, E I (v) represents the initial feature vector corresponding to the node v in the source domain data, represents the initial embedding value corresponding to the node v in the source domain data, E I (i) represents the initial feature vector corresponding to node i in the target domain data, represents the initial embedding value corresponding to node i in the target domain data; Performing neighbor aggregation calculation on the initial embedding values ​​in the source domain data and the target domain data layer by layer through the graph kernel model; Among them, k represents the number of neighbor points in the source domain data, N represents the total number of nodes in the source domain data, represents the embedding value corresponding to the node u in the k-1th layer in the source domain data, represents the neighbor aggregation feature of node v in the kth layer in the source domain data, j represents the number of neighbor points in the target domain data, C represents the total number of nodes in the target domain data, represents the embedding value corresponding to the node f at the j-1th layer in the target domain data, Represents the neighbor aggregation features of the target domain data at the jth layer node i; Based on the result of neighbor aggregation calculation, updating nodes initially embedded in the source domain data and the target domain data respectively; Among them, σ represents the activation function, W represents the weight, and b represents the bias. represents the updated node v in the kth layer of the source domain data, represents the updated node i in the jth layer in the target domain data; By respectively aggregating updated nodes in the source domain data and the target domain data, an embedding representation of a graph is constructed to obtain a source domain graph and a target domain graph respectively; in, represents the updated node v, V in the last layer K of the source domain data G represents the total number of updated nodes in the source domain data, h G represents the source domain graph, represents the updated node i in the last layer J in the target domain data, I H represents the total number of updated nodes in the target domain data, h H represents the target domain graph; Introducing a graph kernel function to calculate a kernel value between the source domain graph and the target domain graph; Among them, φ represents the bandwidth of the graph kernel function.

5. The cross-device bearing fault diagnosis method according to claim 1, characterized in that: The method is based on the graph kernel loss function and calculates the similarity loss between the source domain graph and the target domain graph according to the kernel value, obtains the true distribution probability of the bearing category according to the source domain data and the target domain data based on the cross entropy loss function, and obtains the predicted distribution probability of the bearing category in the process of iteratively updating the nodes in the source domain data and the target domain data through the fully connected layer in the convolutional neural network, so as to calculate the difference between the true distribution probability of the bearing category and the predicted distribution probability of the bearing category, specifically including: Based on a graph kernel loss function, and according to the kernel value, a similarity loss between the source domain graph and the target domain graph is calculated; L=1-k θ ; Among them, L represents the graph kernel loss function; Customize the fault category items of bearings; Based on the definition of the fault category item, obtaining the true distribution probability of the bearing category according to the source domain data and the target domain data; Based on the definition of the fault category item, using the fully connected layer in the convolutional neural network, obtaining the predicted distribution probability of the bearing category in the process of iteratively updating the nodes in the source domain data and the target domain data; Based on the cross entropy loss function, according to the true distribution probability of the bearing category and the predicted distribution probability of the bearing category, the difference between the true distribution probability of the bearing category and the predicted distribution probability of the bearing category is calculated; Among them, P(x a ) indicates that the bearing category in the true distribution is x a The probability of Q(x a ) indicates that the bearing category in the predicted distribution is x a The probability of, F(P,Q) represents the cross entropy loss function.

6. The cross-device bearing fault diagnosis method according to claim 1, characterized in that: The total loss function is calculated by combining the graph kernel loss function and the cross entropy loss function. The total loss function is defined as follows: T = L + F (P, Q); Among them, T represents the total loss function.

7. The cross-device bearing fault diagnosis method according to claim 1, characterized in that: The step of sequentially performing gradient calculation on the graph kernel function, the source domain graph, the target domain graph, the nodes, and the parameters of the graph kernel model based on the total loss function specifically includes: Calculating the gradient of the graph kernel function based on the total loss function; in, Indicates the symbol for partial derivative; Based on the gradient of the graph kernel function and by using the chain rule, respectively calculating the gradients of the source domain graph and the target domain graph; Calculating the gradient of the node based on the gradients of the source domain graph and the target domain graph; Performing gradient calculation on weight parameters and bias parameters of the graph kernel model; in, Represents the node v after the last update of neighbor aggregation, represents the node i after the last update of neighbor aggregation, and R represents transpose.

8. A cross-device bearing fault diagnosis system, characterized in that: include: Acquisition module: used to acquire signal data of the bearing in a known environment and use it as source domain data, and acquire signal data of the bearing in other environments and use it as target domain data; A denoising module; used for performing denoising processing on the source domain data and the target domain data; Extraction module: used for performing feature extraction on the denoised source domain data and the target domain data using a convolutional neural network to obtain an initial feature vector; Graph calculation module: used for treating the initial feature vector as a node, introducing a graph kernel model to iteratively update the nodes in the source domain data and the target domain data respectively, so as to obtain a source domain graph and a target domain graph, and calculating a kernel value between the source domain graph and the target domain graph through a graph kernel function; Loss calculation module: used to calculate the similarity loss between the source domain graph and the target domain graph based on the graph kernel loss function and the kernel value, obtain the true distribution probability of the bearing category based on the source domain data and the target domain data based on the cross entropy loss function, and obtain the predicted distribution probability of the bearing category in the process of iteratively updating the nodes in the source domain data and the target domain data through the fully connected layer in the convolutional neural network, so as to calculate the difference between the true distribution probability of the bearing category and the predicted distribution probability of the bearing category, and calculate the total loss function by combining the graph kernel loss function and the cross entropy loss function; Gradient calculation module: based on the total loss function, sequentially performing gradient calculations on the graph kernel function, the source domain graph, the target domain graph, the nodes, and the parameters of the graph kernel model; Optimization module: used to use the cross entropy loss function to perform supervised learning on the classification results during the iterative update of the nodes in the source domain data and the target domain data, and to jointly optimize the graph core model in combination with the graph core loss function.

9. A storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the cross-device bearing fault diagnosis method as described in any one of claims 1 to 7 is implemented.

10. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the cross-device bearing fault diagnosis method according to any one of claims 1 to 7 is implemented.

Images