High-speed rail bearing weak fault feature extraction method and system

Abstract

The invention discloses a high-speed rail bearing weak fault feature extraction method and system, and the method comprises the steps: obtaining a fault vibration signal of a high-speed rail bearing, carrying out the dimension reduction processing of the fault vibration signal, and obtaining a target fault vibration signal; determining an adaptive Morlet wavelet parameter in the Morlet wavelet activation function according to a DPSO algorithm to obtain a target Morlet wavelet activation function, and setting an automatic encoder according to the target Morlet wavelet activation function to obtain an adaptive depth automatic encoder; nonlinear mapping in the target fault vibration signal is extracted according to the self-adaptive depth automatic encoder, and a target fault vibration sub-signal is obtained; and performing convolution operation on the target fault vibration sub-signal and the synchronous extraction operator, and extracting fault feature information in the target fault vibration sub-signal according to a preset multi-scale analysis method. By analyzing the extracted fault features, fault frequency information is obtained, and the detection precision and robustness of the weak fault features of the high-speed rail bearing are greatly improved.

Description

Technical Field

[0001] The invention belongs to the technical field of fault feature extraction, and in particular relates to a method and system for extracting weak fault features of high-speed rail bearings. Background Art

[0002] Bearings are one of the key components of the train running gear. There are many mechanical structures that make up the train running gear, including gearboxes, wheelsets and other mechanical components. These mechanical components will also generate vibrations during the operation of the train. These vibration signals are mixed with the vibration signals of the wheelset bearings. The acceleration sensor installed on the wheelset axle box collects mixed signals from multiple vibration sources, and it is difficult to directly analyze and process the measurement signals. Summary of the invention

[0003] The present invention provides a method and system for extracting weak fault features of high-speed rail bearings, which are used to solve the technical problem that it is difficult to directly analyze and process measurement signals.

[0004] In a first aspect, the present invention provides a method for extracting characteristics of weak faults of high-speed rail bearings, comprising: Acquire a fault vibration signal of a high-speed rail bearing, and perform dimensionality reduction processing on the fault vibration signal to obtain a target fault vibration signal; Determine adaptive Morlet wavelet parameters in the Morlet wavelet activation function according to the DPSO algorithm to obtain a target Morlet wavelet activation function, and set an autoencoder according to the target Morlet wavelet activation function to obtain an adaptive deep autoencoder; Extracting a nonlinear mapping in the target fault vibration signal according to the adaptive deep autoencoder to obtain a target fault vibration sub-signal; The target fault vibrator signal is convolved with a synchronous extraction operator, and fault feature information in the target fault vibrator signal is extracted according to a preset multi-scale analysis method.

[0005] In a second aspect, the present invention provides a high-speed rail bearing weak fault feature extraction system, comprising: A processing module is configured to obtain a fault vibration signal of a high-speed rail bearing, and perform dimensionality reduction processing on the fault vibration signal to obtain a target fault vibration signal; A setting module is configured to determine adaptive Morlet wavelet parameters in the Morlet wavelet activation function according to the DPSO algorithm, obtain a target Morlet wavelet activation function, and set an autoencoder according to the target Morlet wavelet activation function to obtain an adaptive deep autoencoder; A first extraction module is configured to extract a nonlinear mapping in the target fault vibration signal according to the adaptive deep autoencoder to obtain a target fault vibration sub-signal; The second extraction module is configured to perform a convolution operation on the target fault vibration sub-signal and a synchronous extraction operator, and extract the fault feature information in the target fault vibration sub-signal according to a preset multi-scale analysis method.

[0006] According to a third aspect, an electronic device is provided, comprising: at least one processor, and a memory communicatively connected to the at least one processor, wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can perform the steps of the method for extracting weak fault features of high-speed rail bearings according to any embodiment of the present invention.

[0007] In a fourth aspect, the present invention further provides a computer-readable storage medium having a computer program stored thereon, wherein when the program instructions are executed by a processor, the processor executes the steps of the method for extracting weak fault features of high-speed rail bearings of any embodiment of the present invention.

[0008] The method and system for extracting weak fault features of high-speed rail bearings in this application considers the performance of adaptive deep autoencoder (ADAE) in weak fault feature enhancement and the advantages of multi-scale synchronous extraction transform (MS-SET) in time-frequency analysis. It reduces the dimension of features through multi-scale local linear section analysis (LLTSA) and combines signal reconstruction technology to ensure accurate retention of features. By analyzing the extracted fault features, the fault frequency information is obtained, which greatly improves the detection accuracy and robustness of weak fault features of high-speed rail bearings. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0010] Figure 1 A flowchart of a method for extracting weak fault features of high-speed rail bearings provided by an embodiment of the present invention; Figure 2 A structural block diagram of a high-speed rail bearing weak fault feature extraction system provided by an embodiment of the present invention; Figure 3 It is a schematic diagram of the structure of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0011] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0012] See also Figure 1 , which shows a flow chart of a method for extracting weak fault features of high-speed rail bearings of the present application.

[0013] like Figure 1 As shown, the method for extracting the characteristics of weak faults of high-speed rail bearings specifically includes the following steps: Step S101, acquiring a fault vibration signal of a high-speed railway bearing, and performing dimensionality reduction processing on the fault vibration signal to obtain a target fault vibration signal.

[0014] In this step, assume is a high-dimensional matrix composed of fault vibration signals, where For the first sample, For the second sample, For the samples, and D are the number of samples and sample dimensions respectively, and each sample represents a data point in the high-dimensional feature space; Determine the neighborhood of each sample and select the k nearest neighbor samples based on the Euclidean distance between samples As the Samples The expression for calculating the Euclidean distance between samples is: , In the formula, is the Euclidean distance between two samples, For the The first nearest neighbor sample of a sample, For the The second nearest neighbor sample of a sample, For the The kth nearest neighbor sample of a sample, For the The jth nearest neighbor sample of a sample; For Samples , by minimizing the reconstruction error, we can obtain The local linear representation on the surface defines the objective function as: , In the formula, is the approximation of the sample reconstructed by the weighted sum of the neighborhood samples; By minimizing the reconstruction error, the weight of each sample can be obtained , so that: , In the formula, is an approximation of the weighted and reconstructed samples, is the weight coefficient, For the The jth nearest neighbor sample of a sample; By minimizing the objective function of the reconstruction error, the transformation matrix is ​​solved , which is used to map high-dimensional data to low-dimensional space. Specifically, the goal is to optimize the global transformation so that each data point can retain its local structure in the high-dimensional space after dimensionality reduction: , In the formula, is a high-dimensional matrix, is the transpose symbol; Finally, the high-dimensional matrix X is transformed by the transformation matrix Map to low-dimensional space to get low-dimensional representation : , Non-negative matrix factorization (NMF) is then used to construct the reconstruction model. The goal of NMF is to represent the low-dimensional signal Decomposed into two non-negative matrices and : , Where W is the basis matrix and H is the coefficient matrix; The optimization is performed with the goal of minimizing the reconstruction error. The expression is:

[0015] In the formula, is a low-dimensional signal, is the basis matrix, is the coefficient matrix, is the Frobenius norm; Reconstruct the low-dimensional signal Y back to the high-dimensional space, the expression is: , In the formula, is the reconstructed signal, for The pseudo-inverse matrix of is the transformation matrix.

[0016] Step S102, determining the adaptive Morlet wavelet parameters in the Morlet wavelet activation function according to the DPSO algorithm, obtaining the target Morlet wavelet activation function, and setting the autoencoder according to the target Morlet wavelet activation function to obtain an adaptive deep autoencoder.

[0017] In this step, the modified autoencoder is designed using the Morlet wavelet activation function to establish a reliable mapping hidden in this complex signal. The expression is: , In the formula, is the amplitude, is the amplitude of the wavelet function at time t, For time, is the bandwidth, is the center frequency; Final Output is calculated as: , , In the formula, is the output of hidden node j, is the output value of the final output node i, is the activation function, is the connection weight coefficient, is the number of hidden layer nodes, is the amplitude adjustment coefficient of Morlet wavelet, is the number of input nodes, is the weight coefficient between input node k and hidden node j, is the input value of input node k, is the shift factor of hidden node j, is the scale factor of hidden node j, is the bandwidth parameter of the Morlet wavelet, is the center frequency of the Morlet wavelet; Adding sparse penalty to the cost function, the expression is: , In the formula, is the cost function after introducing the sparse penalty term, The original cost function value is, is the sparse penalty factor, is the sparsity coefficient, is the actual average activation value of the jth neuron in the hidden layer; The training of the autoencoder minimizes the cost function by updating the weights, which is expressed as: , In the formula, is the learning rate.

[0018] It should be noted that, assuming Indicates the current position of particle i Indicates the current speed of particle i; represents the individual optimal position of particle i; Indicates the global optimal position. Initially, the speed and position of each particle are randomly generated. The position is evenly distributed within the preset range, and the speed is randomly generated within the given range. At the same time, the individual optimal position of the particle is set as the initial position of the particle itself. According to the adaptive inertia weight optimization strategy, in the early stage of the search, the particles conduct a wide global search, and in the later stage of the search, the particles concentrate on the vicinity of the local optimal solution for a refined search. The expression of the adaptive inertia weight optimization strategy is: , In the formula, is the initial inertia weight, is the inertia weight at convergence, is the maximum number of iterations, is the current iteration number; In each generation, the expressions for updating the particle's position and velocity are: , , In the formula, is the velocity of particle i in dimension j at the t+1th iteration, is the position of particle i in dimension j at the t+1th iteration, is the velocity of particle i in dimension j at the tth iteration, is the position of particle i in dimension j at the tth iteration, is the individual optimal position of particle i in dimension j at the tth iteration, is the global optimal position of particle i in dimension j at the tth iteration, , are acceleration constants, , are all random numbers uniformly distributed in the interval [0,1]; In the later stage of iteration, simulated annealing is introduced as a local search strategy to further optimize the global optimal solution, where the probability P of accepting a new solution is expressed as: , In the formula, is the current temperature parameter, is the fitness value of the current global optimal solution, is the newly generated fitness value; If the fitness value of particle i in the current iteration is better than its historical optimal value, the individual optimal position of the particle is updated; if the current fitness value of particle i is better than the fitness value of the global optimal position, the global optimal position is updated; When the number of iterations reaches the maximum value or the fitness value change of the global optimal solution is less than the set threshold, the iteration is terminated and the final optimal solution is returned, that is, the adaptive Morlet wavelet parameters in the Morlet wavelet activation function are determined.

[0019] Step S103: extracting nonlinear mapping in the target fault vibration signal according to the adaptive deep autoencoder to obtain a target fault vibration sub-signal.

[0020] In this step, a more accurate nonlinear mapping that captures the hidden time series signal is achieved by stacking some nonlinear regression layers. In particular, the input is fed into the first MAE to obtain the low-level weights W(1) and features h(1). Secondly, h(1) is regarded as the input to train the second MAE, and then the high-level weights W(2) and features h(2) are obtained. Repeat until the last MAE, and then a nonlinear regression layer needs to be added on top to generate the result. The expression is: , , In the formula, is the input of the sample with time series step length m-1, is the highest level feature of the input, and are the weights and biases of the nonlinear regression layer, respectively.

[0021] Step S104: performing a convolution operation on the target fault vibration sub-signal and a synchronization extraction operator, and extracting fault feature information in the target fault vibration sub-signal according to a preset multi-scale analysis method.

[0022] In this step, the scale factor of the multiscale analysis is set , the expression is: , In the formula, is the first scale factor in multi-scale analysis, is the second scale factor in multi-scale analysis, is the nth scale factor in the multi-scale analysis; According to the set scale factor, construct the corresponding window function for each scale , is the scale factor The adjusted window function, To time according to To scale; Perform STFT on the signal of each scale, and use the window function to convolve the input signal to obtain the time-frequency domain , the expression is: , In the formula, For movable windows, is the reconstructed signal, is an imaginary unit, To represent the angular frequency, is the integration variable; make , according to Parseval's theorem, rewritten in the time-frequency domain The target expression is: , In the formula, is the frequency variable, is the complex conjugate, for Fourier transform of calculate Fourier transform of , the expression is: , Assume an amplitude of A and a frequency of A single-component harmonic signal , the Fourier transform result of the single-component harmonic signal is ; Substitute the Fourier transform result of the single-component harmonic signal into the target expression to obtain the STFT result of the single-component harmonic signal, which is expressed as: , In the formula, is the center frequency, is the STFT result of the signal.

[0023] In summary, the result of the harmonic signal after STFT is still The harmonic signal composition of The window function has the characteristic of being very compact in the frequency domain, and the energy of the time-frequency image is concentrated in the frequency domain. In this frequency range, the amplitude of the time-frequency representation is the largest, and the value is ; and the time-frequency analysis results have the best noise robustness.

[0024] For any ,in , the two-dimensional instantaneous frequency of the STFT result for:

[0025] In the formula, is an imaginary unit, for The time derivative of is calculated as: , The new time-frequency expression is only composed of = The time-frequency coefficient of the instantaneous frequency can be calculated, which is SET, and the expression is: , In the formula, That is SEO, For in time and frequency The time-frequency representation value below.

[0026] Will Substitution In, we get: , In the formula, is the frequency variable, is the Dirac δ function.

[0027] In summary, the method of this application takes into account the performance of adaptive deep autoencoder (ADAE) in weak fault feature enhancement and the advantages of multi-scale synchronous extraction transform (MS-SET) in time-frequency analysis. It reduces the dimension of features through multi-scale local linear section analysis (LLTSA) and combines signal reconstruction technology to ensure accurate retention of features. By analyzing the extracted fault features, the fault frequency information is obtained, which greatly improves the detection accuracy and robustness of the weak fault features of high-speed rail bearings.

[0028] See also Figure 2 , which shows a structural block diagram of a high-speed rail bearing weak fault feature extraction system of the present application.

[0029] like Figure 2 As shown, a high-speed railway bearing weak fault feature extraction system 200 includes a processing module 210, a setting module 220, a first extraction module 230 and a second extraction module 240.

[0030] Among them, the processing module 210 is configured to obtain the fault vibration signal of the high-speed railway bearing, and perform dimensionality reduction processing on the fault vibration signal to obtain the target fault vibration signal; the setting module 220 is configured to determine the adaptive Morlet wavelet parameters in the Morlet wavelet activation function according to the DPSO algorithm, obtain the target Morlet wavelet activation function, and set the autoencoder according to the target Morlet wavelet activation function to obtain an adaptive deep autoencoder; the first extraction module 230 is configured to extract the nonlinear mapping in the target fault vibration signal according to the adaptive deep autoencoder to obtain the target fault vibration sub-signal; the second extraction module 240 is configured to perform a convolution operation on the target fault vibration sub-signal and a synchronous extraction operator, and extract the fault feature information in the target fault vibration sub-signal according to a preset multi-scale analysis method.

[0031] It should be understood that Figure 2 Modules and references documented in Figure 1 Therefore, the operations and features described above for the method and the corresponding technical effects are also applicable to Figure 2 The modules in it will not be described in detail here.

[0032] In other embodiments, the embodiments of the present invention further provide a computer-readable storage medium having a computer program stored thereon, wherein when the program instructions are executed by a processor, the processor executes the method for extracting weak fault features of high-speed rail bearings in any of the above method embodiments; As an implementation mode, the computer-readable storage medium of the present invention stores computer-executable instructions, and the computer-executable instructions are configured as follows: Acquire a fault vibration signal of a high-speed rail bearing, and perform dimensionality reduction processing on the fault vibration signal to obtain a target fault vibration signal; Determine adaptive Morlet wavelet parameters in the Morlet wavelet activation function according to the DPSO algorithm to obtain a target Morlet wavelet activation function, and set an autoencoder according to the target Morlet wavelet activation function to obtain an adaptive deep autoencoder; Extracting a nonlinear mapping in the target fault vibration signal according to the adaptive deep autoencoder to obtain a target fault vibration sub-signal; The target fault vibrator signal is convolved with a synchronous extraction operator, and fault feature information in the target fault vibrator signal is extracted according to a preset multi-scale analysis method.

[0033] The computer-readable storage medium may include a program storage area and a data storage area, wherein the program storage area may store an operating system and application programs required for at least one function; the data storage area may store data created according to the use of the high-speed rail bearing weak fault feature extraction system, etc. In addition, the computer-readable storage medium may include a high-speed random access memory, and may also include a memory, such as at least one disk storage device, a flash memory device, or other non-volatile solid-state storage device. In some embodiments, the computer-readable storage medium may optionally include a memory remotely arranged relative to the processor, and these remote memories may be connected to the high-speed rail bearing weak fault feature extraction system via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0034] Figure 3 is a schematic diagram of the structure of an electronic device provided by an embodiment of the present invention, such as Figure 3 As shown, the device includes: a processor 310 and a memory 320. The electronic device may also include: an input device 330 and an output device 340. The processor 310, the memory 320, the input device 330 and the output device 340 may be connected via a bus or other means. Figure 3 The example of the connection via bus is taken. The memory 320 is the above-mentioned computer-readable storage medium. The processor 310 executes various functional applications and data processing of the server by running the non-volatile software programs, instructions and modules stored in the memory 320, that is, the method for extracting weak fault features of high-speed rail bearings in the above-mentioned method embodiment is implemented. The input device 330 can receive input digital or character information, and generate key signal input related to user settings and function control of the high-speed rail bearing weak fault feature extraction system. The output device 340 may include display devices such as display screens.

[0035] The electronic device can execute the method provided by the embodiment of the present invention, and has the functional modules and beneficial effects corresponding to the execution method. For technical details not described in detail in this embodiment, please refer to the method provided by the embodiment of the present invention.

[0036] As an implementation mode, the electronic device is applied to a high-speed rail bearing weak fault feature extraction system, and is used for a client, and includes: at least one processor; and a memory connected to the at least one processor in communication; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor, so that the at least one processor can: Acquire a fault vibration signal of a high-speed rail bearing, and perform dimensionality reduction processing on the fault vibration signal to obtain a target fault vibration signal; Determine adaptive Morlet wavelet parameters in the Morlet wavelet activation function according to the DPSO algorithm to obtain a target Morlet wavelet activation function, and set an autoencoder according to the target Morlet wavelet activation function to obtain an adaptive deep autoencoder; Extracting a nonlinear mapping in the target fault vibration signal according to the adaptive deep autoencoder to obtain a target fault vibration sub-signal; The target fault vibrator signal is convolved with a synchronous extraction operator, and fault feature information in the target fault vibrator signal is extracted according to a preset multi-scale analysis method.

[0037] Through the description of the above implementation modes, those skilled in the art can clearly understand that each implementation mode can be implemented by means of software plus a necessary general hardware platform, or of course by hardware. Based on such an understanding, the above technical solution can essentially or in other words be embodied in the form of a software product that contributes to the prior art. The computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a magnetic disk, an optical disk, etc., and includes a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods of each embodiment or some parts of the embodiment.

[0038] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for extracting weak fault features of high-speed rail bearings, characterized in that: include: Acquire a fault vibration signal of a high-speed rail bearing, and perform dimensionality reduction processing on the fault vibration signal to obtain a target fault vibration signal; Determine adaptive Morlet wavelet parameters in the Morlet wavelet activation function according to the DPSO algorithm to obtain a target Morlet wavelet activation function, and set an autoencoder according to the target Morlet wavelet activation function to obtain an adaptive deep autoencoder; Extracting a nonlinear mapping in the target fault vibration signal according to the adaptive deep autoencoder to obtain a target fault vibration sub-signal; The target fault vibrator signal is convolved with a synchronous extraction operator, and fault feature information in the target fault vibrator signal is extracted according to a preset multi-scale analysis method.

2. A method for extracting weak fault features of high-speed rail bearings according to claim 1, characterized in that: The performing dimensionality reduction processing on the fault vibration signal to obtain a target fault vibration signal comprises: set up is a high-dimensional matrix composed of fault vibration signals, where For the first sample, For the second sample, For the samples, and D are the number of samples and sample dimensions respectively, and each sample represents a data point in the high-dimensional feature space; Determine the neighborhood of each sample and select the k nearest neighbor samples based on the Euclidean distance between samples As the Samples The expression for calculating the Euclidean distance between samples is: ， In the formula, is the Euclidean distance between two samples, For the The first nearest neighbor sample of a sample, For the The second nearest neighbor sample of a sample, For the The kth nearest neighbor sample of a sample, For the The jth nearest neighbor sample of a sample; For Samples , by minimizing the reconstruction error, we can obtain The local linear representation on the surface defines the objective function as: ， In the formula, is the approximation of the sample reconstructed by the weighted sum of the neighborhood samples; The optimization is performed with the goal of minimizing the reconstruction error. The expression is: ， In the formula, is a low-dimensional signal, is the basis matrix, is the coefficient matrix, is the Frobenius norm; Reconstruct the low-dimensional signal Y back to the high-dimensional space, the expression is: ， In the formula, is the reconstructed signal, for The pseudo-inverse matrix of is the transformation matrix.

3. The method for extracting weak fault features of high-speed rail bearings according to claim 1 is characterized in that: The expression of the automatic encoder is: ， In the formula, is the amplitude, is the amplitude of the wavelet function at time t, For time, is the bandwidth, is the center frequency; Determining the adaptive Morlet wavelet parameters in the Morlet wavelet activation function according to the DPSO algorithm to obtain the target Morlet wavelet activation function includes: Assumptions Indicates the current position of particle i represents the current velocity of particle i; represents the individual optimal position of particle i; Indicates the global optimal position. Initially, the speed and position of each particle are randomly generated. The position is evenly distributed within the preset range, and the speed is randomly generated within the given range. At the same time, the individual optimal position of the particle is set as the initial position of the particle itself. According to the adaptive inertia weight optimization strategy, in the early stage of the search, the particles conduct a wide global search, and in the later stage of the search, the particles concentrate on the vicinity of the local optimal solution for a refined search. The expression of the adaptive inertia weight optimization strategy is: ， In the formula, is the initial inertia weight, is the inertia weight at convergence, is the maximum number of iterations, is the current iteration number; In each generation, the expressions for updating the particle's position and velocity are: ， ， In the formula, is the velocity of particle i in dimension j at the t+1th iteration, is the position of particle i in dimension j at the t+1th iteration, is the velocity of particle i in dimension j at the tth iteration, is the position of particle i in dimension j at the tth iteration, is the individual optimal position of particle i in dimension j at the tth iteration, is the global optimal position of particle i in dimension j at the tth iteration, , are acceleration constants, , are all random numbers uniformly distributed in the interval [0,1]; In the later stage of iteration, simulated annealing is introduced as a local search strategy to further optimize the global optimal solution, where the probability P of accepting a new solution is expressed as: ， In the formula, is the current temperature parameter, is the fitness value of the current global optimal solution, is the newly generated fitness value; If the fitness value of particle i in the current iteration is better than its historical optimal value, the individual optimal position of the particle is updated; if the current fitness value of particle i is better than the fitness value of the global optimal position, the global optimal position is updated; When the number of iterations reaches the maximum value or the fitness value change of the global optimal solution is less than the set threshold, the iteration is terminated and the final optimal solution is returned, that is, the adaptive Morlet wavelet parameters in the Morlet wavelet activation function are determined.

4. The method for extracting weak fault features of high-speed rail bearings according to claim 1 is characterized in that: The convolution operation of the target fault vibrator signal and the synchronization extraction operator, and extracting the fault feature information in the target fault vibrator signal according to a preset multi-scale analysis method includes: Setting the scale factor for multiscale analysis , the expression is: ， In the formula, is the first scale factor in multi-scale analysis, is the second scale factor in multi-scale analysis, is the nth scale factor in the multi-scale analysis; According to the set scale factor, construct the corresponding window function for each scale , is the scale factor The adjusted window function, To time according to To scale; Perform STFT on the signal of each scale, and use the window function to convolve the input signal to obtain the time-frequency domain , the expression is: ， In the formula, For movable windows, is the reconstructed signal, is an imaginary unit, To represent the angular frequency, is the integral variable; make , according to Parseval's theorem, rewritten in the time-frequency domain The target expression is: ， In the formula, is the frequency variable, is the complex conjugate, for Fourier transform of calculate Fourier transform of , the expression is: ， Assume an amplitude of A and a frequency of A single-component harmonic signal , the Fourier transform result of the single-component harmonic signal is ; Substitute the Fourier transform result of the single-component harmonic signal into the target expression to obtain the STFT result of the single-component harmonic signal, which is expressed as: ， In the formula, is the center frequency, is the STFT result of the signal.

5. A high-speed rail bearing weak fault feature extraction system, characterized in that: include: A processing module is configured to obtain a fault vibration signal of a high-speed rail bearing, and perform dimensionality reduction processing on the fault vibration signal to obtain a target fault vibration signal; A setting module is configured to determine adaptive Morlet wavelet parameters in the Morlet wavelet activation function according to the DPSO algorithm, obtain a target Morlet wavelet activation function, and set an autoencoder according to the target Morlet wavelet activation function to obtain an adaptive deep autoencoder; A first extraction module is configured to extract a nonlinear mapping in the target fault vibration signal according to the adaptive deep autoencoder to obtain a target fault vibration sub-signal; The second extraction module is configured to perform a convolution operation on the target fault vibration sub-signal and a synchronous extraction operator, and extract the fault feature information in the target fault vibration sub-signal according to a preset multi-scale analysis method.

6. An electronic device, characterized in that: include: At least one processor, and a memory communicatively connected to the at least one processor, wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method described in any one of claims 1 to 4.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the method according to any one of claims 1 to 4 is implemented.

Images