Motor fault diagnosis method based on graph isomorphic network and cross-graph relation learning

Abstract

The invention discloses a motor fault diagnosis method based on a graph isomorphic network and cross-graph relation learning, and the method comprises the following steps: firstly, obtaining multi-source signals of a motor in different health states, carrying out the normalization, and building an original signal graph sample set; secondly, utilizing graph samples of a motor in different health states to train the model, capturing relation characteristics among cross-graph samples through a multi-head self-attention mechanism and a feedforward network, and combining a knowledge distillation technology to obtain an optimized cascade structure of a multilayer graph isomorphic network; and finally, testing a to-be-diagnosed sample by using the trained model, and finally outputting a diagnosis result. According to the method, the noise immunity is enhanced through fusion of multi-sensor signals, cross-graph relation feature extraction and a new dynamic training feedback strategy, and the precision and robustness of motor fault diagnosis are remarkably improved. Experimental results show that the method provided by the invention is superior to a current mainstream fault diagnosis method in the aspects of fault diagnosis accuracy and robustness.

Description

Technical Field

[0001] The present invention belongs to the field of motor fault diagnosis, and specifically is a novel motor fault diagnosis method based on graph isomorphism network and cross-graph relationship learning. Background Art

[0002] Motors have been widely used in industrial drives, electric vehicles, wind power generation and other fields due to their advantages such as high efficiency, high power density and excellent control performance. However, various faults will inevitably occur in the operation of motors, such as bearing faults, stator winding faults and rotor imbalance. If these faults are not diagnosed and handled in time, the performance of the motor will be degraded and even cause serious safety accidents. Therefore, the development of efficient and accurate motor fault diagnosis methods is of great significance to ensure the safe and reliable operation of motors. In recent years, motor fault diagnosis methods based on deep learning have developed rapidly, especially the progress of algorithms such as graph neural networks, which has greatly improved the data analysis capabilities of the model.

[0003] As an important branch of graph neural networks, graph isomorphism networks (GINs) have shown significant advantages in the field of fault diagnosis due to their powerful feature extraction capabilities. GIN aggregates node features by combining multi-layer perceptrons and sum functions, and generates a global representation of the graph through a graph readout function. This design enables it to perform well in tasks such as fault diagnosis that require accurate identification of graph structures. However, most existing research methods are based on the analysis of single sensor signals, which makes it difficult to fully utilize the complementary information of multi-sensor data, and signal acquisition is easily interfered by noise in actual industrial environments. At the same time, existing fault diagnosis methods based on graph isomorphism networks usually focus on a single sample, ignore the association between graph samples, and cannot effectively capture the similarities or differences between samples of different fault types. This means that the diagnostic accuracy in motor fault diagnosis will decrease. Summary of the invention

[0004] The purpose of the present invention is to provide a motor fault diagnosis method based on graph isomorphic network and cross-graph relationship learning in view of the above problems. The method includes obtaining and normalizing the multi-sensor signals of the motor; obtaining the original graph representation of the multi-source signal graph samples through the cascade structure of the multi-layer graph isomorphic network, making full use of the complementary information of the multi-source data; combining the multi-head self-attention mechanism and the feedforward network to learn the cross-graph sample relationship characteristics, learn the similarities and differences between samples of different fault types, and significantly improve the anti-noise ability of the model; introduce knowledge distillation and other technologies to design a new dynamic feedback training strategy, transmit the relationship characteristics of the cross-graph samples to the cascade structure of the multi-layer graph isomorphic network to establish a dynamic feedback loop, so that the model can better understand the relationship characteristics between the cross-graph samples; take the graph samples of various fault states and the corresponding motor fault labels as input, obtain the output of the model and use it as the fault diagnosis result of the motor. The present invention simultaneously considers the characteristics of multiple sensors in the same fault state and the relationship characteristics between graph samples of different fault categories, directly outputs the fault diagnosis result of the motor through an end-to-end training method, solves the limitations of the existing methods in complex industrial environments, significantly improves the diagnostic accuracy and robustness, and provides reliable support for motor fault diagnosis.

[0005] To achieve the above object, the technical solution adopted by the present invention is: a motor fault diagnosis method based on graph isomorphism network and cross-graph relationship learning, comprising the following steps:

[0006] Step 1: Obtain multi-source signals under different health conditions and establish a sample set of original signals;

[0007] Step 2: Data preprocessing: After normalizing the collected multi-source signals, the sensors are used as nodes, the cosine similarities between different sensors are calculated as edges, and graph samples of different fault categories are constructed;

[0008] Step 3: In the training phase, the multi-layer graph isomorphic network model is trained using training samples from graph samples of different fault categories, and the parameters are continuously optimized;

[0009] Step 4: In the test phase, the test samples in the different fault category graph samples are input into the trained multi-layer graph isomorphic network model, the probability distribution of the sample faults is output, and the sample label with the fault category with the highest probability as the diagnosis is obtained;

[0010] Step 5: Output the diagnosis results.

[0011] Furthermore, in step 2, the method for normalizing the collected multi-source signals is Min-Max normalization to obtain motor fault signal data.

[0012] Furthermore, in step 2, the steps of constructing graph samples of different fault categories are as follows: In a graph sample, each sensor is mapped to a node in the graph. The initial features of the node are composed of the time domain signal data collected by the sensor in a fixed period of time. Assuming that there are N sensors deployed in the motor, and each sensor collects M data points in time period T, the feature vector of the i-th sensor can be expressed as:

[0013]

[0014] in, is the sampling value of the i-th sensor at the k-th time point;

[0015] In order to characterize the correlation between sensors, cosine similarity is used as the metric of edge weight; for any two sensors i and j, their edge weight The calculation formula is as follows:

[0016]

[0017] Among them, the element represents the edge weight between sensors i and j. The numerator on the right side of the equation represents the dot product of the two feature vectors, and the denominator is the product of the modulus lengths of the two vectors.

[0018] The value range of cosine similarity is [−1,1]. The closer the value is to 1, the more consistent the signal change trends of the two sensors are; the closer the value is to -1, the opposite trend is; and the value of 0 indicates that there is no significant linear correlation between the two.

[0019] Based on the above definitions of nodes and edges, construct the graph structure:

[0020] Among them, V is the node set, representing all sensors, E is the edge set, representing the connection relationship between sensors, X is the node feature matrix, each row corresponds to the feature vector of a sensor, and W is the edge weight matrix;

[0021] In this way, multi-source signals are converted into graph-structured data, which can effectively capture the correlation between sensors and provide input for subsequent graph isomorphic networks.

[0022] Furthermore, in step 3, in the training phase, the multi-layer graph isomorphic network model is trained using training samples from graph samples of different fault categories, and the specific process includes the following steps:

[0023] The graph sample G is input into the cascade structure of the multi-layer graph isomorphism network to extract the complex relationship between sensors. For the graph sample G, the first The layer node feature update formula is as follows:

[0024]

[0025] in, Representation Node In the Representation after layer, initial state is the raw data point sequence feature of the sensor, is a learnable or fixed scalar, is a multilayer perceptron used to learn nonlinear features from data from sensors and their neighborhood. Representation Node In the The layer representation, For Node The set of neighbor nodes;

[0026] The cascade structure of multi-layer graph isomorphic networks can capture more extensive sensor-related information layer by layer. Each layer of GIN aggregates features from direct neighborhood sensors and gradually expands to more distant neighborhoods, thereby constructing a deeper sensor representation. This structure enables the model to learn rich spatiotemporal feature relationships from sensor networks.

[0027] The original feature map representation of the image sample is obtained by the following formula:

[0028]

[0029] in, It is a simple summation operation, which ensures that the graph representation is invariant to the sensor order, i.e., permutation, and can effectively capture the global structural characteristics of the entire sensor network. Compared with other GNN architectures, the design of GIN enhances the robustness and expressiveness of the model for sensor data by using a simple summation aggregation operation instead of complex normalization or pooling operations, and obtains the final generated graph representation through the cascade structure of a multi-layer graph isomorphic network. It can better capture the global information of the entire sensor network;

[0030] Get the graph representation of each sample Finally, the relational feature extractor composed of a multi-head self-attention mechanism and a feedforward network is used to extract the relational features between samples across graphs and learn the similarities or differences between different samples.

[0031] The multi-head self-attention mechanism is the key to the relational feature extractor. Its core idea is to focus on different relational features through multiple attention heads to calculate the relationship between graphs and graph samples, as follows:

[0032]

[0033] in, To calculate the formula of the attention mechanism, is a normalization function used to generate the attention output, Q, K, V represent the query, key, and value matrices extracted from the graph representation, respectively, and are derived from the input token T to the d-dimensional space through linear projection. is the dimension of the key vector;

[0034] By calculating the dot product of the query and the key, and then performing softmax normalization on the result and multiplying it by the value matrix, a weighted representation is obtained, which highlights the most relevant relationships. This mechanism can assign higher weights to important features and suppress the impact of noise;

[0035] The unique relational features captured by each attention head in the multi-head self-attention mechanism are concatenated together and linearly transformed into a comprehensive relational feature encoding:

[0036]

[0037] in, is the formula for multi-head self-attention, which represents the combination of multiple attention heads. Indicates concatenating the outputs of all attention heads. , , are the outputs from the 1st, 2nd, and hth attention heads respectively, is the output projection matrix;

[0038] The relationship feature representation between graph samples at the lth relationship layer can be expressed as:

[0039]

[0040] in represents the refined relational feature representation, The proposed relational feature extractor is used to extract the relational features between different graph samples. Sample diagram In all sample sets The original graph representation in The union of;

[0041] Furthermore, in step 3, the parameters are continuously optimized, and the specific process includes the following steps:

[0042] A new dynamic feedback training strategy is designed as a key component for continuously optimizing model parameters. In order to minimize the difference between the original graph representation space and the relationship feature space, two classifiers are used to train the model. and Construct a feedback loop mechanism. The relational features from the relational feature extractor are extracted and transmitted to the cascade structure of the multi-layer graph isomorphism network to establish a dynamic feedback loop. The loss function is as follows:

[0043]

[0044] in, Is from and Cross entropy loss for labeled graph samples, is the coefficient of the cross entropy loss function, is the coefficient of the divergence loss function;

[0045] is a Kullback-Leibler divergence loss used to measure the original graph representation and relation-aware graph representation The difference between the predicted probability distributions is given by:

[0046]

[0047] Among them, among them, is the Kullback-Leibler divergence, , yes The predicted probability distribution, , yes The predicted probability distribution, is the temperature parameter of distillation;

[0048] is the feature similarity loss function, which is represented by reducing the original graph and relation-aware representation The L2 distance between them makes the two embeddings more similar, guiding the cascade structure of the multi-layer graph isomorphism network to understand the feature representation of the relational feature extractor. The formula is as follows:

[0049]

[0050] By designing a loss function of a dynamic feedback loop combined with relational feature extraction to continuously optimize parameters of the cascade structure of the multi-layer graph isomorphic network, the ability of the method proposed in the present invention to understand and utilize the relationship between samples of different fault category graphs can be significantly enhanced.

[0051] The present invention proposes a motor fault diagnosis method based on graph isomorphism network and cross-graph relationship learning. The cascade structure of a multi-layer graph isomorphism network is used to extract the original graph feature representation from the graph sample composed of multiple source signals of the motor, and the relationship features between samples of different fault categories are learned through the relationship feature extractor. After multiple iterations of the new training strategy, motor fault diagnosis is achieved and better diagnosis results are obtained. The present invention provides a new idea for motor fault diagnosis. This method emphasizes learning the relationship features of different fault categories from graph samples composed of multiple time domain signals collected by different sensors, using a multi-head self-attention mechanism and a feedforward network to capture the complex relationship between input features, and realizing timely adjustment through a dynamic feedback training strategy, which can more effectively handle noise. In fault diagnosis applications, the present invention shows stronger robustness to noise interference, significantly improves the recognition accuracy of different fault categories in a noisy environment, and is therefore more suitable for fault diagnosis scenarios under strong noise conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] Figure 1 is a flow chart of the motor fault diagnosis method of the present invention;

[0053] Figure 2 It is a radar chart showing the accuracy results of ten experiments of different fault diagnosis methods after adding -10db noise to the original signal;

[0054] Figure 3 It is a radar chart of the accuracy results of ten experiments of different fault diagnosis methods after adding -5db noise to the original signal;

[0055] Figure 4 It is a radar chart of the accuracy results of ten experiments of different fault diagnosis methods after adding 0db noise to the original signal;

[0056] Figure 5 It is a radar chart showing the accuracy results of ten experiments of different fault diagnosis methods after adding 5db noise to the original signal;

[0057] Figure 6 It is a radar chart of the accuracy results of ten experiments of different fault diagnosis methods after adding 10db noise to the original signal. DETAILED DESCRIPTION

[0058] In order for the present invention to construct graph samples from multi-source signals, graph samples containing different health states are used as inputs of the model, and an optimized multi-layer graph isomorphic network cascade structure is obtained after training the model. The purpose of using the trained multi-layer graph isomorphic network model to diagnose the samples to be tested is achieved. The present invention adopts the following embodiments.

[0059] like Figure 1As shown, the steps of a motor fault diagnosis method based on graph isomorphism network and cross-graph relationship learning are as follows:

[0060] Step 1: Obtain multi-source signals of the motor under different health conditions and establish a sample set of original signals;

[0061] Step 2: Data preprocessing: After normalizing the collected multi-source signals, the sensors are used as nodes, the cosine similarities between different sensors are calculated as edges, and graph samples of different fault categories are constructed;

[0062] For the M data points collected by N sensors deployed in the motor within a period T, the feature vector of the i-th sensor can be expressed as:

[0063]

[0064] in, is the sampling value of the i-th sensor at the k-th time point;

[0065] In a preferred embodiment, seven sensors are used, and the raw data from each sensor in each sample includes 1024 data points;

[0066] In order to characterize the correlation between sensors, the present invention adopts cosine similarity as the metric of edge weight; for any two sensors i and j, their edge weight The calculation formula is as follows:

[0067]

[0068] Among them, the element represents the edge weight between sensors i and j. The numerator on the right side of the equation represents the dot product of the two feature vectors, and the denominator is the product of the modulus lengths of the two vectors.

[0069] Based on the above definitions of nodes and edges, construct the graph structure:

[0070]

[0071] Among them, V is the node set, representing all sensors, E is the edge set, representing the connection relationship between sensors, X is the node feature matrix, each row corresponds to the feature vector of a sensor, and W is the edge weight matrix;

[0072] In this way, multi-source signals are converted into graph structure data, providing input for the subsequent cascade structure of multi-layer graph isomorphic networks;

[0073] Step 3: In the training phase, the multi-layer graph isomorphic network model is trained using training samples from graph samples of different fault categories, and the parameters are continuously optimized;

[0074] The graph sample G is input into the cascade structure of the multi-layer graph isomorphism network to extract the complex relationship between sensors. The process of graph isomorphism network feature extraction is as follows:

[0075] For the graph sample G, the graph isomorphic network The layer node feature update formula is as follows:

[0076]

[0077] in, Representation Node In the Representation after layer, initial state is the raw data point sequence feature of the sensor, is a learnable or fixed scalar, is a multilayer perceptron used to learn nonlinear features from data from sensors and their neighborhood. Representation Node In the The layer representation, For Node The set of neighbor nodes;

[0078] The original feature map representation of the image sample is obtained by the following formula:

[0079]

[0080] Get the graph representation of each sample Finally, the relational features between different samples are learned through a relational feature extractor consisting of a multi-head self-attention mechanism and a feedforward network:

[0081]

[0082] in, To calculate the formula of the attention mechanism, is a normalization function used to generate the attention output, Q, K, V represent the query, key, and value matrices extracted from the graph representation, respectively, and are derived from the input token T to the d-dimensional space through linear projection. is the dimension of the key vector, which represents the structure and features of the encoded graph;

[0083] The relational features captured by each head in the multi-head self-attention mechanism are concatenated together and first converted into a comprehensive relational feature encoding:

[0084]

[0085] in, is the formula for multi-head self-attention, which represents the combination of multiple attention heads. Indicates concatenating the outputs of all attention heads. , , are the outputs from the 1st, 2nd, and hth attention heads respectively, is the output projection matrix;

[0086] The relationship feature representation between graph samples at the lth relationship layer can be expressed as:

[0087]

[0088] in, represents the refined relational feature representation, The proposed relational feature extractor is used to extract the relational features between different graph samples. Sample diagram In all sample sets The original graph representation in The union of;

[0089] Through two classifiers and Construct a feedback loop mechanism. The relational features from the relational feature extractor are extracted and transmitted to the cascade structure of the multi-layer graph isomorphism network to establish a dynamic feedback loop. The loss function is as follows:

[0090]

[0091] in, Is from and Cross entropy loss for labeled graph samples, is the coefficient of the cross entropy loss function, is the coefficient of the divergence loss function;

[0092] is a Kullback-Leibler divergence loss used to measure the original graph representation and contains the relational feature graph representation The difference between the predicted probability distributions is given by:

[0093]

[0094] Among them, among them, is the Kullback-Leibler divergence, , yes The predicted probability distribution, , yes The predicted probability distribution, is the temperature parameter of distillation;

[0095] is the feature similarity loss function, which is represented by reducing the original graph and relation-aware representation The L2 distance between them makes the two embeddings more similar, guiding the cascade structure of the multi-layer graph isomorphism network to understand the feature representation of the relational feature extractor. The formula is as follows:

[0096]

[0097] The loss function of the designed dynamic feedback loop combined with relational feature extraction is used to continuously optimize the parameters of the cascade structure of the multi-layer graph isomorphism network.

[0098] Step 4: In the test phase, the test samples in the different fault category graph samples are input into the trained multi-layer graph isomorphic network model, the probability distribution of the sample faults is output, and the fault category with the highest probability is obtained as the predicted sample label;

[0099] Step 5: Output the diagnosis results.

[0100] The present invention is verified by the following experiments.

[0101] In order to verify the effectiveness of the method proposed in the present invention, the working conditions and health status of the test samples were pre-set during the experiment, and five different conditions of the motor under test were established for analysis, including the normal state of the motor, three bearing fault conditions (inner ring fault, outer ring fault, outer ring combined fault) and magnet crack fault condition.

[0102] In order to obtain multi-source signals of the motor under test, a total of 7 signals were collected, including 3 vibration signals, 1 sound pressure signal and 3 current signals. The three-axis vibration accelerometer (x, y, z) was installed on the motor housing, and the microphone was placed 30 cm away from the end cover of the test motor. In the experiment, the speed of the motor under test was 2800RPM and the sampling frequency was 48000Hz. In the following analysis, the samples under different conditions included one sound pressure signal, three acceleration vibration signals and three current signals, and each signal of each sample had 1024 continuous data points.

[0103] In order to verify the effectiveness of the proposed method in motor fault diagnosis, the graph neural network models GCN, GIN, ChebyNet, GraphSage, GAT, HoGCN, and SGCN are used for comparative analysis. In the comparative analysis of this experiment, in order to evaluate the noise resistance of MsEGIN, Gaussian white noise with different signal-to-noise ratios (SNR) is added to the original signal, and the SNRs are 10db, 5db, 0db, −5db, and −10dB respectively. Each state (normal state, bearing inner race fault, outer race fault, magnet fault, and bearing inner and outer race combined failure) includes 200 samples, of which 120 samples are used for training and 80 are used for testing, which means that there are 400 test samples in total. The experiment is repeated ten times, and the final average results of these ten times are reported.

[0104] In the application of motor fault diagnosis, the diagnostic accuracy (Acc), F1 score (F1-score), precision (Precision), and recall (Recall) are of great concern. Therefore, these four indicators are used to quantitatively compare the diagnostic effects of different methods.

[0105] Table 1 compares the evaluation indicators of different methods. method Avg-Acc F1-score Precision Recall GCN 97.00 97.65 97.79 97.65 GIN 98.4 98.39 98.46 98.40 ChebyNet 97.80 97.80 97.93 97.80 GraphSage 91.10 89.49 90.49 91.10 GAT 97.85 97.85 98.02 97.85 HoGCN 86.80 84.76 89.96 86.80 SGCN 81.40 77.15 80.91 81.40 Method of the present invention 99.58 99.60 99.64 99.60 Table 1

[0106] The experiment used multiple graph neural network models for comparative analysis with the method proposed in the present invention. From the results in Table 1, it can be seen that the method proposed in the present invention exhibits the best performance. Compared with the second best method GIN in the table, it achieves an average accuracy improvement of 1.18%, an F1 score improvement of 1.21%, a precision improvement of 1.18%, and a recall improvement of 1.2%. The method of the present invention performs well in multiple key indicators and achieves a significant improvement in average accuracy, reflecting the model's good classification ability for different fault categories. The outstanding performance in precision and recall shows the superiority of the method of the present invention in reducing false positives and missed faults, and can more effectively capture faults of different categories than other models.

[0107] In order to evaluate the anti-noise ability of the proposed method, Gaussian white noise with different signal-to-noise ratios (SNR) was added to the original signal to test the robustness of the proposed method. The average accuracy results of ten experiments of different methods under different noise conditions are shown in Table 2.

[0108] Table 2 shows the average accuracy (%) of ten experiments for different methods at different signal-to-noise ratios: method -10db -5db 0db 5db 10db GCN 37.25 47.45 62.05 72.80 86.70 GIN 50.35 59.15 70.58 83.75 94.95 ChebyNet 36.9 47.7 62.15 71.55 86.25 GraphSage 39.2 48.2 65.4 72.1 91.85 GAT 36.65 46.75 64.85 72.45 85.55 HOGCN 40.6 48.6 65.3 72.45 88.50 SGCN 37.75 46.85 68.9 74.85 86.10 Method of the present invention 79.30 81.65 90.30 91.80 96.35 Table 2

[0109] As shown in Table 2, as the signal-to-noise ratio decreases, the performance of each model decreases. Due to different robustness to noise, the accuracy drop rates of these methods are also different. Compared with other methods, even at -10dB SNR, the method of the present invention can still obtain the best diagnosis results, which proves that the noise resistance of the method of the present invention is better than that of the comparison method. The radar chart of the accuracy of ten experiments of different methods under different noise conditions is shown in Figure 2-Figure 6 .from Figure 2-Figure 6 It can be seen that the accuracy of the method of the present invention is better than other GNN models at all noise levels most of the time. Under extreme noise conditions (SNR=-10), the accuracy of other methods drops sharply, and the performance of the method of the present invention is relatively stable and has better noise resistance. This is all due to the multi-head self-attention mechanism in the relational feature extractor, which dynamically allocates attention weights according to the importance of the input features, which means that the method of the present invention can pay more attention to important features and ignore unimportant noise features. Therefore, the method of the present invention shows stronger robustness when facing noisy data. The experimental results show that compared with other methods, the method proposed in the present invention can accurately identify motor faults under different noise levels, which fully demonstrates that the proposed method has better stability and noise resistance.

Claims

1. A motor fault diagnosis method based on graph isomorphism network and cross-graph relationship learning, characterized in that: The following steps are involved: Step 1: Obtain multi-source signals of the motor in different health states and establish an original signal sample set; Step 2: Data preprocessing: After normalizing the collected multi-source signals, the sensors are used as nodes, the cosine similarities between different sensors are calculated as edges, and graph samples of different fault categories are constructed; Step 3: In the training phase, the multi-layer graph isomorphic network model is trained using training samples from graph samples of different fault categories, and the parameters are continuously optimized; Step 4: In the test phase, the test samples in the different fault category graph samples are input into the trained multi-layer graph isomorphic network model, the probability distribution of the sample faults is output, and the fault category with the highest probability is obtained as the predicted sample label; Step 5: Output the diagnosis results.

2. A motor fault diagnosis method based on graph isomorphism network and cross-graph relationship learning according to claim 1, characterized in that: In the step 2, the method for normalizing the collected multi-source signals is Min-Max normalization to obtain motor fault signal data.

3. A motor fault diagnosis method based on graph isomorphic network and cross-graph relationship learning according to claim 1, characterized in that: The steps to construct graph samples for different fault categories are as follows: In a graph sample, each sensor is mapped to a node in the graph. The initial features of the node are composed of the time domain signal data collected by the sensor in a fixed period of time. Assuming that there are N sensors deployed in the motor, and each sensor collects M data points in time period T, the feature vector of the i-th sensor can be expressed as: ； in, is the sampling value of the i-th sensor at the k-th time point; In order to characterize the correlation between sensors, cosine similarity is used as the metric of edge weight; For any two sensors i and j, their edge weight The calculation formula is as follows: ； Among them, the element represents the edge weight between sensors i and j. The numerator on the right side of the equation represents the dot product of the two feature vectors, and the denominator is the product of the modulus lengths of the two vectors. Based on the above definitions of nodes and edges, graph samples of different fault categories are constructed: ； Among them, V is the node set, representing all sensors, E is the edge set, representing the connection relationship between sensors, X is the node feature matrix, each row corresponds to the feature vector of a sensor, and W is the edge weight matrix.

4. A motor fault diagnosis method based on graph isomorphism network and cross-graph relationship learning according to claim 1, characterized in that: In step 3, in the training phase, the multi-layer graph isomorphic network model is trained using training samples from graph samples of different fault categories. The specific process includes the following steps: The graph sample G is input into the cascade structure of the multi-layer graph isomorphism network to extract the complex relationship between sensors. For the graph sample G, the first The layer node feature update formula is as follows: ； in, Representation Node In the Representation after layer, initial state is the raw data point sequence feature of the sensor, is a learnable or fixed scalar, is a multilayer perceptron that learns nonlinear features from data from sensors and their neighborhood. Representation Node In the The layer representation, For Node The set of neighbor nodes of The original feature map representation of the image sample is obtained by the following formula: ； in, is a simple sum operation; Get the graph representation of each sample Finally, the relational features between different samples are learned through a relational feature extractor consisting of a multi-head self-attention mechanism and a feedforward network: ； in, To calculate the formula of the attention mechanism, is a normalization function used to generate the attention output, Q, K, V represent the query, key, and value matrices extracted from the graph representation, respectively, derived from the input token T to the d-dimensional space through linear projection, is the dimension of the key vector; The relational features captured by each head in the multi-head self-attention mechanism are concatenated together and first converted into a comprehensive relational feature encoding: ； in, is the formula for multi-head self-attention, which represents the combination of multiple attention heads. Indicates concatenating the outputs of all attention heads. , , are the outputs from the 1st, 2nd, and hth attention heads respectively, is the output projection matrix; The relationship feature representation between graph samples at the lth relationship layer is expressed as: ； in, represents the refined relational feature representation, The proposed relational feature extractor is used to extract the relational features between different graph samples. Sample diagram In all sample sets The original graph representation in The union of .

5. The motor fault diagnosis method based on graph isomorphism network and cross-graph relationship learning according to claim 1 is characterized in that: In step 3, the parameters are continuously optimized, and the specific process includes the following steps: Through two classifiers and Construct a feedback loop mechanism. The relational features from the relational feature extractor are extracted and transmitted to the cascade structure of the multi-layer graph isomorphism network to establish a dynamic feedback loop. The loss function is as follows: ； in, Is from and Cross entropy loss for labeled graph samples, is the coefficient of the cross entropy loss function, is the coefficient of the divergence loss function; is a Kullback-Leibler divergence loss used to measure the original graph representation and contains the relational feature graph representation The difference between the predicted probability distributions is given by: ； in, is the Kullback-Leibler divergence, , yes The predicted probability distribution, , yes The predicted probability distribution, is the temperature parameter of distillation; is the feature similarity loss function, which is represented by reducing the original graph and relation-aware representation The L2 distance between them makes the two embeddings more similar, guiding the cascade structure of the multi-layer graph isomorphism network to understand the feature representation of the relational feature extractor. The formula is as follows: ； The loss function of the designed dynamic feedback loop combined with relational feature extraction is used to continuously optimize the parameters of the cascade structure of the multi-layer graph isomorphism network.

Images