A rolling bearing fault diagnosis method and system

Abstract

The present invention relates to a rolling bearing fault diagnosis method and system. The method firstly uses optimized variational modal decomposition to decompose and reconstruct the collected vibration acceleration signal of the rolling bearing; then uses fine composite multi-scale fuzzy entropy to extract features of the reconstructed signal and construct a fault feature set; finally, a gray wolf group optimized support vector machine classifier is used for diagnosis and identification to diagnose the rolling bearing fault type. The method and system are conducive to rolling bearing fault diagnosis and effectively identify the rolling bearing fault type.

Description

Technical Field

[0001] The invention belongs to the technical field of fault diagnosis, and in particular relates to a rolling bearing fault diagnosis method and system. Background Art

[0002] Among the components of rotating machinery, rolling bearings are the most prone to failure. When a rolling bearing fails, it will show nonlinear and complex dynamic characteristics, making the corresponding vibration signal non-stationary and nonlinear. In addition, during the actual collection of bearing vibration signals, it is often affected by the harsh environment on site, and the noise in the collected signal increases and the signal-to-noise ratio decreases. Therefore, how to extract fault features and identify the state from non-stationary, nonlinear, low-signal-to-noise ratio, and complex frequency component vibration signals is the key to rolling bearing fault diagnosis.

[0003] At present, common fault signal denoising methods mainly include: wavelet transform, empirical mode decomposition (EMD) and local mean decomposition. However, wavelet transform needs to select wavelet basis and decomposition level in advance, which lacks adaptability; EMD has limitations such as frequency confusion, over-envelope, under-envelope, and endpoint effect; local mean decomposition has defects such as slow operation speed and signal conflict. Variational mode decomposition (VMD) is a new non-recursive adaptive signal decomposition method developed on the basis of traditional Wiener filtering. Compared with wavelet transform, EMD and local mean decomposition, the signal decomposed by VMD has the characteristics of high accuracy, fast convergence and good robustness, which is very suitable for processing rolling bearing fault signals. However, the signal processing effect of VMD algorithm is closely related to the two parameters of penalty factor and decomposition number. Summary of the invention

[0004] The object of the present invention is to provide a rolling bearing fault diagnosis method and system, which are conducive to rolling bearing fault diagnosis and effectively identify the rolling bearing fault type.

[0005] To achieve the above-mentioned purpose, the technical solution adopted by the present invention is: a rolling bearing fault diagnosis method, firstly, using optimized variational modal decomposition to decompose and reconstruct the collected vibration acceleration signal of the rolling bearing; then using fine composite multi-scale fuzzy entropy to extract features of the reconstructed signal and construct a fault feature set; finally, using gray wolf pack optimized support vector machine classifier for diagnosis and identification, diagnosing the rolling bearing fault type.

[0006] Further, the method comprises the following steps:

[0007] Step S1, using an acceleration sensor to collect vibration acceleration signals of a rolling bearing in a normal state, an inner ring fault state, an outer ring fault state, and a rolling element fault state, and obtaining a time domain signal sample set;

[0008] Step S2, using an optimization algorithm to optimize the mode number K and penalty factor α in the variational mode decomposition algorithm, then decomposing the collected vibration acceleration signal, and reconstructing the obtained intrinsic mode components;

[0009] Step S3, using fine composite multi-scale fuzzy entropy to extract features from the reconstructed signal and construct a fault feature set;

[0010] Step S4: input the fault features into the gray wolf pack optimized support vector machine classifier for training and testing to diagnose the rolling bearing fault type.

[0011] Further, in step S1, the vibration acceleration signal is a driving end vibration acceleration signal.

[0012] Further, in step S2, the mode number K and penalty factor α in the variational mode decomposition algorithm are optimized using an optimization algorithm, including the following steps:

[0013] Step S201, initialize various parameters of the optimization algorithm and the fitness function in the process of optimizing the parameters; the fitness function is defined as the mean of the envelope entropy of all IMF components after VMD decomposition as the fitness value, and the minimum envelope mean entropy is used as the optimization target to obtain the optimal parameters; wherein, the definition of envelope entropy Ep is as follows:

[0014]

[0015] In the formula, p j is the normalized form of a(j), a(j) is the envelope signal of signal x(j) after Hilbert transform, and N is the length of signal x(j);

[0016] Step S202, calculating the fitness value corresponding to the current position of each group member;

[0017] Step S203: Divide the entire group into c groups, select the individual with the smallest fitness value in the group as the group leader; divide the entire group into d regions, select the individual with the smallest fitness value in the region as the regional winner;

[0018] Step S204: Conduct campaign activities, and other members update their positions according to the group leader and the regional winner, as follows:

[0019] when When , the spatial position is updated according to formula (2):

[0020]

[0021] when When , the spatial position is updated according to formula (3):

[0022]

[0023] In the formula, It indicates the position of the jth member of the i-th group when the number of iterations is t; express The kth element in m * Indicates the group leader or regional winner. When iterating, first m * As the leader of the group, the members of the group update their positions according to the leader's position, and then let m * is the regional winner, and the members in the region update their own positions according to the winner's position; r is a random number between 0 and 1;

[0024] Step S205: exchanging positions between group members with a set probability;

[0025] Step S206, calculating the fitness values ​​of group members after the position update, and selecting a new group leader and regional winner;

[0026] Step S207: The winners of each region form a conference group and conduct conference group affairs, and update the position according to formula (4):

[0027]

[0028] In the formula, represents the winner of region a, represents the winner of area b. Let the number of areas be n, then a is a random integer between 1 and n, b = 1, 2, ..., n, a ≠ b; according to The updated new position is r is a random number between 0 and 1; if The fitness value is better than If it is, replace its position, otherwise keep it as it is;

[0029] Step S208, repeat steps S204 to S207 until the iteration termination condition is met, and output the optimal solution, that is, the optimal α and K parameter values.

[0030] Furthermore, the parameters are initialized as follows: the population size is 20, of which the number of groups is 5, the number of regions is 4, the maximum number of iterations is 20, the maximum probability of group exchange is 1, the optimization range of K is [3,10], and the optimization range of α is [200,3000].

[0031] Furthermore, step S4 specifically includes the following steps:

[0032] Step S401: the fault feature set obtained in step S3 is divided into a training sample set and a test sample set according to a set ratio in the number of each fault category, and then the training sample set and the test sample set are normalized;

[0033] Step S402: According to the SVM model, the radial basis function is selected as the kernel function, and the gray wolf pack optimization algorithm is used to define the average correct recognition rate after three-fold crossover of the training samples as the fitness value, and automatically determine the optimal penalty factor and kernel function parameters of the SVM model;

[0034] Step S403: construct an SVM prediction model using the training samples, and use the SVM prediction model to diagnose and identify the test samples to determine the rolling bearing fault type.

[0035] The present invention also provides a rolling bearing fault diagnosis system, comprising a memory, a processor, and computer program instructions stored in the memory and executable by the processor. When the processor executes the computer program instructions, the above-mentioned method steps can be implemented.

[0036] Compared with the prior art, the present invention has the following beneficial effects: a rolling bearing fault diagnosis method and system are provided, which give full play to the advantages of VMD algorithm in rolling bearing vibration signal noise reduction, and introduce optimization algorithm to achieve optimal selection of VMD algorithm parameters, overcoming the phenomenon of poor decomposition effect caused by traditional artificial parameter setting; in addition, the present invention introduces refined composite multi-scale fuzzy entropy (RCMFE) as a fault feature parameter, so that the fault feature is effectively extracted, and thus a stable fault feature set that is easy to classify and identify is constructed; finally, the present invention uses GWO-SVM classifier to diagnose and identify the fault feature set, with fast operation speed and strong generalization ability. Therefore, the present invention has strong practicality and broad application prospects. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 The present invention is a flowchart of a method implementation according to an embodiment of the present invention.

[0038] Figure 2 It is a time domain waveform diagram of the rolling bearing under different states in the embodiment of the present invention.

[0039] Figure 3 Flow chart of optimizing VMD by the optimization algorithm in an embodiment of the present invention.

[0040] Figure 4 This is the result of decomposition and reconstruction of rolling element fault signal in an embodiment of the present invention.

[0041] Figure 5 This is a flow chart of RCMFE in an embodiment of the present invention.

[0042] Figure 6 ] are RCMFE entropy value curves of four states in the embodiment of the present invention.

[0043] Figure 7 This is the GWO-SVM recognition result in the embodiment of the present invention. DETAILED DESCRIPTION

[0044] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0045] It should be noted that the following detailed descriptions are exemplary and are intended to provide further explanation of the present application. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present application belongs.

[0046] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit the exemplary embodiments according to the present application. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprise" and / or "include" are used in this specification, it indicates the presence of features, steps, operations, devices, components and / or combinations thereof.

[0047] This embodiment provides a rolling bearing fault diagnosis method. First, the collected vibration acceleration signal of the rolling bearing is decomposed and reconstructed by optimizing variational modal decomposition. Then, the reconstructed signal is feature extracted by using fine composite multi-scale fuzzy entropy to construct a fault feature set. Finally, a gray wolf pack optimized support vector machine classifier is used for diagnosis and identification to diagnose the rolling bearing fault type.

[0048] like Figure 1 As shown, the method specifically comprises the following steps:

[0049] Step S1: Use an acceleration sensor to collect vibration acceleration signals of the rolling bearing in a normal state, an inner ring fault state, an outer ring fault state, and a rolling element fault state, and obtain a time domain signal sample set (such as Figure 2 shown).

[0050] Step S2: Optimize the mode number K and penalty factor α in the variational mode decomposition (VMD) algorithm using an optimization algorithm, then decompose the collected vibration acceleration signal, and reconstruct the obtained intrinsic mode component (IMF).

[0051] Step S3: extract features of the reconstructed signal using fine composite multi-scale fuzzy entropy to construct a fault feature set.

[0052] Step S4: input the fault features into a Grey Wolf Pack Optimized Support Vector Machine (GWO-SVM) classifier for training and testing to diagnose the rolling bearing fault type.

[0053] In step S1, the vibration acceleration signal is a driving end vibration acceleration signal.

[0054] In step S2, the mode number K and penalty factor α in the variational mode decomposition algorithm are optimized using an optimization algorithm, that is, the implementation process of optimizing the VMD algorithm is as follows:

[0055] The VMD algorithm is an adaptive signal decomposition method that searches for the optimal solution of the variational model through cyclic iteration, determines the frequency center and bandwidth of the intrinsic mode component (IMF), and achieves effective separation of each component. The implementation method of the VMD algorithm is as follows:

[0056] 1) Set n = 0, initialize and λ 1 ;

[0057] 2) Update using the following formula and λ n+1 ;

[0058]

[0059] 3) Given the discrimination accuracy e>0, if Then stop the iteration; otherwise, return to step 2).

[0060] The implementation of the optimization algorithm is as follows:

[0061] 1) Set the parameters of the optimization algorithm: group size S = number of groups p × number of regions a, maximum number of iterations m, maximum probability of group exchange λ, number of parameters d, upper and lower bounds of parameter values ​​ub, lb, and fitness function during the parameter optimization process.

[0062] Use formula (2) to initialize the population S p×a :

[0063] S p×a =rand(p,a)×(ub-lb)+lb (2)

[0064] 2) Calculate the fitness value corresponding to the current position of each group member;

[0065] 3) Divide the entire group into p groups, select the individual with the smallest fitness value in the group as the group leader; divide the entire group into a regions, and select the individual with the smallest fitness value in the region as the regional winner;

[0066] 4) Conduct campaigns, and other members update their positions based on the group leader and regional winners, as follows:

[0067] when When , the spatial position is updated according to formula (3):

[0068]

[0069] when When , update the spatial position according to formula (4):

[0070]

[0071] In the formula, It indicates the position of the jth member of the i-th group when the number of iterations is t; express The kth element in m * Indicates the group leader or regional winner. When iterating, first m * As the leader of the group, the members of the group update their positions according to the leader's position, and then let m * is the regional winner, and the members in the region update their own positions according to the winner's position; r is a random number between 0 and 1;

[0072] 5) Exchange positions between group members with a certain probability;

[0073] 6) Calculate the fitness values ​​of group members after the position update and select new group leaders and regional winners;

[0074] 7) The winners of each region form a conference group and conduct conference group affairs, updating the position according to formula (5):

[0075]

[0076] In the formula, represents the winner of region a, represents the winner of area b. Let the number of areas be n, then a is a random integer between 1 and n, b = 1, 2, ..., n, a ≠ b; according to The updated new position is r is a random number between 0 and 1. The fitness value is better than If it is, replace its position, otherwise keep it as it is;

[0077] 8) Repeat steps 4) to 7) until the iteration termination condition is met and output the optimal solution.

[0078] Based on this, the present invention proposes an optimized VMD algorithm, and its implementation process is as follows:

[0079] 1) Initialize the parameters of the optimization algorithm and the fitness function in the process of optimizing the parameters. The parameters are set as follows: the group size is 20, the number of groups is 5, the number of regions is 4, the maximum number of iterations is 20, the maximum probability of group exchange is 1, the optimization range of K is [3,10], and the optimization range of α is [200,3000]. The fitness function is defined as the mean of the envelope entropy of all IMF components after VMD decomposition, and the minimum envelope mean entropy is used as the optimization target to obtain the optimal parameters. Among them, the definition of envelope entropy Ep is as follows:

[0080]

[0081] In the formula, p j is the normalized form of a(j), where a(j) is the envelope signal of signal x(j) after Hilbert transform.

[0082] 2) Calculate the fitness value corresponding to the current position of each group member.

[0083] 3) Divide the entire group into 5 groups, and select the individual with the smallest fitness value in the group as the group leader; divide the entire group into 4 regions, and select the individual with the smallest fitness value in the region as the regional winner.

[0084] 4) Conduct campaign activities, and other members update their positions according to (3) and (4).

[0085] 5) Exchange positions between group members with a certain probability.

[0086] 6) Calculate the fitness values ​​of group members after the position update and select new group leaders and regional winners.

[0087] 7) The winners of each region form a conference group and conduct conference group affairs, updating the position according to formula (5).

[0088] 8) Repeat steps 4) to 7) until the iteration termination condition is met and output the optimal solution, which is the optimal α and K parameter values.

[0089] like Figure 5 As shown, in step S3, the implementation process of the refined composite multi-scale fuzzy entropy (RCMFE) is as follows:

[0090] 1) For a given time series {x(i), i=1,2,…,N}, its composite coarse-grained sequence is:

[0091]

[0092] Where s is the scale factor.

[0093] 2) Under scale s, construct the m-dimensional vector of each composite coarse-grained sequence as:

[0094]

[0095] Where m is the embedding dimension,

[0096] 3) Define vector X i and X j The distance between them is:

[0097]

[0098] In the formula, p=1,2,…,m-1, i≠j.

[0099] 4) Introduce the fuzzy membership function and define the vector X i and X j The similarity between them is:

[0100] D i,j =exp(-(d i,j / r) n ) (10)

[0101] Where n and r represent the gradient and similarity tolerance of the exponential function boundary, respectively.

[0102] 5) Define the average similarity function in m dimension as:

[0103]

[0104] 6) Similarly, the average similarity function in m+1 dimension It can also be obtained according to formula (11). Under the scale s, there are s coarse-grained sequences. By calculating the m and m+1 dimensional average similarity functions of each coarse-grained sequence, the expression of the fine composite multi-scale fuzzy entropy can be finally obtained:

[0105]

[0106] In the formula,

[0107] In step S4, a Grey Wolf Pack Optimized Support Vector Machine (GWO-SVM) classifier is used to perform fault identification, which specifically includes the following steps:

[0108] 1) The fault feature set obtained in step S3 is divided into a test sample set and a training sample set in a ratio of 4:6 in terms of the number of each fault category. Then the training sample set and the test sample set are normalized.

[0109] 2) According to the SVM model, the radial basis function is selected as the kernel function, and the gray wolf population optimization algorithm is used. The average correct recognition rate after three-fold crossover of the training samples is defined as the fitness value, and the optimal penalty factor and kernel function parameters of the SVM model are automatically determined. Among them, the gray wolf population size of the GWO algorithm is set to 20, and the termination iteration is 100.

[0110] 3) Use the training samples to build an SVM prediction model, and use the SVM prediction model to diagnose and identify the test samples to determine the type of rolling bearing fault.

[0111] This embodiment also provides a rolling bearing fault diagnosis system, including a memory, a processor, and computer program instructions stored in the memory and capable of being executed by the processor. When the processor executes the computer program instructions, the above-mentioned method steps can be implemented.

[0112] In order to verify the effectiveness of the present invention, the rolling bearing fault diagnosis test bench was used to collect experimental data, and 100 groups of vibration acceleration signals of the rolling bearing in normal state, inner ring fault state, outer ring fault state and rolling element fault state were collected. The motor speed was set to 1772r / min and the sampling frequency was f s =12000Hz, the sampling point is set to 1024.

[0113] Then, if Figure 1 As shown, follow the steps below:

[0114] Step 1: Use the acceleration sensor to collect experimental data of the above rotating mechanical equipment to obtain its vibration acceleration signal. The time domain waveforms of the four states of the rolling bearing are as follows: Figure 2 As shown. Figure 2 It can be seen that the various working conditions cannot be effectively distinguished only from the signal time domain waveform, and the signals collected in the experiment have certain noise interference. Therefore, the optimized VMD method proposed in the present invention is used to reduce the noise of the vibration signal.

[0115] Step 2: Use the optimized VMD algorithm to reduce the noise of the vibration signal. The process is as follows: Figure 3 As shown in Figure 1. Taking a set of rolling element fault signals as an example, the optimized VMD algorithm is used to reduce the noise. The optimization algorithm optimizes the VMD parameters to [228,5], and the global optimal fitness value is 2.8693. The above IMF components are reconstructed to obtain the fault signal after noise reduction. The corresponding decomposition and reconstruction results are shown in Figure 1. Figure 4 shown.

[0116] Step 3: In order to fully characterize the fault information of different working conditions, RCMFE is used to extract fault features of 100 groups of samples in each of the above four working conditions. The scale s is set to 25, n = 2, m = 2, r = 0.15SD, and 400 groups of samples can obtain a 25-dimensional feature matrix F 400×25 , the entropy characteristic curves of the four working conditions are as follows Figure 6 shown.

[0117] Step 4: Input the result after feature extraction into the GWO-SVM multi-fault classifier for identification. The training samples and test samples of each type of working condition are randomly allocated in a ratio of 4:6. The average recognition rate after three-fold crossover of the training samples is defined as the fitness value. The gray wolf population size of the GWO algorithm is set to 20, and the termination iteration is 100. The recognition results are as follows: Figure 7 As shown. Figure 7 It can be seen that the proposed “optimized VMD+RCMFE+GWO-SVM” fault diagnosis model has a fault recognition rate of 100% for rolling bearings under different working conditions, and can accurately distinguish the four types of samples, thereby verifying the effectiveness of the model.

[0118] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that include computer-usable program code.

[0119] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0120] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0121] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0122] The above is only a preferred embodiment of the present invention, and does not limit the present invention in other forms. Any technician familiar with the profession may use the above disclosed technical content to change or modify it into an equivalent embodiment with equivalent changes. However, any simple modification, equivalent change and modification made to the above embodiment according to the technical essence of the present invention without departing from the technical solution of the present invention still belongs to the protection scope of the technical solution of the present invention.

Claims

1. A rolling bearing fault diagnosis method, characterized in that: Firstly, the vibration acceleration signal of the rolling bearing is decomposed and reconstructed by using optimized variational mode decomposition. Then, the reconstructed signal is extracted by using fine composite multi-scale fuzzy entropy to construct a fault feature set. Finally, the gray wolf pack optimized support vector machine classifier is used for diagnosis and identification to diagnose the rolling bearing fault type. The rolling bearing fault diagnosis method comprises the following steps: Step S1, using an acceleration sensor to collect vibration acceleration signals of a rolling bearing in a normal state, an inner ring fault state, an outer ring fault state, and a rolling element fault state, and obtaining a time domain signal sample set; Step S2, using an optimization algorithm to optimize the mode number K and penalty factor α in the variational mode decomposition algorithm, then decomposing the collected vibration acceleration signal, and reconstructing the obtained intrinsic mode components; Step S3, using fine composite multi-scale fuzzy entropy to extract features from the reconstructed signal and construct a fault feature set; Step S4, inputting the fault features into the gray wolf pack optimized support vector machine classifier for training and testing, and diagnosing the rolling bearing fault type; In step S2, the mode number K and penalty factor α in the variational mode decomposition algorithm are optimized using an optimization algorithm, including the following steps: Step S201, initialize various parameters of the optimization algorithm and the fitness function in the process of optimizing the parameters; the fitness function is defined as the mean of the envelope entropy of all IMF components after VMD decomposition as the fitness value, and the minimum envelope mean entropy is used as the optimization target to obtain the optimal parameters; wherein, the definition of envelope entropy Ep is as follows: In the formula, p j is the normalized form of a(j), a(j) is the envelope signal of signal x(j) after Hilbert transform, and N is the length of signal x(j); Step S202, calculating the fitness value corresponding to the current position of each group member; Step S203: Divide the entire group into c groups, select the individual with the smallest fitness value in the group as the group leader; divide the entire group into d regions, select the individual with the smallest fitness value in the region as the regional winner; Step S204: Conduct campaign activities, and other members update their positions according to the group leader and the regional winner, as follows: when When , the spatial position is updated according to formula (2): when When , the spatial position is updated according to formula (3): In the formula, It indicates the position of the jth member of the i-th group when the number of iterations is t; express The kth element in m * Indicates the group leader or regional winner. When iterating, first m * As the leader of the group, the members of the group update their positions according to the leader's position, and then let m * is the regional winner, and the members in the region update their own positions according to the winner's position; r is a random number between 0 and 1; Step S205: exchanging positions between group members with a set probability; Step S206, calculating the fitness values ​​of group members after the position update, and selecting a new group leader and regional winner; Step S207: The winners of each region form a conference group and conduct conference group affairs, and update the position according to formula (4): In the formula, represents the winner of region a, represents the winner of area b. Let the number of areas be n, then a is a random integer between 1 and n, b = 1, 2, ..., n, a ≠ b; according to The updated new position is r is a random number between 0 and 1; if The fitness value is better than If it is, replace its position, otherwise keep it as it is; Step S208, repeat steps S204 to S207 until the iteration termination condition is met, and output the optimal solution, that is, the optimal α and K parameter values.

2. A rolling bearing fault diagnosis method according to claim 1, characterized in that: In step S1, the vibration acceleration signal is a driving end vibration acceleration signal.

3. A rolling bearing fault diagnosis method according to claim 1, characterized in that: The parameters are initialized as follows: the population size is 20, the number of groups is 5, the number of regions is 4, the maximum number of iterations is 20, the maximum probability of group exchange is 1, the optimization range of K is [3,10], and the optimization range of α is [200,3000].

4. A rolling bearing fault diagnosis method according to claim 1, characterized in that: Step S4 specifically includes the following steps: Step S401: the fault feature set obtained in step S3 is divided into a training sample set and a test sample set according to a set ratio in the number of each fault category, and then the training sample set and the test sample set are normalized; Step S402: According to the SVM model, the radial basis function is selected as the kernel function, and the gray wolf pack optimization algorithm is used to define the average correct recognition rate after three-fold crossover of the training samples as the fitness value, and automatically determine the optimal penalty factor and kernel function parameters of the SVM model; Step S403: construct an SVM prediction model using the training samples, and use the SVM prediction model to diagnose and identify the test samples to determine the rolling bearing fault type.

5. A rolling bearing fault diagnosis system, characterized in that: The method comprises a memory, a processor and computer program instructions stored in the memory and capable of being executed by the processor. When the processor executes the computer program instructions, the method steps as described in any one of claims 1 to 4 can be implemented.

Images