Bearing fault diagnosis method and device, electronic equipment and storage medium

Abstract

The invention relates to the technical field of nuclear power plant equipment safety, in particular to a bearing fault diagnosis method and device, electronic equipment and a storage medium. According to the bearing fault diagnosis method provided by the embodiment of the invention, target bearing vibration data is acquired; performing multi-scale feature extraction on the target bearing vibration data through the bearing fault diagnosis model to obtain multi-scale bearing vibration features; carrying out bearing state detection on the target bearing based on the multi-scale bearing vibration characteristics to obtain a bearing vibration state; performing feature filtering processing on the multi-scale bearing vibration features to obtain filtered bearing vibration features; and carrying out bearing fault diagnosis on the target bearing according to the multi-scale bearing vibration characteristics, the bearing vibration state and the filtered bearing vibration characteristics to obtain a bearing fault category. Therefore, the accuracy of bearing fault diagnosis can be improved.

Description

Technical Field

[0001] The present application relates to the field of nuclear power bearing detection, and in particular, to a bearing fault diagnosis method, its device, electronic equipment, and storage medium. Background Art

[0002] Bearing fault diagnosis is the key to ensuring the stable operation of key equipment in nuclear power plants. However, with the continuous increase of bearing-related data in nuclear power plants (such as vibration signals of bearings in pumps, fans, and turbines, etc.), it is becoming increasingly difficult to detect bearing faults to achieve early detection and preventive maintenance of bearing faults, and there may be a situation where the detection of equipment faults in nuclear power plants is not timely.

[0003] Currently, bearing fault diagnosis methods usually extract features in bearing vibration signals through signal processing techniques such as Fourier transform and wavelet transform, and perform bearing fault diagnosis on this feature. However, this method usually relies too much on domain expert knowledge and experience, so that the extracted features cannot fully reflect the fault information of the bearing, and this method is easily affected by noise, resulting in low accuracy of bearing fault diagnosis. Therefore, how to improve the accuracy of bearing fault diagnosis is still a difficult problem to be solved in the industry. Summary of the Invention

[0004] The present application aims to at least solve one of the technical problems existing in the prior art. For this reason, the present application provides a bearing fault diagnosis method, its device, electronic equipment, and storage medium, which can improve the accuracy of bearing fault diagnosis.

[0005] According to the bearing fault diagnosis method of the first aspect embodiment of the present application, it includes:

[0006] Obtain target bearing vibration data; wherein, the target bearing vibration data is used to characterize the state and behavior of the target bearing during operation;

[0007] Perform multi-scale feature extraction on the target bearing vibration data through a pre-trained bearing fault diagnosis model to obtain multi-scale bearing vibration features;

[0008] Based on the multi-scale bearing vibration features, perform bearing state detection on the target bearing to obtain a bearing vibration state; wherein, the bearing vibration state is used to characterize the vibration degree of the target bearing;

[0009] Perform feature filtering processing on the multi-scale bearing vibration features to obtain filtered bearing vibration features;

[0010] Perform bearing fault diagnosis on the target bearing according to the multi-scale bearing vibration features, the bearing vibration state, and the filtered bearing vibration features to obtain a bearing fault category.

[0011] According to some embodiments of the present application, the bearing fault diagnosis of the target bearing based on the multi-scale bearing vibration characteristics, the bearing vibration state, and the filtered bearing vibration characteristics to obtain the bearing fault category includes:

[0012] Obtain the output feature weight and output bias term of the multi-scale bearing vibration characteristics;

[0013] Update the bearing vibration state according to the filtered bearing vibration characteristics to obtain the updated bearing vibration state;

[0014] Identify the bearing fault characteristics of the multi-scale bearing vibration characteristics according to the updated bearing vibration state, the output feature weight, and the output bias term to obtain the bearing fault characteristics;

[0015] Perform bearing fault identification on the target bearing according to a preset activation function to obtain the bearing fault category.

[0016] According to some embodiments of the present application, the performing bearing fault identification on the target bearing according to a preset activation function to obtain the bearing fault category includes:

[0017] Calculate the bearing fault category probability of the bearing fault characteristics according to the activation function to obtain the bearing fault category probability data;

[0018] Determine the bearing fault category according to the bearing fault category probability data and the bearing fault characteristics.

[0019] According to some embodiments of the present application, the determining the bearing fault category according to the bearing fault category probability data and the bearing fault characteristics includes:

[0020] Determine the bearing failure category according to the bearing fault category probability data;

[0021] Identify the fault diameter of the target bearing according to the bearing failure category and the bearing fault characteristics to determine the fault diameter corresponding to the target bearing;

[0022] Determine the bearing fault category according to the fault diameter and the bearing failure category.

[0023] According to some embodiments of the present application, the bearing state detection of the target bearing based on the multi-scale bearing vibration characteristics to obtain the bearing vibration state includes:

[0024] Obtain the historical vibration feature state and historical vibration feature weight of the multi-scale bearing vibration characteristics;

[0025] Obtain the input feature weights and input bias terms of the multi-scale bearing vibration characteristics, and perform bearing state identification on the target bearing according to the input feature weights, the input bias terms, the historical vibration feature state, and the historical vibration feature weights to obtain the bearing vibration state.

[0026] According to some embodiments of the present application, the multi-scale feature extraction of the target bearing vibration data by the pre-trained bearing fault diagnosis model to obtain the multi-scale bearing vibration characteristics includes:

[0027] Extract the low-frequency signal characteristics of the target bearing vibration data through the bearing fault diagnosis model to obtain the low-frequency bearing vibration characteristics;

[0028] Extract the high-frequency signal characteristics of the target bearing vibration data to obtain the high-frequency bearing vibration characteristics;

[0029] Fuse the low-frequency bearing vibration characteristics and the high-frequency bearing vibration characteristics to obtain the multi-scale bearing vibration characteristics.

[0030] According to some embodiments of the present application, the extraction of the low-frequency signal characteristics of the target bearing vibration data by the bearing fault diagnosis model to obtain the low-frequency bearing vibration characteristics includes:

[0031] Perform convolution processing on the target bearing vibration data through the bearing fault diagnosis model and a preset low-frequency convolution kernel to obtain a low-frequency bearing vibration feature map;

[0032] Perform max-pooling processing on the low-frequency bearing vibration feature map to obtain the low-frequency bearing vibration characteristics.

[0033] According to some embodiments of the present application, the performing max-pooling processing on the low-frequency bearing vibration feature map to obtain the low-frequency bearing vibration characteristics includes:

[0034] Obtain the pooling window and pooling stride of the low-frequency bearing vibration feature map;

[0035] Obtain the window maximum value in each pooling window of the low-frequency bearing vibration feature map according to the pooling stride, and splice the window maximum values to obtain the low-frequency bearing vibration characteristics.

[0036] According to some embodiments of the present application, the fusing of the low-frequency bearing vibration characteristics and the high-frequency bearing vibration characteristics to obtain the multi-scale bearing vibration characteristics includes:

[0037] Obtain the low-frequency feature elements of the low-frequency bearing vibration characteristics;

[0038] Obtain the high-frequency characteristic elements of the high-frequency bearing vibration characteristics;

[0039] Multiply the low-frequency characteristic elements and the high-frequency characteristic elements element by element to obtain the multi-scale bearing vibration characteristics.

[0040] According to some embodiments of the present application, the obtaining of the target bearing vibration data includes:

[0041] Collect the original bearing vibration data according to a preset original sampling rate;

[0042] Configure a corresponding downsampling rate for the original bearing vibration data;

[0043] Perform downsampling processing on the original bearing vibration data according to the downsampling rate to obtain the target bearing vibration data.

[0044] According to some embodiments of the present application, before the multi-scale feature extraction is performed on the target bearing vibration data through a pre-trained bearing fault diagnosis model to obtain the multi-scale bearing vibration characteristics, it further includes:

[0045] Obtain the original bearing fault diagnosis model and the training sample set; wherein, the training sample set includes a plurality of training bearing vibration data and the corresponding true fault categories for each of the training bearing vibration data;

[0046] Perform multi-scale feature extraction on the plurality of training bearing vibration data through the original bearing fault diagnosis model to obtain the training multi-scale bearing vibration characteristics;

[0047] Perform bearing state detection on the training multi-scale bearing vibration characteristics to obtain the training bearing vibration state;

[0048] Perform feature filtering processing on the training multi-scale bearing vibration characteristics to obtain the training filtered bearing vibration characteristics;

[0049] Perform bearing fault prediction on the target bearing according to the training bearing vibration state, the training filtered bearing vibration characteristics and the training multi-scale bearing vibration characteristics to obtain the predicted fault category;

[0050] Calculate the loss value between the predicted fault category and the true fault category according to a preset loss function;

[0051] Update the model parameter weights of the original bearing fault diagnosis model based on the loss value, and return to perform multi-scale feature extraction on the training bearing vibration data through the original bearing fault diagnosis model until the original bearing fault diagnosis model meets the preset training conditions to obtain the pre-trained bearing fault diagnosis model.

[0052] According to some embodiments of the present application, after diagnosing the bearing fault based on the multi-scale bearing vibration characteristics according to the bearing vibration state and the filtered bearing vibration characteristics and obtaining the bearing fault category, the method further includes:

[0053] Obtaining historical bearing operation data; wherein, the historical bearing operation data includes a plurality of historical bearing fault categories and fault handling measures corresponding to each historical bearing fault category;

[0054] Determining a target bearing fault category that matches the bearing fault category from the plurality of historical bearing fault categories;

[0055] For the target bearing, executing the fault handling measure corresponding to the target bearing fault category.

[0056] According to an embodiment of the second aspect of the present application, a bearing fault diagnosis device includes:

[0057] A bearing vibration data acquisition module, configured to acquire target bearing vibration data; wherein, the target bearing vibration data is used to characterize the state and behavior of the target bearing during operation;

[0058] A multi-scale feature extraction module, configured to perform multi-scale feature extraction on the target bearing vibration data through a pre-trained bearing fault diagnosis model to obtain multi-scale bearing vibration characteristics;

[0059] A bearing state detection module, configured to perform bearing state detection on the target bearing based on the multi-scale bearing vibration characteristics to obtain a bearing vibration state; wherein, the bearing vibration state is used to characterize the vibration degree of the target bearing;

[0060] A bearing feature filtering module, configured to perform feature filtering processing on the multi-scale bearing vibration characteristics to obtain filtered bearing vibration characteristics;

[0061] A bearing fault diagnosis module, configured to perform bearing fault diagnosis on the target bearing according to the multi-scale bearing vibration characteristics, the bearing vibration state, and the filtered bearing vibration characteristics to obtain a bearing fault category.

[0062] In a third aspect, an embodiment of the present application provides an electronic device, including: a memory and a processor, the memory stores a computer program, and when the processor executes the computer program, the bearing fault diagnosis method according to any one of the embodiments of the first aspect of the present application is implemented.

[0063] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, the storage medium stores a program, and when the program is executed by a processor, the bearing fault diagnosis method according to any one of the embodiments of the first aspect of the present application is implemented.

[0064] The bearing fault diagnosis method, device, electronic device, and storage medium according to the embodiments of the present application have at least the following beneficial effects: The bearing fault diagnosis method according to the embodiments of the present application needs to obtain the vibration data of the target bearing; wherein, the vibration data of the target bearing is used to characterize the state and behavior of the target bearing during operation; the pre-trained bearing fault diagnosis model is used to perform multi-scale feature extraction on the vibration data of the target bearing to obtain multi-scale bearing vibration features; the bearing state of the target bearing is detected based on the multi-scale bearing vibration features to obtain the bearing vibration state; wherein, the bearing vibration state is used to characterize the vibration degree of the target bearing; the multi-scale bearing vibration features are subjected to feature filtering processing to obtain filtered bearing vibration features; the bearing fault diagnosis of the target bearing is performed based on the multi-scale bearing vibration features, the bearing vibration state, and the filtered bearing vibration features to obtain the bearing fault category. In this way, the accuracy of bearing fault diagnosis can be improved.

[0065] The additional aspects and advantages of the present application will be partially given in the following description, partially become apparent from the following description, or be understood through the practice of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0066] The above and / or additional aspects and advantages of the present application will become apparent and be readily understood from the description of the embodiments in conjunction with the following drawings, wherein:

[0067] Figure 1 is a schematic flowchart of the bearing fault diagnosis method provided by the embodiment of the present application;

[0068] Figure 2 is Figure 1 a flowchart of step S101 in

[0069] Figure 3 is another schematic flowchart of the bearing fault diagnosis method provided by the embodiment of the present application;

[0070] Figure 4 is Figure 1 a flowchart of step S102 in

[0071] Figure 5 is Figure 4 a flowchart of step S401 in

[0072] Figure 6 is Figure 5 a flowchart of step S502 in

[0073] Figure 7 is Figure 4 a flowchart of step S403 in

[0074] Figure 8 is Figure 1The flowchart of step S103 in

[0075] Figure 9 is Figure 1 The flowchart of step S105 in

[0076] Figure 10 is Figure 9 The flowchart of step S904 in

[0077] Figure 11 is Figure 10 The flowchart of step S1002 in

[0078] Figure 12 Another schematic flowchart of the bearing fault diagnosis method provided by the embodiment of the present application;

[0079] Figure 13 is the schematic structural diagram of the bearing fault diagnosis device provided by the embodiment of the present application;

[0080] Figure 14 is the schematic hardware structure diagram of the electronic device provided by the embodiment of the present application. Specific embodiments

[0081] The embodiments of the present application will be described in detail below. The examples of the embodiments are shown in the accompanying drawings, where the same or similar reference numerals denote the same or similar elements or elements with the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary only for explaining the present application and should not be construed as limiting the present application.

[0082] In the description of the present application, the meaning of several is one or more, the meaning of multiple is more than two, greater than, less than, exceeding, etc. are understood as not including the present number, above, below, within, etc. are understood as including the present number. If there is a description of first and second, it is only for the purpose of distinguishing technical features and should not be construed as indicating or implying relative importance or implicitly indicating the quantity of the indicated technical features or the sequence of the indicated technical features.

[0083] In the description of the present application, it should be understood that the orientation or positional relationship indicated by terms such as up, down, left, right, front, back, etc. is based on the orientation or positional relationship shown in the accompanying drawings, and is only for the convenience of describing the present application and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be construed as limiting the present application.

[0084] In the description of this specification, the descriptions referring to terms such as "one embodiment", "some embodiments", "schematic embodiments", "examples", "specific examples", or "some examples" mean that the specific features, structures, materials, or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present application. In this specification, the schematic expressions of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described can be combined in a suitable manner in any one or more embodiments or examples.

[0085] In the description of the present application, it should be noted that unless otherwise clearly defined, words such as "set", "installed", and "connected" should be understood in a broad sense, and those skilled in the art can reasonably determine the specific meanings of the above words in the present application in combination with the specific content of the technical solution. In addition, the identification of specific steps hereinafter does not represent a limitation on the step sequence and execution logic. The execution sequence and execution logic between each step should be understood and inferred with reference to the content described in the embodiment.

[0086] Bearing fault diagnosis is the key to ensuring the stable operation of key equipment in nuclear power plants. However, with the continuous increase in bearing-related data in nuclear power plants (such as vibration signals of bearings in pumps, fans, and turbines, etc.), it is becoming increasingly difficult to detect bearing faults to achieve early detection and preventive maintenance of bearing faults, and it may occur that the fault detection of nuclear power plant equipment is not timely.

[0087] Currently, bearing fault diagnosis methods usually extract features in bearing vibration signals through signal processing techniques such as Fourier transform and wavelet transform, and perform bearing fault diagnosis on this feature. However, this method usually overly relies on the knowledge and experience of domain experts, making the extracted features unable to fully reflect the fault information of the bearing, and this method is easily affected by noise, resulting in a low accuracy of bearing fault diagnosis. Therefore, how to improve the accuracy of bearing fault diagnosis remains a difficult problem urgently to be solved in the industry.

[0088] Therefore, using a bearing fault diagnosis model to perform bearing fault diagnosis on target bearing vibration data to improve the accuracy of bearing fault diagnosis, without overly relying on the knowledge and experience of domain experts, can comprehensively consider the bearing feature information related to bearing fault diagnosis, which helps to improve the application effect of the bearing fault diagnosis model in bearing fault diagnosis.

[0089] This application aims to solve at least one of the technical problems existing in the prior art. For this purpose, this application proposes a bearing fault diagnosis method and its device, electronic device, and storage medium, which can improve the accuracy of bearing fault diagnosis.

[0090] The following is a further explanation based on the accompanying drawings:

[0091] Reference Figure 1 , the bearing fault diagnosis method according to the embodiments of the present application may include, but is not limited to:

[0092] Step S101, obtaining target bearing vibration data; wherein, the target bearing vibration data is used to characterize the state and behavior of the target bearing during operation;

[0093] Step S102, performing multi-scale feature extraction on the target bearing vibration data through a pre-trained bearing fault diagnosis model to obtain multi-scale bearing vibration features;

[0094] Step S103, performing bearing state detection on the target bearing based on the multi-scale bearing vibration features to obtain a bearing vibration state; wherein, the bearing vibration state is used to characterize the vibration degree of the target bearing;

[0095] Step S104, performing feature filtering processing on the multi-scale bearing vibration features to obtain filtered bearing vibration features;

[0096] Step S105, performing bearing fault diagnosis on the target bearing according to the multi-scale bearing vibration features, the bearing vibration state, and the filtered bearing vibration features to obtain a bearing fault category.

[0097] The bearing fault diagnosis method shown by steps S101 to S105 of the embodiments of the present application needs to obtain target bearing vibration data; wherein, the target bearing vibration data is used to characterize the state and behavior of the target bearing during operation; performing multi-scale feature extraction on the target bearing vibration data through a pre-trained bearing fault diagnosis model to obtain multi-scale bearing vibration features; performing bearing state detection on the target bearing based on the multi-scale bearing vibration features to obtain a bearing vibration state; wherein, the bearing vibration state is used to characterize the vibration degree of the target bearing; performing feature filtering processing on the multi-scale bearing vibration features to obtain filtered bearing vibration features; performing bearing fault diagnosis on the target bearing according to the multi-scale bearing vibration features, the bearing vibration state, and the filtered bearing vibration features to obtain a bearing fault category. In this way, the accuracy of bearing fault diagnosis can be improved.

[0098] Reference Figure 2 , according to some embodiments of the present application, step S101 of obtaining target bearing vibration data may include, but is not limited to:

[0099] Step S201, collecting original bearing vibration data according to a preset original sampling rate;

[0100] Step S202, configuring a corresponding downsampling rate for the original bearing vibration data;

[0101] Step S203: Downsample the original bearing vibration data according to the downsampling rate to obtain the target bearing vibration data.

[0102] In step S101, during the process of bearing fault diagnosis, obtaining high-quality target bearing vibration data provides important data support for subsequent bearing fault diagnosis. This step involves the acquisition and downsampling of bearing vibration signals, and this process can include the following sub-steps.

[0103] In step S201 of some embodiments, specifically, the original bearing vibration data can reflect the physical state and behavior of the bearing during operation. In a nuclear power plant, the bearing is a core component of key rotating equipment (such as pumps, fans, turbines, etc.), and its health condition is directly related to the operation efficiency and safety of the entire nuclear power plant. The original bearing vibration data can be the vibration data collected from the bearings of key equipment such as nuclear reactor coolant pumps, generators, or turbines. The original bearing vibration data can include, but is not limited to, the amplitude, frequency of bearing vibration, and impact pulse signals of bearing defects (such as wear, cracks, spalling, etc.).

[0104] Specifically, the original sampling rate refers to the sampling rate of the original bearing vibration signal, that is, the number of data points collected per second.

[0105] Specifically, the selection of the original sampling rate is usually at least twice the highest frequency component in the bearing vibration signal, which can avoid the aliasing phenomenon of the bearing vibration signal. Ensure that the collected bearing vibration data can fully capture the important features of the bearing vibration signal.

[0106] In step S202 of some embodiments, specifically, downsampling is a method to reduce the data volume without losing the key information of the bearing. The selection of the downsampling rate needs to be based on the requirements for the bearing fault characteristic frequency and the consideration of the possible noise frequencies in the bearing data.

[0107] In step S203 of some embodiments, specifically, the target bearing vibration data refers to the vibration data related to the key information of the bearing sampled from the original bearing vibration data.

[0108] Specifically, the original bearing vibration data can be downsampled by downsampling methods such as simple discard method, averaging method, interpolation method, or direct selection method to obtain the target bearing vibration data. The downsampled data will be used for subsequent fault diagnosis and analysis, such as feature extraction, pattern recognition, or training of machine learning models.

[0109] Specifically, downsampling the original bearing vibration data according to the downsampling rate can reduce the bearing data volume and retain the key information of the bearing, effectively reducing the computational amount of subsequent data feature extraction.

[0110] In the embodiments of the present application illustrated by steps S201 to S203, the collected original vibration data will contain comprehensive information on the operating state of the bearing, including its vibration patterns during normal operation and any abnormal vibrations, which may be caused by bearing faults. By performing downsampling on the original vibration data, it is possible to retain the data related to bearing vibration anomalies and preliminarily remove the vibration data unrelated to bearing faults, facilitating the subsequent improvement of the efficiency of vibration data feature extraction.

[0111] In some more specific embodiments of the present application, if the original sampling rate of the original bearing vibration data is 10 kHz, which means 10,000 bearing vibration data points are collected per second, and if the fault characteristic frequencies of the bearing mainly concentrate below 1 kHz, it can be determined that the downsampling rate is 1 kHz to reduce the amount of collected bearing vibration data while still being able to capture the fault characteristics of the bearing.

[0112] In step S102, in the embodiments of the present application, by obtaining a pre-trained bearing fault diagnosis model, comprehensive bearing vibration characteristics can be captured by extracting bearing vibration characteristics at multiple scales, and the pre-trained entity classification model can identify the fault categories for the multi-scale bearing vibration characteristics, significantly improving the accuracy of bearing fault diagnosis. This process may include several sub-steps of training the bearing fault diagnosis model before.

[0113] Refer to Figure 3 , according to some embodiments of the present application, before performing multi-scale feature extraction on the target bearing vibration data through the pre-trained bearing fault diagnosis model in step S102 to obtain multi-scale bearing vibration characteristics, the bearing fault diagnosis method may include, but is not limited to:

[0114] Step S301, extracting low-frequency signal characteristics from the target bearing vibration data through the bearing fault diagnosis model to obtain low-frequency bearing vibration characteristics;

[0115] Step S302, extracting high-frequency signal characteristics from the target bearing vibration data to obtain high-frequency bearing vibration characteristics;

[0116] Step S303, performing bearing state detection on the training multi-scale bearing vibration characteristics to obtain the training bearing vibration state;

[0117] Step S304, performing feature filtering processing on the training multi-scale bearing vibration characteristics to obtain the training filtered bearing vibration characteristics;

[0118] Step S305, predicting the bearing faults of the target bearing based on the training bearing vibration state, the training filtered bearing vibration characteristics, and the training multi-scale bearing vibration characteristics to obtain the predicted fault categories;

[0119] Step S306: Calculate the loss values of the predicted fault category and the true fault category according to a preset loss function;

[0120] Step S307: Update the model parameter weights of the original bearing fault diagnosis model based on the loss values, and return to perform multi-scale feature extraction on the training bearing vibration data through the original bearing fault diagnosis model until the original bearing fault diagnosis model meets the preset training conditions, obtaining a pre-trained bearing fault diagnosis model.

[0121] In step S301 of some embodiments, specifically, the training sample set includes multiple training bearing vibration data and the true fault category corresponding to each training bearing vibration data. Among them, the multiple training bearing vibration data can be the historical bearing operation vibration data of the target bearing and the bearing operation vibration data of the bearings with the same model as the target bearing in the nuclear power plant.

[0122] Specifically, the true fault category refers to the correct fault type label that matches the multiple training bearing vibration data. The true fault category can be a healthy state, a bearing fault diameter of 0.007, 0.014, or 0.021 inches corresponding to a rolling fault, a bearing fault diameter of 0.007, 0.014, or 0.021 inches corresponding to an inner ring fault, a bearing fault diameter of 0.007, 0.014, or 0.021 inches corresponding to an outer ring fault, etc.

[0123] Specifically, the original bearing fault diagnosis model can be a neural network model jointly composed of a multi-scale convolutional neural network and a long short-term memory network. The multi-scale convolutional neural network is used to extract the vibration characteristics of the bearing vibration data, and the long short-term memory network classifies the bearing faults for the vibration characteristics to obtain the bearing fault category.

[0124] Specifically, the training sample set is the basis for model learning, and the true fault category is the goal of model learning. By obtaining the training sample set, it can provide data support and a judgment criterion for bearing fault diagnosis for subsequent bearing feature extraction and fault diagnosis.

[0125] In step S302 of some embodiments, specifically, the multi-scale bearing vibration characteristics refer to the characteristics that fuse the low-frequency signal and high-frequency signal of the bearing vibration.

[0126] Specifically, different-sized convolutional kernels of the multi-scale convolutional neural network are used to capture bearing vibration characteristics at different scales. Larger convolutional kernels are used to capture the low-frequency characteristics of the bearing vibration data, which are related to the macroscopic state of the bearing, and smaller convolutional kernels are used to identify the high-frequency characteristics of the bearing vibration data, which may be related to the microscopic defects of the bearing, so as to enable the model to consider both the global and local information of the bearing vibration signal at the same time.

[0127] In step S303 of some embodiments, specifically, the training vibration state is used to characterize the vibration degree of the training bearing.

[0128] Specifically, it is determined whether to add the training multi-scale bearing vibration features and hidden state of the previous layer to the current cell state through a long short-term memory network. Among them, through the current cell state, the bearing vibration features that may affect the bearing fault category recognition can be preliminarily screened out, and the bearing vibration features are stored in the current cell state, and the bearing vibration features corresponding to the current cell state are transmitted to the input gate for bearing state detection.

[0129] Specifically, by performing bearing state detection on the training multi-scale bearing vibration features, the training bearing vibration state can be obtained, and the bearing vibration features related to the bearing fault category recognition can be extracted, which helps to improve the accuracy of bearing fault diagnosis subsequently.

[0130] In step S304 of some embodiments, specifically, the features most relevant to bearing fault recognition can be further selected from the training multi-scale bearing vibration features through the forget gate, and at the same time, the features irrelevant to bearing fault recognition such as the noise of the bearing vibration features are removed.

[0131] Specifically, during the training process, the model may learn thousands or even tens of thousands of features, which cover all aspects of the bearing vibration signal. However, not all of these features contribute to fault diagnosis. By performing feature filtering processing on the training multi-scale bearing vibration features, the training filtered bearing vibration features are obtained. By evaluating the importance and relevance of each training feature to bearing fault diagnosis, the features that have the greatest impact on bearing fault classification are selected, which can reduce the computational complexity of the model and improve the efficiency of bearing fault diagnosis.

[0132] In step S305 of some embodiments, specifically, predicting the fault category means the bearing fault type that the bearing fault detection model predicts is most likely to correspond to the training sample set.

[0133] Specifically, the output of the current layer of the long short-term memory network is based on the input at the current moment (i.e., the training multi-scale bearing vibration features) and the output at the previous moment (the training bearing vibration state and the training filtered bearing vibration features). Combining the softmax function to further process and analyze these features, the probabilities of the training bearing belonging to different bearing fault categories are calculated, and the softmax function converts the output of the network into probability values between 0 and 1.

[0134] Specifically, based on the training bearing vibration state, the training filtered bearing vibration characteristics, and the training multi-scale bearing vibration characteristics, bearing fault prediction is performed on the target bearing, realizing automated fault identification and classification, improving the efficiency and accuracy of bearing fault diagnosis, enabling the maintenance team of the nuclear power plant to respond more quickly to potential problems with the bearings, and enhancing the operational safety and reliability of the entire power plant.

[0135] In step S306 of some embodiments, specifically, the loss value is a quantitative metric that measures the difference between the predicted fault category and the true fault category label. The smaller its value, the closer the prediction of the bearing fault diagnosis model is to the actual situation.

[0136] Specifically, the loss function can be the mean squared error function.

[0137] Specifically, by calculating the loss values of the predicted fault category and the true fault category according to the preset loss function, the difference between the predicted fault category of the model and the true fault category label can be measured. Through the calculation of the loss value, it is convenient to subsequently determine the weight of the model parameters, update the model parameter weights, reduce the prediction error of bearing fault diagnosis, and contribute to improving the accuracy of bearing fault diagnosis.

[0138] In step S307 of some embodiments, specifically, after the loss value calculation is completed, the bearing fault diagnosis model enters the backpropagation stage. In this stage, the gradient of the loss function is calculated. According to the calculated gradient, an optimization algorithm (such as gradient descent or Adam algorithm, etc.) is used to update the parameter weights of the model, enabling the bearing fault diagnosis model to learn how to reduce the prediction errors of bearing fault categories and gradually improve the accuracy of bearing fault category prediction.

[0139] Specifically, the preset training condition can be that the loss value is less than the preset loss threshold.

[0140] Specifically, if the original bearing fault diagnosis model meets the preset training conditions, a pre-trained entity classification model is obtained, which can be used for actual bearing fault diagnosis tasks. Through the process of updating the bearing fault diagnosis model parameters, the bearing fault diagnosis model learns how to extract comprehensive bearing vibration characteristics from the training sample set and learns how to perform accurate bearing fault diagnosis based on the bearing vibration characteristics, contributing to improving the accuracy of the bearing fault diagnosis model in identifying bearing fault categories.

[0141] In some embodiments of the present application, by performing multi-scale feature extraction on the target bearing vibration data through the pre-trained bearing fault diagnosis model, it is possible to capture bearing vibration characteristics at different frequencies, comprehensively focus on the bearing vibration characteristics, and capture subtle changes in the bearing vibration signal, contributing to improving the accuracy of the model in capturing bearing vibration characteristics. In step S102, this process includes the following key operations:

[0142] Refer to Figure 4 , according to some embodiments of the present application, in step S102, a pre-trained bearing fault diagnosis model is used to perform multi-scale feature extraction on the target bearing vibration data to obtain multi-scale bearing vibration features, which may include, but are not limited to:

[0143] Step S401, using the bearing fault diagnosis model to perform low-frequency signal feature extraction on the target bearing vibration data to obtain low-frequency bearing vibration features;

[0144] Step S402, performing high-frequency signal feature extraction on the target bearing vibration data to obtain high-frequency bearing vibration features;

[0145] Step S403, fusing the low-frequency bearing vibration features and the high-frequency bearing vibration features to obtain multi-scale bearing vibration features.

[0146] In some embodiments of the present application, performing low-frequency signal feature extraction on the target bearing vibration data through the bearing fault diagnosis model is a key step. The model can deeply analyze the low-frequency components of the bearing vibration signal through this step, which helps to improve the model's ability to capture the macroscopic operating state of the bearing vibration features or the long-term trend of the bearing vibration. In step S401, this process includes the following key operations:

[0147] Refer to Figure 5 , according to some embodiments of the present application, in step S401, a bearing fault diagnosis model is used to perform low-frequency signal feature extraction on the target bearing vibration data to obtain low-frequency bearing vibration features, which may include, but are not limited to:

[0148] Step S501, using the bearing fault diagnosis model and a preset low-frequency convolution kernel to perform convolution processing on the target bearing vibration data to obtain a low-frequency bearing vibration feature map;

[0149] Step S502, performing max-pooling processing on the low-frequency bearing vibration feature map to obtain low-frequency bearing vibration features.

[0150] In step S501 of some embodiments, specifically, the low-frequency bearing vibration feature map refers to a visual representation of the low-frequency features extracted from the bearing vibration signal, and these feature maps reflect the distribution and intensity of the bearing vibration signal in the low-frequency range.

[0151] Specifically, the pre-trained bearing fault diagnosis model also includes a multi-scale convolutional neural network and a long short-term memory network. The different-sized low-frequency convolution kernels of the multi-scale convolutional neural network slide on the target bearing vibration data according to the preset convolution step length, and perform element-wise product summation on the data region covered by the low-frequency convolution kernel to determine the low-frequency bearing vibration feature map.

[0152] For example, in the bearing fault diagnosis of a nuclear power plant, the sizes of the low-frequency convolution kernels are 20×20 and 10×10. The 20x20 low-frequency convolution kernel slides on the vibration data with a stride of 2. In each sliding area, the 20x20 low-frequency convolution kernel performs an element-wise product summation with the corresponding part of the vibration data to generate an eigenvalue. The element-wise product summation operation is repeated for the remaining target bearing vibration data areas, and the first low-frequency convolution feature map is determined based on all the eigenvalues. The corresponding second low-frequency convolution feature map is also obtained through the 10x10 low-frequency convolution kernel, and the first low-frequency convolution feature map and the second low-frequency convolution feature map are fused to obtain the low-frequency bearing vibration feature map.

[0153] Specifically, by performing convolution processing on the target bearing vibration data through the low-frequency convolution kernel, a low-frequency bearing vibration feature map containing different scales can be generated, enabling the bearing fault diagnosis model to extract low-frequency features from the target bearing vibration data that are related to the target bearing and range from coarse to fine, providing data support for the bearing fault diagnosis.

[0154] Step S502 of some embodiments refers to Figure 6 According to some embodiments of the present application, step S502 performs max pooling processing on the low-frequency bearing vibration feature map to obtain the low-frequency bearing vibration features, which may include, but are not limited to:

[0155] Step S601, obtaining the pooling window and pooling stride of the low-frequency bearing vibration feature map;

[0156] Step S602, obtaining the window maximum value in each pooling window of the low-frequency bearing vibration feature map according to the pooling stride, and splicing the window maximum values to obtain the low-frequency bearing vibration features.

[0157] In some embodiments of the present application, by performing max pooling processing on the low-frequency bearing vibration feature map, it helps to extract and strengthen the key low-frequency features in the bearing vibration signal. At the same time, the data dimension of the bearing vibration features is reduced, the computational complexity during feature extraction by the model is decreased, and the efficiency of low-frequency bearing vibration feature extraction is improved. This process is carried out in step S505 and may include the following sub-steps.

[0158] For step S601 of some embodiments, specifically, the pooling window defines the area size for performing the pooling operation on the low-frequency bearing vibration feature map; the pooling stride determines the interval of window movement.

[0159] For step S602 of some embodiments, specifically, by finding the maximum value within each pooling window, this value represents the strongest signal of the bearing vibration features within that window.

[0160] Further, after performing max pooling, the maximum values obtained from each pooling window are concatenated to form a new feature vector, which constitutes the low-frequency bearing vibration feature. If the inner ring fault of the target bearing appears as specific peaks in the low-frequency vibration feature map, max pooling will ensure that these peaks are retained in the final feature vector.

[0161] The embodiments of the present application provided via steps S601 to S602 help to highlight the significant features in the bearing vibration signal while suppressing unimportant fluctuations and noises, enabling the model to focus more on those low-frequency signal components that are most crucial for bearing fault diagnosis.

[0162] The embodiments of the present application provided via steps S501 to S502 enable the bearing fault diagnosis model to effectively extract low-frequency features from the target bearing vibration data, providing global data support for bearing fault diagnosis.

[0163] In some more specific embodiments of the present application, the size of the low-frequency bearing vibration feature map is 100x100. If the pooling window is 2x2 and the pooling stride is 2, then the maximum value in each 2x2 region will be selected to represent the bearing vibration feature of that region. And the max pooling operation will traverse each 2x2 region of the low-frequency bearing vibration feature map to obtain four new feature values, and the largest one is selected as the representative of that region, so that the 100x100 feature map is converted into a 50x50 low-frequency bearing vibration feature map, and each value is the maximum value of its corresponding region.

[0164] In step S402 of some embodiments, specifically, the high-frequency bearing vibration feature map refers to the visual representation of the high-frequency features extracted from the bearing vibration signal, and these feature maps reflect the distribution and intensity of the bearing vibration signal in the high-frequency range.

[0165] Specifically, the high-frequency convolution kernel of the multi-scale convolutional neural network slides on the target bearing vibration data according to a preset convolution stride, and performs element-wise product summation on the data region covered by the high-frequency convolution kernel to determine the high-frequency bearing vibration feature map.

[0166] For example, in the bearing fault diagnosis of a nuclear power plant, if the size of the high-frequency convolution kernel is 6x6, the 6x6 convolution kernel slides on the vibration data with a stride of 2. In each sliding region, the 6x6 convolution kernel performs element-wise product summation with the corresponding part of the vibration data to generate a feature value, and the element-wise product summation operation is repeated for the remaining target bearing vibration data regions, and the high-frequency bearing vibration features are determined according to all the feature values.

[0167] Specifically, the high-frequency features are related to local damage or instantaneous changes of the target bearing. If there are impacts of rolling elements or minute cracks in bearing components in the target bearing, by using a smaller convolution kernel for feature extraction, the details of minute cracks in the target bearing can be captured, providing further data support for subsequent bearing fault diagnosis.

[0168] Referring Figure 7 , according to some embodiments of the present application, step S403 fuses the low-frequency bearing vibration features and the high-frequency bearing vibration features to obtain multi-scale bearing vibration features, which may include, but are not limited to:

[0169] Step S701, obtaining low-frequency feature elements of the low-frequency bearing vibration features;

[0170] Step S702, obtaining high-frequency feature elements of the high-frequency bearing vibration features;

[0171] Step S703, multiplying the low-frequency feature elements and the high-frequency feature elements element by element to obtain multi-scale bearing vibration features.

[0172] In some embodiments of the present application, fusing the low-frequency bearing vibration features and the high-frequency bearing vibration features enables the model to consider both the macroscopic and microscopic information of the target bearing vibration signal simultaneously. By integrating the bearing vibration features extracted from different frequency ranges into a unified feature representation, it provides more comprehensive data feature support for subsequent fault diagnosis. This process is carried out in step S403 and may include the following sub-steps.

[0173] In step S701 of some embodiments, specifically, the low-frequency feature elements are the feature elements extracted by a low-frequency convolution kernel, reflecting the distribution and intensity of the bearing vibration signal in the low-frequency range, and the low-frequency feature elements may reflect the overall operating state and long-term change trend of the bearing, such as the normal operating state of the target bearing, the foundation loosening state of the target bearing, etc.

[0174] In step S702 of some embodiments, specifically, the high-frequency feature elements are the feature elements extracted by a high-frequency convolution kernel, reflecting the rapid changes and impact features of the bearing in the bearing vibration signal, usually related to local damage or defects of the bearing, such as cracks, spalls generated by bearing rolling, or minute impacts between bearing components.

[0175] In step S703 of some embodiments, specifically, the process of multiplying element by element involves multiplying each element in the low-frequency feature vector by the corresponding element in the high-frequency feature vector to determine the multi-scale bearing vibration features.

[0176] Specifically, if the target bearing generates specific high-frequency vibration signals due to inner ring faults, and the macroscopic effects of such faults are also reflected in the low-frequency characteristics, then by fusing these characteristics through element-wise multiplication, these low-frequency and high-frequency abnormal vibration signals can be captured simultaneously. The fused multi-scale bearing vibration characteristics provide a more comprehensive description of the bearing vibration characteristics for the model, enabling the model to more accurately identify and classify different bearing faults.

[0177] In the embodiments of the present application shown in steps S701 to S703, when diagnosing bearing faults through multi-scale bearing vibration characteristics, it is not necessary to overly rely on domain expert knowledge and experience. Instead, it provides a comprehensive bearing information basis and effectively removes the noise effects unrelated to bearing faults, which helps the subsequent model to more effectively identify the health status and potential faults of the bearing.

[0178] In the embodiments of the present application shown in steps S401 to S403, the low-frequency convolution kernel is used to capture the low-frequency characteristics of the bearing vibration data, which are related to the overall operating state of the bearing, and the high-frequency convolution kernel is used to identify the high-frequency characteristics of the bearing vibration data, which are related to the microscopic defects of the bearing, and can provide global and local data support for subsequent bearing fault diagnosis.

[0179] Referring to Figure 8 , according to some embodiments of the present application, step S103 performs bearing state detection on the target bearing based on multi-scale bearing vibration characteristics to obtain the bearing vibration state, which may include, but is not limited to:

[0180] Step S801, obtaining the historical vibration characteristic state and historical vibration characteristic weight of the multi-scale bearing vibration characteristics;

[0181] Step S802, obtaining the input characteristic weight and input bias term of the multi-scale bearing vibration characteristics, and performing bearing state identification on the target bearing according to the input characteristic weight, input bias term, historical vibration characteristic state, and historical vibration characteristic weight to obtain the bearing vibration state.

[0182] In some embodiments of the present application, performing state detection on the target bearing based on multi-scale bearing vibration characteristics involves identifying the characteristics related to the bearing fault categories through multi-scale characteristics extracted from the vibration signals. In step S103, this process includes the following key operations:

[0183] Step S801 of some embodiments. Specifically, the historical vibration feature state refers to the vibration feature state of the previous moment of the long short-term memory network, that is, the hidden state of the previous moment; the historical vibration feature weight represents the importance of the vibration feature state of the previous moment. Among them, the output of the hidden state is the conversion of the cell state information within the current time step, and the hidden state has passed through the screening of the output gate, which determines which information is important and can be passed to the next part of the network or used for the final bearing fault category recognition.

[0184] Furthermore, if the historical vibration feature state may include the vibration modes of the bearing in the past few time steps, while the historical vibration feature weight represents the importance of these modes in the bearing state recognition, that is, if the model learns that a specific low-frequency feature frequently appears in the normal operating state of the bearing, then this specific feature will be given a higher weight in the historical vibration feature state.

[0185] Step S802 of some embodiments. Specifically, the input feature weight represents the importance of the multi-scale bearing vibration features for bearing fault diagnosis.

[0186] Specifically, the input bias term is used to help the model converge faster. Even if all the model input features are zero, the neural network can still produce a non-zero output, which increases the non-linear characteristics of the long short-term memory network and improves the speed of model convergence.

[0187] Specifically, the long short-term memory network also includes an input gate. Through the input gate, it can be determined which features in the collected multi-scale bearing vibration features (such as vibration frequency, amplitude or mode) will be transmitted to the current cell state for storage, facilitating subsequent feature analysis of the output gate for this current cell state.

[0188] Furthermore, the input of the input gate is composed of the hidden state of the previous moment and the multi-scale bearing vibration features. By analyzing the hidden state and the multi-scale bearing vibration features through the input gate, it can be determined whether the hidden state of the previous moment and the multi-scale bearing vibration need to be retained in the current cell state. Among them, the current cell state transfer is a way of information transfer within the long short-term memory network, which transfers information within the network along the time steps. Its main function is to store the long-term dependency information of the target bearing features, and the output of the cell state reflects the accumulation of all relevant information within the current time step, including the new information output by the input gate, the previously stored information output by the forget gate, and the updated state by these information.

[0189] Specifically, if the vibration frequency and vibration mode of the target bearing indicate that there may be a fault, then the vibration frequency and vibration mode are used as the bearing vibration state.

[0190] Through steps S801 to S802 shown in the embodiments of the present application, bearing condition detection based on multi-scale bearing vibration characteristics can help the model initially screen out information about the bearing health condition, help the model promptly identify and respond to potential bearing faults, and also contribute to improving the reliability of the operation of nuclear power plant equipment.

[0191] In step S104 of some embodiments, specifically, the long short-term memory network further includes a forgetting gate, and through the forgetting gate, features irrelevant to bearing fault diagnosis can be further filtered.

[0192] Specifically, when the forgetting gate performs feature filtering processing, it is achieved through the joint action of the forgetting feature weight, the forgetting bias term, the historical vibration feature state, and the historical vibration feature weight.

[0193] Specifically, if the model learns that some high-frequency features are highly correlated with the early fault signs of the bearing, while some low-frequency features are irrelevant to the bearing fault, then in the feature filtering stage, the model will tend to retain these high-frequency features and "forget" or reduce the weights of those low-frequency features. This helps the model focus on the features that are most helpful for identifying bearing faults, facilitating subsequent improvement of the accuracy and efficiency of bearing fault diagnosis.

[0194] Refer to Figure 9 , according to some embodiments of the present application, step S105 performs bearing fault diagnosis on the target bearing based on the multi-scale bearing vibration characteristics, the bearing vibration state, and the filtered bearing vibration characteristics to obtain the bearing fault category, which may include, but is not limited to:

[0195] Step S901, obtaining the output feature weight and the output bias term of the multi-scale bearing vibration characteristics;

[0196] Step S902, updating the bearing vibration state according to the filtered bearing vibration characteristics to obtain the updated bearing vibration state;

[0197] Step S903, identifying the bearing fault features from the multi-scale bearing vibration characteristics according to the updated bearing vibration state, the output feature weight, and the output bias term to obtain the bearing fault features;

[0198] Step S904, performing bearing fault identification on the target bearing according to the preset activation function to obtain the bearing fault category.

[0199] In some embodiments of the present application, performing fault diagnosis on the target bearing by combining the multi-scale bearing vibration characteristics, the bearing vibration state, and the filtered bearing vibration characteristics is a comprehensive analysis process. This process aims to accurately identify the current fault category of the bearing through the advanced feature extraction and bearing fault identification capabilities of the deep learning model. In step S105, this process includes several key operations:

[0200] For step S901 of some embodiments, specifically, the long short-term memory network further includes an output gate, and the output feature weight represents the importance of the multi-scale bearing vibration features analyzed by the output gate for bearing fault diagnosis.

[0201] Specifically, the output bias term is further used to help the model converge faster. Even if all the input features of the model are zero, the neural network can still produce a non-zero output, which increases the non-linear characteristics of the long short-term memory network and improves the convergence speed of the model.

[0202] For step S902 of some embodiments, specifically, the updated bearing vibration state can be combined with the cell state layer in the network through the long short-term memory network. The cell state layer is responsible for maintaining and updating the feature state of the bearing in the long short-term memory network and can capture the long-term dependencies in the bearing vibration signal.

[0203] Specifically, the forget gate can determine the multi-scale bearing vibration features that need to be filtered, and the input gate determines the influence of the multi-scale bearing vibration features input at the current moment on the current cell state. The output gate will control the bearing vibration state that needs to be updated to determine the final fault diagnosis result.

[0204] Specifically, if the high-frequency vibration mode related to the inner ring fault of the bearing is highlighted in the filtered bearing vibration features, the relevant features are stored in the cell state to update the bearing vibration state. The updated bearing vibration state reflects the current condition of the bearing after considering the filtered features and may include the wear degree or fault severity of the bearing, etc.

[0205] Specifically, updating the bearing vibration state according to the filtered bearing vibration features takes into account the influence of the bearing vibration features at different times on bearing fault diagnosis, that is, it considers the time characteristics of the bearing, and can capture the characteristic changes of the target bearing from the normal state to the fault state based on the changes of the bearing vibration features at different times.

[0206] For step S903 of some embodiments, specifically, the bearing fault feature represents the features related to the target bearing fault.

[0207] Specifically, the output gate performs a weighted sum of the multi-scale bearing vibration features, the learned output feature weight, and the bias term, which can further strengthen the features highly related to the bearing fault and at the same time suppress the unimportant multi-scale bearing vibration features, enabling the model to extract the most useful information for fault diagnosis from the multi-scale bearing vibration features and form the bearing fault feature.

[0208] Further, if the output gate recognizes the rolling bearing vibration characteristics related to the bearing rolling element fault, during the fault feature recognition process, an important weight is assigned to the rolling bearing vibration characteristics, so as to accurately extract the bearing fault characteristics related to the bearing fault.

[0209] In some more specific embodiments of the present application, the bearing condition detection can be represented by the following formula:

[0210] i t = σ(W i x t + V i h t-1 + b i )

[0211] Wherein, i t represents the bearing vibration state at time t, σ represents the sigmoid activation function, W i represents the input feature weight of the bearing vibration state, x t represents the multi-scale bearing vibration characteristics at time t, V i represents the historical vibration feature weight of the bearing vibration state, h t-1 represents the historical vibration feature state at time t-1, b i represents the input bias term of the bearing vibration state.

[0212] Specifically, the feature filtering process can be represented by the following formula:

[0213] f t = σ(W f x t + V f h t-1 + b f )

[0214] Wherein, f t represents the filtered bearing vibration characteristics at time t, σ represents the sigmoid activation function, W f represents the forgetting feature weight of the filtered bearing vibration characteristics, x t represents the multi-scale bearing vibration characteristics at time t, V f represents the historical vibration feature weight of the filtered bearing vibration characteristics, h t-1 represents the historical vibration feature state at time t-1, b f represents the forgetting bias term of the filtered bearing vibration characteristics.

[0215] Specifically, the bearing fault characteristics can be represented by the following formula:

[0216] o t = σ(W o x t + Vo h t-1 +b o )

[0217] Among them, o t represents the bearing fault feature at time t, σ represents the sigmoid activation function, W o represents the output feature weight of the bearing fault feature, x t represents the multi-scale bearing vibration feature at time t, V o represents the historical vibration feature weight of the bearing fault feature, h t-1 represents the historical vibration feature state at time t-1, b o represents the output bias term of the bearing fault feature.

[0218] Specifically, the cell state can be represented by the following formula:

[0219] c t = f t * c t-1 + i t * tanh(W c x t + V c h t-1 + b c )

[0220] Among them, c t represents the cell state at time t, f t represents the filtered bearing vibration feature at time t, c t-1 represents the cell state at time t-1, tanh represents the hyperbolic tangent activation function, W c represents the cell feature weight of the cell state, x t represents the multi-scale bearing vibration feature at time t, V c represents the historical vibration feature weight of the cell state, h t-1 represents the historical vibration feature state at time t-1, b c represents the cell bias term of the cell state.

[0221] Specifically, the hidden state can be represented by the following formula:

[0222] h t = o t * tanh(c t )

[0223] Among them, h t represents the hidden state at time t, o t represents the bearing fault feature at time t, tanh represents the hyperbolic tangent activation function, c t represents the cell state at time t.

[0224] Step S904 of some embodiments, refer to Figure 10 , according to some embodiments of the present application, in step S904, bearing fault identification is performed on the target bearing according to a preset activation function to obtain the bearing fault category, which may include, but is not limited to:

[0225] Step S1001, calculate the bearing fault category probability according to the bearing fault characteristics by the activation function to obtain bearing fault category probability data;

[0226] Step S1002, determine the bearing fault category according to the bearing fault category probability data and the bearing fault characteristics.

[0227] In some embodiments of the present application, using a preset activation function to perform fault identification on the target bearing can predict the fault category of the target bearing through the non-linear transformation of the activation function. In step S904, this process includes several key operations:

[0228] Step S1001 of some embodiments, specifically, the preset activation function may be the Softmax activation function.

[0229] Specifically, the Softmax function can be used to determine the bearing fault category probability data belonging to different bearing failure categories.

[0230] In some more specific embodiments of the present application, the bearing fault category probability data can be determined by the following formula:

[0231]

[0232] where q(z j ) represents the bearing fault category probability data of the jth bearing fault characteristic, z j represents the jth bearing fault characteristic, and z k represents the kth bearing failure category.

[0233] For example, if the bearing fault characteristics of the turbine are activated by the Softmax function, the bearing fault category probability data of 0.65 for the inner ring fault of the turbine bearing, 0.1 for the rolling element fault, 0.2 for the outer ring fault, and 0.05 for the normal state can be obtained.

[0234] Step S1002 of some embodiments, refer to Figure 11 , according to some embodiments of the present application, in step S1002, determine the bearing fault category according to the bearing fault category probability data and the bearing fault characteristics, which may include, but is not limited to:

[0235] Step S1101, determine the bearing failure category according to the bearing fault category probability data;

[0236] Step S1102: Identify the fault diameter of the target bearing based on the bearing failure category and bearing fault characteristics to determine the fault diameter corresponding to the target bearing.

[0237] Step S1103: Determine the bearing fault category based on the fault diameter and bearing failure category.

[0238] In some embodiments of the present application, determining the bearing fault category is a process based on the bearing fault failure category and bearing fault diameter identification. In step S1002, this process includes several key operations:

[0239] In step S1101 of some embodiments, specifically, the bearing failure category refers to the initial fault category that occurs in the target bearing during actual operation, which may include rolling failure category, outer ring failure category, inner ring failure category, and normal state category.

[0240] Specifically, the bearing failure category can take the bearing fault corresponding to the maximum bearing fault category probability data in the bearing fault category probability data as the bearing failure category.

[0241] Specifically, if the fault category probability data of the turbine bearing includes inner ring fault 0.65, rolling element fault 0.1, outer ring fault 0.2, and normal state 0.05, then the inner ring fault 0.65 is taken as the inner ring failure category of the target bearing.

[0242] In step S1102 of some embodiments, specifically, the fault diameter refers to the bearing wear diameter size after the bearing fails. It includes sizes of 0.007, 0.014, and 0.021 inches.

[0243] Specifically, the Softmax function can be further used to determine the fault diameter belonging to different bearing failure categories.

[0244] In step S1103 of some embodiments, specifically, the bearing fault categories include the category with a rolling fault diameter of 0.007, the category with a rolling fault diameter of 0.014, the category with a rolling fault of 0.021, the category with an inner ring fault diameter of 0.007, the category with an inner ring fault diameter of 0.014, the category with an inner ring fault of 0.021, the category with an outer ring fault diameter of 0.007, the category with an outer ring fault diameter of 0.014, the category with an outer ring fault of 0.021, and the normal state category.

[0245] Specifically, if the fault diameter corresponding to the bearing failure category is determined through the bearing fault diagnosis model, then the bearing failure category and its corresponding fault diameter are taken as the final bearing fault category.

[0246] Through steps S1101 to S1103 shown in the embodiments of the present application, not only can the failure categories of the target bearing be identified, but also the bearing fault diameters corresponding to different failure categories can be further determined, so as to determine the final bearing fault category, achieving accurate identification of 10 different bearing fault types and significantly improving the accuracy of bearing fault diagnosis.

[0247] Through steps S1001 to S1002 shown in the embodiments of the present application, the model can automatically extract key information from bearing fault features and accurately identify different fault categories of the bearing in combination with the activation function, significantly improving the accuracy of bearing fault diagnosis, which is of great significance for realizing early bearing fault detection and preventive maintenance strategies.

[0248] In some more specific embodiments of the present application, if the bearing fault features of the turbine are activated through the Softmax function, the bearing fault category probability data of 0.65 for the inner ring fault of the turbine bearing, 0.1 for the rolling element fault, 0.2 for the outer ring fault, and 0.05 for the normal state can be obtained. If the bearing failure category is determined to be the inner ring failure category, the inner ring failure category is input into the Softmax activation function, and the category probability for a fault diameter of 0.007 inches is 0.25, the category probability for a fault diameter of 0.014 inches is 0.05, and the category probability for a fault diameter of 0.021 inches is 0.7. Then, the fault diameter is determined to be 0.021 inches, and the category with an inner ring fault diameter of 0.014 is used as the bearing fault category of the target bearing.

[0249] Refer to Figure 12 According to some embodiments of the present application, after step S105 performs bearing fault diagnosis on the multi-scale bearing vibration features based on the bearing vibration state and filters the bearing vibration features to obtain the bearing fault category, the bearing fault diagnosis method may further include, but is not limited to:

[0250] Step S1201, obtaining bearing historical operation data; wherein, the bearing historical operation data includes multiple historical bearing fault categories and the fault handling measures corresponding to each historical bearing fault category;

[0251] Step S1202, determining the target bearing fault category that matches the bearing fault category from multiple historical bearing fault categories;

[0252] Step S1203, for the target bearing, executing the fault handling measures corresponding to the target bearing fault category.

[0253] In some embodiments of the present application, after the target bearing fault category is determined, it is necessary to further confirm the risk situation of the existence of this bearing fault category to achieve predictive maintenance of the target bearing. After step S1005, this process includes several key operations:

[0254] For step S1201 of some embodiments, specifically, the historical operation data of the bearing includes multiple historical bearing fault categories and corresponding fault handling measures for each historical bearing fault category.

[0255] Specifically, the historical bearing fault categories also include the category with a rolling fault diameter of 0.007, the category with a rolling fault diameter of 0.014, the category with a rolling fault of 0.021, the category with an inner ring fault diameter of 0.007, the category with an inner ring fault diameter of 0.014, the category with an inner ring fault of 0.021, the category with an outer ring fault diameter of 0.007, the category with an outer ring fault diameter of 0.014, the category with an outer ring fault of 0.021, and the normal state category.

[0256] Specifically, if the historical bearing has faults such as wear, cracks, or improper lubrication, it is necessary to further confirm whether the fault belongs to a rolling fault, an inner ring fault, or an outer ring fault, and further identify the bearing fault diameter corresponding to the fault to determine the final historical bearing fault category.

[0257] Specifically, the fault handling measures can include adjusting the bearing operation parameters, replacing the lubricant, cleaning the fault location, performing maintenance procedures, and replacing the bearing, etc.

[0258] Specifically, the historical operation data of the bearing can be obtained from the enterprise database of the nuclear power plant.

[0259] For step S1202 of some embodiments, specifically, the target bearing fault category refers to the fault category that matches the bearing fault category.

[0260] Specifically, first extract the historical vibration characteristics of the bearing historical operation data, compare the current vibration characteristics with the historical vibration characteristics to find the most similar matching characteristics, determine the bearing fault category of the current vibration characteristics according to the matching characteristics, and screen out the faults that match the bearing fault category from the historical bearing fault categories as the target bearing fault category.

[0261] Furthermore, finding the most similar matching characteristics can be determined by the Euclidean distance algorithm, the cosine similarity algorithm, or the Jaccard similarity algorithm.

[0262] Specifically, determining the target bearing fault category that matches the bearing fault category from multiple historical bearing fault categories can determine the bearings of the nuclear power plant equipment that need to be maintained, which helps to take effective bearing maintenance measures subsequently to prevent the further development of faults.

[0263] Step S1203 of some embodiments. Specifically, the fault handling measures refer to the solutions taken for different bearing fault categories, which may include, but are not limited to, replacement of bearing lubricant, cleaning of the bearing, alignment adjustment of the bearing, or replacement of the bearing, etc.

[0264] Specifically, the fault handling measures may also include preventive maintenance activities and physical operations, such as preventive maintenance like regular inspections and replacement of lubricants, and physical operations like replacement of components and change of the operating settings of the bearing machine, etc.

[0265] Specifically, after implementing the fault handling measures, the operating state of the target bearing needs to be re-evaluated. By using the bearing fault diagnosis model again to analyze the vibration data of the bearing, it is confirmed whether the bearing fault has been resolved. If the fault has not been resolved, it is necessary to return to the step of performing multi-scale feature extraction on the vibration data of the target bearing through the pre-trained bearing fault diagnosis model, re-analyze the bearing data, and select new handling measures until the bearing fault has been resolved.

[0266] Through steps S1201 to S1203 provided by the embodiments of the present application, by taking corresponding fault handling measures for the target bearing, it is of great significance for realizing early fault detection and preventive maintenance of nuclear power plant equipment, and can reduce the risk of unexpected shutdown of nuclear power plant equipment through predictive maintenance, effectively ensuring the stable operation of nuclear power plant equipment.

[0267] According to some more specific embodiments of the present application, for example, in a nuclear power plant, when abnormal vibration occurs in the target bearing, the abnormal vibration data will be recorded and compared with the vibration data in the historical fault database. If the abnormal vibration of the target bearing is most similar to the fault data of insufficient inner ring lubrication in the historical records, the fault category of the target bearing will be determined as an inner ring fault, and the lubricant will be replaced or adjusted to improve the bearing lubrication conditions.

[0268] It should be noted that in the embodiments of the present application, first, by obtaining the vibration data of the target bearing, the real-time dynamic behavior of the bearing including normal operation and potential faults can be captured, and through the pre-trained bearing fault diagnosis model to perform multi-scale feature extraction on the vibration data of the target bearing, the local and global information of the bearing vibration signal can be captured, providing a rich data basis for fault diagnosis; second, based on the multi-scale bearing vibration characteristics, the bearing state of the target bearing is detected to obtain the bearing vibration state, and the multi-scale bearing vibration characteristics are subjected to feature filtering processing to obtain the filtered bearing vibration characteristics, which can remove the noise characteristics irrelevant to the target bearing and retain the bearing vibration characteristics related to bearing fault diagnosis; finally, based on the multi-scale bearing vibration characteristics, the bearing vibration state and the filtered bearing vibration characteristics, the bearing fault diagnosis of the target bearing is performed, which can accurately identify different bearing fault types and improve the accuracy of bearing fault diagnosis.

[0269] Referring to Figure 13 , the bearing fault diagnosis device according to the second aspect embodiment of the present application may include, but is not limited to:

[0270] A bearing vibration data acquisition module 1301, configured to acquire target bearing vibration data; wherein, the target bearing vibration data is used to characterize the state and behavior of the target bearing during operation;

[0271] A multi-scale feature extraction module 1302, configured to perform multi-scale feature extraction on the target bearing vibration data through a pre-trained bearing fault diagnosis model to obtain multi-scale bearing vibration features;

[0272] A bearing state detection module 1303, configured to perform bearing state detection on the target bearing based on the multi-scale bearing vibration features to obtain a bearing vibration state; wherein, the bearing vibration state is used to characterize the vibration degree of the target bearing;

[0273] A bearing feature filtering module 1304, configured to perform feature filtering processing on the multi-scale bearing vibration features to obtain filtered bearing vibration features;

[0274] A bearing fault diagnosis module 1305, configured to perform bearing fault diagnosis on the target bearing according to the multi-scale bearing vibration features, the bearing vibration state, and the filtered bearing vibration features to obtain a bearing fault category.

[0275] It can be seen that the content in the above-mentioned bearing fault diagnosis method embodiments is applicable to the embodiments of this bearing fault diagnosis device. The functions specifically implemented by the embodiments of this bearing fault diagnosis device are the same as those of the above-mentioned bearing fault diagnosis method embodiments, and the beneficial effects achieved are also the same as those of the above-mentioned bearing fault diagnosis method embodiments.

[0276] Referring to Figure 14 , Figure 14 schematically shows the hardware structure of an electronic device in another embodiment. The electronic device includes:

[0277] A processor 1401, which can be implemented in a general-purpose CPU (Central Processing Unit), a microprocessor, an application-specific integrated circuit (ASIC), or one or more integrated circuits, etc., and is configured to execute relevant programs to implement the technical solutions provided by the embodiments of the present application;

[0278] The memory 1402 can be implemented in the form of a read-only memory (ROM), a static storage device, a dynamic storage device, or a random access memory (RAM), etc. The memory 1402 can store an operating system and other application programs. When implementing the technical solutions provided in the embodiments of this specification through software or firmware, the relevant program codes are stored in the memory 1402 and are called by the processor 1401 to execute the bearing fault diagnosis method of the embodiments of this application;

[0279] The input / output interface 1403 is used to implement information input and output;

[0280] The communication interface 1404 is used to implement communication interaction between this device and other devices. It can achieve communication through wired means (such as USB, network cable, etc.) or wireless means (such as mobile network, WI-FI, Bluetooth, etc.);

[0281] The bus 1405 transmits information between various components of the device (such as the processor 1401, the memory 1402, the input / output interface 1403, and the communication interface 1404);

[0282] Among them, the processor 1401, the memory 1402, the input / output interface 1403, and the communication interface 1404 achieve communication connections with each other inside the device through the bus 1405.

[0283] The embodiments of this application also provide a computer program product, which includes a computer program. The processor of the computer device reads and executes this computer program, so that the computer device executes to implement the above-mentioned bearing fault diagnosis method.

[0284] Terms such as "first", "second", "third", "fourth", etc. (if any) in the specification of this disclosure and the above-mentioned drawings are used to distinguish similar objects and do not necessarily need to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged under appropriate circumstances, so that the embodiments of this disclosure described here can be implemented in an order other than those illustrated or described here. In addition, the terms "include" and "contain" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or device that includes a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but may include other steps or units not clearly listed or inherent to these processes, methods, products, or devices.

[0285] It should be understood that in this disclosure, "at least one (item)" means one or more, and "a plurality" means two or more. "And / or" is used to describe the association relationship of associated objects, indicating that there can be three relationships. For example, "A and / or B" can mean: only A exists, only B exists, and both A and B exist simultaneously. Here, A and B can be singular or plural. The character " / " generally indicates that the associated objects before and after are in an "or" relationship. "At least one (one)" or similar expressions below refer to any combination of these items, which can include, but are not limited to, any combination of single items (ones) or plural items (ones). For example, at least one (one) of a, b, or c can mean: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, and c can be single or multiple.

[0286] It should be understood that in the description of the embodiments of this application, the meaning of a plurality (or multiple items) is more than two. Understandings such as greater than, less than, exceeding, etc. do not include the present number, and understandings such as above, below, within, etc. include the present number.

[0287] In several embodiments provided in this disclosure, it should be understood that the disclosed systems, devices, and methods can be implemented in other ways. For example, the device embodiments described above are only illustrative. For example, the division of units is only a logical function division. In actual implementation, there can be other division methods. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed couplings or direct couplings or communication connections to each other can be through some interfaces, and the indirect couplings or communication connections of devices or units can be in electrical, mechanical, or other forms.

[0288] The units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they can be located in one place or distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0289] In addition, the functional units in each embodiment of this disclosure can be integrated into one processing unit, or each unit can exist physically alone, or two or more units can be integrated into one unit. The above-mentioned integrated units can be implemented in the form of hardware or in the form of software functional units.

[0290] If an integrated unit 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 disclosure, in essence, or the part that contributes to the prior art, or all or part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and may include, but is not limited to, a number of instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods of the various embodiments of the present disclosure. The foregoing storage medium includes: various media that can store program codes, such as USB flash drives, mobile hard disks, read-only memories (ROM), random access memories (RAM), magnetic disks, or optical discs.

[0291] It should also be understood that the various embodiments provided in the embodiments of the present application can be combined arbitrarily to achieve different technical effects.

[0292] The above is a specific description of the embodiments of the present disclosure, but the present disclosure is not limited to the above embodiments. Those skilled in the art can also make various equivalent deformations or substitutions without departing from the spirit of the present disclosure, and these equivalent deformations or substitutions are all included within the scope defined by the claims of the present disclosure.

Claims

1. A bearing fault diagnosis method, characterized in that: include: Acquire target bearing vibration data; wherein the target bearing vibration data is used to characterize the state and behavior of the target bearing during operation; Performing multi-scale feature extraction on the target bearing vibration data through a pre-trained bearing fault diagnosis model to obtain a multi-scale bearing vibration feature; Performing bearing state detection on the target bearing based on the multi-scale bearing vibration characteristics to obtain a bearing vibration state; wherein the bearing vibration state is used to characterize the vibration degree of the target bearing; Performing feature filtering processing on the multi-scale bearing vibration features to obtain filtered bearing vibration features; A bearing fault diagnosis is performed on the target bearing according to the multi-scale bearing vibration characteristics, the bearing vibration state and the filtered bearing vibration characteristics to obtain a bearing fault category.

2. The method according to claim 1, characterized in that: The performing bearing fault diagnosis on the target bearing according to the multi-scale bearing vibration characteristics, the bearing vibration state and the filtered bearing vibration characteristics to obtain a bearing fault category includes: Obtaining output feature weights and output bias items of the multi-scale bearing vibration features; updating the bearing vibration state according to the filtered bearing vibration characteristics to obtain an updated bearing vibration state; According to the updated bearing vibration state, the output feature weight, and the output bias item, bearing fault feature identification is performed on the multi-scale bearing vibration feature to obtain a bearing fault feature; Bearing fault identification is performed on the target bearing according to a preset activation function to obtain the bearing fault category.

3. The method according to claim 2, characterized in that The step of performing bearing fault identification on the target bearing according to a preset activation function to obtain the bearing fault category includes: Calculate the probability of a bearing fault category for the bearing fault feature according to the activation function to obtain bearing fault category probability data; The bearing fault category is determined according to the bearing fault category probability data and the bearing fault characteristics.

4. The method according to claim 3, characterized in that The step of determining the bearing fault category according to the bearing fault category probability data and the bearing fault characteristics comprises: Determining a bearing failure category based on the bearing failure category probability data; performing fault diameter identification on the target bearing according to the bearing failure category and the bearing fault characteristics to determine a fault diameter corresponding to the target bearing; The bearing fault category is determined according to the fault diameter and the bearing failure category.

5. The method according to claim 1, characterized in that The performing bearing state detection on the target bearing based on the multi-scale bearing vibration characteristics to obtain the bearing vibration state includes: Obtaining a historical vibration feature state and a historical vibration feature weight of the multi-scale bearing vibration feature; The input feature weight and input bias term of the multi-scale bearing vibration feature are obtained, and the bearing state of the target bearing is identified according to the input feature weight, the input bias term, the historical vibration feature state and the historical vibration feature weight to obtain the bearing vibration state.

6. The method according to claim 1, characterized in that The multi-scale feature extraction of the target bearing vibration data by the pre-trained bearing fault diagnosis model to obtain the multi-scale bearing vibration feature includes: Extracting low-frequency signal features from the target bearing vibration data using the bearing fault diagnosis model to obtain low-frequency bearing vibration features; Extracting high-frequency signal features from the target bearing vibration data to obtain high-frequency bearing vibration features; The low-frequency bearing vibration characteristics and the high-frequency bearing vibration characteristics are fused to obtain the multi-scale bearing vibration characteristics.

7. The method according to claim 6, characterized in that The step of extracting low-frequency signal features from the target bearing vibration data using the bearing fault diagnosis model to obtain low-frequency bearing vibration features includes: The target bearing vibration data is convolved by the bearing fault diagnosis model and a preset low-frequency convolution kernel to obtain a low-frequency bearing vibration characteristic diagram; The low-frequency bearing vibration feature map is subjected to maximum pooling processing to obtain the low-frequency bearing vibration feature.

8. The method according to claim 7, characterized in that The performing maximum pooling processing on the low-frequency bearing vibration feature map to obtain the low-frequency bearing vibration feature includes: Obtaining a pooling window and a pooling step size of the low-frequency bearing vibration characteristic map; The window maximum value of the low-frequency bearing vibration characteristic graph in each pooling window is obtained according to the pooling step size, and the window maximum values ​​are spliced ​​to obtain the low-frequency bearing vibration characteristic.

9. The method according to claim 6, characterized in that The step of fusing the low-frequency bearing vibration feature and the high-frequency bearing vibration feature to obtain the multi-scale bearing vibration feature includes: Obtaining low-frequency characteristic elements of the low-frequency bearing vibration characteristics; Obtaining high-frequency characteristic elements of the high-frequency bearing vibration characteristics; The low-frequency characteristic element and the high-frequency characteristic element are multiplied element by element to obtain the multi-scale bearing vibration characteristic.

10. The method according to claim 1, characterized in that The step of obtaining target bearing vibration data comprises: Collecting raw bearing vibration data according to a preset raw sampling rate; Configuring a corresponding downsampling rate for the original bearing vibration data; The original bearing vibration data is downsampled according to the downsampling rate to obtain the target bearing vibration data.

11. The method according to claim 1, characterized in that: Before extracting multi-scale features from the target bearing vibration data using the pre-trained bearing fault diagnosis model to obtain multi-scale bearing vibration features, the method further includes: Obtaining the original bearing fault diagnosis model and training sample set; wherein the training sample set includes a plurality of training bearing vibration data and a real fault category corresponding to each of the training bearing vibration data; Performing multi-scale feature extraction on the plurality of training bearing vibration data through the original bearing fault diagnosis model to obtain training multi-scale bearing vibration features; Performing bearing state detection on the training multi-scale bearing vibration characteristics to obtain a training bearing vibration state; Performing feature filtering processing on the training multi-scale bearing vibration features to obtain training filtered bearing vibration features; Perform bearing fault prediction on the target bearing according to the training bearing vibration state, the training filtered bearing vibration feature and the training multi-scale bearing vibration feature to obtain a predicted fault category; Calculating the loss values ​​of the predicted fault category and the actual fault category according to a preset loss function; The model parameter weights of the original bearing fault diagnosis model are updated based on the loss value, and the multi-scale feature extraction of the training bearing vibration data is performed by the original bearing fault diagnosis model until the original bearing fault diagnosis model meets the preset training conditions, thereby obtaining the pre-trained bearing fault diagnosis model.

12. The method according to claim 1, characterized in that After performing bearing fault diagnosis on the multi-scale bearing vibration characteristics according to the bearing vibration state and the filtered bearing vibration characteristics to obtain the bearing fault category, the method further includes: Acquire historical operation data of the bearing; wherein the historical operation data of the bearing includes a plurality of historical bearing fault categories and a fault handling measure corresponding to each of the historical bearing fault categories; determining a target bearing fault category matching the bearing fault category from a plurality of historical bearing fault categories; For the target bearing, the fault handling measure corresponding to the target bearing fault category is executed.

13. A bearing fault diagnosis device, characterized in that: include: A bearing vibration data acquisition module is used to acquire target bearing vibration data; wherein the target bearing vibration data is used to characterize the state and behavior of the target bearing during operation; A multi-scale feature extraction module, used to extract multi-scale features from the target bearing vibration data through a pre-trained bearing fault diagnosis model to obtain multi-scale bearing vibration features; A bearing state detection module, used to perform bearing state detection on the target bearing based on the multi-scale bearing vibration characteristics to obtain a bearing vibration state; wherein the bearing vibration state is used to characterize the vibration degree of the target bearing; A bearing feature filtering module, used to perform feature filtering processing on the multi-scale bearing vibration features to obtain filtered bearing vibration features; A bearing fault diagnosis module is used to perform bearing fault diagnosis on the target bearing according to the multi-scale bearing vibration characteristics, the bearing vibration state and the filtered bearing vibration characteristics to obtain a bearing fault category.

14. An electronic device, characterized in that: include: A memory and a processor, wherein the memory stores a computer program, and the processor implements the bearing fault diagnosis method according to any one of claims 1 to 12 when executing the computer program.

15. A computer-readable storage medium, characterized in that: The storage medium stores a program, and the program is executed by a processor to implement the bearing fault diagnosis method according to any one of claims 1 to 12.

Images