Lightweight bearing fault diagnosis method based on multi-scale feature fusion

Abstract

A lightweight bearing fault diagnosis method based on multi-scale feature fusion relates to a bearing fault diagnosis method.Bearing vibration signals are collected and input into a trained fault diagnosis model for fault diagnosis, the fault diagnosis model comprises a multi-layer feature extraction module and an output layer, the feature extraction module adopts a residual structure, and the output layer adopts an output layer; in the main branch of the residual structure, the input data is firstly subjected to point-by-point convolution processing, the data subjected to the point-by-point convolution processing is divided into two paths, one path is reserved, and the other path is subjected to channel-by-channel hole convolution processing and then is subjected to channel dimension splicing with the reserved path of data subjected to the point-by-point convolution processing; a channel attention mechanism is introduced into a residual structure auxiliary branch to process input data, and then the input data are fused with data processed by a main branch to serve as output of a feature extraction module. By means of complementation of channel-by-channel cavity convolution and point-by-point convolution, noise in the cavity can be filtered, remote induction correlation characteristics of the signal and difference characteristics between different channels are captured, and the fault diagnosis accuracy is improved.

Description

Technical Field

[0001] The present invention relates to a bearing fault diagnosis method, in particular to a lightweight bearing fault diagnosis method based on multi-scale feature fusion. Background Art

[0002] As the core component of industrial machinery, rolling bearings are widely used in rotating equipment such as motors, engines and gearboxes. However, rolling bearings are prone to failure due to long-term operation in complex environments such as high speed and heavy load. According to statistics, 45%-55% of mechanical equipment damage is related to bearings. Once a rolling bearing fails, it can cause property losses at the least and casualties at the worst. Therefore, it is extremely important to propose a bearing fault diagnosis method with high accuracy and strong feasibility.

[0003] At present, deep learning is widely used in the field of mechanical fault diagnosis due to its ability of adaptive feature learning. In actual industrial environments, mechanical equipment often operates in complex environments. The signals collected by sensors are inevitably mixed with a lot of noise and nonlinear interference. In order to improve the noise resistance and diagnostic accuracy of the model, the existing technology uses multi-head attention to fully consider the characteristics of rolling bearing faults to design attention modules, enhance the features related to the fault and ignore irrelevant features. However, due to the addition of multi-head attention, the FLOPs of the model also reaches tens of millions. There is also a bearing fault feature extraction network based on dynamic convolution and depthwise separable convolution, and the spatial attention mechanism (SAM) is used to adjust the weight of the output feature map, so that the model pays more attention to useful information. However, the module proposed by this method is too complex and costly to be deployed in practice.

[0004] Although the methods used in existing technologies have achieved certain results, they still have shortcomings in balancing recognition accuracy and model lightweight. Summary of the invention

[0005] The technical problem to be solved by the present invention is to improve the anti-noise interference capability of the model and to provide a lightweight bearing fault diagnosis method based on multi-scale feature fusion.

[0006] The technical solution adopted by the present invention to solve the above technical problems is: a lightweight bearing fault diagnosis method based on multi-scale feature fusion, collecting vibration signals during the operation of the bearing, and inputting them into a trained fault diagnosis model for fault diagnosis. The fault diagnosis model includes a multi-layer feature extraction module and an output layer that outputs the fault type according to the extracted features. The feature extraction module adopts a residual structure. The main branch of the residual structure first performs point-by-point convolution processing on the input data, and the data after point-by-point convolution processing is divided into two paths, one of which is retained, and the other is processed by channel-by-channel hole convolution and then spliced ​​with the retained one path processed by point-by-point convolution in the channel dimension; a channel attention mechanism is introduced in the secondary branch of the residual structure to process the input data, and then it is fused with the data processed by the main branch as the output of the feature extraction module.

[0007] Furthermore, a wide kernel convolution layer is used to perform data compression and channel expansion on the collected original vibration signal, and then it is input into the feature extraction module for feature extraction.

[0008] The convolution kernel size of the wide kernel convolution layer is 16, and the number of convolution kernel channels is 128.

[0009] The channel-by-channel atrous convolution of the feature extraction module adopts large-kernel channel-by-channel atrous convolution, and the convolution kernel size is 1×63.

[0010] Furthermore, the channel-by-channel atrous convolutions in the multi-layer feature extraction module have different dilation rates.

[0011] The fault diagnosis model includes 4 layers of feature extraction modules connected in sequence, and the expansion rates of channel-by-channel hole convolutions in the 4 layers of feature extraction modules are set to 1, 2, 5 and 1 respectively.

[0012] In the sub-branch of the residual structure of the feature extraction module, ECA attention is used to calculate channel weights to obtain the importance of input data, and then weights are assigned to each channel to obtain weighted data.

[0013] In the feature extraction module, the data after point-by-point convolution is processed using the PReLU activation function, and after batch normalization by the BN layer, it is divided into two paths for subsequent processing.

[0014] In the feature extraction module, the data after channel-by-channel atrous convolution is processed using the PReLU activation function, and after batch normalization by the BN layer, it is spliced ​​with the retained data in the channel dimension.

[0015] The output layer includes a global average pooling layer, a fully connected layer and a Softmax layer.

[0016] The beneficial effects of the present invention are as follows: in the feature extraction module of the fault diagnosis model, feature extraction of different spatial positions, different distances and different channels is realized by point-by-point convolution, channel-by-channel dilated convolution and channel attention mechanism, and fully integrated, so as to learn features more efficiently and improve the model's ability to resist noise interference. The results of point-by-point convolution are integrated with the results of channel-by-channel dilated convolution, and the channel-by-channel dilated convolution is used to complement the point-by-point convolution, so as to filter the noise in the hole, capture the correlation characteristics of the long-distance induction of the signal and the difference characteristics between different channels, and improve the accuracy of fault diagnosis. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 It is a structural schematic diagram of the fault diagnosis model of the present invention.

[0018] Figure 2 It is a structural schematic diagram of the feature extraction module of the present invention.

[0019] Figure 3 It is a line chart of recognition accuracy of different methods with different signal-to-noise ratios.

[0020] Figure 4 It is a t-SNE dimension reduction graph of the feature graph of the multi-layer feature extraction module of the present invention. DETAILED DESCRIPTION

[0021] The technical solution of the present invention is described clearly and completely below in conjunction with the accompanying drawings and specific implementation methods. The specific contents listed in the following embodiments are not limited to the technical features necessary for the technical problems to be solved by the technical solutions recorded in the claims. At the same time, the enumerated embodiments are only part of the present invention, not all embodiments.

[0022] The lightweight bearing fault diagnosis method based on multi-scale feature fusion of the present invention relies on a trained fault diagnosis model. The input data of the fault diagnosis model is the vibration signal collected during the bearing movement process. The vibration signal is feature extracted and processed by the fault diagnosis model, and the fault type is output through its output layer.

[0023] The fault diagnosis model of the present invention includes multiple layers of feature extraction modules for extracting features from input data to learn fault features in depth. The structure of the feature extraction module is as follows: Figure 2As shown in the figure, the residual structure is adopted, which is a lightweight multi-scale hole residual module, named MSL Block. The residual structure of the feature extraction module has two branches, and the main branch includes a point-by-point convolution layer Conv (1×1) and a channel-by-channel hole convolution layer. In the main branch, the data with a dimension of C×W is first processed by the point-by-point convolution layer Conv (1×1) to learn the instantaneous features of the signal at close range and the features of different spatial positions. At the same time, the dimension becomes C / 2×W, and then it is divided into two paths. One path of data is retained first, and the other path of data is sent to the channel-by-channel hole convolution layer for processing. This operation performs hole and channel-by-channel convolution at the same time during convolution, learns the difference features of different channels and long-distance correlation features, increases the receptive field of network feature extraction, and fully captures the time correlation information of the signal. The data processed by the channel-by-channel hole convolution is spliced ​​with the previously retained data processed by the point-by-point convolution in the channel dimension to achieve multi-angle feature fusion. In the secondary branch, the channel attention mechanism is used to process the input data. Specifically, ECA attention can be used to calculate the channel weights, obtain the importance of the input data, enhance the model's sensitivity to important features, and then assign weights to each channel to obtain weighted data. After the calculation of the main branch and the secondary branch is completed, the residual method is used to further fuse the fault feature information extracted by the two branches, enriching the fault feature information while alleviating network degradation, and improving the model's ability to extract multi-angle fault features.

[0024] The channel-by-channel atrous convolution adopts a large-kernel channel-by-channel atrous convolution, and the convolution kernel size can be 1×63. The large-kernel channel-by-channel convolution significantly increases the receptive field of the convolution layer, and retains the phase information of the signal, avoiding the loss of spatial position features in deep downsampling. Combining the atrous convolution with the large-kernel channel-by-channel convolution can further improve the receptive field of the convolution layer. At the same time, the role of the large-kernel channel-by-channel atrous convolution is complementary to the point-by-point convolution, which can filter the noise in the hole, capture the long-distance induced correlation characteristics of the signal, and the difference characteristics between different channels.

[0025] Figure 1In the illustrated embodiment, the fault diagnosis model adopts a 4-layer feature extraction module (MSL Block). The fault diagnosis model now uses a wide-core convolution layer to perform data compression and channel expansion on the collected original vibration signal to better capture the low-frequency components in the signal, reduce the impact of high-frequency noise on fault diagnosis, and lay a solid foundation for subsequent feature learning. The convolution kernel size of the wide-core convolution layer is 16, and the number of convolution kernel channels is 128. Wide-core convolution can effectively compress the input signal through a larger receptive field and reduce the redundant information of the data. Channel expansion through wide-core convolution can increase the number of channels of the feature map, so that the network can learn more diverse features. The feature map after channel expansion can capture features of different frequencies and scales, enhancing the expression ability of the model. Therefore, using a wide-core convolution layer to compress the original vibration signal and expand the channel can not only effectively reduce the interference of high-frequency noise, but also increase the diversity of features through channel expansion. This processing method not only improves the anti-noise ability of the model, but also reduces the computational complexity, making the model more lightweight and efficient. The data processed by the wide-core convolution layer is input into the feature extraction module for feature extraction. In this embodiment, four layers of feature extraction modules are stacked to process the data in sequence. The channel-by-channel atrous convolutions in different feature extraction modules may have different dilation rates, for example Figure 1 In the illustrated embodiment, the expansion rate of the first MSL Block is 1, the expansion rate of the second MSL Block is set to 2, the expansion rate of the third MSL Block is 5, and the expansion rate of the fourth MSL Block is 1. By setting different expansion rates at different layers, the problem of grid effect caused by stacking multiple layers of dilated convolution can be avoided, which effectively alleviates information loss and improves network performance.

[0026] Figure 1 The data processing process of the second MSL Block is shown in the figure, and the other MSL Blocks are similar. In the figure, the data dimension of the second MSL Block is C×W. After the data of dimension C×W is processed by point-by-point convolution with convolution kernel size k=1 in the main branch, the data dimension becomes C / 2×W. On the one hand, the data is retained, and on the other hand, it is processed by channel-by-channel dilated convolution with convolution kernel k=63. The processing result is concatenated with the retained data to form data of dimension C×W. The data originally input to the MSLBlock is processed by the ECA attention mechanism in the secondary branch, and then added and fused with the data processed by the main branch as the output of the MSL Block.

[0027] like Figure 2As shown in the figure, in the feature extraction module, the data after point-by-point convolution is processed using the PReLU activation function, and after batch normalization by the BN layer, it is divided into two paths for subsequent processing. The data after channel-by-channel atrous convolution is also processed using the PReLU activation function, and after batch normalization by the BN layer, it is concatenated with the retained data in the channel dimension. The PReLU activation function is used to improve the nonlinear fitting ability of the model, and the BN layer is used to batch normalize the input of each layer to prevent the gradient from disappearing.

[0028] like Figure 1 As shown in the figure, the output layer of the fault diagnosis model includes a global average pooling layer (GAP), a fully connected layer (Linear) and a SoftMax layer. After feature extraction by the multi-layer feature extraction module, the learned features are globally average pooled, classified using the fully connected layer and the Softmax layer, and finally the fault type is output.

[0029] The CWRU bearing dataset is used to train and verify the fault diagnosis model. The dataset can be divided into 10 fault types according to the fault location and damage degree. According to the different loads, it is divided into 4 datasets, A, B, C, and D. Among them, datasets A, B, and C are datasets under loads of 1HP, 2HP, and 3HP, respectively. Dataset D is a dataset mixed with 0-3 loads. In this dataset, the vibration signal is randomly overlapped and sampled with a step size of 207 and a sample length of 1024. The number of single-class samples in each dataset is 500, with a total of 5000 samples. The ratio of training set, validation set, and test set is 8:1:1.

[0030] The experimental data parameter statistics are shown in Table 1 below: The following adopts the above Figure 1 The fault diagnosis model of the embodiment performs a fault diagnosis experiment to test the accuracy of the fault diagnosis.

[0031] First, an ablation experiment is performed, and the fault diagnosis model of the present invention is used as the baseline model. Six other comparison cases are set, namely, no dilated convolution, no point-by-point convolution, no large-kernel channel-by-channel dilated convolution, no ECA attention mechanism (retaining residual connection), no ECA attention mechanism and residual connection, and ReLU instead of PReLU. The experiment is performed using dataset D, and the experimental results are shown in Table 2 below: As can be seen from Table 2, no matter which module is missing in the fault diagnosis model, the performance will be reduced, and as the noise increases, the performance of the comparison model will be more seriously reduced. At the same time, it can be seen from the table that when the large-core per-channel dilated convolution is not used, the performance of the model is more obviously reduced. This shows that when doing bearing fault diagnosis based on vibration signals, long-distance correlation features are very important. The huge receptive field brought by the large-core per-channel dilated convolution is very helpful for long-distance feature extraction. At the same time, the dilated convolution can effectively filter noise and bring performance improvement.

[0032] In order to verify the anti-noise ability and complex working condition diagnosis ability of the fault diagnosis model of the present invention, WDCNN, MA1DCNN, DRSN-CW, and MIXCNN2 are introduced for comparison. By adding Gaussian white noise with different signal-to-noise ratios to the CWRU data sets A, B, C, and D, the noise interference in the actual working environment is simulated, and the fault recognition accuracy of different models in a high noise environment is explored, which also further proves the feature extraction ability of the method proposed in this paper. The specific results are shown in Table 3 below: In order to more intuitively show the accuracy of different models, a line graph of the fault recognition accuracy of different models for the CWRU dataset D under different noise environments is drawn, as shown in Figure 3 shown.

[0033] From Table 3 and Figure 3 It can be seen that compared with several other methods, the present invention can show better fault identification performance when facing the complex working condition data set D (mixed load). At the same time, when facing data sets A, B, and C that are not affected by load changes, the recognition accuracy of the present invention is higher.

[0034] In order to better understand the training process and feature extraction capability of the fault diagnosis model of the present invention, the t-SNE dimensionality reduction algorithm is used to reduce the dimension of the output feature map of the middle layer of the model. The results are as follows: Figure 4 As shown in the figure, after the first feature extraction module (MSL Block1), the ten categories tend to separate. After the second and third feature extraction modules (MSL Block2 and MSL Block3), there are obvious boundary areas between different categories, but the observation points in the same category are still relatively scattered. After the fourth feature extraction module (MSL Block4), different categories are completely separated, and the observation points in the same category are closely clustered. This shows that the fault diagnosis model of the present invention is very effective in extracting features from raw data and has good anti-noise interference capabilities.

[0035] The following is an analysis of model complexity. By comparing the parameter quantity and FLOPs with other bearing fault diagnosis models or lightweight models, the lightweight degree of the fault diagnosis model of the present invention is proved, wherein the parameter quantity can reflect the size of the model's video memory occupation, and FLOPs can reflect the model's computational complexity. The specific comparison results are shown in Table 4 below: It can be seen from Table 4 and Table 3 that although the number of parameters and FLOPs of WDCNN are lower than those of the present invention, its accuracy is poor. MA1DCNN and DRSN-CW did not consider lightweight when designing, and due to the existence of some special modules in their network structure, the computational complexity of the model is also very high. MIXCNN2 is a bearing fault diagnosis model with mixed information. This model uses deep separable convolution to reduce the number of parameters, and its accuracy is also good in a noisy environment. The number of parameters of the present invention is 33% less than it, and the FLOPs is 27% less than it. In addition, the accuracy and noise resistance of the present invention are higher than MIXCNN2.

[0036] The above description of the specific implementation mode is only used to help understand the technical concept and core idea of ​​the present invention. Although the technical solution is described and illustrated using a specific preferred embodiment, it should not be understood as a limitation of the present invention itself. Those skilled in the art may make various changes in form and details without departing from the technical concept of the present invention. These easily conceived changes or substitutions should all be included in the protection scope of the present invention.

Claims

1. A lightweight bearing fault diagnosis method based on multi-scale feature fusion collects vibration signals during the operation of the bearing and inputs them into a trained fault diagnosis model for fault diagnosis. The fault diagnosis model includes a multi-layer feature extraction module and an output layer that outputs the fault type according to the extracted features, and is characterized in that: The feature extraction module adopts a residual structure. In the main branch of the residual structure, the input data is firstly processed by point-by-point convolution. The data after the point-by-point convolution processing is divided into two paths, one of which is retained, and the other is processed by channel-by-channel hole convolution and then spliced ​​with the retained one path processed by point-by-point convolution in the channel dimension; a channel attention mechanism is introduced in the secondary branch of the residual structure to process the input data, and then it is fused with the data processed by the main branch as the output of the feature extraction module.

2. The lightweight bearing fault diagnosis method based on multi-scale feature fusion according to claim 1, characterized in that: The wide kernel convolution layer is used to perform data compression and channel expansion on the collected original vibration signal, and then it is input into the feature extraction module for feature extraction.

3. The lightweight bearing fault diagnosis method based on multi-scale feature fusion according to claim 2 is characterized in that: The convolution kernel size of the wide-core convolution layer is 16, and the number of convolution kernel channels is 128.

4. The lightweight bearing fault diagnosis method based on multi-scale feature fusion according to claim 1, characterized in that: The channel-by-channel atrous convolution of the feature extraction module adopts large-kernel channel-by-channel atrous convolution, and the convolution kernel size is 1×63.

5. The lightweight bearing fault diagnosis method based on multi-scale feature fusion according to claim 1, characterized in that: The channel-by-channel atrous convolutions in the multi-layer feature extraction modules have different dilation rates.

6. The lightweight bearing fault diagnosis method based on multi-scale feature fusion according to claim 5, characterized in that: The fault diagnosis model includes 4 layers of feature extraction modules connected in sequence, and the expansion rates of channel-by-channel hole convolutions in the 4 layers of feature extraction modules are set to 1, 2, 5 and 1 respectively.

7. The lightweight bearing fault diagnosis method based on multi-scale feature fusion according to claim 1, characterized in that: In the sub-branch of the residual structure of the feature extraction module, ECA attention is used to calculate channel weights to obtain the importance of input data, and then weights are assigned to each channel to obtain weighted data.

8. The lightweight bearing fault diagnosis method based on multi-scale feature fusion according to claim 1, characterized in that: In the feature extraction module, the data after point-by-point convolution is processed using the PReLU activation function, and after batch normalization by the BN layer, it is divided into two paths for subsequent processing.

9. The lightweight bearing fault diagnosis method based on multi-scale feature fusion according to claim 8, characterized in that: In the feature extraction module, the data after channel-by-channel atrous convolution is processed using the PReLU activation function, and after batch normalization by the BN layer, it is spliced ​​with the retained data in the channel dimension.

10. The lightweight bearing fault diagnosis method based on multi-scale feature fusion according to claim 1, characterized in that: The output layer includes a global average pooling layer, a fully connected layer and a Softmax layer.

Images