Rolling bearing fault diagnosis method, device, equipment and medium based on multiple algorithms

Abstract

The present invention discloses a rolling bearing fault diagnosis method, apparatus, device, and medium based on multiple algorithms, relating to the technical field of mechanical fault diagnosis. The method comprises: obtaining and processing vibration signals of a target rolling bearing of unknown fault type under different loads to obtain a two-dimensional spectrogram of the target rolling bearing vibration signal; inputting the two-dimensional spectrogram into a rolling bearing fault diagnosis model to obtain fault diagnosis data for the target rolling bearing; wherein the rolling bearing fault diagnosis model comprises an input module, an implicit module, and an output module; the implicit module comprises an improved convolutional layer, an improved pooling layer, and a fully connected layer connected in sequence; the improved convolutional layer is a convolutional layer embedded in a compression-excitation network; and the improved pooling layer is a pooling layer embedded in a width learning algorithm. The present invention can better adapt to complex and changing industrial scenarios, and has higher diagnostic accuracy and stronger generalization capability.

Description

Technical Field

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

[0002] Rotating machinery and equipment are a vital component of the manufacturing industry. In the context of intelligent manufacturing, rotating machinery and equipment are developing towards intelligence, informationization, and automation. Since rolling bearings in rotating machinery and equipment operate continuously at high speeds during production, they are often accompanied by wear and aging, which is a contributing factor to safety accidents. Therefore, rolling bearing fault diagnosis is crucial to ensuring the reliable operation of rotating machinery and equipment. Rolling bearing fault diagnosis techniques include analyzing signals such as temperature, vibration, ultrasound, and electrostatic discharge. Vibration signals are the most commonly used for signal analysis due to their ease of acquisition.

[0003] Currently, the key steps in data-driven intelligent fault diagnosis methods are fault feature extraction and fault state classification. These methods typically use shallow machine learning algorithms such as support vector machines, artificial neural networks, and random forests to identify fault states. However, these methods rely on specific domain knowledge and are unable to learn deep features with insufficient generalization capabilities, making them incapable of handling scenarios with high integration and complex operating conditions.

[0004] In recent years, deep learning algorithms have been widely used in the field of mechanical fault diagnosis due to their powerful automatic learning capabilities and high classification accuracy. Deep learning algorithms can better adapt to large amounts of high-dimensional, heterogeneous data, and the nonlinear processing units within their deep architectures enable more accurate feature extraction and abstract modeling. Convolutional Neural Networks (CNNs) are particularly well-represented in the field of rolling bearing fault diagnosis due to their unique features, including translation-invariant classification, weight sharing, and convolutional computation. Summary of the Invention

[0005] In view of this, the present invention provides a rolling bearing fault diagnosis method, device, equipment and medium based on multiple algorithms, which can better adapt to complex and changeable industrial scenarios and have higher diagnostic accuracy and stronger generalization ability.

[0006] To achieve the above object, the present invention provides the following solutions:

[0007] In a first aspect, the present invention provides a rolling bearing fault diagnosis method based on multiple algorithms, comprising:

[0008] Obtain vibration signals of target rolling bearings with unknown fault types under different load conditions;

[0009] Processing the target rolling bearing vibration signal to obtain a two-dimensional spectrum diagram of the target rolling bearing vibration signal;

[0010] Inputting the two-dimensional stereo frequency spectrum of the target rolling bearing vibration signal into a rolling bearing fault diagnosis model to obtain fault diagnosis data of the target rolling bearing; the fault diagnosis data includes the fault type and the fault location;

[0011] In which, the rolling bearing fault diagnosis model is obtained by training an improved convolutional neural network with sample data; the sample data includes a two-dimensional stereo frequency spectrum of a sample rolling bearing vibration signal and a corresponding fault diagnosis data label; the improved convolutional neural network includes an input module, an implicit module and an output module; the implicit module includes an improved convolutional layer, an improved pooling layer and a fully connected layer connected in sequence; the improved convolutional layer is a convolutional layer embedded with a compression-excitation network; the improved pooling layer is a pooling layer embedded with a width learning algorithm.

[0012] In a second aspect, the present invention provides a rolling bearing fault diagnosis device based on multiple algorithms, comprising:

[0013] A vibration signal acquisition module, used to acquire vibration signals of target rolling bearings with unknown fault types under different load conditions;

[0014] a two-dimensional stereo spectrum diagram determination module, configured to process the target rolling bearing vibration signal to obtain a two-dimensional stereo spectrum diagram of the target rolling bearing vibration signal;

[0015] a fault diagnosis data determination module, configured to input the two-dimensional stereo frequency spectrum of the target rolling bearing vibration signal into a rolling bearing fault diagnosis model to obtain fault diagnosis data of the target rolling bearing; the fault diagnosis data includes a fault type and a fault location;

[0016] In which, the rolling bearing fault diagnosis model is obtained by training an improved convolutional neural network with sample data; the sample data includes a two-dimensional stereo frequency spectrum of a sample rolling bearing vibration signal and a corresponding fault diagnosis data label; the improved convolutional neural network includes an input module, an implicit module and an output module; the implicit module includes an improved convolutional layer, an improved pooling layer and a fully connected layer connected in sequence; the improved convolutional layer is a convolutional layer embedded with a compression-excitation network; the improved pooling layer is a pooling layer embedded with a width learning algorithm.

[0017] In a third aspect, the present invention provides an electronic device comprising a memory and a processor, wherein the memory is used to store a computer program, and the processor runs the computer program to enable the electronic device to execute the rolling bearing fault diagnosis method based on multiple algorithms according to the first aspect.

[0018] In a fourth aspect, the present invention provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the rolling bearing fault diagnosis method based on multiple algorithms described in the first aspect.

[0019] According to the specific embodiments provided by the present invention, the present invention discloses the following technical effects:

[0020] 1. The effectiveness of training models by combining convolutional neural networks, compression-excitation networks, and width learning can not only extract deeper features from the channel dimension, but also retain the characteristics of width learning that uses ridge regression to effectively update the model, thereby better adapting to complex and changing industrial scenarios.

[0021] 2. An attention mechanism is introduced in the feature extraction process to adaptively allocate the channel weights of the feature image, focus on important features, reduce the interference of redundant features and noise, improve the accuracy of rolling bearing diagnosis and the generalization ability of the method, and improve the robustness and stability of fault diagnosis.

[0022] 3. Using the two-dimensional spectrum diagram as the model input, compared with several typical fault diagnosis methods, this method has higher diagnostic accuracy for rolling bearing vibration signals. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0024] Figure 1 Schematic diagram of the flow of the rolling bearing fault diagnosis method based on multiple algorithms of the present invention;

[0025] Figure 2 Schematic diagram of the training process of the rolling bearing fault diagnosis model of the present invention;

[0026] Figure 3 Schematic diagram of the conversion from a one-dimensional vibration signal to a two-dimensional spectrum diagram of the present invention;

[0027] Figure 4 It is a two-dimensional spectrum diagram of the vibration signal of the rolling bearing sample of the present invention;

[0028] Figure 5 Schematic diagram of the implementation process of the improved convolutional layer of the present invention;

[0029] Figure 6The histogram of the comparison between SECNN-BLS of the present invention and other eight fault diagnosis models on the MFPT dataset;

[0030] Figure 7 This is a schematic structural diagram of a rolling bearing fault diagnosis device based on multiple algorithms of the present invention. DETAILED DESCRIPTION

[0031] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0032] The attention mechanism is a key concept in deep learning algorithms. It adaptively extracts features based on the importance of different feature information, thereby improving the contribution of important features in the model. While deep learning algorithms possess powerful feature extraction capabilities, they are difficult to update on demand due to high training costs, limitations on sample data volume, and fixed structures. The emergence of the Broad Learning System (BLS) has significantly improved these shortcomings. First proposed in 2018, BLS consists of an input layer, a feature layer, an enhancement layer, and an output layer. It enriches deep learning algorithms and expands the structure of neurons in convolutional neural networks. Incremental learning is the core of BLS, and its advantage lies in its ability to cope with ever-increasing data without requiring recalculation. Furthermore, BLS can efficiently update the model, and its simple linear structure reduces computational complexity and workload, making it easy to process small batches of data. The simple linear structure makes BLS easy to reshape and integrate. Since its introduction, numerous BLS variants have been derived and widely applied in many fields. However, in constructing the feature layer, BLS uses randomly generated weights and biases, which can easily lead to insufficient feature extraction from samples, affecting the accuracy of fault diagnosis.

[0033] In view of this, the present invention provides a rolling bearing fault diagnosis method, device, equipment and medium based on CNN, BLS and Squeeze-and-Excitation Networks (SENet), which can better adapt to complex and changeable industrial scenarios and have higher diagnostic accuracy and stronger generalization ability.

[0034] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0035] Example 1

[0036] like Figure 1 As shown, an embodiment of the present invention provides a rolling bearing fault diagnosis method based on multiple algorithms, which includes the following steps.

[0037] Step 100: Obtain vibration signals of a target rolling bearing with unknown fault types under different load conditions.

[0038] Step 200: Process the target rolling bearing vibration signal to obtain a two-dimensional spectrum diagram of the target rolling bearing vibration signal.

[0039] Step 300: Input the two-dimensional spectrum diagram of the target rolling bearing vibration signal into a rolling bearing fault diagnosis model to obtain fault diagnosis data of the target rolling bearing; the fault diagnosis data includes the fault type and the fault location.

[0040] The training process of the rolling bearing fault diagnosis model is as follows: Figure 2 As shown, the rolling bearing fault diagnosis model is obtained by training an improved convolutional neural network with sample data; the sample data includes a two-dimensional stereo frequency spectrum of a sample rolling bearing vibration signal and a corresponding fault diagnosis data label; the improved convolutional neural network includes an input module, an implicit module and an output module; the implicit module includes an improved convolutional layer, an improved pooling layer and a fully connected layer connected in sequence; the improved convolutional layer is a convolutional layer embedded with a compression-excitation network; the improved pooling layer is a pooling layer embedded with a width learning algorithm.

[0041] As a preferred implementation, step 200 described in the embodiment of the present invention specifically includes:

[0042] The target rolling bearing vibration signal is processed by a short-time Fourier transform algorithm to obtain a two-dimensional spectrum diagram of the target rolling bearing vibration signal.

[0043] As a preferred implementation, the process of determining the sample data according to the embodiment of the present invention is as follows:

[0044] S1: Convert the time series signal in the sample rolling bearing dataset into the original vibration signal in the time domain.

[0045] S2: Set the sampling parameters of the sliding window overlapping sampling method according to the sampling frequency and load conditions of the sample rolling bearing data set.

[0046] S3: Use the sliding window overlapping sampling method to perform multiple short-time Fourier transforms (STFTs) on the original vibration signal to obtain the following Figure 3The two-dimensional spectrum diagram of the sample rolling bearing vibration signal is shown, and based on the two-dimensional spectrum diagram of the sample rolling bearing vibration signal, a two-dimensional stereo spectrum diagram of the sample rolling bearing vibration signal is constructed.

[0047] The calculation process of STFT can be expressed as:

[0048]

[0049] Where z(t) is the original vibration signal, g(tu) is the window function, and t is time.

[0050] S4: Determine the fault diagnosis data label corresponding to the original vibration signal as the fault diagnosis data label corresponding to the two-dimensional spectrum diagram of the sample rolling bearing vibration signal.

[0051] Figure 4 is the two-dimensional spectrum diagram of the sample rolling bearing vibration signal; Figure 4 (a) is the two-dimensional spectrum diagram of the vibration signal of the reference rolling bearing. Figure 4 (b) is the two-dimensional spectrum diagram of the rolling bearing vibration signal when the inner ring fails. Figure 4 (c) in the figure is a two-dimensional spectrum diagram of the rolling bearing vibration signal when the outer ring is faulty.

[0052] As a preferred embodiment, the implementation process of the improved convolutional layer described in the embodiment of the present invention is as follows: Figure 5 As shown in the figure, different colors represent different weights. A squeeze-excitation network (SENet) is embedded in a convolutional neural network. Its main principle is to introduce an attention mechanism to image channels during the convolution process of the CNN. For a feature map with a given dimension of c1×w×h, a normal convolution transform is performed to obtain a feature map with dimensions of c2×w×h.

[0053] The improved convolutional layer is used to:

[0054] A convolution operation is performed on the two-dimensional stereo spectrogram to obtain a feature image with dimensions of c1×w×h.

[0055] A convolution operation is performed on the feature image with a dimension of c1×w×h to obtain a feature image with a dimension of c2×w×h.

[0056] Through the compression operation in the compression-excitation network, a feature image with a dimension of c2×w×h is compressed using a spatial size of w×h to obtain a feature image with a dimension of c2×1×1; wherein the compression operation is determined by the following formula.

[0057]

[0058] u c Represents a feature image with dimension c2×1×1, V c Represents a feature image with dimensions c2×w×h.

[0059] The feature image with the dimension of c2×1×1 is processed through the excitation operation in the compression-excitation network, and the weight of each channel in the feature image with the dimension of c2×1×1 is determined.

[0060] Specifically, the feature image with dimension c2×1×1 is processed through two fully connected operations and activation functions to obtain the channel weight matrix z. The channel weight matrix z has c2 channel weights, and its processing expression is:

[0061] z=F ex (u c ,W)=σ(W2δ(W1u));

[0062] Among them, z is the channel weight, W1 is the dimensionality reduction parameter of the first fully connected operation, W2 is the dimension recovery parameter of the second fully connected layer, σ is the sigmoid function of weight normalization, and δ is the ReLU function.

[0063] Through the excitation operation in the compression-excitation network, different weights are assigned to the c2 channels according to the importance of each channel according to the channel weight matrix z, and then the weight of each channel in the feature image with dimension c2×1×1 is determined, that is, the target channel weight matrix z c .

[0064] According to the weight of each channel in the feature image with dimension c2×1×1 and the feature image with dimension c2×w×h, the output result of the improved convolution layer is obtained, specifically:

[0065] The target channel weight matrix z c and a feature image V with dimensions c2×w×h c Multiply by function F scale , get the output result X' of the improved convolutional layer c , and its processing expression is:

[0066] X' c =F scale (V c ,z c )=z c V c .

[0067] As a preferred embodiment, the improved pooling layer described in the embodiment of the present invention is used for:

[0068] Perform a pooling operation on the output result of the improved convolutional layer to obtain the output result after the pooling operation;

[0069] The width learning algorithm is used to process the output result after the pooling operation to obtain the output result of the improved pooling layer.

[0070] As a preferred implementation, the training process of the rolling bearing fault diagnosis model according to the embodiment of the present invention is as follows:

[0071] Step (1): Divide the sample data into training set and test set.

[0072] Step (2): Load the convolution layer of the pre-built improved convolutional neural network, input the sample data in the training set into the improved convolutional neural network, set 16 convolution kernels of size 5×5, the stride length is 1, and use ReLu as the activation function for convolution operation.

[0073] Step (3): Input the feature image obtained by the convolution operation into the compression-excitation network, calculate the importance of different channels in the feature image, assign corresponding weights, and multiply the channel feature image by the weights and pass it to the global maximum pooling layer. The window size of the compression-excitation network is set to 2, and the stride length is set to 2.

[0074] Step (4): The high-level feature image representation obtained from the pooling layer is input into the width learning algorithm and mapped to the feature node. The enhanced nodes are generated through the nonlinear function to form an enhanced matrix to train the fault classifier. Finally, the effectiveness of the rolling bearing fault diagnosis model is verified using the test set samples.

[0075] In order to verify the feasibility of the embodiment of the present invention, an experiment is conducted on 20 types of labeled data samples in the rolling bearing dataset (MFPT) to verify the effectiveness of the rolling bearing fault diagnosis model.

[0076] Step 1: Analyze and process the MFPT dataset.

[0077] The MFPT dataset includes rolling element bearing vibration signals collected from 20 different health status categories, including three normal conditions with constant load, three outer race fault conditions with constant load, seven outer race fault conditions with varying loads, and seven inner race fault conditions with varying loads. Specific parameters are shown in Table 1.

[0078] Table 1 Parameters of the MFPT dataset

[0079]

[0080] Step 2: Apply the rolling bearing fault diagnosis model provided by the embodiment of the present invention to the MFPT dataset to verify its performance.

[0081] The rolling bearing fault diagnosis model provided by the embodiment of the present invention is compared with several shallow learning algorithms including support vector machine (SVM) and back propagation neural network (BPNN) and typical deep learning algorithms including autoencoder (AE), traditional CNN and related variants. The results are shown in Table 2. The rolling bearing fault diagnosis model provided by the embodiment of the present invention has the advantage of high training and testing accuracy.

[0082] Table 2 Comparison results of rolling bearing fault diagnosis model and other methods

[0083]

[0084] Step 3: For each type of fault, the performance of the rolling bearing fault diagnosis model is evaluated using four performance indicators: accuracy, precision, recall, and F1-score. These performance indicators are easy to calculate and effective. In addition, when these two performance indicators are observed simultaneously, the rolling bearing fault diagnosis model provided by the embodiment of the present invention has more reference values. Here, a commonly used measurement curve, the receiver operating characteristic curve (ROC), is selected, and the corresponding false positive rate (FPR) and true positive rate (TPR) are plotted, and the area under the receiver operating characteristic curve (AU) is calculated.

[0085] The results in Table 3 demonstrate that the rolling bearing fault diagnosis model provided by the present invention achieves classification accuracy exceeding 90% for 17 inner and outer race faults, with most reaching 95%. For the I-3 fault category, the accuracy reaches 99%. The recognition accuracy and recall rates for each fault category are both above 90%. The F1-score and AU-ROC values ​​are both close to 1, demonstrating that the rolling bearing fault diagnosis model provided by the present invention is feasible and effective.

[0086] Table 3 Performance indicators of various types of rolling bearing faults

[0087]

[0088]

[0089] The rolling bearing fault diagnosis model provided by the embodiment of the present invention is represented by SECNN-BLS.

[0090] Figure 6 The figure below compares SECNN-BLS with eight other fault diagnosis models on the MFPT dataset. Obviously, SECNN-BLS has advantages in all fault diagnosis indicators.

[0091] Example 2

[0092] In order to execute the method corresponding to the above-mentioned embodiment 1 and achieve the corresponding functions and technical effects, a rolling bearing fault diagnosis device based on multiple algorithms is provided below.

[0093] like Figure 7 As shown, the rolling bearing fault diagnosis device based on multiple algorithms provided by the embodiment of the present invention includes:

[0094] The vibration signal acquisition module 1 is used to acquire the vibration signal of the target rolling bearing with unknown fault type under different load conditions.

[0095] The two-dimensional stereo spectrum diagram determining module 2 is used to process the target rolling bearing vibration signal to obtain a two-dimensional stereo spectrum diagram of the target rolling bearing vibration signal.

[0096] The fault diagnosis data determination module 3 is used to input the two-dimensional stereo spectrum diagram of the target rolling bearing vibration signal into the rolling bearing fault diagnosis model to obtain the fault diagnosis data of the target rolling bearing; the fault diagnosis data includes the fault type and the fault location.

[0097] In which, the rolling bearing fault diagnosis model is obtained by training an improved convolutional neural network with sample data; the sample data includes a two-dimensional stereo frequency spectrum of a sample rolling bearing vibration signal and a corresponding fault diagnosis data label; the improved convolutional neural network includes an input module, an implicit module and an output module; the implicit module includes an improved convolutional layer, an improved pooling layer and a fully connected layer connected in sequence; the improved convolutional layer is a convolutional layer embedded with a compression-excitation network; the improved pooling layer is a pooling layer embedded with a width learning algorithm.

[0098] Example 3

[0099] An embodiment of the present invention provides an electronic device including a memory and a processor, wherein the memory is used to store a computer program, and the processor runs the computer program to enable the electronic device to execute the rolling bearing fault diagnosis method based on multiple algorithms of embodiment one.

[0100] Optionally, the above-mentioned electronic device may be a server.

[0101] In addition, an embodiment of the present invention further provides a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, the rolling bearing fault diagnosis method based on multiple algorithms of the first embodiment is implemented.

[0102] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. Reference can be made to the common and similar parts between the various embodiments. For the systems disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple, and the relevant parts can be referred to the method description.

[0103] This document uses specific examples to illustrate the principles and implementation methods of the present invention. The above examples are only intended to help understand the method and core concept of the present invention. At the same time, those skilled in the art will find that the specific implementation methods and application scopes may vary based on the concept of the present invention. In summary, the contents of this specification should not be construed as limiting the present invention.

Claims

1. A rolling bearing fault diagnosis method based on multiple algorithms, characterized in that: include: Obtain vibration signals of target rolling bearings with unknown fault types under different load conditions; Processing the target rolling bearing vibration signal to obtain a two-dimensional spectrum diagram of the target rolling bearing vibration signal; Inputting the two-dimensional stereo frequency spectrum of the target rolling bearing vibration signal into a rolling bearing fault diagnosis model to obtain fault diagnosis data of the target rolling bearing; the fault diagnosis data includes the fault type and the fault location; In which, the rolling bearing fault diagnosis model is obtained by training an improved convolutional neural network with sample data; the sample data includes a two-dimensional stereo frequency spectrum of a sample rolling bearing vibration signal and a corresponding fault diagnosis data label; the improved convolutional neural network includes an input module, an implicit module and an output module; the implicit module includes an improved convolutional layer, an improved pooling layer and a fully connected layer connected in sequence; the improved convolutional layer is a convolutional layer embedded with a compression-excitation network; the improved pooling layer is a pooling layer embedded with a width learning algorithm.

2. The rolling bearing fault diagnosis method based on multiple algorithms according to claim 1, characterized in that: Processing the target rolling bearing vibration signal to obtain a two-dimensional spectrum diagram of the target rolling bearing vibration signal specifically includes: The target rolling bearing vibration signal is processed by a short-time Fourier transform algorithm to obtain a two-dimensional spectrum diagram of the target rolling bearing vibration signal.

3. The rolling bearing fault diagnosis method based on multiple algorithms according to claim 1, characterized in that: The process of determining the sample data is as follows: Convert the time series signal in the sample rolling bearing dataset into the original vibration signal in the time domain; The sampling parameters of the sliding window overlapping sampling method are set according to the sampling frequency and load conditions of the sample rolling bearing data set; The sliding window overlapping sampling method is used to perform multiple short-time Fourier transforms on the original vibration signal to obtain a two-dimensional stereo spectrum of the sample rolling bearing vibration signal; The fault diagnosis data label corresponding to the original vibration signal is determined as the fault diagnosis data label corresponding to the two-dimensional spectrum diagram of the sample rolling bearing vibration signal.

4. The rolling bearing fault diagnosis method based on multiple algorithms according to claim 1, characterized in that: The improved convolutional layer is used to: Perform convolution operation on the two-dimensional stereo spectrum map to obtain a feature image with dimensions of c1×w×h; Perform a convolution operation on the feature image with a dimension of c1×w×h to obtain a feature image with a dimension of c2×w×h; Through the compression operation in the compression-excitation network, the feature image with the dimension of c2×w×h is compressed using the spatial size of w×h to obtain a feature image with the dimension of c2×1×1; The feature image with the dimension of c2×1×1 is processed through the excitation operation in the compression-excitation network, and the weight of each channel in the feature image with the dimension of c2×1×1 is determined; The output result of the improved convolutional layer is obtained according to the weight of each channel in the feature image with dimension c2×1×1 and the feature image with dimension c2×w×h.

5. The rolling bearing fault diagnosis method based on multiple algorithms according to claim 4, characterized in that: In the aspect of processing the feature image having a dimension of c2×1×1 through the excitation operation in the compression-excitation network and determining the weight of each channel in the feature image having a dimension of c2×1×1, the improved convolutional layer is further used to: Processing a feature image having a dimension of c2×1×1 through two fully connected operations and an activation function to obtain a channel weight matrix z; the channel weight matrix z has c2 channel weights; Through the excitation operation in the compression-excitation network, different weights are assigned to the c2 channels according to the channel weight matrix z, and then the weight of each channel in the feature image with dimension c2×1×1 is determined.

6. The rolling bearing fault diagnosis method based on multiple algorithms according to claim 4, characterized in that: In the convolution operation, 16 convolution kernels of size 5×5 and a stride of length 1 are set; in the excitation operation and the compression operation, a window of size 2 and a stride of length 2 are set.

7. The rolling bearing fault diagnosis method based on multiple algorithms according to claim 1, characterized in that: The improved pooling layer is used to: Perform a pooling operation on the output result of the improved convolutional layer to obtain the output result after the pooling operation; The width learning algorithm is used to process the output result after the pooling operation to obtain the output result of the improved pooling layer.

8. A rolling bearing fault diagnosis device based on multiple algorithms, characterized in that: include: A vibration signal acquisition module, used to acquire vibration signals of target rolling bearings with unknown fault types under different load conditions; a two-dimensional stereo spectrum diagram determination module, configured to process the target rolling bearing vibration signal to obtain a two-dimensional stereo spectrum diagram of the target rolling bearing vibration signal; a fault diagnosis data determination module, configured to input the two-dimensional stereo frequency spectrum of the target rolling bearing vibration signal into a rolling bearing fault diagnosis model to obtain fault diagnosis data of the target rolling bearing; the fault diagnosis data includes a fault type and a fault location; In which, the rolling bearing fault diagnosis model is obtained by training an improved convolutional neural network with sample data; the sample data includes a two-dimensional stereo frequency spectrum of a sample rolling bearing vibration signal and a corresponding fault diagnosis data label; the improved convolutional neural network includes an input module, an implicit module and an output module; the implicit module includes an improved convolutional layer, an improved pooling layer and a fully connected layer connected in sequence; the improved convolutional layer is a convolutional layer embedded with a compression-excitation network; the improved pooling layer is a pooling layer embedded with a width learning algorithm.

9. An electronic device, characterized in that: The electronic device comprises a memory and a processor, wherein the memory is used to store a computer program, and the processor runs the computer program to enable the electronic device to execute the rolling bearing fault diagnosis method based on multiple algorithms according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that It stores a computer program, which, when executed by a processor, implements the rolling bearing fault diagnosis method based on multiple algorithms as described in any one of claims 1 to 7.

Images