Fault diagnosis method for motor bearing based on wpt-1d cnn

Abstract

The present application relates to the technical field of bearing fault diagnosis, and particularly provides a motor bearing fault diagnosis method based on WPT-1DCNN, aiming to solve the problem of low accuracy of existing bearing fault diagnosis methods. To this end, the motor bearing fault diagnosis method based on WPT-1DCNN comprises: obtaining a vibration signal of a motor bearing to be detected; performing WPT transformation on the vibration signal of the motor bearing to be detected, and determining a frequency domain feature vector of the vibration signal of the motor bearing to be detected after multi-layer wavelet packet decomposition; inputting the vibration signal of the motor bearing to be detected and the frequency domain feature vector of the vibration signal of the motor bearing to be detected into a pre-trained double-branch fault diagnosis model, and outputting a fault classification result of the vibration signal of the motor bearing to be detected by the double-branch fault diagnosis model. The diagnosis method can further improve the accuracy of bearing fault diagnosis and has good generalization.

Description

Technical Field

[0001] This invention relates to the field of bearing fault diagnosis technology, and specifically provides a fault diagnosis method and apparatus, storage medium and control device for motor bearings based on WPT-1DCNN. Background Technology

[0002] With the increasing level of automation, the performance requirements for mechanical equipment are also rising. Bearings are a major component of mechanical equipment. Because they operate in harsh natural environments and are subject to various unpredictable factors, they are highly susceptible to failure at some point. Once a failure occurs, it will cause delays and economic losses to some extent, and in severe cases, may even lead to personal injury or death. Therefore, fault diagnosis technology for mechanical equipment is receiving increasing attention and research. If timely judgments can be made and feedback provided to relevant personnel for handling, losses can be minimized.

[0003] When a bearing fails, it generates vibration signals containing a large amount of impact noise. These vibration signals exhibit significant time-varying characteristics, thus displaying obvious nonlinear behavior. If the fault features can be accurately extracted, bearing fault identification can be effectively performed. Existing fault diagnosis methods for motor bearings involve extracting time-domain or frequency-domain features, and then using convolutional neural networks to diagnose the fault based on these features. However, this method cannot simultaneously utilize the large amount of frequency-domain and time-domain information contained in the bearing vibration signal. Furthermore, traditional one-dimensional convolutional neural networks lack sufficient depth in data mining of vibration signals. Therefore, existing bearing fault diagnosis methods suffer from problems such as high data loss, low accuracy, and low generalization.

[0004] Accordingly, there is a need in the field for a new fault diagnosis solution for motor bearings to address the above-mentioned problems. Summary of the Invention

[0005] To overcome the above-mentioned shortcomings, this invention is proposed to provide a method, apparatus, storage medium, and control device for diagnosing motor bearings based on WPT-1DCNN, which solves or at least partially solves the technical problems of low accuracy and low generalization of existing bearing fault diagnosis methods.

[0006] In a first aspect, the present invention provides a fault diagnosis method for motor bearings based on WPT-1DCNN, the method comprising:

[0007] Obtain the vibration signal of the bearing of the motor to be tested;

[0008] The vibration signal of the motor bearing to be tested is subjected to WPT transformation, and after multi-layer wavelet packet decomposition, the frequency domain feature vector of the vibration signal of the motor bearing to be tested is extracted.

[0009] The vibration signal of the motor bearing to be tested and its frequency domain feature vector are input into a pre-trained dual-branch fault diagnosis model, which outputs the fault classification result of the vibration signal of the motor bearing to be tested.

[0010] In one technical solution of the above-mentioned fault diagnosis method for motor bearings based on WPT-1DCNN, the vibration signal of the motor bearing to be tested is subjected to WPT transform, and after determining the multi-layer wavelet packet decomposition, the frequency domain feature vector of the vibration signal of the motor bearing to be tested is extracted, including:

[0011] The vibration signal of the bearing of the motor to be tested is decomposed into 8 sub-bands by three-level wavelet packet decomposition.

[0012] The frequency domain feature vector H of the vibration signal of the motor bearing to be detected is determined by multiplying the maximum amplitude of each of the eight sub-frequency bands by the frequency corresponding to the maximum amplitude of each of the eight sub-frequency bands. k ,for:

[0013]

[0014] Among them, A k The maximum amplitude value in each sub-band; f k The frequency corresponding to the maximum amplitude value in each sub-band; k=1,2,3,4,5,6,7,8.

[0015] In one technical solution of the above-mentioned fault diagnosis method for motor bearings based on WPT-1DCNN, the vibration signal of the motor bearing to be tested and its frequency domain feature vector are input into a pre-trained dual-branch fault diagnosis model. The dual-branch fault diagnosis model outputs the fault classification result of the vibration signal of the motor bearing to be tested, including:

[0016] The vibration signal of the motor bearing to be tested is input into the 1DCNN convolutional layer of the dual-branch fault diagnosis model, and the 1DCNN convolutional layer outputs a 128*t-dimensional feature vector.

[0017] The 128*t-dimensional feature vector is input into the multi-head self-attention mechanism layer of the dual-branch fault diagnosis model, and the multi-head self-attention mechanism layer outputs a 128*1-dimensional time-domain feature vector K of the vibration signal of the motor bearing to be detected.

[0018] The frequency domain feature vector H of the vibration signal of the motor bearing to be tested in an 8*1 dimensional matrix is ​​obtained. k The time-domain feature vector K of the vibration signal of the motor bearing to be tested (128*1 dimensional) is input into the dual-branch feature fusion layer of the dual-branch fault diagnosis model, and the dual-branch feature fusion layer outputs the time-frequency domain feature vector of the vibration signal of the motor bearing to be tested (136*1 dimensional).

[0019] The time-frequency domain feature vector of the vibration signal of the motor bearing under test, which is 136*1 dimensional, is input into the fully connected layer of the dual-branch fault diagnosis model. The fully connected layer outputs the fault diagnosis signal of the vibration signal of the motor bearing under test, which is 4*1 dimensional.

[0020] The fault diagnosis signal of the vibration signal of the motor bearing under test in 4*1 dimension is input into the Softmax layer of the dual-branch fault diagnosis model. The Softmax layer outputs the fault classification result of the vibration signal of the motor bearing under test.

[0021] In one technical solution of the above-mentioned fault diagnosis method for motor bearings based on WPT-1DCNN, the method further includes:

[0022] The dual-branch fault diagnosis model is trained to obtain a pre-trained dual-branch fault diagnosis model.

[0023] In one technical solution of the above-mentioned fault diagnosis method for motor bearings based on WPT-1DCNN, the step of training the dual-branch fault diagnosis model to obtain a pre-trained dual-branch fault diagnosis model includes:

[0024] Obtain the dataset of raw vibration signals of motor bearings, and divide the dataset into training set and test set according to a certain ratio;

[0025] The dual-branch fault diagnosis model is trained using the training set. The training process includes: performing WPT transform on the original vibration signals of the motor bearings in the training set, and after determining the multi-layer wavelet packet decomposition, extracting the frequency domain feature vector of the original vibration signals of the motor bearings in the training set; inputting the frequency domain feature vector of the original vibration signals of the motor bearings in the training set and the original vibration signals of the motor bearings in the training set into the dual-branch fault diagnosis model being trained; the dual-branch fault diagnosis model outputs the fault classification result of the original vibration signals of the motor bearings in the training set.

[0026] By comparing the fault classification results of the original vibration signals of the motor bearings in the training set with the fault labels of the original vibration signals of the motor bearings in the training set, the loss function of the dual-branch fault diagnosis model is calculated.

[0027] When the loss function reaches the requirement of model optimization, training of the two-branch fault diagnosis model is stopped;

[0028] The trained fault diagnosis model is tested using the test set. After the test is completed, a trained dual-branch fault diagnosis model is obtained.

[0029] In one technical solution of the above-mentioned fault diagnosis method for motor bearings based on WPT-1DCNN, the loss function is the cross-entropy loss function.

[0030] In one technical solution of the above-mentioned fault diagnosis method for motor bearings based on WPT-1DCNN, the 1DCNN convolutional layer in the trained dual-branch fault diagnosis model includes an input layer, a first convolutional layer, a first pooling layer, a second convolutional layer, and a second pooling layer; wherein,

[0031] The first convolutional layer is set to 1 input channel, 64 output channels, stride 1, and edge padding 4 layers; the first pooling layer is a max pooling layer; the second convolutional layer is set to 64 input channels, 128 output channels, stride 1, and edge padding 2 layers; the second pooling layer is an average pooling layer.

[0032] In a second aspect, the present invention provides a fault diagnosis device for motor bearings based on WPT-1DCNN, the device comprising:

[0033] The acquisition module is used to acquire the vibration signal of the bearing of the motor to be tested;

[0034] The frequency domain feature extraction module is used to perform WPT transformation on the vibration signal of the motor bearing to be detected, and after determining the multi-layer wavelet packet decomposition, extract the frequency domain feature vector of the vibration signal of the motor bearing to be detected.

[0035] The diagnostic module is used to input the vibration signal of the motor bearing under test and the frequency domain feature vector of the vibration signal of the motor bearing under test into a pre-trained dual-branch fault diagnosis model, and the dual-branch fault diagnosis model outputs the fault classification result of the vibration signal of the motor bearing under test.

[0036] In a third aspect, a control device is provided, comprising a processor and a storage device, the storage device being adapted to store a plurality of program codes, the program codes being adapted to be loaded and run by the processor to perform the method described in any of the above-described technical solutions of the fault diagnosis method for motor bearings based on WPT-1DCNN.

[0037] In a fourth aspect, a computer-readable storage medium is provided, wherein a plurality of program codes are stored therein, the program codes being adapted to be loaded and run by a processor to perform the method described in any of the above-described technical solutions of the fault diagnosis method for motor bearings based on WPT-1DCNN.

[0038] The above-described technical solutions of the present invention have at least one or more of the following beneficial effects:

[0039] In implementing the technical solution of this invention, a fault diagnosis method for motor bearings based on WPT-1DCNN is proposed. This method includes: acquiring the vibration signal of the motor bearing to be tested; performing WPT transform on the vibration signal of the motor bearing to be tested, and after determining multi-layer wavelet packet decomposition, extracting the frequency domain feature vector of the vibration signal of the motor bearing to be tested; inputting the vibration signal of the motor bearing to be tested and the frequency domain feature vector of the vibration signal of the motor bearing to be tested into a pre-trained bi-branch fault diagnosis model, and the bi-branch fault diagnosis model outputting the fault classification result of the vibration signal of the motor bearing to be tested. This prediction method, by employing a bi-branch fault diagnosis model that integrates frequency domain feature vectors and time domain feature vectors, enables the model to effectively improve the accuracy of bearing fault diagnosis and has good generalization ability. Attached Figure Description

[0040] The disclosure of this invention will become more readily understood with reference to the accompanying drawings. It will be readily understood by those skilled in the art that these drawings are for illustrative purposes only and are not intended to limit the scope of protection of this invention. Furthermore, similar numbers in the drawings are used to denote similar components, wherein:

[0041] Figure 1 This is a schematic flowchart of the main steps of a fault diagnosis method for motor bearings based on WPT-1DCNN according to an embodiment of the present invention.

[0042] Figure 2 This is a schematic diagram of the main steps of step S102 according to an embodiment of the present invention;

[0043] Figure 3 This is a schematic diagram of the main steps of step S103 according to an embodiment of the present invention;

[0044] Figure 4 This is a schematic diagram of the 1DCNN structure;

[0045] Figure 5 This is a schematic diagram illustrating the principle of the self-attention mechanism layer;

[0046] Figure 6 This is a schematic diagram of the CWRU bearing test bench;

[0047] Figure 7 This is a schematic diagram of the accuracy variation curves of the training set and the test set according to an embodiment of the present invention;

[0048] Figure 8 This is a schematic diagram of the loss function variation curves of the training set and the test set according to an embodiment of the present invention;

[0049] Figure 9 This is a schematic diagram of the main structure of a control device according to an embodiment of the present invention. Detailed Implementation

[0050] Some embodiments of the present invention will now be described with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are merely illustrative of the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0051] In the description of this invention, "module" and "processor" can include hardware, software, or a combination of both. A module can include hardware circuitry, various suitable sensors, communication ports, memory, and may also include software components, such as program code, or a combination of software and hardware. A processor can be a central processing unit, microprocessor, image processor, digital signal processor, or any other suitable processor. The processor has data and / or signal processing capabilities. The processor can be implemented in software, in hardware, or a combination of both. Non-transitory computer-readable storage media includes any suitable medium capable of storing program code, such as magnetic disks, hard disks, optical disks, flash memory, read-only memory, random access memory, etc. The term "A and / or B" means all possible combinations of A and B, such as only A, only B, or A and B. The terms "at least one A or B" or "at least one of A and B" have a similar meaning to "A and / or B" and can include only A, only B, or A and B. The singular terms "a" or "this" can also include plural forms.

[0052] See appendix Figure 1 , Figure 1 This is a schematic flowchart illustrating the main steps of a motor bearing fault diagnosis method based on WPT-1DCNN according to an embodiment of the present invention. Figure 1 As shown, the fault diagnosis method for motor bearings based on WPT-1DCNN in this embodiment of the invention mainly includes the following steps S101-S103.

[0053] Step S101: Obtain the vibration signal of the bearing of the motor to be tested.

[0054] In this embodiment, under actual working conditions, the vibration sensor is set at the detection point of the motor bearing to be tested. When the vibration sensor receives the instruction to collect the vibration signal of the motor bearing, the one-dimensional vibration signal of the motor bearing to be tested is collected by the vibration sensor installed at the motor drive end.

[0055] Step S102: Perform WPT transform on the vibration signal of the motor bearing to be tested, and after determining the multi-layer wavelet packet decomposition, extract the frequency domain feature vector of the vibration signal of the motor bearing to be tested.

[0056] In this embodiment, the principle of WPT transform is to decompose the signal into multiple wavelet packets, each containing different frequency components, thereby enabling better analysis of the signal characteristics.

[0057] In order to better characterize the frequency domain feature vector of the vibration signal of the motor bearing to be tested, after verification by a large amount of experimental data, this embodiment of the invention selects to perform three-level wavelet packet decomposition on the vibration signal of the motor bearing to be tested, and obtains eight sub-frequency bands after decomposition of the vibration signal of the motor bearing to be tested.

[0058] In one embodiment of the present invention, such as Figure 2 As shown, step S102 further includes steps S1021-S1022:

[0059] Step S1021: Perform 3-level wavelet packet decomposition on the vibration signal of the motor bearing to be detected to obtain 8 sub-frequency bands after decomposition.

[0060] Step S1022: Determine the frequency domain feature vector H of the vibration signal of the motor bearing to be detected in 8*1 dimensions based on the product of the maximum amplitude of each of the 8 sub-frequency bands and the frequency corresponding to the maximum amplitude of each of the 8 sub-frequency bands. k ,for:

[0061]

[0062] Among them, A k The maximum amplitude value in each sub-band; f k The frequency corresponding to the maximum amplitude value in each sub-band; k=1,2,3,4,5,6,7,8.

[0063] In this embodiment, to eliminate redundant information and noise interference, the product of the maximum amplitude and its corresponding frequency in each sub-band is selected to characterize the most significant feature in each sub-band. This feature selection not only greatly saves computational resources but also yields more robust features, as expressed by the following formula:

[0064]

[0065] Among them, H k This represents the maximum amplitude A of a certain sub-band. k and its corresponding frequency f k The features obtained by multiplication are given, where the subscript k represents different frequency bands. In this embodiment, k = 1, 2, 3, 4, 5, 6, 7, 8. H is calculated within each frequency band. k Furthermore, they are combined and uniformly represented as an 8*1 dimensional frequency domain feature vector. This is the frequency domain feature vector of the vibration signal of the motor bearing to be tested.

[0066] Step S103: Input the vibration signal of the motor bearing to be tested and the frequency domain feature vector of the vibration signal of the motor bearing to be tested into the pre-trained dual-branch fault diagnosis model, and the dual-branch fault diagnosis model outputs the fault classification result of the vibration signal of the motor bearing to be tested.

[0067] In one embodiment of the present invention, such as Figure 3 As shown, step S103 may further include steps S1031-S1035:

[0068] Step S1031: Input the vibration signal of the motor bearing to be tested into the 1DCNN convolutional layer of the dual-branch fault diagnosis model. The 1DCNN convolutional layer outputs a 128*t-dimensional feature vector.

[0069] In a specific example, the dual-branch fault diagnosis model includes a 1DCNN convolutional layer, a multi-head self-attention mechanism layer, a dual-branch feature fusion layer, a fully connected layer, and a softmax layer. The 1DCNN convolutional layer is a one-dimensional convolutional neural network, and its structure is as follows: Figure 4 As shown, it includes an input layer, hidden layers, and one or more one-dimensional convolutional layers, one or more pooling layers, a fully connected layer, and an output layer. The one-dimensional vibration signal of the motor bearing to be detected is input into the input layer. After passing through the one-dimensional convolutional layer and the pooling layer, the output is the temporal feature extraction of the one-dimensional vibration signal of the motor bearing to be detected. The 1DCNN convolutional layer can automatically perform feature extraction through training and determine the specific number of one-dimensional convolutional layers and pooling layers included in the hidden layer to achieve the best effect of temporal feature extraction.

[0070] Therefore, the one-dimensional vibration signal of the motor bearing to be detected is input into the time-domain branch composed of 1DCNN convolutional layers, and the one-dimensional vibration signal X of the motor bearing to be detected is then processed. in Temporal feature extraction is performed, and after passing through a one-dimensional convolutional layer and a pooling layer, the extracted data features X are obtained. 1DCNN ,for:

[0071]

[0072] in, Indicates the first b represents the weight of the node corresponding to the layer; b is the bias.

[0073] Step S1032: Input the 128*t-dimensional feature vector into the multi-head self-attention mechanism layer of the dual-branch fault diagnosis model. The multi-head self-attention mechanism layer outputs the 128*1-dimensional time-domain feature vector K of the vibration signal of the motor bearing to be detected.

[0074] In a specific example, the self-attention mechanism layer is an attention mechanism that associates different positions within a single sequence. By focusing on all positions within the same sequence, it calculates the response at each position. Essentially, it highlights important information about the target object through a series of attention weights while suppressing irrelevant information, thus associating global information and efficiently capturing the connection between global and local information. A schematic diagram of the self-attention mechanism is shown below. Figure 5 As shown, by Figure 5 It can be seen that the basic steps of the self-attention mechanism are:

[0075] N sequences from a single sample An adaptive linear mapping is performed, transforming the vector into three vectors of length d. ;

[0076] vector The query matrix Q, index matrix K, and content matrix V are synthesized in parallel; the attention weights are obtained by calculating the product of Q and K. The calculation formula is as follows, where S is the set of weights among N sequences;

[0077]

[0078] Standardizing S yields The formula is as follows:

[0079]

[0080] Use the Softmax function to convert the weights into probabilistic form The formula is as follows:

[0081]

[0082] Obtain the weighted matrix Z:

[0083]

[0084] The entire process of self-attention can be summarized as follows:

[0085]

[0086] In this embodiment, after extracting the temporal features of the one-dimensional vibration signal of the motor bearing to be detected, the feature X output by the 1DCNN convolutional layer is... 1DCNN The input is fed into a multi-head self-attention mechanism layer for feature correlation analysis, and the final output is a 128*1 dimensional temporal feature vector K of the vibration signal of the motor bearing to be detected, which is:

[0087]

[0088] in, N is the length of the input sequence; , , The features X represent the output of the 1DCNN convolutional layer, respectively. 1DCNN The corresponding query matrix, index matrix, and content matrix.

[0089] Step S1033: Extract the frequency domain feature vector H of the vibration signal of the bearing of the motor to be tested from the 8*1 dimensional vector. k The time-domain feature vector K of the vibration signal of the motor bearing to be tested (128*1 dimensional) is input into the dual-branch feature fusion layer of the dual-branch fault diagnosis model. The dual-branch feature fusion layer outputs the time-frequency domain feature vector of the vibration signal of the motor bearing to be tested (136*1 dimensional).

[0090] In a specific example, the frequency domain feature vector and the time domain feature vector are fused to obtain a 136*1 dimensional time-frequency domain feature vector X. out ,for

[0091]

[0092] Among them, a m Let m be the m-th vector in the 136*1-dimensional time-frequency domain feature vector, where m = 1, 2, ..., 136.

[0093] Step S1034: Input the time-frequency domain feature vector of the vibration signal of the motor bearing to be tested (136*1 dimensional) into the fully connected layer of the dual-branch fault diagnosis model. The fully connected layer outputs the fault diagnosis signal of the vibration signal of the motor bearing to be tested (4*1 dimensional).

[0094] Step S1035: Input the fault diagnosis signal of the vibration signal of the motor bearing to be tested (4*1 dimensional) into the Softmax layer of the dual-branch fault diagnosis model. The Softmax layer outputs the fault classification result of the vibration signal of the motor bearing to be tested.

[0095] In this embodiment, the time-frequency domain feature vector of the vibration signal of the motor bearing under test (136*1 dimensional) is input into the fully connected layer of the dual-branch fault diagnosis model. The fault diagnosis signal of the vibration signal of the motor bearing under test (4*1 dimensional) is obtained through the fully connected layer. Then, the weight of each channel in the fault diagnosis signal of the vibration signal of the motor bearing under test (4*1 dimensional) is obtained through the Softmax layer. The weights are mapped to the (0,1) numerical space. The sum of the weights of each channel feature is 1. Each channel represents a bearing state, which includes four types: inner ring fault, outer ring fault, cage fault, and normal state. In a specific example, if the results are P=0.94 for outer ring fault, P=0.05 for inner ring fault, P=0.01 for cage fault, and P=0.00 for normal state, it means that the motor bearing under test is in the outer ring fault state.

[0096] In this embodiment, two branches are input to a pre-trained dual-branch fault diagnosis model. The first branch inputs an 8*1 dimensional frequency domain feature vector of the vibration signal of the motor bearing to be tested. The second branch inputs the vibration signal of the motor bearing to be tested. In the second branch, the vibration signal of the motor bearing to be tested passes through a 1DCNN convolutional layer and a multi-head self-attention mechanism layer in the dual-branch fault diagnosis model to extract a 128*1 dimensional temporal feature vector of the vibration signal of the motor bearing to be tested. Then, the 8*1 dimensional frequency domain feature vector of the vibration signal of the motor bearing to be tested input from the first branch and the 128*1 dimensional feature vector extracted from the second branch are combined. A 136*1 dimensional time-domain feature vector is input into the bi-branch feature fusion layer of the bi-branch fault diagnosis model for feature fusion. The bi-branch feature fusion layer outputs a 136*1 dimensional time-frequency domain feature vector of the vibration signal of the motor bearing to be tested. Then, the 136*1 dimensional time-frequency domain feature vector is input into the fully connected layer of the bi-branch fault diagnosis model. The fully connected layer outputs a 4*1 dimensional fault diagnosis signal of the vibration signal of the motor bearing to be tested. Finally, the 4*1 dimensional fault diagnosis signal is input into the Softmax layer of the bi-branch fault diagnosis model. The Softmax layer outputs the fault classification result of the vibration signal of the motor bearing to be tested.

[0097] In one embodiment of the present invention, before using the pre-trained dual-branch fault diagnosis model, the method further includes:

[0098] Step S100: Train the dual-branch fault diagnosis model to obtain a pre-trained dual-branch fault diagnosis model.

[0099] In one embodiment of the present invention, step S100 further includes steps S1001-S1005:

[0100] Step S1001: Obtain the dataset of the original vibration signal of the motor bearing, and divide the dataset into a training set and a test set according to a certain ratio.

[0101] In this embodiment, the dataset selected is bearing vibration data from the CWRU (Case Western Reserve University) bearing data center. The CWRU dataset uses SKF6205 bearings, with faults introduced through electrical discharge machining (EDM). The fault diameters are 0.18 mm, 0.36 mm, and 0.53 mm, respectively. Figure 6 As shown, the CWRU bearing test bench is powered by a 2-horsepower motor. Figure 6 Left side), torque sensor / decoder ( Figure 6 (Middle connection point), power tester () Figure 6 (right side) and control electronics ( Figure 6 The motor consists of components such as (not shown in the image). The rated output power of the motor is 2.2kW, and the sampling frequencies are 12kHz and 48kHz. In order to obtain bearing data for different faults, spark counting was used to manually process the inner ring, outer ring, and cage, which included four different types of vibration signals: normal state, inner ring fault, outer ring fault, and cage fault.

[0102] In one specific example, the experimental motor rotated at 1797 r / min, and the data was collected at a sampling frequency of 12 kHz. The vibration data in the experiment were collected using an accelerometer mounted on the drive end of the motor housing. In this embodiment, data collected from the SKF6205 bearing at the drive end was used, and the relevant bearing parameters are shown in Table 1.

[0103] Table 1

[0104]

[0105] In this example, motor bearing failures are categorized into four types: outer ring failure, inner ring failure, cage failure, and normal condition. The diameters of the three failure types are set to 0.007, 0.014, and 0.021 inches (1 inch = 25.4 mm), respectively. The bearings are then reinstalled on the motor, with the motor load set to 0, 1, 2, and 3 hp (1 hp = 735.50 W) to correspond to different motor speeds. Vibration acceleration signals of the bearings are collected using sensors, thus obtaining a sample dataset of bearings operating under various conditions. The relevant bearing failure information is shown in Table 2.

[0106] Table 2

[0107]

[0108] The collected dataset consists of four folders: one folder for normal data and three folders for faulty data. Each folder contains 1000 samples, for a total of 4000 samples. These samples are divided into training and test sets in an 8:2 ratio. Specifically, 800 samples from the normal dataset are used for training and 200 samples for testing; 800 samples from the inner ring fault dataset are used for training and 200 samples for testing; 800 samples from the outer ring fault dataset are used for training and 200 samples for testing; and 800 samples from the cage fault dataset are used for training and 200 samples for testing.

[0109] Step S1002: The dual-branch fault diagnosis model is trained using the training set. The training process includes: Step S10021: The original vibration signals of the motor bearings in the training set are subjected to WPT transform, and after determining the multi-layer wavelet packet decomposition, the frequency domain feature vector of the original vibration signals of the motor bearings in the training set is extracted; Step S10022: The frequency domain feature vector of the original vibration signals of the motor bearings in the training set and the original vibration signals of the motor bearings in the training set are input into the dual-branch fault diagnosis model being trained, and the dual-branch fault diagnosis model outputs the fault classification result of the original vibration signals of the motor bearings in the training set.

[0110] In this embodiment, the wavelet packet decomposition is determined to have 3 layers, and the 8*1 dimensional frequency domain feature vector corresponding to the original vibration signal of the motor bearing in the training set is extracted. , where A k训 To train the maximum amplitude value in each sub-band of the original vibration signal of the motor bearing; f k训 The frequency corresponding to the maximum amplitude value in each sub-band of the original vibration signal of the training motor bearing; k=1,2,3,4,5,6,7,8.

[0111] In a specific example, step S10022 further includes:

[0112] The vibration signal of the motor bearing in the training set is input into the 1DCNN convolutional layer of the dual-branch fault diagnosis model being trained. The 1DCNN convolutional layer outputs a 128*t-dimensional feature vector.

[0113] The 128*t-dimensional feature vector is input into the multi-head self-attention mechanism layer of the dual-branch fault diagnosis model being trained. The multi-head self-attention mechanism layer outputs a 128*1-dimensional time-domain feature vector K of the vibration signal of the motor bearing in the training set. 训 ;

[0114] The frequency domain feature vector H of the original vibration signal of the motor bearing in the 8*1 dimensional training set is obtained. k训The time-domain feature vector K of the original vibration signal of the motor bearing in the 128*1 dimensional training set. 训 The input is fed into the dual-branch feature fusion layer of the dual-branch fault diagnosis model being trained, and the dual-branch feature fusion layer outputs a 136*1-dimensional time-frequency domain feature vector of the original vibration signal of the motor bearing in the training set.

[0115] The time-frequency domain feature vector of the original vibration signal of the motor bearing in the training set of 136*1 dimensional is input into the fully connected layer of the dual-branch fault diagnosis model being trained, and the fully connected layer outputs the fault diagnosis signal of the original vibration signal of the motor bearing in the training set of 4*1 dimensional.

[0116] The fault diagnosis signal of the original vibration signal of the motor bearing in the 4*1 dimensional training set is input into the Softmax layer of the dual-branch fault diagnosis model being trained. The Softmax layer outputs the fault classification result of the original vibration signal of the motor bearing in the training set.

[0117] Step S1003: Compare the fault classification results of the original vibration signals of the motor bearings in the training set and the fault labels of the original vibration signals of the motor bearings in the training set with the output of the dual-branch fault diagnosis model, and calculate the loss function of the dual-branch fault diagnosis model.

[0118] In a specific example, fault types are labeled, for example, normal state is set to "0", outer ring fault is set to "1", inner ring fault is set to "2", and cage fault is set to "3". The fault labels corresponding to the original vibration signals in the training set are known. During the training of the two-branch fault diagnosis model, the fault classification results of the original vibration signals of the motor bearings in the training set output by the two-branch fault diagnosis model are compared with the fault labels of the original vibration signals of the motor bearings in the training set, and the loss function of the two-branch fault diagnosis model is calculated. When the loss function does not meet the condition of the model being optimal, the two-branch fault diagnosis model continues to be trained.

[0119] In one embodiment of the present invention, the loss function is the cross-entropy loss function.

[0120] Step S1004: When the loss function meets the requirements of model optimization, stop training the dual-branch fault diagnosis model.

[0121] As training iterates, the parameters of the two-branch fault diagnosis model are continuously updated, and the model becomes increasingly optimized. When the loss function meets the requirements for model optimization, the training process ends, and training of the two-branch fault diagnosis model is stopped.

[0122] Step S1005: Test the trained fault diagnosis model using the test set. After the test is completed, the trained dual-branch fault diagnosis model is obtained.

[0123] In a specific example, the design of the two-branch fault diagnosis model is adjusted based on its performance on the training set to obtain the final two-branch fault diagnosis model. The trained two-branch fault diagnosis model is then tested on the test set to obtain the model's accuracy, which is then compared with other methods. Experimental results show that the two-branch fault diagnosis model proposed in this embodiment can learn the inherent correspondence between data and classification labels through the training set, and thereby infer the bearing fault location corresponding to the test data.

[0124] In one embodiment of the present invention, the 1DCNN convolutional layer in the trained dual-branch fault diagnosis model includes an input layer, a first convolutional layer, a first pooling layer, a second convolutional layer, and a second pooling layer; wherein,

[0125] The first convolutional layer is set to 1 input channel, 64 output channels, stride 1, and edge padding 4 layers; the first pooling layer is a max pooling layer; the second convolutional layer is set to 64 input channels, 128 output channels, stride 1, and edge padding 2 layers; the second pooling layer is an average pooling layer.

[0126] In this embodiment, after training the dual-branch fault diagnosis model using a training set and validating it using a test set, the parameter settings for each layer of the trained dual-branch fault diagnosis model are as follows:

[0127] Perform a three-level wavelet packet transform, and set the parameter for the number of levels of the corresponding wavelet packet decomposition to 3;

[0128] The 1DCNN convolutional layer is configured with two 1D convolutional layers and two pooling layers. The first convolutional layer is configured with 1 input channel, 64 output channels, a stride of 1, and 4 layers of edge padding. The first pooling layer is a max-time pooling layer. The second convolutional layer is configured with 64 input channels, 128 output channels, a stride of 1, and 2 layers of edge padding. The second pooling layer is an average-time pooling layer.

[0129] The vibration signal data after passing through the 1DCNN convolutional layer is divided into three parts and input into the multi-head attention mechanism layer to form Q, K, V matrices. After a series of calculations and processing, the final output is a 128*1-dimensional feature.

[0130] Dropout layer. To effectively prevent overfitting and speed up training, the Dropout coefficient is set to 0.5.

[0131] Batch Normalization layer. Normalizes the input data, which can effectively improve training speed. The parameters set for each convolutional layer are: Batch_size is set to 64; momentum used for dynamic mean and dynamic variance is 0.1; to ensure numerical stability (the denominator cannot approach or take 0), the value eps is added to the denominator to be 0.00001.

[0132] Loss function. The loss function used in this embodiment is the cross-entropy loss function. The cross-entropy loss function has been described in great detail in the prior art, and will not be repeated here.

[0133] Adam optimizer. As an adaptive algorithm, Adam can further improve the training speed of the model. The model proposed in this embodiment sets the parameters of Adam as follows: =0.99; =0.999.

[0134] To more clearly illustrate the parameters of each layer of the 1DCNN convolutional layer used in this embodiment, Table 3 details the parameter settings of the proposed dual-branch fault diagnosis model, including layer type, convolutional kernel size, activation function, and dimensions of input and output data.

[0135] Table 3

[0136] Layer type kernel size Activation function Input Dimensions Output Dimension Input layer - - - (1,2048) One-dimensional convolution 9 ReLU (1,2048) (64,2048) Batch Norm layer 1 - - (64,2048) (64,2048) Dropout layer 1 - - (64,2048) (64,2048) One-dimensional convolution 5 ReLU (64,2048) (128,2048) Batch Norm layer 2 - - (128,2048) (128,2048) Dropout layer 2 - - (128,2048) (128,2048) Self-attention mechanism layer - - (128,2048) 128 Time pooling layer - - (128,2048) 128 Fully connected layer - Softmax 4 4

[0137] After setting the dual-branch fault diagnosis model based on the above parameters, training the model began. The number of iterations was set to 20, and the batch size was set to 64, meaning 64 samples were fed into the model at once for training. The optimization process of the model was recorded, and the changes in accuracy and loss function with the number of iterations (Epochs) during the model training process were recorded. Figure 7 and Figure 8 middle. Figure 7 Curve 1 represents the accuracy of the training set, and curve 2 represents the accuracy of the test set. Figure 8 Curve 3 represents the loss function for the training set, and curve 4 represents the loss function for the test set. (In summary...) Figure 7 and Figure 8 As can be seen, due to the influence of Dropout, the accuracy curve of the test set exhibits a spiral upward trend, while the loss function curve exhibits a spiral downward trend. From Figure 7It is evident that the training set stabilized after minor fluctuations and changes, but the accuracy of the test set fluctuated significantly in the first 10 iterations, increasing from 0.9 to over 0.98. From the 11th iteration onwards, although the accuracy fluctuated somewhat, it remained within the range of 0.98-0.99, ultimately achieving a result higher than 0.99. Through analysis of... Figure 8 Observations show that the loss function generally exhibits a downward trend. Although there are peaks on the test set at the 7th, 12th, 15th, and 18th iterations due to Dropout, the loss function remains below 0.07. For the training set, the loss function has remained around 0 since the second iteration, indicating that the network has stabilized by the second iteration. Overall, observations... Figure 7 and Figure 8 As can be seen from the curve changes, the model proposed in this embodiment has good stability.

[0138] It's important to note that judging a model's quality isn't about focusing on the accuracy on the training set, but rather on whether its performance on the test set approaches the level of the training set. Figure 7 and 8 As can be seen, the test set accuracy and the final value of the loss function both reached a high level close to that of the training set, fully demonstrating that the model proposed in this embodiment has excellent generalization ability, that is, it has good classification ability even for unfamiliar data. Furthermore, it can be seen that our two-branch model achieved extremely high accuracy after a very short number of iterations. Then the training curve gradually flattens out, and the training gains decrease accordingly; therefore, we stopped training after 20 epochs (iterations).

[0139] Based on steps S101-S103 above, a fault diagnosis method for motor bearings based on WPT-1DCNN is proposed. This method includes: acquiring the vibration signal of the motor bearing to be tested; performing WPT transform on the vibration signal of the motor bearing to be tested, and after determining multi-layer wavelet packet decomposition, extracting the frequency domain feature vector of the vibration signal of the motor bearing to be tested; inputting the vibration signal of the motor bearing to be tested and the frequency domain feature vector of the vibration signal of the motor bearing to be tested into a pre-trained bi-branch fault diagnosis model, and the bi-branch fault diagnosis model outputting the fault classification result of the vibration signal of the motor bearing to be tested. This prediction method, by employing a bi-branch fault diagnosis model that integrates frequency domain feature vectors and time domain feature vectors, effectively improves the accuracy of bearing fault diagnosis and has good generalization ability.

[0140] It should be noted that although the steps in the above embodiments are described in a specific order, those skilled in the art will understand that in order to achieve the effects of the present invention, different steps do not necessarily have to be executed in such an order. They can be executed simultaneously (in parallel) or in other orders, and these variations are all within the scope of protection of the present invention.

[0141] Furthermore, the present invention also provides a fault diagnosis device for motor bearings based on WPT-1DCNN.

[0142] See appendix Figure 9 , Figure 9 This is a main structural block diagram of a motor bearing fault diagnosis device based on WPT-1DCNN according to an embodiment of the present invention. Figure 9 As shown, the motor bearing fault diagnosis device based on WPT-1DCNN in this embodiment of the invention mainly includes an acquisition module 11, a frequency domain feature extraction module 12, and a diagnosis module 13. In some embodiments, one or more of the acquisition module 11, the frequency domain feature extraction module 12, and the diagnosis module 13 can be combined into a single module. In some embodiments, the acquisition module 11 can be configured to acquire the vibration signal of the motor bearing to be tested. The frequency domain feature extraction module 12 can be configured to perform WPT transform on the vibration signal of the motor bearing to be tested, and after determining the multi-layer wavelet packet decomposition, extract the frequency domain feature vector of the vibration signal of the motor bearing to be tested. The diagnosis module 13 can be configured to input the vibration signal of the motor bearing to be tested and the frequency domain feature vector of the vibration signal of the motor bearing to be tested into a pre-trained dual-branch fault diagnosis model, and the dual-branch fault diagnosis model outputs the fault classification result of the vibration signal of the motor bearing to be tested.

[0143] In one implementation, a description of the specific functions can be found in steps S101-S103.

[0144] The aforementioned fault diagnosis device for motor bearings based on WPT-1DCNN is used to perform... Figure 1 The embodiments of the motor bearing fault diagnosis method based on WPT-1DCNN shown are similar in technical principle, technical problem solved and technical effect. Those skilled in the art can clearly understand that, for the sake of convenience and brevity, the specific working process and related descriptions of the motor bearing fault diagnosis device based on WPT-1DCNN can be found in the embodiments of the motor bearing fault diagnosis method based on WPT-1DCNN, and will not be repeated here.

[0145] Those skilled in the art will understand that all or part of the processes in the method of the above embodiment of the present invention can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable file, or some intermediate form. The computer-readable storage medium can include any entity or device capable of carrying the computer program code, a medium, a USB flash drive, a portable hard drive, a magnetic disk, an optical disk, a computer memory, a read-only memory, a random access memory, an electrical carrier signal, a telecommunication signal, and a software distribution medium, etc. It should be noted that the content included in the computer-readable storage medium can be appropriately added or removed according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, the computer-readable storage medium does not include electrical carrier signals and telecommunication signals.

[0146] Furthermore, the present invention also provides a control device. In one embodiment of the control device according to the present invention, the control device includes a processor and a storage device. The storage device can be configured to store a program for executing the WPT-1DCNN-based fault diagnosis method for motor bearings described in the above-described method embodiments. The processor can be configured to execute the program in the storage device, which includes, but is not limited to, the program for executing the WPT-1DCNN-based fault diagnosis method for motor bearings described in the above-described method embodiments. For ease of explanation, only the parts related to the embodiments of the present invention are shown; for specific technical details not disclosed, please refer to the method section of the embodiments of the present invention. This control device can be a control device device comprising various electronic devices.

[0147] Furthermore, the present invention also provides a computer-readable storage medium. In one embodiment of the computer-readable storage medium according to the present invention, the computer-readable storage medium can be configured to store a program that executes the WPT-1DCNN-based fault diagnosis method for motor bearings described in the above-described method embodiments. This program can be loaded and run by a processor to implement the above-described WPT-1DCNN-based fault diagnosis method for motor bearings. For ease of explanation, only the parts related to the embodiments of the present invention are shown; for specific technical details not disclosed, please refer to the method section of the embodiments of the present invention. The computer-readable storage medium can be a storage device comprising various electronic devices. Optionally, in the embodiments of the present invention, the computer-readable storage medium is a non-transitory computer-readable storage medium.

[0148] Furthermore, it should be understood that since the various modules are only provided to illustrate the functional units of the device of the present invention, the physical devices corresponding to these modules may be the processor itself, or a part of the processor's software, a part of its hardware, or a combination of software and hardware. Therefore, the number of modules shown in the figures is merely illustrative.

[0149] Those skilled in the art will understand that the various modules in the device can be adaptively split or combined. Such splitting or combining of specific modules will not cause the technical solution to deviate from the principles of the present invention; therefore, the technical solutions after splitting or combining will fall within the protection scope of the present invention.

[0150] The technical solution of the present invention has been described above with reference to the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after such changes or substitutions will all fall within the scope of protection of the present invention.

Claims

1. A method for fault diagnosis of motor bearing based on WPT-1DCNN, characterized in that, The method comprises the following steps: obtaining a vibration signal of a motor bearing to be detected; performing WPT transformation on the vibration signal of the motor bearing to be detected, and determining a frequency domain feature vector of the vibration signal of the motor bearing to be detected after multi-layer wavelet packet decomposition; inputting the vibration signal of the motor bearing to be detected and the frequency domain feature vector of the vibration signal of the motor bearing to be detected into a pre-trained double-branch fault diagnosis model, and outputting a fault classification result of the vibration signal of the motor bearing to be detected by the double-branch fault diagnosis model. The WPT transformation is performed on the vibration signal of the motor bearing to be detected, and after multi-layer wavelet packet decomposition, the frequency domain feature vector of the vibration signal of the motor bearing to be detected is extracted, comprising: performing 3-layer wavelet packet decomposition on the vibration signal of the motor bearing to be detected to obtain 8 sub-frequency bands after decomposition; determining the 8*1-dimensional frequency domain feature vector H of the vibration signal of the motor bearing to be detected according to the product of the maximum amplitude of each sub-frequency band in the 8 sub-frequency bands and the frequency corresponding to the maximum amplitude of each sub-frequency band in the 8 sub-frequency bands k For: Wherein, A k is the maximum amplitude in each sub-frequency band; f k is the frequency corresponding to the maximum amplitude in each sub-frequency band; k=1, 2, 3, 4, 5, 6, 7, 8; The vibration signal of the motor bearing to be detected and the frequency domain feature vector of the vibration signal of the motor bearing to be detected are input to the pre-trained double-branch fault diagnosis model, and the double-branch fault diagnosis model outputs the fault classification result of the vibration signal of the motor bearing to be detected, including: inputting the vibration signal of the motor bearing to be detected to the 1DCNN convolution layer of the double-branch fault diagnosis model, and the 1DCNN convolution layer outputs a feature vector of 128*t dimensions; inputting the feature vector of 128*t dimensions to the multi-head self-attention mechanism layer of the double-branch fault diagnosis model, and the multi-head self-attention mechanism layer outputs a time domain feature vector K of the vibration signal of the motor bearing to be detected of 128*1 dimensions; inputting the frequency domain feature vector H of the vibration signal of the motor bearing to be detected of 8*1 dimensions to the double-branch feature fusion layer of the double-branch fault diagnosis model, and the double-branch feature fusion layer outputs a time-frequency domain feature vector of the vibration signal of the motor bearing to be detected of 136*1 dimensions; inputting the time-frequency domain feature vector of the vibration signal of the motor bearing to be detected of 136*1 dimensions to the full connection layer of the double-branch fault diagnosis model, and the full connection layer outputs a fault diagnosis signal of the vibration signal of the motor bearing to be detected of 4*1 dimensions; inputting the fault diagnosis signal of the vibration signal of the motor bearing to be detected of 4*1 dimensions to the Softmax layer of the double-branch fault diagnosis model, and the Softmax layer outputs the fault classification result of the vibration signal of the motor bearing to be detected. k The vibration signal of the motor bearing to be detected and the frequency domain feature vector of the vibration signal of the motor bearing to be detected are input to the pre-trained double-branch fault diagnosis model, and the double-branch fault diagnosis model outputs the fault classification result of the vibration signal of the motor bearing to be detected, including: inputting the vibration signal of the motor bearing to be detected to the 1DCNN convolution layer of the double-branch fault diagnosis model, and the 1DCNN convolution layer outputs a feature vector of 128*t dimensions; inputting the feature vector of 128*t dimensions to the multi-head self-attention mechanism layer of the double-branch fault diagnosis model, and the multi-head self-attention mechanism layer outputs a time domain feature vector K of the vibration signal of the motor bearing to be detected of 128*1 dimensions; inputting the frequency domain feature vector H of the vibration signal of the motor bearing to be detected of 8*1 dimensions to the double-branch feature fusion layer of the double-branch fault diagnosis model, and the double-branch feature fusion layer outputs a time-frequency domain feature vector of the vibration signal of the motor bearing to be detected of 136*1 dimensions; inputting the time-frequency domain feature vector of the vibration signal of the motor bearing to be detected of 136*1 dimensions to the full connection layer of the double-branch fault diagnosis model, and the full connection layer outputs a fault diagnosis signal of the vibration signal of the motor bearing to be detected of 4*1 dimensions; inputting the fault diagnosis signal of the vibration signal of the motor bearing to be detected of 4*1 dimensions to the Softmax layer of the double-branch fault diagnosis model, and the Softmax layer outputs the fault classification result of the vibration signal of the motor bearing to be detected.

2. The WPT-1D CNN based motor bearing fault diagnosis method of claim 1, wherein, The method further comprises the following steps: training the double-branch fault diagnosis model to obtain the pre-trained double-branch fault diagnosis model.

3. The WPT-1D CNN based motor bearing fault diagnosis method of claim 2, wherein, The training of the double-branch fault diagnosis model to obtain the pre-trained double-branch fault diagnosis model comprises the following steps: obtaining a data set of original vibration signals of motor bearings, and dividing the data set into a training set and a test set at a certain ratio; training the double-branch fault diagnosis model from the training set, and the training process comprises the following steps: performing WPT transformation on the original vibration signals of the motor bearings in the training set, and determining a frequency domain feature vector of the original vibration signals of the motor bearings in the training set after multi-layer wavelet packet decomposition; inputting the frequency domain feature vector of the original vibration signals of the motor bearings in the training set and the original vibration signals of the motor bearings in the training set into the double-branch fault diagnosis model being trained, and outputting a fault classification result of the original vibration signals of the motor bearings in the training set by the double-branch fault diagnosis model; comparing the fault classification result of the original vibration signals of the motor bearings in the training set output by the double-branch fault diagnosis model with a fault label of the original vibration signals of the motor bearings in the training set, and calculating a loss function of the double-branch fault diagnosis model; stopping the training of the double-branch fault diagnosis model when the loss function meets the requirement of the optimal model; testing the trained fault diagnosis model from the test set, and obtaining the pre-trained double-branch fault diagnosis model after the testing is completed.

4. The WPT-1D CNN based motor bearing fault diagnosis method of claim 3, wherein, The loss function is a cross-entropy loss function.

5. The WPT-1D CNN based motor bearing fault diagnosis method of claim 3, wherein, In the 1DCNN convolution layer of the pre-trained double-branch fault diagnosis model, an input layer, a first convolution layer, a first pooling layer, a second convolution layer, and a second pooling layer are included, wherein the first convolution layer is set to have 1 input channel, 64 output channels, a step length of 1, and an edge padding of 4 layers; the first pooling layer is a maximum time pooling layer; the second convolution layer is set to have 64 input channels, 128 output channels, a step length of 1, and an edge padding of 2 layers; and the second pooling layer is an average time pooling layer.

6. A WPT-1DCNN-based motor bearing fault diagnosis device, characterized in that, The method comprises the following steps: obtaining a vibration signal of a motor bearing to be detected; The frequency domain feature extraction module is configured to perform WPT transformation on the vibration signal of the motor bearing to be detected, and determine a frequency domain feature vector of the vibration signal of the motor bearing to be detected after multi-layer wavelet packet decomposition; the frequency domain feature vector of the vibration signal of the motor bearing to be detected after the WPT transformation and the multi-layer wavelet packet decomposition includes: performing 3-layer wavelet packet decomposition on the vibration signal of the motor bearing to be detected to obtain 8 sub-frequency bands after the decomposition; determining an 8*1-dimensional frequency domain feature vector H of the vibration signal of the motor bearing to be detected according to the product of the maximum amplitude of each of the 8 sub-frequency bands and the frequency corresponding to the maximum amplitude of each of the 8 sub-frequency bands k , wherein A , wherein A k is the maximum amplitude in each sub-frequency band; f k is the frequency corresponding to the maximum amplitude in each sub-frequency band; k = 1, 2, 3, 4, 5, 6, 7, 8 The diagnostic module is used for inputting the vibration signal of the motor bearing to be detected and the frequency domain feature vector of the vibration signal of the motor bearing to be detected into a pre-trained double-branch fault diagnosis model, and the double-branch fault diagnosis model outputs a fault classification result of the vibration signal of the motor bearing to be detected. The inputting of the vibration signal of the motor bearing to be detected and the frequency domain feature vector of the vibration signal of the motor bearing to be detected into the pre-trained double-branch fault diagnosis model and the outputting of the fault classification result of the vibration signal of the motor bearing to be detected by the double-branch fault diagnosis model include: inputting the vibration signal of the motor bearing to be detected into a 1DCNN convolution layer of the double-branch fault diagnosis model, and the 1DCNN convolution layer outputs a 128*t-dimensional feature vector; inputting the 128*t-dimensional feature vector into a multi-head self-attention mechanism layer of the double-branch fault diagnosis model, and the multi-head self-attention mechanism layer outputs a 128*1-dimensional time domain feature vector K of the vibration signal of the motor bearing to be detected; inputting an 8*1-dimensional frequency domain feature vector H of the vibration signal of the motor bearing to be detected into a double-branch feature fusion layer of the double-branch fault diagnosis model, and the double-branch feature fusion layer outputs a 136*1-dimensional time-frequency domain feature vector of the vibration signal of the motor bearing to be detected; inputting the 136*1-dimensional time-frequency domain feature vector of the vibration signal of the motor bearing to be detected into a full connection layer of the double-branch fault diagnosis model, and the full connection layer outputs a 4*1-dimensional fault diagnosis signal of the vibration signal of the motor bearing to be detected; inputting the 4*1-dimensional fault diagnosis signal of the vibration signal of the motor bearing to be detected into a Softmax layer of the double-branch fault diagnosis model, and the Softmax layer outputs a fault classification result of the vibration signal of the motor bearing to be detected. k The inputting of the vibration signal of the motor bearing to be detected and the frequency domain feature vector of the vibration signal of the motor bearing to be detected into the pre-trained double-branch fault diagnosis model and the outputting of the fault classification result of the vibration signal of the motor bearing to be detected by the double-branch fault diagnosis model include: inputting the vibration signal of the motor bearing to be detected into a 1DCNN convolution layer of the double-branch fault diagnosis model, and the 1DCNN convolution layer outputs a 128*t-dimensional feature vector; inputting the 128*t-dimensional feature vector into a multi-head self-attention mechanism layer of the double-branch fault diagnosis model, and the multi-head self-attention mechanism layer outputs a 128*1-dimensional time domain feature vector K of the vibration signal of the motor bearing to be detected; inputting an 8*1-dimensional frequency domain feature vector H of the vibration signal of the motor bearing to be detected into a double-branch feature fusion layer of the double-branch fault diagnosis model, and the double-branch feature fusion layer outputs a 136*1-dimensional time-frequency domain feature vector of the vibration signal of the motor bearing to be detected; inputting the 136*1-dimensional time-frequency domain feature vector of the vibration signal of the motor bearing to be detected into a full connection layer of the double-branch fault diagnosis model, and the full connection layer outputs a 4*1-dimensional fault diagnosis signal of the vibration signal of the motor bearing to be detected; inputting the 4*1-dimensional fault diagnosis signal of the vibration signal of the motor bearing to be detected into a Softmax layer of the double-branch fault diagnosis model, and the Softmax layer outputs a fault classification result of the vibration signal of the motor bearing to be detected.

7. A control device comprising a processor and a storage device, the storage device being adapted to store a plurality of program codes, characterized in that, The program code is adapted to be loaded and run by the processor to execute the WPT-1DCNN-based fault diagnosis method of the motor bearing according to any one of claims 1 to 5.

8. A computer readable storage medium having stored therein a plurality of program codes, characterized in that, The program code is adapted to be loaded and run by the processor to execute the WPT-1DCNN-based fault diagnosis method of the motor bearing according to any one of claims 1 to 5.

Images