Bearing fault diagnosis method and system

Abstract

The invention provides a bearing fault diagnosis method and system, and relates to the technical field of bearing fault detection, and the method comprises the steps: collecting bearing vibration data; the bearing vibration signals are converted into a wavelet two-dimensional time-frequency image, an angle field two-dimensional time-frequency image and a transfer field two-dimensional time-frequency image through a continuous wavelet transform algorithm, a Grubrum angle field coding algorithm and a Markov transfer field algorithm; fusing the wavelet two-dimensional time-frequency image, the angular field two-dimensional time-frequency image and the transfer field two-dimensional time-frequency image to generate a fused time-frequency image; constructing a bearing fault detection model with a Swinin-T channel, a mixed attention channel and a feature fusion module; inputting the bearing vibration data as a training set into a bearing fault detection model until the predicted loss value of the bearing fault detection model is within a preset predicted loss value interval; acquiring a vibration signal of the to-be-detected bearing; and inputting a vibration signal of the bearing to be detected into the trained bearing fault detection model, and outputting a predicted bearing fault type.

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 and system. Background Art

[0002] Bearing fault diagnosis refers to the process of identifying potential faults and preventing equipment damage by monitoring and analyzing the operating status and performance of bearings. Common faults include wear, cracks, shedding and insufficient lubrication, and the bearing fault diagnosis method is to identify potential failure modes such as rolling element failure, inner or outer ring wear, etc. by collecting and analyzing these signals.

[0003] With the improvement of industrial automation, bearing fault diagnosis has become crucial. By timely discovering and diagnosing bearing faults, equipment downtime can be reduced, equipment reliability and stability can be improved, equipment service life can be extended, personnel safety and environmental protection can be guaranteed, economic benefits and operational safety can be improved, which is of great significance to industrial production and equipment management.

[0004] However, traditional bearing fault diagnosis methods rely on traditional signal processing, but as the data scale increases and the fault mode becomes more complicated, these methods are no longer sufficient to meet the challenge of feature extraction. With the development of machine learning, support vector machines and random forests have been introduced to extract bearing features, but experts are required to manually design features, which makes it difficult to capture deep data features. Deep learning methods have attracted attention for their ability to automatically extract deep features, but existing models such as convolutional neural networks and recurrent neural networks focus on extracting single features such as local feature extraction, making it difficult to capture global information, affecting diagnostic accuracy. Summary of the invention

[0005] In order to solve the technical problem that the deep learning methods in the prior art have attracted attention due to their ability to automatically extract deep-level features, but the existing models such as convolutional neural networks and recurrent neural networks focus on local feature extraction and are difficult to capture global information, thus affecting the accuracy of diagnosis, the present invention provides a bearing fault diagnosis method and system.

[0006] The technical solution provided by the embodiment of the present invention is as follows:

[0007] First aspect

[0008] A bearing fault diagnosis method provided by an embodiment of the present invention includes:

[0009] S1: collecting bearing vibration data, wherein the bearing vibration data includes a bearing vibration signal and a bearing fault type corresponding to the bearing vibration signal;

[0010] S2: The bearing vibration signal is converted into a wavelet two-dimensional time-frequency image, an angular field two-dimensional time-frequency image and a transfer field two-dimensional time-frequency image by using the continuous wavelet transform algorithm, the Gram angular field coding algorithm and the Markov transfer field algorithm respectively;

[0011] S3: Fusing the wavelet 2D time-frequency image, the angular field 2D time-frequency image and the transfer field 2D time-frequency image to generate a fused time-frequency image;

[0012] S4: Construct a bearing fault detection model with a Swin-T channel, a hybrid attention channel and a feature fusion module, wherein both the Swin-T channel and the hybrid attention channel are connected to the feature fusion module;

[0013] S5: inputting the bearing vibration data as a training set into the bearing fault detection model to train the bearing fault detection model until the predicted loss value of the bearing fault detection model is within a preset predicted loss value range;

[0014] S6: Acquire the vibration signal of the bearing to be detected;

[0015] S7: Input the vibration signal of the bearing to be detected into the trained bearing fault detection model, and output the predicted bearing fault type.

[0016] Second aspect

[0017] An embodiment of the present invention provides a bearing fault diagnosis system, comprising:

[0018] processor;

[0019] A memory having computer-readable instructions stored therein, wherein when the computer-readable instructions are executed by the processor, the bearing fault diagnosis method according to the first aspect is implemented.

[0020] The third aspect

[0021] An embodiment of the present invention provides a computer-readable storage medium, on which a computer program is stored. When the program is executed by a processor, the bearing fault diagnosis method of the first aspect is implemented.

[0022] The beneficial effects brought about by the technical solution provided by the embodiment of the present invention include at least:

[0023] In the present invention, by collecting bearing vibration signals and bearing fault types corresponding to the bearing vibration signals, a variety of algorithms such as continuous wavelet transform algorithm, Gram angular field coding algorithm and Markov transfer field algorithm are then used to generate different time-frequency images, and these images are fused to fully capture the multi-scale and multi-angle feature information in the bearing vibration signal. Subsequently, a bearing fault detection model with a Swin-T channel, a hybrid attention channel and a feature fusion module is established to further extract local features and global features, and the feature fusion module is used to improve the expression ability of the model and the fault detection accuracy. Finally, the accurate prediction of the fault type is achieved by predicting the bearing vibration signal to be detected. This method can effectively enhance the perception and recognition capabilities of complex vibration signals and enhance the extraction of deep-level data features through the fusion of multiple algorithms, multiple channels and multiple features, thereby improving the accuracy and robustness of fault diagnosis, reducing maintenance costs, and improving the safety and reliability of equipment. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of 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 creative work.

[0025] Figure 1 A schematic diagram of a flow chart of a bearing fault diagnosis method provided by an embodiment of the present invention;

[0026] Figure 2 A structural schematic diagram of a bearing fault detection model provided by an embodiment of the present invention;

[0027] Figure 3 A structural schematic diagram of a bearing fault diagnosis system provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0028] The technical solution of the present invention is described below in conjunction with the accompanying drawings.

[0029] In the embodiments of the present invention, words such as "exemplarily" and "for example" are used to indicate examples, illustrations or explanations. Any embodiment or design described as "example" in the present invention should not be interpreted as being more preferred or more advantageous than other embodiments or designs. Specifically, the use of the word "example" is intended to present the concept in a specific way. In addition, in the embodiments of the present invention, the meaning expressed by "and / or" can be both, or it can be either of the two.

[0030] In order to make the technical problems, technical solutions and advantages to be solved by the present invention more clear, a detailed description will be given below with reference to the accompanying drawings and specific embodiments.

[0031] Reference Manual Attached Figure 1 , showing a flow chart of a bearing fault diagnosis method provided by an embodiment of the present invention.

[0032] An embodiment of the present invention provides a bearing fault diagnosis method, which can be implemented by a bearing fault diagnosis device, which can be a terminal or a server. The processing flow of the bearing fault diagnosis method can include the following steps:

[0033] S1: Collect bearing vibration data, wherein the bearing vibration data includes a bearing vibration signal and a bearing fault type corresponding to the bearing vibration signal.

[0034] Among them, the bearing vibration signal refers to the vibration information generated by the bearing during operation collected by the sensor, usually expressed in the form of a time series. The bearing fault type is the classification label of the possible fault type obtained after analyzing these signals.

[0035] It should be noted that by collecting bearing vibration signals and corresponding fault type data, a complete data set is formed, which provides rich basic data for subsequent model training. This data collection process can capture the characteristic information of the bearing under different fault conditions, thereby providing diversified and comprehensive input for training an efficient and accurate fault diagnosis model, ensuring that the model can better identify and classify different fault types in practical applications, and improve the accuracy and reliability of diagnosis.

[0036] S2: The bearing vibration signal is converted into wavelet two-dimensional time-frequency image, angular field two-dimensional time-frequency image and transfer field two-dimensional time-frequency image by using continuous wavelet transform algorithm, Gram angular field coding algorithm and Markov transfer field algorithm respectively.

[0037] Among them, the continuous wavelet transform algorithm is a tool for transforming signals from the time domain to the time-frequency domain. It can provide localized characteristics of signals at different times and frequencies. The Gram angular field coding algorithm captures the angular relationship information of the time series by mapping the time series data to the polar coordinate angle difference on a two-dimensional plane. The Markov transition field algorithm is based on the Markov chain and represents the transition probability relationship of the time series as a matrix to capture the state transition characteristics of the sequence. The wavelet two-dimensional time-frequency image refers to the image calculated by the continuous wavelet transform algorithm, the angular field two-dimensional time-frequency image refers to the image calculated by the Gram angular field coding algorithm, and the transfer field two-dimensional time-frequency image refers to the image calculated by the Markov transition field algorithm.

[0038] It should be noted that three different algorithms are used to convert the bearing vibration signal, and three two-dimensional time-frequency images are generated, which describe the vibration signal from different angles and frequency domain characteristics. This multi-angle and multi-method processing can capture the changing characteristics of the signal more comprehensively, enhance the diversity and effectiveness of time-frequency images in expressing bearing fault characteristics, and thus provide more abundant information for subsequent model training and feature extraction, and improve the accuracy and robustness of diagnosis.

[0039] In a possible implementation, S2 specifically includes sub-steps S201 to S203:

[0040] S201: Convert the bearing vibration signal into a wavelet two-dimensional time-frequency image through a continuous wavelet transform algorithm.

[0041] S202: Convert the bearing vibration signal into an angular field two-dimensional time-frequency image by using a Gram angular field coding algorithm.

[0042] S203: Convert the bearing vibration signal into a two-dimensional time-frequency image of the transfer field through a Markov transfer field algorithm.

[0043] In a possible implementation, the wavelet two-dimensional time-frequency image is calculated as follows:

[0044]

[0045] Among them, CWT f (a, b) represents the continuous wavelet transform result of the bearing vibration signal at scale a and position b, that is, the generated wavelet two-dimensional time-frequency image, a represents the scale factor, b represents the translation factor, f(t) represents the input bearing vibration signal, represents the scaled and translated wavelet basis function, represents the normalization factor, represents the complex conjugate of the wavelet basis function after scaling and translation, and dt represents the differential element in the integral.

[0046] In a possible implementation manner, the angular field two-dimensional time-frequency image is specifically calculated as follows:

[0047]

[0048] Among them, GADF represents the matrix representing the two-dimensional time-frequency image of the angular field, α p Represents the polar coordinate angle of the pth point in the time series data, α q Represents the polar coordinate angle of the qth point in the time series data, p = 1, ..., m, m represents α p The total number of polar coordinate angles, q = 1, ..., n, n represents α q The total number of polar coordinate angles, sin(αm -α n ) represents α m and α n The sine of the difference.

[0049] In a possible implementation manner, the two-dimensional time-frequency image of the transfer field is calculated as follows:

[0050]

[0051] Where M represents the matrix representing the two-dimensional time-frequency image of the transfer field, m ij Represents the element in the i-th row and j-th column of the matrix, i = 1, 2, …, n, j = 1, 2, …, n, n represents the total number of rows and columns.

[0052] S3: Fusing the wavelet two-dimensional time-frequency image, the angle field two-dimensional time-frequency image and the transfer field two-dimensional time-frequency image to generate a fused time-frequency image.

[0053] Among them, the fused time-frequency image is a new image generated through this fusion process, which contains the time-frequency feature information extracted by various algorithms such as continuous wavelet transform, Gram angle field coding, and Markov transfer field.

[0054] It should be noted that by fusing the time-frequency images generated by different algorithms, we can fully combine the advantages of each algorithm in feature extraction to form a comprehensive image containing a variety of information. This can make up for the limitations of a single algorithm in capturing specific signal features, ensure that the fused image has higher expressive power and information density, and provide more comprehensive and detailed feature information for subsequent model input, thereby effectively improving the recognition ability and accuracy of the fault detection model.

[0055] In a possible implementation, S3 specifically includes:

[0056] The wavelet two-dimensional time-frequency image, the angle field two-dimensional time-frequency image and the transfer field two-dimensional time-frequency image are gray-scaled to obtain the corresponding grayscale images.

[0057] The grayscale images corresponding to the three methods are fused as RGB channels respectively, and finally the required two-dimensional fused time-frequency image containing rich feature information is obtained.

[0058] Among them, the RGB three channels refer to the red channel (Red), green channel (Green) and blue channel (Blue) of the image. The grayscale image generated by continuous wavelet transform is mapped to the red channel, the grayscale image generated by Gram angle field coding is mapped to the green channel, and the grayscale image generated by Markov transfer field is mapped to the blue channel. By fusing these grayscale images as RGB three channels, the features extracted by different algorithms can be combined into a color fused image. This image also contains the feature information extracted by three different algorithms, thereby improving the expression ability of information.

[0059] Reference Manual Attached Figure 2 , showing a structural schematic diagram of the bearing fault detection model provided by the present invention.

[0060] Figure 2 In the paper, the bearing fault detection model includes a Swin-T channel, a hybrid attention channel and a feature fusion module, wherein the Swin-T channel includes a single convolutional layer unit, a ReLU activation unit, a maximum pooling unit, a sliding window unit and a sliding window attention mechanism unit, and the hybrid attention channel includes a channel attention unit and a spatial attention unit.

[0061] S4: Construct a bearing fault detection model with a Swin-T channel, a hybrid attention channel and a feature fusion module, where both the Swin-T channel and the hybrid attention channel are connected to the feature fusion module.

[0062] Among them, the Swin-T channel is based on a part of the Swin Transformer architecture, which is used to extract local features of the image through a sliding window mechanism while considering local contextual information. The hybrid attention channel combines the channel attention and spatial attention mechanisms to capture the global features of the image more broadly. The feature fusion module is used to integrate local features and global features to generate a feature representation that contains information from both. The bearing fault detection model refers to a model for detecting bearing faults built using algorithms such as deep learning.

[0063] It should be noted that by introducing the Swin-T channel and the hybrid attention channel, the fused time-frequency image can be feature extracted from the local and global levels respectively. This dual channel design can establish a more comprehensive feature representation between local details and global structures, improving the model's ability to perceive complex features. By further integrating these two types of features through the feature fusion module, the fault diagnosis model can improve its ability to identify various types of faults and enhance the generalization performance and robustness of the model.

[0064] In one possible implementation, the Swin-T channel includes a single convolutional layer unit, a ReLU activation unit, a maximum pooling unit, a sliding window unit, and a sliding window attention mechanism unit;

[0065] The Swin-T channel is used to extract local features. The extraction process includes:

[0066] Input fused time-frequency image.

[0067] The local features of the fused time-frequency image are extracted through two-dimensional convolution operation, sliding window operation and sliding window attention mechanism operation in the Swin Transformer Block module:

[0068] X = MaxPool(ReLU(Conv2d(x)))

[0069] Among them, X represents local features, MaxPool represents the maximum pooling operation, ReLU represents the activation function, Conv2d represents the two-dimensional convolution operation, and x represents the fused time-frequency image.

[0070] In one possible implementation, the hybrid attention channel includes a channel attention unit and a spatial attention unit;

[0071] The mixed attention channel is used to extract global features. The extraction process specifically includes:

[0072] Input fused time-frequency image.

[0073] The tensor dimensions of the input fused time-frequency image are rearranged through the rearrangement operation in the channel attention unit so that the channel dimension is moved to the last dimension.

[0074] Among them, the channel attention unit is part of the hybrid attention channel, which is used to focus on the importance of different channels (feature layers). It enhances the expression of important features by adjusting the weights of each channel. The rearrangement operation refers to the reordering of the dimensions of the input multidimensional data. The tensor dimension refers to the dimension of the tensor in each direction (such as width, height, channel, etc.).

[0075] The features at each spatial position of the rearranged fused time-frequency image tensor are regarded as a feature vector.

[0076] Among them, the feature vector refers to a one-dimensional vector composed of the eigenvalues ​​on all channels at each spatial position. The feature vector can be understood as a collection of all channel values ​​at a spatial position, which together describe the image features at that position.

[0077] A spatial attention unit including the first convolutional layer and the second convolutional layer is constructed, and a batch normalization word unit and a ReLU activation subunit are connected between the first convolutional layer and the second convolutional layer in sequence.

[0078] Among them, the first convolutional layer and the second convolutional layer are the basic components of the convolutional neural network (CNN), which extracts local features of the input data through convolution operations. The spatial attention unit is a mechanism for guiding the model's attention on the spatial dimensions (i.e., height and width) of the feature map. Batch Normalization (BN) is a technology used to accelerate neural network training and improve stability. It normalizes the features in each small batch and keeps the mean and variance of the output within a stable range, thereby accelerating the convergence of the model and avoiding gradient vanishing or exploding. ReLU (Rectified Linear Unit) is one of the most commonly used activation functions in deep learning. It sets negative numbers in the input to 0 and keeps positive numbers unchanged.

[0079] The second convolutional layer restores the number of channels of the fused time-frequency image tensor to the desired number.

[0080] The global features of the fused time-frequency image tensor are extracted based on batch normalized word units.

[0081] It should be noted that by rearranging the dimensions of the tensor and placing the channel dimension at the end, it is convenient to independently analyze the features of each spatial position. This operation can effectively capture important information at the spatial position. Subsequently, the combination of convolutional layers with batch normalization and activation units not only improves the stability and nonlinear expression ability of feature extraction, but also enables the model to focus on important features at different spatial positions. Finally, by restoring the number of channels through the second convolutional layer, important feature information between different channels can be retained and integrated, thereby enhancing the model's understanding of global features and ultimately improving the accuracy and robustness of diagnosis.

[0082] In one possible implementation, the feature fusion module fuses the local features extracted by the Swin-T channel and the global features extracted by the hybrid attention channel:

[0083] F map =concat(αF ST ,βF HAM )

[0084] Among them, F map represents the fusion feature, F ST represents the local features extracted by the Swin-T channel, F HAM represents the global features extracted by the mixed attention channel, α represents the attention weight corresponding to the local features, and β represents the attention weight corresponding to the global features.

[0085] S5: Input the bearing vibration data as a training set into the bearing fault detection model to train the bearing fault detection model until the predicted loss value of the bearing fault detection model is within a preset predicted loss value range.

[0086] Among them, the training set refers to the bearing vibration data set used to train the model, including vibration signals and corresponding fault type labels. The data in the training set is used to allow the model to learn the fault characteristics corresponding to different vibration signals, so as to accurately identify faults. The prediction loss value refers to the difference measure between the model's prediction and the actual value, which is usually calculated through a certain loss function. A smaller prediction loss value indicates that the model has a higher prediction accuracy.

[0087] It should be noted that those skilled in the art can set the size of the preset predicted loss value interval according to actual needs, and the present invention is not limited here. Specifically, the predicted loss value interval is to monitor whether the loss value in the training process tends to be flat.

[0088] In a possible implementation, the fused time-frequency image is divided into a training set, a validation set, and a test set according to a preset ratio, and the preset ratio is specifically 7:2:1.

[0089] Among them, the training set is used to train the model. Through the training data, the model continuously adjusts internal parameters (such as the weights of the neural network) to learn the patterns and features in the data. The training set is the main data source for model learning. The validation set is used to evaluate the performance of the model during the training process to adjust the model's hyperparameters and prevent overfitting. The validation set provides a data source independent of the training set for performance checks during training. The test set is used to evaluate the final performance of the model. The data of the test set is completely independent of the training set and the validation set, simulating the performance of the model in the actual environment.

[0090] S6: Obtain the vibration signal of the bearing to be detected.

[0091] Among them, the bearing vibration signal to be detected refers to the bearing vibration data that needs to be diagnosed for fault, which is usually new data collected from actual application scenarios. These signal data have no labels, that is, it is impossible to know in advance whether there is a fault or the specific fault type.

[0092] It should be noted that by using a fully trained and optimized model, the operating status of the bearing can be effectively monitored and diagnosed in real time, thereby improving the safety and operating efficiency of the equipment and reducing the need for human intervention.

[0093] S7: Input the vibration signal of the bearing to be detected into the trained bearing fault detection model, and output the predicted bearing fault type.

[0094] Among them, the trained bearing fault detection model refers to a bearing fault detection model that has undergone data learning and parameter optimization in the previous training process. The model has the ability to extract features from the input vibration signal and classify it. The prediction of the bearing fault type is the output result of the model on the input vibration signal, that is, the classification prediction of the bearing's operating status or specific fault type.

[0095] It should be noted that by inputting the vibration signal to be detected into the trained detection model, the fault type of the bearing can be automatically output. In this way, potential problems of the bearing can be discovered in time at an early stage, thereby effectively preventing the occurrence of faults, reducing maintenance costs and improving the safety and reliability of the equipment.

[0096] The beneficial effects brought about by the technical solution provided by the embodiment of the present invention include at least:

[0097] In the present invention, by collecting bearing vibration signals and bearing fault types corresponding to the bearing vibration signals, a variety of algorithms such as continuous wavelet transform algorithm, Gram angular field coding algorithm and Markov transfer field algorithm are then used to generate different time-frequency images, and these images are fused to fully capture the multi-scale and multi-angle feature information in the bearing vibration signal. Subsequently, a bearing fault detection model with a Swin-T channel, a hybrid attention channel and a feature fusion module is established to further extract local features and global features, and the feature fusion module is used to improve the expression ability of the model and the fault detection accuracy. Finally, the accurate prediction of the fault type is achieved by predicting the bearing vibration signal to be detected. This method can effectively enhance the perception and recognition capabilities of complex vibration signals and enhance the extraction of deep-level data features through the fusion of multiple algorithms, multiple channels and multiple features, thereby improving the accuracy and robustness of fault diagnosis, reducing maintenance costs, and improving the safety and reliability of equipment.

[0098] Reference Manual Attached Figure 3 , showing a structural schematic diagram of a bearing fault diagnosis system provided by the present invention.

[0099] The present invention further provides a bearing fault diagnosis system 20, which is applied to the above-mentioned bearing fault diagnosis method, comprising:

[0100] Processor 201.

[0101] The memory 202 stores computer-readable instructions. When the computer-readable instructions are executed by the processor 201, the bearing fault diagnosis method of the method embodiment is implemented.

[0102] The bearing fault diagnosis system 20 provided by the present invention can execute the above-mentioned bearing fault diagnosis method and achieve the same or similar technical effects. To avoid repetition, the present invention will not go into details.

[0103] The beneficial effects brought about by the technical solution provided by the embodiment of the present invention include at least:

[0104] In the present invention, by collecting bearing vibration signals and bearing fault types corresponding to the bearing vibration signals, a variety of algorithms such as continuous wavelet transform algorithm, Gram angular field coding algorithm and Markov transfer field algorithm are then used to generate different time-frequency images, and these images are fused to fully capture the multi-scale and multi-angle feature information in the bearing vibration signal. Subsequently, a bearing fault detection model with a Swin-T channel, a hybrid attention channel and a feature fusion module is established to further extract local features and global features, and the feature fusion module is used to improve the expression ability of the model and the fault detection accuracy. Finally, the accurate prediction of the fault type is achieved by predicting the bearing vibration signal to be detected. This method can effectively enhance the perception and recognition capabilities of complex vibration signals and enhance the extraction of deep-level data features through the fusion of multiple algorithms, multiple channels and multiple features, thereby improving the accuracy and robustness of fault diagnosis, reducing maintenance costs, and improving the safety and reliability of equipment.

[0105] It should be understood that the processor in the embodiment of the present invention may be a central processing unit (CPU), and the processor may also be 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. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0106] It should also be understood that the memory in the embodiments of the present invention may be a volatile memory or a non-volatile memory, or may include both volatile and non-volatile memories. Among them, the non-volatile memory may be a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), or a flash memory. The volatile memory may be a random access memory (RAM), which is used as an external cache. By way of example and not limitation, many forms of random access memory (RAM) are available, such as static RAM (SRAM), dynamic random access memory (DRAM), synchronous DRAM (SDRAM), double data rate synchronous dynamic random access memory (DDR SDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous link DRAM (SLDRAM), and direct rambus RAM (DR RAM).

[0107] The above embodiments can be implemented in whole or in part by software, hardware (such as circuits), firmware or any other combination. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When a computer instruction or computer program is loaded or executed on a computer, a process or function according to an embodiment of the present invention is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. Computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, computer instructions can be transmitted from one website, computer, server or data center to another website, computer, server or data center by wired (such as infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server or data center that contains one or more available media sets. The available medium can be a magnetic medium (for example, a floppy disk, a hard disk, a tape), an optical medium (for example, a DVD), or a semiconductor medium. The semiconductor medium can be a solid-state hard disk.

[0108] It should be understood that the term "and / or" in this article is only a description of the association relationship of associated objects, indicating that there can be three relationships. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. A and B can be singular or plural. In addition, the character " / " in this article generally indicates that the associated objects before and after are in an "or" relationship, but it may also indicate an "and / or" relationship. Please refer to the context for specific understanding.

[0109] In the present invention, "at least one" means one or more, and "more than one" means two or more. "At least one of the following" or similar expressions refers to any combination of these items, including any combination of single or plural items. For example, at least one of a, b, or c can be represented by: a, b, c, ab, ac, bc, or abc, where a, b, c can be single or multiple.

[0110] It should be understood that in various embodiments of the present invention, the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0111] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0112] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described equipment, devices and units can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0113] In the several embodiments provided by the present invention, it should be understood that the disclosed devices, apparatuses and methods can be implemented in other ways. For example, the device embodiments described above are only schematic, for example, the division of units is only a logical function division, and there may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another device, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0114] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0115] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0116] If the function 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 this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the methods of various embodiments of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk, etc., various media that can store program codes.

[0117] An embodiment of the present invention provides a computer-readable storage medium on which a computer program is stored. When the program is executed by a processor, a bearing fault diagnosis method as described in the method embodiment is implemented.

[0118] A computer-readable storage medium provided by the present invention can implement the steps and effects of the bearing fault diagnosis method of the above method embodiment. To avoid repetition, the present invention will not go into details.

[0119] The beneficial effects brought about by the technical solution provided by the embodiment of the present invention include at least:

[0120] In the present invention, by collecting bearing vibration signals and bearing fault types corresponding to the bearing vibration signals, a variety of algorithms such as continuous wavelet transform algorithm, Gram angular field coding algorithm and Markov transfer field algorithm are then used to generate different time-frequency images, and these images are fused to fully capture the multi-scale and multi-angle feature information in the bearing vibration signal. Subsequently, a bearing fault detection model with a Swin-T channel, a hybrid attention channel and a feature fusion module is established to further extract local features and global features, and the feature fusion module is used to improve the expression ability of the model and the fault detection accuracy. Finally, the accurate prediction of the fault type is achieved by predicting the bearing vibration signal to be detected. This method can effectively enhance the perception and recognition capabilities of complex vibration signals and enhance the extraction of deep-level data features through the fusion of multiple algorithms, multiple channels and multiple features, thereby improving the accuracy and robustness of fault diagnosis, reducing maintenance costs, and improving the safety and reliability of equipment.

[0121] The above are only specific embodiments of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art who is familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

[0122] There are a few points to note:

[0123] (1) The drawings of the embodiments of the present invention only relate to the structures related to the embodiments of the present invention, and other structures may refer to the general design.

[0124] (2) For the sake of clarity, in the drawings used to describe the embodiments of the present invention, the thickness of the layers or regions is exaggerated or reduced, that is, these drawings are not drawn according to the actual scale. It is understood that when an element such as a layer, film, region or substrate is referred to as being "on" or "under" another element, the element may be "directly" "on" or "under" the other element or there may be intermediate elements.

[0125] (3) In the absence of conflict, the embodiments of the present invention and the features therein may be combined with each other to obtain new embodiments.

[0126] The above are only specific embodiments of the present invention, but the protection scope of the present invention is not limited thereto. The protection scope of the present invention shall be based on the protection scope of the claims.

Claims

1. A bearing fault diagnosis method, characterized in that: include: S1: collecting bearing vibration data, wherein the bearing vibration data includes a bearing vibration signal and a bearing fault type corresponding to the bearing vibration signal; S2: using a continuous wavelet transform algorithm, a Gram angular field coding algorithm and a Markov transfer field algorithm to convert the bearing vibration signal into a wavelet two-dimensional time-frequency image, an angular field two-dimensional time-frequency image and a transfer field two-dimensional time-frequency image respectively; S3: fusing the wavelet two-dimensional time-frequency image, the angle field two-dimensional time-frequency image and the transfer field two-dimensional time-frequency image to generate a fused time-frequency image; S4: constructing a bearing fault detection model having a Swin-T channel, a hybrid attention channel and a feature fusion module, wherein the Swin-T channel and the hybrid attention channel are both connected to the feature fusion module; S5: inputting the bearing vibration data as a training set into the bearing fault detection model to train the bearing fault detection model until the predicted loss value of the bearing fault detection model is within a preset predicted loss value range; S6: Acquire the vibration signal of the bearing to be detected; S7: Input the bearing vibration signal to be detected into the trained bearing fault detection model, and output the predicted bearing fault type.

2. The bearing fault diagnosis method according to claim 1, characterized in that: The S2 specifically includes: S201: converting the bearing vibration signal into a wavelet two-dimensional time-frequency image by a continuous wavelet transform algorithm; S202: converting the bearing vibration signal into an angular field two-dimensional time-frequency image by using a Gram angular field coding algorithm; S203: Convert the bearing vibration signal into a two-dimensional time-frequency image of the transfer field by using a Markov transfer field algorithm.

3. The bearing fault diagnosis method according to claim 2, characterized in that: The calculation method of the wavelet two-dimensional time-frequency image is specifically as follows: Among them, CWT f (a, b) represents the continuous wavelet transform result of the bearing vibration signal at scale a and position b, that is, the generated wavelet two-dimensional time-frequency image, a represents the scale factor, b represents the translation factor, f(t) represents the input bearing vibration signal, represents the scaled and translated wavelet basis function, represents the normalization factor, represents the complex conjugate of the wavelet basis function after scaling and translation, and dt represents the differential element in the integral.

4. The bearing fault diagnosis method according to claim 2, characterized in that: The calculation method of the angular field two-dimensional time-frequency image is specifically as follows: Among them, GADF represents the matrix representing the two-dimensional time-frequency image of the angular field, α p Represents the polar coordinate angle of the pth point in the time series data, α q Represents the polar coordinate angle of the qth point in the time series data, p = 1,,m, m represents α p The total number of polar coordinate angles, q = 1,,n, n represents α q The total number of polar coordinate angles, sin(α m -α n ) represents α m and α n The sine of the difference.

5. The bearing fault diagnosis method according to claim 2, characterized in that: The calculation method of the two-dimensional time-frequency image of the transfer field is specifically as follows: Where M represents the matrix representing the two-dimensional time-frequency image of the transfer field, m ij Represents the element in the i-th row and j-th column of the matrix, i = 1, 2,, n, j = 1, 2,, n, n represents the total number of rows and columns.

6. The bearing fault diagnosis method according to claim 1, characterized in that: The Swin-T channel includes a single convolutional layer unit, a ReLU activation unit, a maximum pooling unit, a sliding window unit, and a sliding window attention mechanism unit; The Swin-T channel is used to extract local features, and the extraction process specifically includes: Inputting the fused time-frequency image; The local features of the fused time-frequency image are extracted through two-dimensional convolution operation and sliding window operation and sliding window attention mechanism operation in the Swin Transformer Block module: X = MaxPool(ReLU(Conv2d(x))) Among them, X represents local features, MaxPool represents the maximum pooling operation, ReLU represents the activation function, Conv2d represents the two-dimensional convolution operation, and x represents the fused time-frequency image.

7. The bearing fault diagnosis method according to claim 1, characterized in that: The hybrid attention channel includes a channel attention unit and a spatial attention unit; The hybrid attention channel is used to extract global features, and the extraction process specifically includes: Inputting the fused time-frequency image; Rearranging the tensor dimension of the input fused time-frequency image through a rearrangement operation in the channel attention unit so that the channel dimension is moved to the last dimension; The feature at each spatial position of the rearranged fused time-frequency image tensor is regarded as a feature vector; Constructing a spatial attention unit including a first convolutional layer and a second convolutional layer, wherein a batch normalization word unit and a ReLU activation subunit are sequentially connected between the first convolutional layer and the second convolutional layer; The number of channels of the fused time-frequency image tensor is restored to a desired number through the second convolutional layer; The global features of the fused time-frequency image tensor are extracted according to the batch normalization word unit.

8. The bearing fault diagnosis method according to claim 1, characterized in that: The feature fusion module fuses the local features extracted by the Swin-T channel and the global features extracted by the hybrid attention channel: F map =concat(αF ST ,βF HAM ) Among them, F map represents the fusion feature, F ST represents the local features extracted by the Swin-T channel, F HAM represents the global features extracted by the mixed attention channel, α represents the attention weight corresponding to the local features, and β represents the attention weight corresponding to the global features.

9. A bearing fault diagnosis method system, characterized in that: include: processor; A memory having computer-readable instructions stored thereon, wherein when the computer-readable instructions are executed by the processor, the attack bypass blocking method based on network abnormal behavior analysis as described in any one of claims 1 to 8 is implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the bearing fault diagnosis method according to any one of claims 1 to 8 is implemented.

Images