A bearing fault diagnosis method, device, equipment and storage medium

Abstract

The present invention discloses a bearing fault diagnosis method, device, equipment and storage medium, the method comprising: collecting time series data of the bearing; using Gram angular field to process the time series data, generating Gram difference field map and Gram sum field map, inputting the bearing fault diagnosis model, and obtaining the bearing fault diagnosis result; wherein the bearing fault diagnosis model adopts a convolutional neural network, the model is used to: perform parallel convolution on the Gram difference field map and the Gram sum field map, and use sliding convolution to replace the pooling operation, perform feature extraction and feature aggregation, and obtain fault features; based on the fault features, the CBAM attention mechanism is used to adaptively adjust the channel and spatial weights of the feature map to generate the fault diagnosis result. The present invention can improve the comprehensiveness and accuracy of bearing fault diagnosis.

Description

Technical Field

[0001] The present invention relates to the technical field of bearing fault detection, and in particular to a bearing fault diagnosis method, device, equipment and storage medium based on a CBAM attention mechanism of Gram angular field and parallel convolution. Background Art

[0002] As industrial technology develops towards high precision, high efficiency and high automation, the requirements for bearing performance and reliability are becoming increasingly stringent. However, the working environment of bearings is complex and changeable, coupled with a variety of potential factors, such as high temperature, heavy load, high-speed operation, material fatigue, poor lubrication and improper installation, which lead to frequent bearing failures. These failures may cause equipment shutdown in serious cases, causing huge economic losses and safety hazards to enterprises. Therefore, timely and accurate diagnosis of bearing failures is crucial.

[0003] In the prior art, the traditional method based on one-dimensional time series analysis is usually used for fault diagnosis. However, one-dimensional data is difficult to intuitively display the time dependence and change information in the time series, and it is often difficult to effectively extract and represent the deep features in the data when processing one-dimensional time series data, which limits the ability to extract and identify fault features. Secondly, in the process of feature extraction, the traditional convolutional neural network structure often uses a pooling layer to reduce the spatial size of the feature map. Although this step can reduce the amount of calculation and memory usage, it may also cause the loss and blurring of key feature information. In addition, when processing fault data, the prior art often lacks an attention mechanism for key fault features, which makes the model easily disturbed when processing background noise and irrelevant information, thereby affecting the diagnosis effect. Summary of the invention

[0004] In order to solve the above technical problems, the present invention provides a bearing fault diagnosis method, device, equipment and storage medium, which improves the comprehensiveness and accuracy of bearing fault diagnosis by adopting Gramian Angular Field (GAF), CBAM (Convolutional Block Attention Module) self-attention mechanism, parallel convolution and sliding convolution for feature extraction.

[0005] An embodiment of the present invention provides a bearing fault diagnosis method, comprising:

[0006] Collect time series data of bearings;

[0007] Processing the time series data using a Gram angle field to generate a Gram difference field diagram and a Gram sum field diagram;

[0008] Inputting the Gram difference field diagram and the Gram sum field diagram into a bearing fault diagnosis model to obtain a bearing fault diagnosis result;

[0009] The bearing fault diagnosis model adopts a convolutional neural network, which is used to:

[0010] Convolving the Gram difference field map and the Gram sum field map in parallel, and using sliding convolution to replace the pooling operation to perform feature extraction and feature aggregation to obtain fault features;

[0011] Based on the fault features, the CBAM attention mechanism is used to adaptively adjust the channel and spatial weights of the feature map to generate the fault diagnosis result.

[0012] As an improvement of the above scheme, the time series data is processed by using the Gram angle field to generate the Gram difference field map and the Gram sum field map, including:

[0013] Normalizing the time series data, and converting the normalized time series data into polar coordinates;

[0014] generating a Gram sum field diagram based on the cosine sum between the time series data;

[0015] A Gram difference field map is generated according to the angle difference between the time series data.

[0016] As an improvement of the above scheme, the CBAM attention mechanism includes a channel attention mechanism, which is used to:

[0017] By calculating the global average pooling and global maximum pooling of each channel, the global information of each channel is obtained;

[0018] The global information is fused through a fully connected layer to generate channel attention weights to weight each channel.

[0019] As an improvement of the above scheme, the CBAM attention mechanism also includes a spatial attention mechanism, which is used to:

[0020] By performing one-dimensional convolution on each channel, the local information of each pixel is obtained;

[0021] Global average pooling and global maximum pooling are performed on the local information to generate spatial attention weights to weight each pixel.

[0022] As an improvement of the above solution, the bearing fault diagnosis model includes: a sampling module, a spatial attention module and a channel attention module;

[0023] The structure of the sampling module is convolution layer, regularization layer, sliding convolution layer, ReLu layer, convolution layer, regularization layer, sliding convolution layer, ReLu layer, convolution layer and regularization layer in sequence;

[0024] The spatial attention module includes: a convolutional layer and a sigmoid layer;

[0025] The channel attention module includes: an adaptive average pooling layer, an adaptive maximum pooling layer, a convolutional layer, a ReLu layer and a convolutional layer.

[0026] The embodiment of the present invention further provides a bearing fault diagnosis device, comprising:

[0027] Data acquisition module, used to collect time series data of bearings;

[0028] A Gram angle field module, used for processing the time series data using the Gram angle field to generate a Gram difference field diagram and a Gram sum field diagram;

[0029] A fault diagnosis module, used for inputting the Gram difference field diagram and the Gram sum field diagram into a bearing fault diagnosis model to obtain a bearing fault diagnosis result;

[0030] The bearing fault diagnosis model adopts a convolutional neural network, which is used to:

[0031] Convolving the Gram difference field map and the Gram sum field map in parallel, and using sliding convolution to replace the pooling operation to perform feature extraction and feature aggregation to obtain fault features;

[0032] Based on the fault features, the CBAM attention mechanism is used to adaptively adjust the channel and spatial weights of the feature map to generate the fault diagnosis result.

[0033] Furthermore, the Gram angle field module is specifically used for:

[0034] Normalizing the time series data, and converting the normalized time series data into polar coordinates;

[0035] generating a Gram sum field diagram based on the cosine sum between the time series data;

[0036] A Gram difference field map is generated according to the angle difference between the time series data.

[0037] Furthermore, the bearing fault diagnosis model includes: a sampling module, a spatial attention module and a channel attention module;

[0038] The structure of the sampling module is convolution layer, regularization layer, sliding convolution layer, ReLu layer, convolution layer, regularization layer, sliding convolution layer, ReLu layer, convolution layer and regularization layer in sequence;

[0039] The spatial attention module includes: a convolutional layer and a sigmoid layer;

[0040] The channel attention module includes: an adaptive average pooling layer, an adaptive maximum pooling layer, a convolutional layer, a ReLu layer and a convolutional layer.

[0041] An embodiment of the present invention further provides a computer device, comprising a processor and a memory, wherein a computer program is stored in the memory and the computer program is configured to be executed by the processor, and when the processor executes the computer program, the bearing fault diagnosis method described in any one of the above items is implemented.

[0042] An embodiment of the present invention further provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, wherein when the computer program is running, the device where the computer-readable storage medium is located is controlled to execute any one of the above-mentioned bearing fault diagnosis methods.

[0043] Compared with the prior art, the beneficial effects of a bearing fault diagnosis method, device, equipment and storage medium provided by an embodiment of the present invention are: by adopting the Gram angular field, one-dimensional time series data is converted into two-dimensional image data, the time dependence and change information in the time series are retained, which is beneficial to the feature extraction of the convolutional neural network; by adopting a parallel convolution method, the Gram difference field map and the Gram sum field map are processed simultaneously to extract their fault features respectively, which can make full use of the information in different types of images generated by the Gram angular field and improve the comprehensiveness and accuracy of feature extraction; by replacing the traditional pooling operation with a sliding convolution, the problems of feature loss and feature blurring are avoided and more detail information is retained; by adding a CBAM attention mechanism to the back end of the model, the attention information of the two dimensions of channel and space can be considered at the same time, and the weight of the feature map can be adaptively adjusted, so that the model can pay more attention to important feature information and ignore irrelevant background noise, further improving the performance of the fault diagnosis model. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 It is a flow chart of a bearing fault diagnosis method provided by an embodiment of the present invention;

[0045] Figure 2 is a Gram diagram under each fault mode provided by an embodiment of the present invention;

[0046] Figure 3 is a structural schematic diagram of a bearing fault diagnosis model provided by an embodiment of the present invention;

[0047] Figure 4 is an iterative curve diagram of the accuracy of each model provided by the embodiment of the present invention;

[0048] Figure 5 is a bar chart of the average accuracy of each model provided by the embodiment of the present invention;

[0049] Figure 6 It is a confusion matrix diagram of a bearing fault diagnosis model provided by an embodiment of the present invention;

[0050] Figure 7 is a characteristic visualization diagram of a bearing fault diagnosis model provided by an embodiment of the present invention;

[0051] Figure 8 It is a structural schematic diagram of a bearing fault diagnosis device provided by an embodiment of the present invention;

[0052] Fig. 9 It is a structural schematic diagram of a computer device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0053] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. 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 creative work are within the scope of protection of the present invention.

[0054] See also Figure 1 , Figure 1 1 is a flow chart of a bearing fault diagnosis method provided by an embodiment of the present invention. The bearing fault diagnosis method comprises:

[0055] S1: Collect time series data of bearings;

[0056] S2: Processing the time series data using the Gram angle field to generate a Gram difference field diagram and a Gram sum field diagram;

[0057] S3: inputting the Gram difference field diagram and the Gram sum field diagram into a bearing fault diagnosis model to obtain a bearing fault diagnosis result;

[0058] The bearing fault diagnosis model adopts a convolutional neural network, which is used to:

[0059] Convolving the Gram difference field map and the Gram sum field map in parallel, and using sliding convolution to replace the pooling operation to perform feature extraction and feature aggregation to obtain fault features;

[0060] Based on the fault features, the CBAM attention mechanism is used to adaptively adjust the channel and spatial weights of the feature map to generate the fault diagnosis result.

[0061] Specifically, the embodiment of the present invention uses a bearing accelerated life test data set for model training and testing, which uses LDK UER204 rolling bearing as the test bearing. The data set includes three working conditions, each of which has five faulty bearings. The data set is obtained from a bearing accelerated life test bench, which consists of components such as an AC motor, a motor speed controller, a rotating shaft, a support bearing, a digital force display, vertical and horizontal acceleration sensors, a hydraulic loading system, and a test bearing.

[0062] For the collected bearing data, the data is first cut into groups of 2500 sample points, and the segmented data is transformed into a Gram angular field to generate the corresponding Gram diagram; the obtained data is shuffled and the data set is divided into a training set and a test set in a ratio of 8:2. Specifically, the detailed information of the data set is shown in Table 1, and the GASF (Gram sum field diagram) and GADF (Gram difference field diagram) corresponding to each fault mode are shown in Figure 2 shown.

[0063] Table 1

[0064]

[0065] As one of the optional embodiments, the using of the Gram angle field to process the time series data to generate the Gram difference field diagram and the Gram sum field diagram includes:

[0066] Normalizing the time series data, and converting the normalized time series data into polar coordinates;

[0067] generating a Gram sum field diagram based on the cosine sum between the time series data;

[0068] A Gram difference field map is generated according to the angle difference between the time series data.

[0069] Specifically, the Gram angular field is a method for converting time series data into images. It mainly maps the time series into a polar coordinate system and uses trigonometric functions to generate two types of images: GASF (Gram Sum Field Diagram) and GADF (Gram Difference Field Diagram).

[0070] First, the time series data is converted into Normalized to interval, get normalized data :

[0071]

[0072] Then, the normalized data is mapped to the Cartesian coordinate system according to the following formula to obtain the polar angle at each time: and polar diameter :

[0073]

[0074] in, is the timestamp, is the regularization constant.

[0075] Furthermore, the Gram angle field defines two unique forms of inner products with penalty terms to eliminate the influence of Gaussian noise. They are defined as follows:

[0076]

[0077]

[0078] According to the above two different inner product definition forms, GASF and GADF can be obtained respectively:

[0079]

[0080]

[0081] See also Figure 3 , Figure 3 It is a structural schematic diagram of a bearing fault diagnosis model provided by an embodiment of the present invention. Further, after obtaining the Gram angular field diagrams GASF and GADF, they are input into the bearing fault diagnosis model. First, GASF and GADF are processed in parallel by using parallel convolution. Through the sliding and operation of multiple convolution kernels on the image, the key fault features in the sum field and the difference field can be extracted at the same time, and the comprehensive capture of features can be achieved; further, the sliding convolution method is used to replace the pooling operation in the convolution network, and feature aggregation is performed by step-by-step sliding convolution, which retains the local features of the image and avoids information loss and blur.

[0082] Furthermore, a CBAM attention mechanism is added to the backend of the model. Through the spatial attention module and the channel attention module, the channel and spatial weights of the feature map are adaptively adjusted, so that the model can focus on more important fault features and ignore irrelevant background noise, further improving the accuracy and robustness of fault diagnosis and achieving more reliable fault detection and identification.

[0083] Specifically, CBAM is an attention mechanism for CNN, which aims to enhance the model's attention to important features, thereby improving network performance. CBAM mainly focuses on two key aspects: channel attention and spatial attention.

[0084] As one of the optional embodiments, the CBAM attention mechanism includes a channel attention mechanism, and the channel attention mechanism is used to:

[0085] By calculating the global average pooling and global maximum pooling of each channel, the global information of each channel is obtained;

[0086] The global information is fused through a fully connected layer to generate channel attention weights to weight each channel.

[0087] Specifically, channel attention obtains the global information of each channel by calculating the global average pooling and global maximum pooling of each channel; then, these global information are fused together through a fully connected layer to generate a channel attention weight for weighting each channel. The embodiment of the present invention adopts channel attention, which helps the model pay more attention to the channels that contribute to the task during the learning process.

[0088] As one of the optional embodiments, the CBAM attention mechanism also includes a spatial attention mechanism, and the spatial attention mechanism is used to:

[0089] By performing one-dimensional convolution on each channel, the local information of each pixel is obtained;

[0090] Global average pooling and global maximum pooling are performed on the local information to generate spatial attention weights to weight each pixel.

[0091] Specifically, spatial attention obtains local information of each pixel by performing a one-dimensional convolution operation on each channel, and then generates a spatial attention weight by performing global average pooling and global maximum pooling on these local information. This weight is used to weight each pixel in order to emphasize important areas in the image.

[0092] The embodiment of the present invention combines channel attention and spatial attention, and CBAM can effectively capture the relationship between channels and space in the input feature map, thereby improving the performance of the model.

[0093] As one of the optional embodiments, the bearing fault diagnosis model includes: a sampling module, a spatial attention module and a channel attention module;

[0094] The structure of the sampling module is convolution layer, regularization layer, sliding convolution layer, ReLu layer, convolution layer, regularization layer, sliding convolution layer, ReLu layer, convolution layer and regularization layer in sequence;

[0095] The spatial attention module includes: a convolutional layer and a sigmoid layer;

[0096] The channel attention module includes: an adaptive average pooling layer, an adaptive maximum pooling layer, a convolutional layer, a ReLu layer and a convolutional layer.

[0097] Specifically, the specific structural parameters of the bearing fault diagnosis model are shown in Table 2:

[0098] Table 2

[0099]

[0100] Furthermore, the training set is used to train the bearing fault diagnosis model, and the test set is used to test the model to analyze the model performance. In order to illustrate that the present invention has a better effect on bearing fault diagnosis than the prior art, the models in the prior art: ResNet, MnasNet, Vgg16 are selected for comparative analysis with the model of the present invention. All models are written in Python and built using the Pytorch framework. The cross entropy loss function is used for loss measurement, and the Adam algorithm is used for parameter optimization, and the learning rate is set to 0.01. During the training process, the data batch size is set to 64, the number of training rounds is set to 100, and the learning rate is defined as 0.001.

[0101] See also Figure 4 and Figure 5 , Figure 4 is an iterative curve diagram of the accuracy of each model provided in the embodiment of the present invention, Figure 5 It is a bar chart of the average accuracy of each model tested 10 times provided by the embodiment of the present invention. It can be seen that the model proposed in the embodiment of the present invention has the fastest convergence speed, which can basically reach convergence after 20 iterations, can achieve 100% diagnostic accuracy and has good stability; while the other three models have a slower convergence speed, and are inferior to the model proposed in the embodiment of the present invention in terms of diagnostic accuracy and stability, among which Vgg16 has a diagnostic accuracy of only 93.5% and the worst stability.

[0102] For further information, see Figure 6 and Figure 7 , Figure 6 : is a confusion matrix diagram of a bearing fault diagnosis model provided by an embodiment of the present invention. It can be seen that the model proposed by the embodiment of the present invention has a good classification effect in each fault mode, and the values ​​on the diagonal are all 1.00. Figure 7 This is a feature visualization diagram of a bearing fault diagnosis model provided by an embodiment of the present invention. It can be seen that the fault features are compactly clustered with clear boundaries, the feature distances are far under different fault types, and the feature extraction effect is good, which further proves the diagnostic advantage of the model proposed in the embodiment of the present invention.

[0103] Furthermore, in order to verify the effectiveness of the improved model provided by the embodiment of the present invention compared with the original model, an ablation experiment is used for experimental verification. The parallel convolution is changed to a single-channel convolution, and the CBAM attention mechanism is removed for fault diagnosis. The ablation experiment results are shown in Table 3.

[0104] Table 3

[0105]

[0106] It can be seen from the experimental results that the accuracy of diagnosis using a single channel will be significantly reduced, and the stability will also deteriorate. In addition, the performance of the model without the CBAM module will further decrease. Therefore, this result further proves the superiority of the bearing fault diagnosis method provided by the embodiment of the present invention.

[0107] The embodiment of the present invention improves the accuracy of the diagnosis model by enhancing the features of fault samples, solves the problem of imbalance in training samples, and significantly improves the accuracy of bearing fault diagnosis; by adopting the Gram angular field, one-dimensional time series data is converted into two-dimensional image data, the time dependency and change information in the time series are retained, which is beneficial to the feature extraction of the convolutional neural network; by adopting a parallel convolution method, the Gram difference field map and the Gram sum field map are processed simultaneously to extract their fault features respectively, which can make full use of the information in different types of images generated by the Gram angular field and improve the comprehensiveness and accuracy of feature extraction; by replacing the traditional pooling operation with sliding convolution, the problems of feature loss and feature blurring are avoided and more detail information is retained; by adding the CBAM attention mechanism to the back end of the model, the attention information of the two dimensions of channel and space are considered at the same time, and the weight of the feature map is adaptively adjusted, so that the model can pay more attention to important feature information and ignore irrelevant background noise, thereby further improving the performance of the fault diagnosis model.

[0108] Correspondingly, the present invention also provides a bearing fault diagnosis device, which can implement all the processes of the bearing fault diagnosis method in the above embodiment.

[0109] See also Figure 8 , Figure 8 : is a structural schematic diagram of a bearing fault diagnosis device provided by an embodiment of the present invention. The bearing fault diagnosis device comprises:

[0110] Data acquisition module 801, used to collect time series data of bearings;

[0111] A Gram angle field module 802 is used to process the time series data using the Gram angle field to generate a Gram difference field diagram and a Gram sum field diagram;

[0112] A fault diagnosis module 803 is used to input the Gram difference field diagram and the Gram sum field diagram into a bearing fault diagnosis model to obtain a bearing fault diagnosis result;

[0113] The bearing fault diagnosis model adopts a convolutional neural network, which is used to:

[0114] Convolving the Gram difference field map and the Gram sum field map in parallel, and using sliding convolution to replace the pooling operation to perform feature extraction and feature aggregation to obtain fault features;

[0115] Based on the fault features, the CBAM attention mechanism is used to adaptively adjust the channel and spatial weights of the feature map to generate the fault diagnosis result.

[0116] Preferably, the Gram angle field module 802 is specifically used for:

[0117] Normalizing the time series data, and converting the normalized time series data into polar coordinates;

[0118] generating a Gram sum field diagram based on the cosine sum between the time series data;

[0119] A Gram difference field map is generated according to the angle difference between the time series data.

[0120] Preferably, the bearing fault diagnosis model comprises: a sampling module, a spatial attention module and a channel attention module;

[0121] The structure of the sampling module is convolution layer, regularization layer, sliding convolution layer, ReLu layer, convolution layer, regularization layer, sliding convolution layer, ReLu layer, convolution layer and regularization layer in sequence;

[0122] The spatial attention module includes: a convolutional layer and a sigmoid layer;

[0123] The channel attention module includes: an adaptive average pooling layer, an adaptive maximum pooling layer, a convolutional layer, a ReLu layer and a convolutional layer.

[0124] Preferably, the CBAM attention mechanism includes a channel attention mechanism, and the channel attention mechanism is used to:

[0125] By calculating the global average pooling and global maximum pooling of each channel, the global information of each channel is obtained;

[0126] The global information is fused through a fully connected layer to generate channel attention weights to weight each channel.

[0127] Preferably, the CBAM attention mechanism further includes a spatial attention mechanism, and the spatial attention mechanism is used to:

[0128] By performing one-dimensional convolution on each channel, the local information of each pixel is obtained;

[0129] Global average pooling and global maximum pooling are performed on the local information to generate spatial attention weights to weight each pixel.

[0130] In specific implementation, the working principle, control process and technical effect achieved by the bearing fault diagnosis device provided in the embodiment of the present invention are the same as those of the bearing fault diagnosis method in the above embodiment, and will not be repeated here.

[0131] See also Fig. 9 , Fig. 9 9 is a structural block diagram of a computer device provided in an embodiment of the present invention, wherein the computer device comprises: a processor 901, a memory 902, and a computer program stored in the memory 902 and executable on the processor 901. When the processor 901 executes the computer program, the steps in the above-mentioned bearing fault diagnosis method embodiment are implemented. Alternatively, when the processor 901 executes the computer program, the functions of each module / unit in the above-mentioned device embodiments are implemented.

[0132] Exemplarily, the computer program may be divided into one or more modules / units, which are stored in the memory 902 and executed by the processor 901 to implement the present invention. The one or more modules / units may be a series of computer program instruction segments capable of implementing specific functions, and the instruction segments are used to describe the execution process of the computer program in the computer device.

[0133] The computer device may include, but is not limited to, a processor 901 and a memory 902. Those skilled in the art will appreciate that the schematic diagram is merely an example of a computer device and does not constitute a limitation on the computer device, and may include more or fewer components than shown in the diagram, or may combine certain components, or different components, for example, the computer device may also include an input / output device, a network access device, a bus, etc.

[0134] The processor 901 may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor, etc. The processor 901 is the control center of the computer device, and uses various interfaces and lines to connect various parts of the entire computer device.

[0135] The memory 902 can be used to store the computer program and / or module. The processor 901 implements various functions of the computer device by running or executing the computer program and / or module stored in the memory 902 and calling the data stored in the memory 902. The memory 902 can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, an application required for at least one function (such as a sound playback function, an image playback function, etc.), etc.; the data storage area can store data created according to the use of the mobile phone (such as audio data, a phone book, etc.), etc. In addition, the memory 902 can include a high-speed random access memory, and can also include a non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (SecureDigital, SD) card, a flash card (Flash Card), at least one disk storage device, a flash memory device, or other volatile solid-state storage devices.

[0136] Wherein, if the module / unit integrated in the computer device is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on such an understanding, the present invention implements all or part of the processes in the above-mentioned embodiment method, and can also be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium, and the computer program can implement the steps of the above-mentioned various method embodiments when executed by the processor 901. Wherein, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording medium, U disk, mobile hard disk, disk, optical disk, computer memory, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), electric carrier signal, telecommunication signal and software distribution medium, etc.

[0137] An embodiment of the present invention further provides a computer-readable storage medium, which includes a stored computer program, wherein when the computer program is running, the device where the computer-readable storage medium is located is controlled to execute the bearing fault diagnosis method described in any of the above embodiments.

[0138] The embodiments of the present invention provide a bearing fault diagnosis method, device, equipment and storage medium, which have the following beneficial effects: by adopting the Gram angular field, one-dimensional time series data is converted into two-dimensional image data, the time dependence and change information in the time series are retained, which is beneficial to the feature extraction of the convolutional neural network; by adopting the parallel convolution method, the Gram difference field map and the Gram sum field map are processed simultaneously, and their fault features are extracted respectively, which can make full use of the information in different types of images generated by the Gram angular field, and improve the comprehensiveness and accuracy of feature extraction; by replacing the traditional pooling operation with sliding convolution, the problems of feature loss and feature blurring are avoided, and more detail information is retained; by adding the CBAM attention mechanism to the back end of the model, the attention information of the two dimensions of channel and space are considered at the same time, and the weight of the feature map is adaptively adjusted, so that the model can pay more attention to important feature information and ignore irrelevant background noise, further improving the performance of the fault diagnosis model.

[0139] The above is a preferred embodiment of the present invention. It should be pointed out that a person skilled in the art can make several improvements and modifications without departing from the principle of the present invention. These improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A bearing fault diagnosis method, characterized in that: include: Collect time series data of bearings; Processing the time series data using a Gram angle field to generate a Gram difference field diagram and a Gram sum field diagram; Inputting the Gram difference field diagram and the Gram sum field diagram into a bearing fault diagnosis model to obtain a bearing fault diagnosis result; The bearing fault diagnosis model adopts a convolutional neural network, which is used to: Convolving the Gram difference field map and the Gram sum field map in parallel, and using sliding convolution to replace the pooling operation to perform feature extraction and feature aggregation to obtain fault features; Based on the fault features, the CBAM attention mechanism is used to adaptively adjust the channel and spatial weights of the feature map to generate the fault diagnosis result; The bearing fault diagnosis model includes: a sampling module, a spatial attention module and a channel attention module; The structure of the sampling module is convolution layer, regularization layer, sliding convolution layer, ReLu layer, convolution layer, regularization layer, sliding convolution layer, ReLu layer, convolution layer and regularization layer in sequence; The spatial attention module includes: a convolutional layer and a sigmoid layer; The channel attention module includes: an adaptive average pooling layer, an adaptive maximum pooling layer, a convolutional layer, a ReLu layer and a convolutional layer.

2. The bearing fault diagnosis method according to claim 1, characterized in that: The step of using the Gram angle field to process the time series data to generate a Gram difference field diagram and a Gram sum field diagram includes: Normalizing the time series data, and converting the normalized time series data into polar coordinates; generating a Gram sum field diagram based on the cosine sum between the time series data; A Gram difference field map is generated according to the angle difference between the time series data.

3. The bearing fault diagnosis method according to claim 1, characterized in that: The CBAM attention mechanism includes a channel attention mechanism, which is used to: By calculating the global average pooling and global maximum pooling of each channel, the global information of each channel is obtained; The global information is fused through a fully connected layer to generate channel attention weights to weight each channel.

4. The bearing fault diagnosis method according to claim 3, characterized in that: The CBAM attention mechanism also includes a spatial attention mechanism, which is used to: By performing one-dimensional convolution on each channel, the local information of each pixel is obtained; Global average pooling and global maximum pooling are performed on the local information to generate spatial attention weights to weight each pixel.

5. A bearing fault diagnosis device, characterized in that: include: Data acquisition module, used to collect time series data of bearings; A Gram angle field module, used for processing the time series data using the Gram angle field to generate a Gram difference field diagram and a Gram sum field diagram; A fault diagnosis module, used for inputting the Gram difference field diagram and the Gram sum field diagram into a bearing fault diagnosis model to obtain a bearing fault diagnosis result; The bearing fault diagnosis model adopts a convolutional neural network, which is used to: Convolving the Gram difference field map and the Gram sum field map in parallel, and using sliding convolution to replace the pooling operation to perform feature extraction and feature aggregation to obtain fault features; Based on the fault features, the CBAM attention mechanism is used to adaptively adjust the channel and spatial weights of the feature map to generate the fault diagnosis result; The bearing fault diagnosis model includes: a sampling module, a spatial attention module and a channel attention module; The structure of the sampling module is convolution layer, regularization layer, sliding convolution layer, ReLu layer, convolution layer, regularization layer, sliding convolution layer, ReLu layer, convolution layer and regularization layer in sequence; The spatial attention module includes: a convolutional layer and a sigmoid layer; The channel attention module includes: an adaptive average pooling layer, an adaptive maximum pooling layer, a convolutional layer, a ReLu layer and a convolutional layer.

6. The bearing fault diagnosis device according to claim 5, characterized in that: The Gram angle field module is specifically used for: Normalizing the time series data, and converting the normalized time series data into polar coordinates; generating a Gram sum field diagram based on the cosine sum between the time series data; A Gram difference field map is generated according to the angle difference between the time series data.

7. A computer device, characterized in that: It comprises a processor and a memory, wherein a computer program is stored in the memory and the computer program is configured to be executed by the processor, and when the processor executes the computer program, the bearing fault diagnosis method according to any one of claims 1 to 4 is implemented.

8. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, wherein when the device where the computer-readable storage medium is located executes the computer program, the bearing fault diagnosis method according to any one of claims 1 to 4 is implemented.

Images