Rolling bearing fault diagnosis method and device, electronic equipment and storage medium

Abstract

The invention discloses a rolling bearing fault diagnosis method and device, electronic equipment and a storage medium. The rolling bearing fault diagnosis method mainly comprises the following steps of obtaining a sample data set; performing data preprocessing on the sample data set, and dividing the preprocessed sample data set into a training data set and a test data set; constructing a fault diagnosis model, and training the fault diagnosis model by adopting the training data set; inputting the test data set into the trained fault diagnosis model to obtain a fault diagnosis classification result; through combination of the convolutional neural network and the bidirectional long-short-term memory network, adaptive feature extraction and rolling bearing fault diagnosis under high-noise and complex working conditions are realized.

Description

Technical Field

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

[0002] Rotating machinery is widely used in industrial equipment such as aircraft, high-speed railways, and ships. Due to its frequent rotation and unstable load, key components such as bearings are prone to various faults. Therefore, in order to ensure the safe operation of rotating machinery, fault diagnosis and identification are indispensable technologies in the application. The rapid development of the field of artificial intelligence has greatly promoted the rolling bearing fault diagnosis technology. At present, the research on rolling bearing fault diagnosis technology mainly focuses on two directions: machine learning and deep learning. In the field of rolling bearing fault diagnosis driven by machine learning, Zhou Fuqiang (Zhou Fuqiang. Research on the application of bearing fault diagnosis based on improved XGBoost in big data environment [D]. Shenyang University, 2021.) used local mean decomposition (LMD) technology for signal denoising and proposed a dimensionality reduction technology based on kernel principal component analysis (KPCA). Bai Guangchen (Bai Guangchen. Research on the application of big data in rolling bearing fault diagnosis [D]. Beijing University of Posts and Telecommunications, 2017.) improved the accuracy of neural network fault diagnosis by optimizing and adjusting the back propagation (BP) neural network. Yao Na (Yao Na. Wind turbine rolling bearing fault diagnosis based on machine learning [D]. Northeast Electric Power University, 2023.) proposed an innovative data enhancement strategy and combined it with the K-means clustering algorithm. Su Yanchen et al. (Su Yanchen, Li Jiguang, Zhou Bo, et al. Research on rolling bearing fault diagnosis based on gated cyclic residual network [J]. Railway Locomotive and Rolling Stock, 2023, 43(3): 57-63.) proposed a gated recurrent unit (GRU) network architecture combined with a residual network to reduce the loss of time series information and alleviate the performance degradation problem caused by the deep network structure. In the field of rolling bearing fault diagnosis driven by deep learning, Zhu Jiyang (Zhu Jiyang. Research on rolling bearing fault diagnosis method based on one-dimensional convolution transfer learning [D]. North University of China, 2023.) proposed a research on rolling bearing fault diagnosis method based on one-dimensional convolution transfer learning. Zhang Tianyuan (Zhang Tianyuan. Research on rolling bearing fault diagnosis and RUL prediction methods based on deep learning [D]. North University of China, 2023.) used S transform to preprocess the bearing vibration signal and constructed the SAM-MSCDN feature extraction model. Overall, although the existing rolling bearing fault diagnosis methods can monitor the rolling bearing status and identify faults to a certain extent, they still have some shortcomings, such as difficulty in feature extraction, low robustness, and limited model generalization ability. Summary of the invention

[0003] In order to solve the deficiencies in the prior art, the present invention provides a rolling bearing fault diagnosis method, device, electronic device and storage medium. The fault diagnosis model constructed by the present invention comprises two branches, a feature fusion module for weighted fusion of features processed by the two branches, and a classification layer, wherein the first branch of the two branches is used to extract key time features, and comprises a one-dimensional convolutional neural network and a bidirectional long short-term memory network, wherein the one-dimensional convolutional neural network comprises a one-dimensional (1D) convolutional layer and a one-dimensional maximum pooling layer; the second branch is used to extract key spatial features, and comprises a two-dimensional (2D) convolutional neural network, and the two-dimensional convolutional neural network comprises two two-dimensional convolutional layers and two two-dimensional maximum pooling layers. Through the fault diagnosis model, the fault features of the rolling bearing under complex working conditions can be automatically extracted and identified, thereby realizing adaptive fault feature extraction under multi-dimensional signals and rolling bearing fault diagnosis under high noise and complex working conditions.

[0004] In order to achieve the above object, the specific scheme adopted by the present invention is: On the one hand, the present invention discloses a rolling bearing fault diagnosis method, which mainly comprises the following steps: Step S1, obtaining vibration signals of rolling bearings under multiple working conditions through public data sets, labeling and segmenting the collected vibration signals, and obtaining a sample data set of vibration signals; Step S2, preprocessing the sample data set, and dividing the preprocessed sample data set into a training data set and a test data set; Step S3: construct a fault diagnosis model and use a training data set to train the fault diagnosis model; Step S4: input the test data set into the trained fault diagnosis model to obtain the fault diagnosis classification result; Wherein, in step S3, the fault diagnosis model includes two branches, a feature fusion module for weighted fusion of features processed by the two branches, and a classification layer, wherein the first branch of the two branches is used to extract key time features, including a one-dimensional convolutional neural network and a bidirectional long short-term memory network, wherein the one-dimensional convolutional neural network includes a one-dimensional convolutional layer and a one-dimensional maximum pooling layer; the second branch is used to extract key spatial features, including a two-dimensional convolutional neural network, and the two-dimensional convolutional neural network includes two two-dimensional convolutional layers and two two-dimensional maximum pooling layers; the specific method for training the fault diagnosis model using the training data set is: Step S31, using a one-dimensional convolutional neural network to extract local features of the training data set, then inputting the local features into a bidirectional long short-term memory network to extract key time features in the signal, and the extracted key time features are processed through a flattening layer; Step S32, mapping the training data set into a two-dimensional image through Gram matrix transformation, inputting the two-dimensional image into a two-dimensional convolutional neural network, extracting key spatial features in the signal, and processing the extracted key spatial features through a flattening layer; Step S33, weighted fusion of the features processed by the flattening layer in step S31 and step S32; Step S34: Input the weighted fused features into the classification layer, thus completing the training of the entire model.

[0005] Furthermore, step S1 specifically includes the following steps: Step S1 specifically includes the following steps: obtaining vibration signals of rolling bearings at the same sampling frequency, different loads, and different speeds through the Southeast University wind power transmission system simulation test bench, taking 500 data points as a sample data, and assigning sample data labels according to the rolling bearing state, and combining multiple sample data to obtain a sample data set, 80% of the sample data set is divided out as a training data set, and the remaining 20% ​​is used as a test data set.

[0006] Furthermore, the sampling frequency is 1365HZ, and the rolling bearing includes three states: inner ring 0.5mm crack, inner ring 0.8mm crack and normal state.

[0007] Further, step S2 specifically includes the following steps: using a Kalman filter to denoise the signal, which calculates the Kalman gain by continuously updating the covariance of the system state estimate, combining the filter error covariance at the previous moment and the prediction error covariance at the current moment; in the continuous filtering process of the system state, new measurement data is introduced at each step to provide additional information, and the discrete form of the Kalman filter linear state equation is as follows: x k+1 =Ax k +Bu k +w k y k =Hx k +v k Where: x k and x k+1 Represent the state vectors at the kth moment and the k+1th moment respectively; w k is the system noise vector; u k is the system input; y k is the system output, that is, the observed value; v k is the measurement noise vector; A is the state transfer matrix; B is the input matrix; H is the output matrix; The Kalman filter updates the state estimate in real time, and the update process follows the following formula: Where: It represents the state estimate without considering the observed value at the kth moment, that is, the prior state estimate at the kth moment; It uses the observation value y at the k-1th moment k-1 The state estimate obtained after is the posterior state estimate at the k-1th moment; P k is the prior estimation error covariance matrix; P k-1 is the posterior estimation error covariance matrix, and Q represents the covariance matrix of the process noise.

[0008] Furthermore, in step S33, the weights of the features output by the two branches are both 0.5.

[0009] Furthermore, the bidirectional long short-term memory network includes an input layer, a forward propagation layer, a back propagation layer and an output layer, wherein the calculation formula of the output layer is as follows: In the formula, y t represents the output result of the final output layer, which is calculated by superimposing the output values ​​of the two hidden layers at each time step. g represents the activation function, U represents the weight matrix, which is used to linearly transform the forward and backward hidden state vectors, ht represents the hidden state vector at time step t, and c is the bias term.

[0010] Furthermore, the function used in the classification layer is the softmax activation function, which converts the outputs of multiple neurons into the (0,1) interval for processing multi-classification problems; its formula is as follows: Where: q(x i ) represents the i-th element x in the input vector i The corresponding output probability, x is the logical input of the function layer, which produces a C-dimensional probability distribution; the function converts the output of the last layer of the network into a probability value that sums to 1 through a specific mathematical formula.

[0011] On the other hand, the present invention discloses a rolling bearing fault diagnosis device, the device comprising: A vibration signal acquisition module, used to acquire vibration signals of rolling bearings under multiple working conditions; A signal space feature extraction module, used to extract key time features in the signal using the first branch; A signal time feature extraction module, used to extract key spatial features in the signal using the second branch; The rolling bearing fault diagnosis module is used to construct a fault diagnosis model and use the fault diagnosis model to perform fault diagnosis on the rolling bearing.

[0012] In another aspect, the present invention discloses an electronic device, comprising: one or more processors; A storage device for storing one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors implement the above-mentioned rolling bearing fault diagnosis method.

[0013] In a final aspect, the present invention discloses a storage medium comprising computer executable instructions, which are used to perform the above rolling bearing fault diagnosis method when executed by a computer processor.

[0014] Beneficial effects:

[0015] (1) The rolling bearing fault diagnosis method disclosed in the present invention automatically extracts and identifies the fault features of rolling bearings under complex working conditions by combining a convolutional neural network with a bidirectional long short-term memory network, thereby realizing adaptive fault feature extraction under multi-dimensional signals and rolling bearing fault diagnosis under high noise and complex working conditions.

[0016] (2) The rolling bearing fault diagnosis method disclosed in the present invention uses a Kalman filter, which can separate useful information from noise in the signal and achieve signal noise reduction.

[0017] (3) The rolling bearing fault diagnosis method disclosed in the present invention can effectively and adaptively extract spatial features due to its 1D-2D-convolutional neural network.

[0018] (4) In the rolling bearing fault diagnosis method disclosed in the present invention, the bidirectional long short-term memory network captures the time-dependent characteristics of the signal with its advantages in time series analysis. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 It is a flow chart of the present invention.

[0020] Figure 2 It is a structural diagram of the fault diagnosis model in the present invention.

[0021] Figure 3 It is a predicted fault classification result diagram of the present invention.

[0022] Figure 4 It is the confusion matrix result diagram of the present invention. DETAILED DESCRIPTION

[0023] The technical solution of the present invention will be clearly and completely described below in conjunction with specific embodiments. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0024] The present invention discloses a rolling bearing fault diagnosis method, device, electronic equipment and storage medium. Figure 1 As shown, the main steps include: obtaining a sample data set, preprocessing the sample data set, and dividing the preprocessed sample data set into a training data set and a test data set; building a fault diagnosis model, and using the training data set to train the fault diagnosis model; inputting the test data set into the trained fault diagnosis model to obtain the fault diagnosis classification result; through the combination of convolutional neural network and bidirectional long short-term memory network, adaptive feature extraction and rolling bearing fault diagnosis under high noise and complex working conditions are realized. Each step is described in detail below.

[0025] Step S1, obtaining a sample data set: obtaining vibration signals of rolling bearings under multiple working conditions through public data sets, labeling and segmenting the collected vibration signals, and obtaining a sample data set of vibration signals.

[0026] The vibration signals of rolling bearings at the same sampling frequency, different loads and different speeds are obtained through the wind power transmission system simulation test bench of Southeast University. The sampling frequency is 1365HZ. The rolling bearings have three states: 0.5mm crack on the inner ring, 0.8mm crack on the inner ring and normal state. 500 data points are taken as a sample data, and the sample data labels are assigned according to the rolling bearing state. Multiple sample data are combined to obtain a sample data set.

[0027] Step S2: preprocess the fault data set, and divide the preprocessed sample data set into a training data set and a test data set.

[0028] The sample data set is preprocessed, that is, outliers are removed, duplicate values ​​are integrated, missing values ​​are filled, and signal noise reduction is performed on the data set. The signal noise reduction process is: the Kalman filter is used to denoise the signal, which continuously updates the covariance of the system state estimate, combines the filter error covariance at the previous moment and the prediction error covariance at the current moment, to calculate the Kalman gain. In the continuous filtering process of the system state, new measurement data is introduced at each step to provide additional information. The discrete form of the Kalman filter linear state equation is as follows: x k+1 =Ax k +Bu k +w k y k =Hx k +v k Where: x k and x k+1 Represent the state vectors at the kth moment and the k+1th moment respectively; w k is the system noise vector; u k is the system input; y k is the system output, that is, the observed value; v k is the measurement noise vector; A is the state transfer matrix; B is the input matrix; H is the output matrix.

[0029] The Kalman filter updates the state estimate in real time, and the update process follows the following formula: Where: It represents the state estimate without considering the observed value at the kth moment, that is, the prior state estimate at the kth moment; It uses the observation value y at the k-1th moment k-1 The state estimate obtained after is the posterior state estimate at the k-1th moment; is the prior estimation error covariance matrix; P k-1 is the posterior estimation error covariance matrix, and Q represents the covariance matrix of the process noise.

[0030] Step S3: construct a fault diagnosis model and use the training data set to train the fault diagnosis model.

[0031] Step S31: Figure 2 As shown, the fault diagnosis model includes two branches, a feature fusion module for weighted fusion of features processed by the two branches, and a classification layer, wherein the first branch of the two branches is used to extract key time features, including a one-dimensional convolutional neural network and a bidirectional long short-term memory network; the second branch is used to extract key spatial features, including a two-dimensional convolutional neural network.

[0032] The preprocessed training set is input into two branches respectively to extract different features from the vibration signal. Specifically: First branch: The preprocessed training data set is passed through a 1D convolutional neural network to extract the local features of the signal. The extracted features are then input into a bidirectional long short-term memory network to extract the key time features in the signal. The extracted features are processed through a flattening layer to meet subsequent needs. The 1D convolutional neural network contains a one-dimensional convolution layer and a one-dimensional maximum pooling layer. Second branch: The preprocessed training data set is transformed by the Gram matrix to map it into a two-dimensional image, which is then input into a two-dimensional convolutional neural network to extract key spatial features from the signal. The extracted features are processed through a flattening layer to meet subsequent needs. The two-dimensional convolutional neural network contains two two-dimensional convolutional layers and two two-dimensional maximum pooling layers.

[0033] The bidirectional LSTM network in the first branch contains two complete sets of LSTM network parameters, one of which processes the forward information of the sequence (from the past to the future), and the other processes the reverse information of the sequence (from the future to the past). The forward and backward mathematical expressions of the bidirectional LSTM network are as follows: Where h represents the calculation result of the hidden layer of the long short-term memory network; x t represents the input data at time step t; w represents the input weight of the long short-term memory network; v represents the weight at time step t-1; h t-1 represents the output of the previous moment, b represents the bias term of the long short-term memory network layer; f represents the activation function.

[0034] The bidirectional long short-term memory network includes an input layer, a forward propagation layer, a back propagation layer, and an output layer, where the calculation formula of the output layer is as follows: In the formula, y t represents the output result of the final output layer, which is calculated by superimposing the output values ​​of the two hidden layers at each time step. g represents the activation function, U represents the weight matrix, which is used to linearly transform the forward and backward hidden state vectors, and h t represents the hidden state vector at time step t, and c is the bias term.

[0035] Step S32: weighted fusion of the important features extracted in step S3, with the weight of the features output by each branch being 0.5.

[0036] Step S33: Input the extracted final features into the classification layer to obtain the predicted bearing state. The function used in the classification layer is the softmax activation function, which converts the outputs of multiple neurons into the (0,1) interval for processing multi-classification problems. The formula is as follows: Where: q(x i ) represents the i-th element x in the input vector iThe corresponding output probability, x is the logical input of the function layer, which produces a C-dimensional probability distribution. The function converts the output of the last layer of the network into a probability value that sums to 1 through a specific mathematical formula. The category of the sample is determined by the label with the highest probability value.

[0037] The mean of the cross entropy function between the predicted bearing state and the sample data label of the test data set is calculated as the loss function of the fault diagnosis model, and the model parameters are optimized by back propagation through the Adam stochastic gradient descent algorithm until the maximum training round set is reached. The number of iterations is 150, the batch size is 32, the Adam optimizer is used, the learning rate is 0.01, the regularization parameter is 0.001, and the back propagation algorithm is used to iteratively update the parameters of the fault diagnosis model until the number of iterations is reached, which means that the training of the rolling bearing fault diagnosis model is completed, and step S4 is entered.

[0038] Step S4: input the test data set into the trained fault diagnosis model, and calculate the diagnostic accuracy of the test data set.

[0039] Step S41: After the parameters of the fault diagnosis model stop updating, a trained fault diagnosis model is obtained, and the test data set is input into the trained fault diagnosis model to obtain a characteristic parameter matrix.

[0040] Step S42: Input the feature parameter matrices obtained in step S41 into the flattening layer for processing to meet subsequent needs, then perform weighted fusion on the flattened features of the two branches, and then input the final features into the classification layer, output the bearing state category of the identified test data set, compare the bearing state category of the identified test data set with the sample data label of the test data set, and calculate the diagnostic recognition accuracy of this rolling bearing fault.

[0041] In order to determine the optimal number of iterations, the present invention studies the number of iterations, and the final result is shown in Table 1.

[0042] Table 1 Test results of different iteration numbers Iterations Time(s) Accuracy loss Learning Rate 1 0 23.44％ 1.8110 0.01 50 1 96.09％ 0.1564 0.01 100 3 95.31％ 0.1234 0.01 150 4 96.88％ 0.0919 0.0001

[0043] Table 1 shows the time, accuracy, loss and basic learning rate of each iteration when the data set is input into the model for testing. In the first iteration, the model's small batch accuracy is low (23.44%) and the loss value is high (1.8110), which indicates that the model has just started training and has not learned enough information. By the 50th iteration, the small batch accuracy has increased significantly to 96.09% and the loss value has decreased to 0.1546, which indicates that the model has learned effective features and has good prediction performance. In the 100th iteration, the small batch accuracy has decreased slightly (95.31%) and the loss value has increased (0.1234), which may be due to the model starting to overfit the training data. In the 150th iteration, the small batch accuracy has increased again to 96.88%, and the accuracy has reached the highest level. At the same time, the basic learning rate is reduced to 0.0001, which is to make more detailed weight adjustments in the later stages of training to avoid overfitting.

[0044] In order to evaluate the performance of different models in the rolling bearing fault diagnosis task, the present invention uses a unified data set to test three model architectures, including a one-dimensional convolutional neural network-two-dimensional convolutional neural network model (abbreviated as 1DCNN-2DCNN), an LSTM model, and the fault diagnosis model of the present invention. In the experiment, all models were trained for 150 iterations. The experimental results are summarized in Table 2.

[0045] Table 2 Accuracy test results of each model Iterations Model Name Accuracy 150 1DCNN-2DCNN 89.72％ 150 LSTM 90.58％ 150 Fault diagnosis model in the present invention 96.88％

[0046] Table 2 shows the accuracy of each model on the test set. Specifically, when the iteration cycle reaches 150, the accuracy of the 1DCNN-2DCNN model and the LSTM model are 92.67% and 90.58% respectively. In contrast, the fault diagnosis model of the present invention, which combines the 1D-2D-CNN and BILSTM algorithms, shows a higher accuracy of 96.88%.

[0047] The experimental results show that combining the deep feature extraction capability of 1DCNN-2DCNN with the strength of BiLSTM in processing time series data can effectively improve the performance of the model in rolling bearing fault diagnosis tasks. This combined model not only enhances the ability to capture the deep features of vibration signals, but also significantly improves the understanding of time dependence, thus achieving significant progress in the accuracy and robustness of fault diagnosis.

[0048] In order to more intuitively display the accuracy of the proposed model for each classification, the predicted fault classification result diagram and the confusion matrix of the fault classification results are plotted. Figure 3The figure shows the results of the model on the fault classification task. The figure uses a graphical representation to intuitively present the correspondence between the predicted results and the actual labels. In this figure, the horizontal axis represents the number of the predicted sample, and the vertical axis represents the classification result. In order to distinguish the predicted value from the actual value, two different graphic symbols are used: the triangle is used to represent the fault category predicted by the model, and the asterisk (star symbol) represents the actual fault category of the sample. When the predicted value is consistent with the actual value, the triangle and the asterisk overlap in the figure, indicating that the fault classification is correct. According to Figure 3 The results show that the fault diagnosis accuracy of the model of the present invention on the test set reaches 96.88%. Figure 4 The confusion matrix of the test set is shown, which represents the actual fault type on the vertical axis (ordinate) and the fault type diagnosed by the model on the horizontal axis (abscissa). Figure 4 The data shows that the model achieved an accuracy rate of 86.7%, 100% and 96.2% respectively in the classification tasks of three different fault types: normal bearings, 0.8 mm inner ring faults and 0.5 mm inner ring faults.

[0049] The preferred embodiments of the present invention are described in detail above; however, the present invention is not limited to the specific details in the above embodiments. Within the technical concept of the present invention, various equivalent transformations can be made to the technical solutions of the present invention, and these equivalent transformations all belong to the protection scope of the present invention.

Claims

1. A rolling bearing fault diagnosis method, characterized in that: The main steps are as follows: Step S1, obtaining vibration signals of rolling bearings under multiple working conditions through public data sets, labeling and segmenting the collected vibration signals, and obtaining a sample data set of vibration signals; Step S2, preprocessing the sample data set, and dividing the preprocessed sample data set into a training data set and a test data set; Step S3: construct a fault diagnosis model and use a training data set to train the fault diagnosis model; Step S4: input the test data set into the trained fault diagnosis model to obtain the fault diagnosis classification result; Wherein, in step S3, the fault diagnosis model includes two branches, a feature fusion module for weighted fusion of features processed by the two branches, and a classification layer, wherein the first branch of the two branches is used to extract key time features, including a one-dimensional convolutional neural network and a bidirectional long short-term memory network, wherein the one-dimensional convolutional neural network includes a one-dimensional convolutional layer and a one-dimensional maximum pooling layer; the second branch is used to extract key spatial features, including a two-dimensional convolutional neural network, and the two-dimensional convolutional neural network includes two two-dimensional convolutional layers and two two-dimensional maximum pooling layers; the specific method for training the fault diagnosis model using the training data set is: Step S31, using a one-dimensional convolutional neural network to extract local features of the training data set, then inputting the local features into a bidirectional long short-term memory network to extract key time features in the signal, and the extracted key time features are processed through a flattening layer; Step S32, mapping the training data set into a two-dimensional image through Gram matrix transformation, inputting the two-dimensional image into a two-dimensional convolutional neural network, extracting key spatial features in the signal, and processing the extracted key spatial features through a flattening layer; Step S33, weighted fusion of the features processed by the flattening layer in step S31 and step S32; Step S34: Input the weighted fused features into the classification layer, thus completing the training of the entire model.

2. A rolling bearing fault diagnosis method according to claim 1, characterized in that: Step S1 specifically includes the following steps: obtaining the vibration signals of the rolling bearing at the same sampling frequency, different loads, and different speeds through the Southeast University wind power transmission system simulation test bench, taking 500 data points as a sample data, and assigning sample data labels according to the rolling bearing state, combining multiple sample data to obtain a sample data set, 80% of the sample data set is divided out as a training data set, and the remaining 20% ​​is used as a test data set.

3. A rolling bearing fault diagnosis method according to claim 2, characterized in that: The sampling frequency is 1365HZ, and the rolling bearing has three states: inner ring crack of 0.5mm, inner ring crack of 0.8mm and normal state.

4. A rolling bearing fault diagnosis method according to claim 1, characterized in that: Step S2 specifically includes the following steps: using a Kalman filter to denoise the signal, which calculates the Kalman gain by continuously updating the covariance of the system state estimate, combining the filter error covariance at the previous moment and the prediction error covariance at the current moment; in the continuous filtering process of the system state, new measurement data is introduced at each step to provide additional information. The discrete form of the Kalman filter linear state equation is as follows: x k+1 =Ax k +Bu k +w k y k =Hx k +v k Where: x k and x k+1 Represent the state vectors at the kth moment and the k+1th moment respectively; w k is the system noise vector; u k is the system input; y k is the system output, that is, the observed value; v k is the measurement noise vector; A is the state transfer matrix; B is the input matrix; H is the output matrix; The Kalman filter updates the state estimate in real time, and the update process follows the following formula: Where: It represents the state estimate without considering the observed value at the kth moment, that is, the prior state estimate at the kth moment; It uses the observation value y at the k-1th moment k-1 The state estimate obtained after is the posterior state estimate at the k-1th moment; P k is the prior estimation error covariance matrix; P k-1 is the posterior estimation error covariance matrix, and Q represents the covariance matrix of the process noise.

5. A rolling bearing fault diagnosis method according to claim 1, characterized in that: In step S33, the weights of the features output by the two branches are both 0.

5.

6. A rolling bearing fault diagnosis method according to claim 1, characterized in that: The bidirectional long short-term memory network includes an input layer, a forward propagation layer, a back propagation layer, and an output layer, where the calculation formula of the output layer is as follows: In the formula, y t represents the output result of the final output layer, which is calculated by superimposing the output values ​​of the two hidden layers at each time step. g represents the activation function, U represents the weight matrix, which is used to linearly transform the forward and backward hidden state vectors, and h t represents the hidden state vector at time step t, and c is the bias term.

7. A rolling bearing fault diagnosis method according to claim 1, characterized in that: The function used in the classification layer is the softmax activation function, which converts the output of multiple neurons into the (0,1) interval to handle multi-classification problems; The formula is as follows: Where: q(x i ) represents the i-th element x in the input vector i The corresponding output probability, x is the logical input of the function layer, which produces a C-dimensional probability distribution; the function converts the output of the last layer of the network into a probability value that sums to 1 through a specific mathematical formula.

8. A rolling bearing fault diagnosis device, characterized in that: The device comprises: A vibration signal acquisition module, used to acquire vibration signals of rolling bearings under multiple working conditions; A signal space feature extraction module, used to extract key time features in the signal using the first branch; A signal time feature extraction module, used to extract key spatial features in the signal using the second branch; The rolling bearing fault diagnosis module is used to construct a fault diagnosis model and use the fault diagnosis model to perform fault diagnosis on the rolling bearing.

9. An electronic device, characterized in that: include: one or more processors; A storage device for storing one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors implement the rolling bearing fault diagnosis method described in any one of claims 1-7.

10. A storage medium containing computer executable instructions, characterized in that: The computer executable instructions are used to execute the rolling bearing fault diagnosis method as described in any one of claims 1-7 when executed by a computer processor.

Images