Rolling bearing fault diagnosis method and system

Abstract

The invention provides a rolling bearing fault diagnosis method and system, and the method comprises the steps: collecting historical wheel set bearing signals with different fault types in real time, carrying out the feature extraction of the historical wheel set bearing signals based on a preset algorithm, and generating a corresponding fault sample according to a plurality of feature values extracted in real time; generating a corresponding training set, a verification set and a test set according to the fault sample, and training a preset CNN model through the training set, the verification set and the test set based on a preset rule to generate a corresponding bearing fault diagnosis model in real time; and acquiring an actual wheel set bearing signal correspondingly generated by the wheel set bearing in real time, and outputting a confusion matrix corresponding to the actual wheel set bearing signal in real time through the bearing fault diagnosis model so as to judge whether the wheel set bearing has a fault or not in real time according to the confusion matrix. According to the invention, whether the bearing has a fault can be objectively and accurately judged, and the working efficiency is correspondingly improved.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and particularly relates to a rolling bearing fault diagnosis method and system. Background Art

[0002] With the progress of technology and the rapid development of productivity, people have developed various types of bearings, which have been widely used in many fields, effectively improving the existing productivity and work efficiency at the same time.

[0003] Among them, existing rolling bearings, as important mechanical components, are prone to various faults under complex working conditions, which may lead to corresponding accidents. Based on this, it is necessary to perform real-time diagnosis during the actual operation of rolling bearings to determine whether a rolling bearing has failed in real time.

[0004] Furthermore, in the process of diagnosing rolling bearings in the prior art, most of them will collect vibration signals of mechanical equipment in real time and perform feature extraction processing on the vibration signals collected in real time to determine whether a rolling bearing has failed according to the processing results. However, in the actual application process, due to the weak fault features in the vibration signals and the difficulty of effectively and comprehensively extracting fault information by existing linear signal processing methods, the accuracy of fault identification is correspondingly reduced, and the work efficiency is correspondingly reduced. Summary of the Invention

[0005] Based on this, the purpose of the present invention is to provide a rolling bearing fault diagnosis method and system to solve the problem that it is difficult to comprehensively extract fault information of rolling bearings in the prior art, resulting in a reduction in fault identification accuracy.

[0006] The first aspect of the embodiment of the present invention proposes: A rolling bearing fault diagnosis method, wherein the method includes: Collect historical axle box bearing signals with different fault types in real time, and perform feature extraction processing on the historical axle box bearing signals based on a preset algorithm to generate corresponding fault samples according to a number of feature values extracted in real time; Generate a corresponding training set, validation set, and test set according to the fault samples, and train a preset CNN model through the training set, the validation set, and the test set based on preset rules to generate a corresponding bearing fault diagnosis model in real time; Obtain the actual axle box bearing signals generated by the axle box bearing in real time, and output a confusion matrix corresponding to the actual axle box bearing signals in real time through the bearing fault diagnosis model to determine whether the axle box bearing has failed in real time according to the confusion matrix.

[0007] The beneficial effects of the present invention are as follows: By collecting historical axle box bearing signals of different fault types in real time, corresponding fault samples can be generated comprehensively. Based on this, a training set, a validation set, and a test set for subsequent training are correspondingly made, and the preset CNN model can be trained comprehensively immediately through the current training set, validation set, and test set, and a required bearing fault diagnosis model can be obtained. Based on this, in the actual application process, only the real-time collected actual axle box bearing signals need to be correspondingly input into the current bearing fault diagnosis model, and it can objectively and accurately determine whether the current bearing has a fault, correspondingly improving the work efficiency.

[0008] Further, the step of performing feature extraction processing on the historical axle box bearing signals based on a preset algorithm to generate corresponding fault samples according to a plurality of feature values extracted in real time includes: When the historical axle box bearing signals are obtained in real time, real-time parsing processing is performed on the historical axle box bearing signals to detect the original time series contained therein in real time; Performing generalized composite coarse-graining processing on the original time series to generate a corresponding target time series in real time; Generating a plurality of the feature values according to the preset algorithm and the target time series, and generating the fault samples according to the plurality of feature values in real time, and each of the feature values is unique.

[0009] Further, the step of generating a plurality of the feature values according to the preset algorithm and the target time series includes: When the target time series is obtained in real time, a target scale factor adapted to the target time series is matched in real time in a preset database; Calculating a plurality of target probability values generated by the target time series in a discrete mode when the target scale factor is in real time through the preset algorithm; Setting the plurality of target probability values as the plurality of feature values, and performing integration processing on the plurality of feature values in real time to generate the fault samples correspondingly.

[0010] Further, the expression of the algorithm for performing generalized composite coarse-graining processing on the original time series to generate a corresponding target time series in real time is:

[0011] Among them, represents the target time series, s represents the target scale factor, b represents the sequence position, and x b represents the sequence value, It represents the average value of sequence values, h represents a constant, and j represents a sequence factor.

[0012] Further, the step of training the preset CNN model through the training set, the validation set, and the test set based on a preset rule to generate a corresponding bearing fault diagnosis model in real time includes: When the training set is obtained in real time, input the training set into the first convolutional layer of the preset CNN model correspondingly, so that the first convolutional layer outputs a first feature map correspondingly; Input the first feature map into the first pooling layer correspondingly, so that the first pooling layer outputs a first pooled feature map correspondingly, and input the first pooled feature map into the second convolutional layer correspondingly, so that the second convolutional layer outputs a second feature map correspondingly; Input the second feature map into the second pooling layer correspondingly, so that the second pooling layer outputs a second pooled feature map correspondingly, and generate the bearing fault diagnosis model according to the second pooled feature map.

[0013] Further, the step of generating the bearing fault diagnosis model according to the second pooled feature map includes: When the second pooled feature map is obtained in real time, input the second pooled feature map into the fully connected layer of the preset CNN model correspondingly, so that the fully connected layer outputs a corresponding one-dimensional vector; Input the one-dimensional vector into the hidden layer and the output layer correspondingly, so that the output layer outputs a perception vector corresponding to the training set in real time, and construct the bearing fault diagnosis model in real time according to the perception vector.

[0014] Further, the step of constructing the bearing fault diagnosis model in real time according to the perception vector includes: When the perception vector is obtained in real time, add a corresponding activation function and the perception vector to the perception vector in real time to generate a corresponding initial diagnosis function; Test and verify the initial diagnosis function through the test set and the validation set in sequence to construct the bearing fault diagnosis model correspondingly.

[0015] The second aspect of the embodiment of the present invention proposes: A rolling bearing fault diagnosis system, wherein the system includes: An extraction module, configured to collect historical axle box bearing signals with different fault types in real time, and perform feature extraction processing on the historical axle box bearing signals based on a preset algorithm to generate corresponding fault samples according to a number of feature values extracted in real time; A training module, configured to generate a corresponding training set, validation set, and test set according to the fault samples, and train a preset CNN model based on preset rules through the training set, the validation set, and the test set to generate a corresponding bearing fault diagnosis model in real time; A judgment module, configured to obtain in real time the actual wheel pair bearing signal generated by the wheel pair bearing, and output in real time a confusion matrix corresponding to the actual wheel pair bearing signal through the bearing fault diagnosis model, so as to judge in real time whether the wheel pair bearing fails according to the confusion matrix.

[0016] Further, the extraction module is specifically configured to: When the historical wheel pair bearing signal is obtained in real time, perform real-time parsing processing on the historical wheel pair bearing signal to detect in real time the original time series contained therein; Perform generalized composite coarse-graining processing on the original time series to generate a corresponding target time series in real time; Generate a plurality of the eigenvalue according to the preset algorithm and the target time series, and generate the fault samples in real time according to the plurality of the eigenvalue, and each of the eigenvalue is unique.

[0017] Further, the extraction module is specifically configured to: When the target time series is obtained in real time, match in real time a target scale factor adapted to the target time series in a preset database; Calculate in real time a plurality of target probability values generated by the target time series when the target scale factor is in a discrete mode through the preset algorithm; Set the plurality of target probability values as the plurality of eigenvalue, and perform integration processing on the plurality of eigenvalue in real time to generate the fault samples correspondingly.

[0018] Further, the algorithm expression for performing generalized composite coarse-graining processing on the original time series to generate a corresponding target time series in real time is:

[0019] Wherein, represents the target time series, s represents the target scale factor, b represents the sequence position, x b represents the sequence value, represents the average value of the sequence values, h represents a constant, and j represents a sequence factor.

[0020] Further, the training module is specifically configured to: When the training set is obtained in real time, input the training set into the first convolutional layer of the preset CNN model correspondingly, so that the first convolutional layer outputs a first feature map correspondingly; Input the first feature map into the first pooling layer correspondingly, so that the first pooling layer outputs a first pooled feature map correspondingly, and input the first pooled feature map into the second convolutional layer correspondingly, so that the second convolutional layer outputs a second feature map correspondingly; Input the second feature map into the second pooling layer correspondingly, so that the second pooling layer outputs a second pooled feature map correspondingly, and generate the bearing fault diagnosis model according to the second pooled feature map.

[0021] Further, the training module is specifically configured to: When the second pooled feature map is obtained in real time, input the second pooled feature map into the fully connected layer of the preset CNN model correspondingly, so that the fully connected layer outputs a corresponding one-dimensional vector; Input the one-dimensional vector into the hidden layer and the output layer correspondingly, so that the output layer outputs a perception vector corresponding to the training set in real time, and construct the bearing fault diagnosis model in real time according to the perception vector.

[0022] Further, the training module is specifically configured to: When the perception vector is obtained in real time, add a corresponding activation function and the perception vector to the perception vector in real time to generate a corresponding initial diagnosis function in real time; Test and verify the initial diagnosis function in turn through the test set and the validation set to construct the bearing fault diagnosis model correspondingly.

[0023] The third aspect of the embodiments of the present invention proposes: A computer includes a memory, a processor, and a computer program stored on the memory and executable on the processor. Wherein, when the processor executes the computer program, the rolling bearing fault diagnosis method described above is implemented.

[0024] The fourth aspect of the embodiments of the present invention proposes: A readable storage medium stores a computer program thereon. Wherein, when the program is executed by a processor, the rolling bearing fault diagnosis method described above is implemented.

[0025] The additional aspects and advantages of the present invention will be partly given in the following description, partly will become obvious from the following description, or be understood through the practice of the present invention. Description of the Drawings

[0026] Figure 1Flowchart of the rolling bearing fault diagnosis method provided by the first embodiment of the present invention; Figure 2 Block diagram of the structure of the rolling bearing fault diagnosis system provided by the third embodiment of the present invention.

[0027] The following specific embodiments will further illustrate the present invention in conjunction with the above-mentioned drawings. Specific Embodiments

[0028] To facilitate the understanding of the present invention, the present invention will be described more comprehensively below with reference to the relevant drawings. Several embodiments of the present invention are given in the drawings. However, the present invention can be implemented in many different forms and is not limited to the embodiments described herein. On the contrary, these embodiments are provided to make the disclosure of the present invention more thorough and comprehensive.

[0029] It should be noted that when an element is referred to as being "fixed to" another element, it can be directly on the other element or there can also be an intermediate element. When an element is considered to be "connected" to another element, it can be directly connected to the other element or there may be an intermediate element at the same time. The terms "vertical", "horizontal", "left", "right" and similar expressions used herein are for illustrative purposes only.

[0030] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the technical field to which the present invention belongs. The terms used in the description of the present invention herein are only for the purpose of describing specific embodiments and are not intended to limit the present invention. The term "and / or" used herein includes any and all combinations of one or more of the related listed items.

[0031] Please refer to Figure 1 , which shows the rolling bearing fault diagnosis method provided by the first embodiment of the present invention. The rolling bearing fault diagnosis method provided in this embodiment can quickly and accurately determine whether there is a problem with the bearing, correspondingly improving the work efficiency.

[0032] Specifically, this embodiment provides: A rolling bearing fault diagnosis method, specifically including the following steps: Step S10, collect historical axle box bearing signals with different fault types in real time, and perform feature extraction processing on the historical axle box bearing signals based on a preset algorithm to generate corresponding fault samples according to a number of feature values extracted in real time; Step S20, generate a corresponding training set, validation set, and test set according to the fault samples, and train a preset CNN model through the training set, the validation set, and the test set based on a preset rule to generate a corresponding bearing fault diagnosis model in real time; Step S30: Obtain the actual axle bearing signal generated by the axle bearing in real time, and output the corresponding confusion matrix corresponding to the actual axle bearing signal in real time through the bearing fault diagnosis model, so as to judge whether the axle bearing fails according to the confusion matrix in real time.

[0033] Specifically, in this embodiment, it should be noted first that in order to objectively and accurately judge whether there is a problem with the bearing, it is necessary to understand the working state of the bearing and the corresponding working information in real time. In addition, in order to automatically complete the determination of the bearing working state, it is necessary to further construct a bearing fault diagnosis model adapted to each type of bearing. Based on this, for the convenience of implementation, the present invention will set up a server in the background. Specifically, this server can collect historical axle bearing signals with different fault types that have already been generated in real time. At the same time, in order to improve the accuracy of subsequent model training, it is necessary to further analyze the current historical axle bearing information. Preferably, the present invention will further perform feature extraction processing on the current historical axle bearing signal, and can further extract several required feature values inside the current historical axle bearing signal in real time. At the same time, based on the current feature values, the required fault samples can be created in real time for subsequent processing.

[0034] Furthermore, after obtaining the required fault samples in real time through the above steps, the current fault samples will be further split. Preferably, the present invention will split the current fault samples into the required training set, validation set, and test set according to the ratio of 7:1:2. Based on this, the pre-set CNN model can be trained correspondingly, and the required bearing fault diagnosis model can be trained according to the pre-set rules. On this basis, the present invention will receive the actual axle bearing signal generated by the existing bearing during actual operation in real time through this bearing fault diagnosis model. At the same time, this bearing fault diagnosis model can output the required confusion matrix. Based on this, it can be directly judged whether there is a problem with the current bearing according to this confusion matrix, so as to objectively and effectively judge whether there is a problem with the bearing, and correspondingly improve the work efficiency.

[0035] Second Embodiment Furthermore, the step of performing feature extraction processing on the historical axle bearing signal based on a preset algorithm to generate corresponding fault samples according to several feature values extracted in real time includes: When the historical axle bearing signal is obtained in real time, perform real-time analysis processing on the historical axle bearing signal to detect the original time series contained inside the historical axle bearing signal in real time; Perform generalized composite coarse-graining processing on the original time series to generate a corresponding target time series in real time; Generate a number of the eigenvalue according to the preset algorithm and the target time series, and generate the fault sample according to a number of the eigenvalue in real time, and each of the eigenvalue is unique.

[0036] Further, the step of generating a number of the eigenvalue according to the preset algorithm and the target time series includes: When the target time series is obtained in real time, match a target scale factor adapted to the target time series in a preset database in real time; Calculate a number of target probability values generated by the target time series when the target scale factor is in a discrete mode through the preset algorithm in real time; Set a number of the target probability values as a number of the eigenvalue, and perform integration processing on a number of the eigenvalue in real time to generate the fault sample correspondingly.

[0037] Further, the expression of the algorithm for performing generalized composite coarse-graining processing on the original time series to generate a corresponding target time series in real time is:

[0038] Among them, represents the target time series, s represents the target scale factor, b represents the sequence position, x b represents the sequence value, represents the average value of the sequence value, h represents a constant, and j represents the sequence factor.

[0039] Further, the steps of training a preset CNN model through the training set, the validation set, and the test set based on preset rules to generate a corresponding bearing fault diagnosis model in real time include: When the training set is obtained in real time, input the training set into the first convolutional layer of the preset CNN model correspondingly, so that the first convolutional layer outputs a first feature map correspondingly; Input the first feature map into the first pooling layer, so that the first pooling layer outputs a first pooled feature map correspondingly, and input the first pooled feature map into the second convolutional layer, so that the second convolutional layer outputs a second feature map correspondingly; Input the second feature map into the second pooling layer, so that the second pooling layer outputs a second pooled feature map correspondingly, and generate the bearing fault diagnosis model according to the second pooled feature map.

[0040] Further, the step of generating the bearing fault diagnosis model according to the second pooled feature map includes: When the second pooled feature map is obtained in real time, input the second pooled feature map into the fully connected layer of the preset CNN model, so that the fully connected layer outputs a corresponding one-dimensional vector; Input the one-dimensional vector into the hidden layer and the output layer, so that the output layer outputs a perception vector corresponding to the training set in real time, and construct the bearing fault diagnosis model in real time according to the perception vector.

[0041] Further, the step of constructing the bearing fault diagnosis model in real time according to the perception vector includes: When the perception vector is obtained in real time, add a corresponding activation function and perception vector to the perception vector in real time to generate a corresponding initial diagnosis function in real time; Test and verify the initial diagnosis function in turn through the test set and the validation set to construct the bearing fault diagnosis model correspondingly.

[0042] In addition, in this embodiment, it should also be noted that after the required historical wheel pair bearing signals are obtained in real time through the above steps, in order to further objectively and effectively train the required bearing diagnosis model, the current historical wheel pair bearing signals will be further analyzed. It should be pointed out that since the existing signals are generated within a certain time threshold, based on this, the present invention will further detect the original time series adapted to the current signal. It should be noted that the original time series contains a series of useful numerical values, and the required eigenvalue can be further extracted from the current original time series. Based on this, the present invention will further perform generalized composite coarse-graining processing on the current original time series, that is, adjust the original time series to a standard format and further generate the required target time series. Based on this, the present invention will further calculate a number of probability values generated when the current target time series is in the discrete mode of the above scale factor, that is, the accuracy of the real-time judgment result. At the same time, the present invention will further calculate the probability mean value corresponding to the current number of probability values. It should be pointed out that the present invention will set the above number of probability values as the required number of eigenvalues for subsequent processing.

[0043] Further, after separately producing the required training set, validation set, and test set through the above steps, the present invention will start the actual training. Preferably, the present invention will first input the current training set into the first convolutional layer of the above-mentioned preset CNN model, and the current first convolutional layer can correspondingly output the required first feature map. Among them, the size of the first feature map is 14x1, the number of convolutional kernels is 16, the size is 1x1, the stride is 1, and the padding method is "valid". Based on this, the current first feature map will be further input into the first pooling layer. Correspondingly, the first pooling layer can output a first pooled feature map with a size of 14x1 and the number of channels unchanged, and the output form of the first pooled feature map is None×14×16. Based on this, the current first pooled feature map will be further input into the second convolutional layer, and the second feature map can be correspondingly output. At the same time, the current second feature map will be input into the second pooling layer, and the corresponding second pooled feature map can be further output. It should be noted that the size of the second pooled feature map is the same as the size of the above-mentioned first pooled feature map. On this basis, the present invention will further input the current second pooled feature map into the fully connected layer, and the current fully connected layer can correspondingly output the required one-dimensional vector. Specifically, the size of the one-dimensional vector is 224x1. Based on this, the current one-dimensional vector is sequentially input into the hidden layer and the output layer, and the current output layer can finally output a perception vector with a size of None×7. Based on this, the current perception vector is further verified and perceived through the above-mentioned validation set and test set respectively, and the required bearing fault diagnosis model can be finally obtained, and the subsequent diagnosis process can be correspondingly completed, thereby correspondingly improving the work efficiency.

[0044] Please refer to Figure 2 , the third embodiment of the present invention provides: A rolling bearing fault diagnosis system, wherein the system includes: An extraction module, configured to collect historical axle box bearing signals with different fault types in real time, and perform feature extraction processing on the historical axle box bearing signals based on a preset algorithm to generate corresponding fault samples according to a plurality of feature values extracted in real time; A training module, configured to generate a corresponding training set, validation set, and test set according to the fault samples, and train a preset CNN model through the training set, the validation set, and the test set based on a preset rule to generate a corresponding bearing fault diagnosis model in real time; A judgment module, configured to obtain the actual axle box bearing signal generated by the axle box bearing in real time, and output a confusion matrix corresponding to the actual axle box bearing signal in real time through the bearing fault diagnosis model, so as to judge in real time whether the axle box bearing has a fault according to the confusion matrix.

[0045] Further, the extraction module is specifically configured to: When the historical axle box bearing signal is obtained in real time, perform real-time parsing and processing on the historical axle box bearing signal to detect in real time the original time series contained therein; Perform generalized composite coarse-graining processing on the original time series to generate a corresponding target time series in real time; Generate a number of the eigenvalue according to the preset algorithm and the target time series, and generate the fault sample in real time according to the number of the eigenvalue, and each of the eigenvalue is unique.

[0046] Further, the extraction module is specifically configured to: When the target time series is obtained in real time, match in real time in the preset database a target scale factor adapted to the target time series; Calculate in real time by the preset algorithm a number of target probability values generated when the target scale factor of the target time series is in a discrete mode; Set the number of the target probability values as the number of the eigenvalue, and perform integration processing on the number of the eigenvalue in real time to generate the fault sample correspondingly.

[0047] Further, the algorithm expression for performing generalized composite coarse-graining processing on the original time series to generate a corresponding target time series in real time is:

[0048] Wherein, represents the target time series, s represents the target scale factor, b represents the sequence position, x b represents the sequence value, represents the average value of the sequence value, h represents a constant, and j represents the sequence factor.

[0049] Further, the training module is specifically configured to: When the training set is obtained in real time, input the training set into the first convolutional layer of the preset CNN model correspondingly, so that the first convolutional layer outputs a first feature map correspondingly; Input the first feature map into the first pooling layer correspondingly, so that the first pooling layer outputs a first pooled feature map correspondingly, and input the first pooled feature map into the second convolutional layer correspondingly, so that the second convolutional layer outputs a second feature map correspondingly; Input the corresponding second feature map into the second pooling layer, so that the second pooling layer outputs a corresponding second pooled feature map, and generate the bearing fault diagnosis model according to the second pooled feature map.

[0050] Further, the training module is specifically configured to: When the second pooled feature map is obtained in real time, input the second pooled feature map into the fully connected layer of the preset CNN model, so that the fully connected layer outputs a corresponding one-dimensional vector; Input the one-dimensional vector into the hidden layer and the output layer, so that the output layer outputs a perception vector corresponding to the training set in real time, and construct the bearing fault diagnosis model in real time according to the perception vector.

[0051] Further, the training module is specifically configured to: When the perception vector is obtained in real time, add a corresponding activation function and the perception vector to the perception vector in real time to generate a corresponding initial diagnosis function in real time; Test and verify the initial diagnosis function in sequence through the test set and the validation set, so as to construct the bearing fault diagnosis model correspondingly.

[0052] The fourth embodiment of the present invention provides a computer, including a memory, a processor, and a computer program stored on the memory and executable on the processor. Wherein, when the processor executes the computer program, the rolling bearing fault diagnosis method as described above is implemented.

[0053] The fifth embodiment of the present invention provides a readable storage medium, on which a computer program is stored. Wherein, when the program is executed by a processor, the rolling bearing fault diagnosis method as described above is implemented.

[0054] In summary, the rolling bearing fault diagnosis method and system provided by the above embodiments of the present invention can objectively and accurately determine whether there is a problem with the bearing, correspondingly improving the work efficiency.

[0055] It should be noted that the above-mentioned each module can be a functional module or a program module, which can be implemented by software or hardware. For the modules implemented by hardware, the above-mentioned each module can be located in the same processor; or the above-mentioned each module can also be located in different processors in any combination form.

[0056] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a definitional sequence of executable instructions for implementing logical functions, which can be embodied in any computer-readable medium for use by or in connection with an instruction execution system, apparatus, or device, such as a computer-based system, a system including a processor, or other systems that can fetch and execute instructions from the instruction execution system, apparatus, or device. As used in this specification, a "computer-readable medium" can be any device that can contain, store, communicate, propagate, or transport the program for use by or in connection with the instruction execution system, apparatus, or device.

[0057] More specific examples (a non-exhaustive list) of the computer-readable medium include the following: an electrical connection (electronic device) having one or more wirings, a portable computer diskette (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disc read-only memory (CDROM). Additionally, the computer-readable medium can even be paper or other suitable medium on which the program can be printed, as the program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpretation, or otherwise processing as appropriate, and then storing it in a computer memory.

[0058] It should be understood that various parts of the present invention can be implemented by hardware, software, firmware, or a combination thereof. In the above-described embodiments, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, any one or a combination of the following techniques well known in the art can be used: discrete logic circuits having logic gates for implementing logical functions on data signals, application specific integrated circuits having appropriate combinational logic gates, programmable gate arrays (PGAs), field programmable gate arrays (FPGAs), etc.

[0059] In the description of this specification, the descriptions referring to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples", etc., 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 invention. In this specification, the schematic representations 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 any one or more embodiments or examples in a suitable manner.

[0060] The above-described embodiments merely represent several implementation manners of the present invention. The description thereof is relatively specific and detailed, but it should not be construed as a limitation on the scope of the present invention. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present invention, several modifications and improvements can still be made, and these all fall within the protection scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the appended claims.

Claims

1. A rolling bearing fault diagnosis method, characterized in that: The method comprises: Collect historical wheelset bearing signals with different fault types in real time, and perform feature extraction processing on the historical wheelset bearing signals based on a preset algorithm, so as to generate corresponding fault samples according to a number of feature values ​​extracted in real time; Generate a corresponding training set, a validation set and a test set according to the fault samples, and train a preset CNN model through the training set, the validation set and the test set based on preset rules to generate a corresponding bearing fault diagnosis model in real time; The actual wheelset bearing signal corresponding to the wheelset bearing is acquired in real time, and a confusion matrix corresponding to the actual wheelset bearing signal is output in real time through the bearing fault diagnosis model, so as to determine whether the wheelset bearing is faulty in real time according to the confusion matrix.

2. The rolling bearing fault diagnosis method according to claim 1, characterized in that: The step of performing feature extraction processing on the historical wheelset bearing signal based on a preset algorithm to generate corresponding fault samples according to a number of feature values ​​extracted in real time includes: When the historical wheelset bearing signal is acquired in real time, the historical wheelset bearing signal is analyzed and processed in real time to detect the original time series contained in the internal correspondence of the historical wheelset bearing signal in real time; Performing generalized composite coarse-graining processing on the original time series to generate a corresponding target time series in real time; A plurality of characteristic values ​​are correspondingly generated according to the preset algorithm and the target time series, and the fault samples are correspondingly generated according to the plurality of characteristic values ​​in real time, and each characteristic value is unique.

3. The rolling bearing fault diagnosis method according to claim 2, characterized in that: The step of generating a plurality of the characteristic values ​​according to the preset algorithm and the target time series comprises: When the target time series is acquired in real time, a target scale factor adapted to the target time series is matched in real time in a preset database; Calculate in real time through the preset algorithm a number of target probability values ​​corresponding to the target time series when the target scale factor is in a discrete mode; The target probability values ​​are correspondingly set as the characteristic values, and the characteristic values ​​are integrated in real time to generate the fault samples accordingly.

4. The rolling bearing fault diagnosis method according to claim 3, characterized in that: The algorithm expression for performing generalized composite coarse-graining processing on the original time series to generate the corresponding target time series in real time is: in, represents the target time series, s represents the target scale factor, b represents the sequence position, x b Represents a sequence value, represents the average of the sequence values, h represents a constant, and j represents the sequence factor.

5. The rolling bearing fault diagnosis method according to claim 1, characterized in that: The step of training the preset CNN model through the training set, the validation set and the test set based on the preset rules to generate the corresponding bearing fault diagnosis model in real time includes: When the training set is acquired in real time, the training set is input into the first convolution layer of the preset CNN model, so that the first convolution layer outputs a first feature map; Input the first feature map to the inside of the first pooling layer, so that the first pooling layer outputs the first pooling feature map, and input the first pooling feature map to the inside of the second convolutional layer, so that the second convolutional layer outputs the second feature map; The second feature map is input into the second pooling layer so that the second pooling layer outputs the second pooling feature map, and the bearing fault diagnosis model is generated according to the second pooling feature map.

6. The rolling bearing fault diagnosis method according to claim 5, characterized in that: The step of generating the bearing fault diagnosis model according to the second pooling feature map includes: When the second pooled feature map is acquired in real time, the second pooled feature map is input into the fully connected layer of the preset CNN model so that the fully connected layer outputs a corresponding one-dimensional vector; The one-dimensional vector is input into the hidden layer and the output layer accordingly, so that the output layer outputs the perception vector corresponding to the training set in real time, and the bearing fault diagnosis model is constructed in real time according to the perception vector.

7. The rolling bearing fault diagnosis method according to claim 6, characterized in that: The step of constructing the bearing fault diagnosis model in real time according to the perception vector comprises: When the perception vector is acquired in real time, a corresponding activation function and the perception vector are added to the perception vector in real time to generate a corresponding initial diagnosis function in real time; The initial diagnostic function is tested and verified in turn through the test set and the verification set to construct the bearing fault diagnosis model accordingly.

8. A rolling bearing fault diagnosis system, characterized in that: The system comprises: An extraction module is used to collect historical wheelset bearing signals with different fault types in real time, and perform feature extraction processing on the historical wheelset bearing signals based on a preset algorithm, so as to generate corresponding fault samples according to a number of feature values ​​extracted in real time; A training module, used to generate a corresponding training set, a verification set and a test set according to the fault samples, and train a preset CNN model through the training set, the verification set and the test set based on preset rules to generate a corresponding bearing fault diagnosis model in real time; The judgment module is used to obtain the actual wheelset bearing signal generated by the wheelset bearing in real time, and output the confusion matrix corresponding to the actual wheelset bearing signal in real time through the bearing fault diagnosis model, so as to judge whether the wheelset bearing is faulty in real time according to the confusion matrix.

9. A computer comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the rolling bearing fault diagnosis method according to any one of claims 1 to 7 is implemented.

10. A readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the rolling bearing fault diagnosis method as described in any one of claims 1 to 7 is implemented.

Images