Motor bearing fault diagnosis method, device, equipment and medium

Abstract

The present application discloses a motor bearing fault diagnosis method, apparatus, equipment, and medium, relating to the technical field of motor bearing fault diagnosis. The method comprises: obtaining a vibration signal and a current signal of a motor bearing, performing feature extraction on the vibration signal and the current signal respectively to obtain a first time domain feature, a first frequency domain feature, and a first energy feature of the vibration signal and a second time domain feature, a second frequency domain feature, and a second energy feature of the current signal, performing feature fusion and feature dimensionality reduction on the first time domain feature, the first frequency domain feature, the first energy feature, the second time domain feature, the second frequency domain feature, and the second energy feature to obtain fused and reduced dimensionality features, using the fused and reduced dimensionality features as input, and using a fault diagnosis model to determine whether the motor bearing has a fault and the type of fault if a fault exists. The present application can improve the accuracy of fault diagnosis.

Description

Technical Field

[0001] The present application relates to the technical field of motor bearing fault diagnosis, and in particular to a motor bearing fault diagnosis method, device, equipment and medium. Background Art

[0002] With the rapid development of high-speed train technology, traction motors, as the core power system of high-speed trains, have a significant impact on the reliability and safety of the entire train. Bearings are critical components in traction motors, and failures can not only degrade motor performance but, in severe cases, lead to serious safety accidents. Therefore, accurately diagnosing traction motor bearing failures is crucial for ensuring safe operation of high-speed trains and reducing maintenance costs.

[0003] Traditional motor bearing fault diagnosis methods primarily rely on the analysis of a single signal, such as a vibration signal or current signal. However, in complex operating environments, a single signal often fails to fully reflect the health status of the motor bearing and is easily affected by noise and interference, resulting in inaccurate fault diagnosis results. Multi-signal fusion technology has attracted widespread attention in the field of fault diagnosis. By integrating multiple sensor signals, it can provide more comprehensive and reliable fault information, effectively improving the accuracy and robustness of fault diagnosis. However, current multi-signal fusion methods often directly use multiple signals as model inputs, and the model is used to perform fault diagnosis, resulting in low fault diagnosis accuracy. Summary of the Invention

[0004] The purpose of this application is to provide a motor bearing fault diagnosis method, device, equipment and medium, which can improve the fault diagnosis accuracy.

[0005] To achieve the above objectives, this application provides the following solutions.

[0006] In a first aspect, the present application provides a motor bearing fault diagnosis method, the motor bearing fault diagnosis method comprising:

[0007] Obtain vibration signals and current signals of motor bearings;

[0008] performing feature extraction on the vibration signal and the current signal respectively to obtain a first time domain feature, a first frequency domain feature, and a first energy feature of the vibration signal, and a second time domain feature, a second frequency domain feature, and a second energy feature of the current signal; the first energy feature and the second energy feature are both energy entropy;

[0009] Performing feature fusion and feature dimensionality reduction on the first time domain feature, the first frequency domain feature, the first energy feature, the second time domain feature, the second frequency domain feature, and the second energy feature to obtain a fused and reduced dimensionality feature;

[0010] The fused dimensionality-reduced features are used as input, and a fault diagnosis model is used to determine whether a motor bearing has a fault and the fault type when a fault exists. The fault diagnosis model is a support vector machine, and the hyperparameters of the support vector machine are determined by an improved dung beetle optimization algorithm. The improved dung beetle optimization algorithm is an algorithm obtained by improving the dung beetle optimization algorithm. The improvement includes: using Logistic chaotic mapping to initialize the population, using the golden sine strategy to update the position update formula of the dung beetle rolling behavior to obtain a first update formula, using a multi-objective selection strategy to update the position update formula of the dung beetle breeding behavior to obtain a second update formula, using an adaptive step size and differential evolution strategy to update the position update formula of the dung beetle foraging behavior to obtain a third update formula, and using a dynamic weight strategy to update the position update formula of the dung beetle stealing behavior to obtain a fourth update formula.

[0011] In a second aspect, the present application provides a motor bearing fault diagnosis device, the motor bearing fault diagnosis device comprising: a vibration sensor, a current sensor, and a processor;

[0012] The vibration sensor is installed on the motor bearing, and the vibration sensor is used to collect the vibration signal of the motor bearing;

[0013] The current sensor is installed on the motor bearing, and is used to collect the current signal of the motor bearing;

[0014] The processor is communicatively connected to the vibration sensor and the current sensor respectively, and the processor is used to execute the above-mentioned motor bearing fault diagnosis method.

[0015] In a third aspect, the present application provides a computer device comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the above-mentioned motor bearing fault diagnosis method.

[0016] In a fourth aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, which implements the above-mentioned motor bearing fault diagnosis method when executed by a processor.

[0017] According to the specific embodiments provided in this application, this application has the following technical effects:

[0018] The present application provides a motor bearing fault diagnosis method, device, equipment and medium, which obtains the vibration signal and current signal of the motor bearing, performs feature extraction on the vibration signal and the current signal respectively, obtains the first time domain feature, the first frequency domain feature and the first energy feature of the vibration signal and the second time domain feature, the second frequency domain feature and the second energy feature of the current signal, performs feature fusion and feature dimensionality reduction on the first time domain feature, the first frequency domain feature, the first energy feature, the second time domain feature, the second frequency domain feature and the second energy feature, obtains the fused and reduced dimensionality feature, uses the fused and reduced dimensionality feature as input, and uses a fault diagnosis model to determine whether the motor bearing has a fault and the type of fault when a fault exists. By introducing feature extraction, feature fusion and feature dimensionality reduction, the present application can fully fuse the features of the vibration signal and the features of the current signal to obtain the fused and reduced dimensionality feature, and subsequently inputs the fused and reduced dimensionality feature into the fault diagnosis model to perform fault diagnosis, thereby fully considering the correlation between the vibration signal and the current signal to perform fault diagnosis, which can improve the accuracy of fault diagnosis. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0020] Figure 1 This is an application environment diagram of a motor bearing fault diagnosis method provided in Example 1 of the present application.

[0021] Figure 2 A flow chart of a motor bearing fault diagnosis method provided in Example 1 of the present application.

[0022] Figure 3 A schematic diagram of the technical route of a motor bearing fault diagnosis method provided in Example 1 of the present application.

[0023] Figure 4 A detailed flowchart of a motor bearing fault diagnosis method provided in Example 1 of the present application.

[0024] Figure 5 This is a time-frequency domain schematic diagram of variational modal decomposition of a vibration signal provided in Example 1 of the present application; wherein, Figure 5 a1, a2, a3, a4, a5, and a6 are the time domain diagrams of the six intrinsic mode function components obtained by performing variational modal decomposition on the vibration signal. The horizontal axis is time and the vertical axis is amplitude. Figure 5b1, b2, b3, b4, b5, and b6 in the figure are frequency domain diagrams of the six intrinsic mode function components obtained by performing variational modal decomposition on the vibration signal. The horizontal axis is the frequency and the vertical axis is the amplitude.

[0025] Figure 6 This is a time-frequency domain diagram of variational modal decomposition of a current signal provided in Example 1 of the present application; wherein, Figure 6 a1, a2, a3, a4, a5, a6, and a7 are the time domain diagrams of the seven intrinsic mode function components obtained by performing variational modal decomposition on the current signal. The abscissa is time and the ordinate is amplitude. Figure 6 b1, b2, b3, b4, b5, b6, and b7 are frequency domain diagrams of the seven intrinsic mode function components obtained by performing variational modal decomposition on the current signal. The horizontal axis is the frequency and the vertical axis is the amplitude.

[0026] Figure 7 A schematic flow chart of the improved dung beetle optimization algorithm provided in Example 1 of the present application.

[0027] Figure 8 Schematic diagram of the classification prediction results of the test set provided in Example 1 of the present application.

[0028] Figure 9 Schematic diagram of the classification confusion matrix of the test set provided in Example 1 of the present application.

[0029] Figure 10 A schematic diagram of the structure of a computer device provided in Example 3 of the present application. DETAILED DESCRIPTION

[0030] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0031] Example 1.

[0032] The motor bearing fault diagnosis method provided in the embodiment of the present application can be applied to Figure 1In the application environment shown, the terminal communicates with the server via a network. The data storage system can store data that the server needs to process. The data storage system can be set up separately, integrated on the server, or placed on a cloud or other server. The terminal can send a pending fault diagnosis request to the server. After the server receives the pending fault diagnosis request, it obtains the vibration signal and current signal of the motor bearing for the pending fault diagnosis request. It performs feature extraction on the vibration signal and current signal to obtain the first time domain feature, first frequency domain feature, and first energy feature of the vibration signal, and the second time domain feature, second frequency domain feature, and second energy feature of the current signal. It then performs feature fusion and feature dimensionality reduction on the first time domain feature, first frequency domain feature, first energy feature, second time domain feature, second frequency domain feature, and second energy feature to obtain fused and reduced features. Using the fused and reduced features as input, the fault diagnosis model is used to determine whether the motor bearing is faulty and, if so, the fault type. The server can provide feedback to the terminal regarding the fault diagnosis result obtained in response to the fault diagnosis request, including whether the motor bearing is faulty and, if so, the fault type.

[0033] In addition, in some embodiments, the motor bearing fault diagnosis method can also be implemented independently by a server or a terminal. For example, the terminal can directly process the fault diagnosis request to be processed, or the server can obtain the fault diagnosis request to be processed from the data storage system and process the fault diagnosis request to be processed.

[0034] In an exemplary embodiment, Figure 2 As shown, a motor bearing fault diagnosis method is provided. The method is executed by a computer device, specifically, it can be executed by a computer device such as a terminal or a server alone, or it can be executed by a terminal and a server together. In the embodiment of the present application, the method is applied to Figure 1 The following steps are used as an example to illustrate the server.

[0035] Step S1, obtaining a vibration signal and a current signal of a motor bearing.

[0036] Step S2: performing feature extraction on the vibration signal and the current signal respectively to obtain a first time domain feature, a first frequency domain feature, and a first energy feature of the vibration signal, and a second time domain feature, a second frequency domain feature, and a second energy feature of the current signal; the first energy feature and the second energy feature are both energy entropy.

[0037] Step S3: performing feature fusion and feature dimensionality reduction on the first time domain feature, the first frequency domain feature, the first energy feature, the second time domain feature, the second frequency domain feature, and the second energy feature to obtain a fused and reduced dimensionality feature.

[0038] Step S4, using the fused dimensionality reduction features as input, and using a fault diagnosis model to determine whether the motor bearing has a fault and the fault type when a fault exists; the fault diagnosis model is a support vector machine, and the hyperparameters of the support vector machine are determined by an improved dung beetle optimization algorithm, and the improved dung beetle optimization algorithm is an algorithm obtained by improving the dung beetle optimization algorithm, and the improvement includes: using Logistic chaos mapping to initialize the population, using the golden sine strategy to update the position update formula of the dung beetle rolling behavior to obtain a first update formula, using a multi-objective selection strategy to update the position update formula of the dung beetle breeding behavior to obtain a second update formula, using an adaptive step size and differential evolution strategy to update the position update formula of the dung beetle foraging behavior to obtain a third update formula, and using a dynamic weight strategy to update the position update formula of the dung beetle stealing behavior to obtain a fourth update formula.

[0039] By implementing the above-mentioned steps S1 to S4, this embodiment provides a motor bearing fault diagnosis method, first performing feature extraction on the vibration signal and the current signal respectively to obtain the first time domain feature, the first frequency domain feature and the first energy feature of the vibration signal and the second time domain feature, the second frequency domain feature and the second energy feature of the current signal, and then performing feature fusion and feature dimensionality reduction on the first time domain feature, the first frequency domain feature, the first energy feature, the second time domain feature, the second frequency domain feature and the second energy feature to obtain the fused and reduced dimensionality feature, and finally using the fused and reduced dimensionality feature as input, a fault diagnosis model is used to determine whether there is a fault in the motor bearing and the type of fault when a fault exists. By introducing feature extraction, feature fusion and feature dimensionality reduction, the features of the vibration signal and the features of the current signal can be fused before the vibration signal and the current signal are input into the fault diagnosis model, so that when the vibration signal and the current signal are used for fault diagnosis, the correlation between the vibration signal and the current signal is considered, and the fault diagnosis accuracy can be improved.

[0040] The following, combined Figure 3 and Figure 4 , the motor bearing fault diagnosis method used in this embodiment is introduced in detail, including the following steps.

[0041] Step 1: Collect the vibration signal and current signal of the motor bearing.

[0042] This embodiment uses an acceleration sensor and a current sensor to collect the vibration signal and current signal of the motor bearing, respectively. Specifically, a three-axis acceleration sensor can be used to collect the vibration signal of the motor bearing, and a current transformer can be used to collect the current signal of the motor bearing. The collected vibration signal and current signal are then time-aligned and downsampled (i.e., down-sampled) to balance signal quality and computational cost and construct the original data set.

[0043] At this time, in this embodiment, the vibration signal and current signal of the motor bearing are obtained.

[0044] Step 2: Extract the time domain and frequency domain features of the vibration and current signals, construct a time-frequency domain dataset, decompose the vibration and current signals through variational mode decomposition (VMD), and calculate the energy entropy of the intrinsic mode function (IMF) components obtained by decomposition to construct an energy entropy dataset.

[0045] This embodiment extracts the typical time domain features and typical frequency domain features of the collected vibration signal and current signal to obtain the time domain features of the vibration signal and current signal, such as mean, kurtosis, effective value and crest factor. At the same time, the frequency domain features of the vibration signal and current signal, such as main frequency of the spectrum, harmonic energy ratio, power spectrum density and total harmonic distortion, are obtained to construct the typical time domain and frequency domain data sets corresponding to the vibration signal. (including the time domain characteristics and frequency domain characteristics of vibration signals) and typical time domain and frequency domain data sets corresponding to current signals (including the time domain characteristics and frequency domain characteristics of the current signal), and Composed of time-frequency domain data sets.

[0046] This embodiment uses a dual-index comprehensive evaluation method based on reconstruction accuracy and modal independence to analyze and obtain the most suitable VMD decomposition level for each of the vibration signal and the current signal. Based on the VMD decomposition level, VMD decomposition is performed on the vibration signal and the current signal respectively to obtain multiple intrinsic mode function components corresponding to each of the vibration signal and the current signal. The energy entropy of the multiple intrinsic mode function components obtained by decomposition is then calculated to construct a VMD decomposition energy entropy dataset for the vibration signal. (including energy entropy of vibration signals) and current signal VMD decomposition energy entropy data set (including the energy entropy of the current signal), and Composition energy entropy data set.

[0047] Energy entropy H The calculation expression is:

[0048] (1);

[0049] In formula (1), K is the number of eigenmode function components; For the The energy proportion of the eigenmode function component is equal to The total energy is the sum of the energies of each eigenmode function component.

[0050] The method for finding the most suitable number of VMD decomposition layers is a dual-index comprehensive evaluation method based on reconstruction accuracy and modal independence. In this dual-index comprehensive evaluation method based on reconstruction accuracy and modal independence, the root mean square error (RMSE) and the mean cross-correlation coefficient (MCC) are combined to select the number of VMD decomposition layers. K , specifically by traversing different VMD decomposition layers K , calculate each K The root mean square error RMSE between the reconstructed signal and the original signal under the value and the average cross-correlation coefficient MCC between each IMF component are used to select the optimal value according to the comprehensive performance of these two indicators. K The specific steps for finding the most suitable number of VMD decomposition layers using a dual-index comprehensive evaluation method based on reconstruction accuracy and modal independence are as follows.

[0051] (1) Set the number of VMD decomposition layers K The search range is, for example, from 2 to 20.

[0052] (2) Take multiple values within the search range according to the preset step size to obtain multiple K Value, for each K Value, in this K Under the value, the original signal Perform VMD decomposition and get K IMF components , put this K The IMF components are added together to reconstruct the original signal and the reconstructed signal is obtained. .

[0053] Among them, the reconstructed signal The calculation formula is:

[0054] (2);

[0055] In formula (2), For the IMF components.

[0056] (3) Calculate the reconstructed signal and the original signal The RMSE between K RMSE under the value.

[0057] The calculation formula of the root mean square error RMSE is:

[0058] (3);

[0059] In formula (3), for K RMSE under the value; is the signal length, i.e. the number of time points of the original signal; The original signal At the time point The value of To reconstruct the signal At the time point value.

[0060] (4) Use the average mutual correlation coefficient MCC to calculate the correlation between each IMF component and obtain the K The MCC under the value of , measures whether there is redundancy between modes.

[0061] The calculation formula of the average mutual correlation coefficient MCC is:

[0062] (4);

[0063] In formula (4), for K MCC under value; For the IMF components Hedi IMF components The Pearson correlation coefficient between .

[0064] The calculation formula of Pearson correlation coefficient is:

[0065] (5);

[0066] In formula (5), For the IMF components Hedi IMF components The covariance between IMF components Hedi IMF components The joint change trend between For the IMF components The standard deviation of IMF components the degree of volatility; For the IMF components The standard deviation of IMF components degree of fluctuation.

[0067] The formulas for calculating covariance and standard deviation are:

[0068] (6);

[0069] In formula (6), For the IMF components at time The value of For the The mean of the IMF components, ; For the IMF components at time The value of For the The mean of the IMF components, .

[0070] (5) Construct a weighted comprehensive scoring function to calculate each K The comprehensive score under the value, select the one with the smallest comprehensive score K The value is taken as the most suitable VMD decomposition layer number.

[0071] The weighted comprehensive scoring function is:

[0072] (7);

[0073] In formula (7), for K Comprehensive score under value; is the first weight; for K Normalized RMSE under value; is the second weight; for K Normalized MCC under the value. and According to user needs, when more emphasis is placed on reconstruction accuracy, increase ; When more emphasis is placed on modal independence, increase .

[0074] Normalize RMSE and MCC respectively to obtain normalized RMSE and normalized MCC. The calculation formula is:

[0075] (8);

[0076] In formula (8), For all K The minimum value of RMSE under the value; For all K The maximum value of RMSE under the value; For all K The minimum value of MCC under the value; For all K The maximum value of MCC under the value.

[0077] As an emerging signal processing technology, variational mode decomposition (VMD) has the core idea of decomposing a given signal into a series of intrinsic mode function components with specific center frequencies, while minimizing the sum of the estimated bandwidths of each eigenmode function component.

[0078] To ensure the stability of the decomposition results, the constrained variational model is introduced, which is expressed as follows:

[0079] (9);

[0080] In formula (9), Represents the set of components of each intrinsic mode function; Represents a set of center frequencies of each eigenmode function component; K is the number of VMD decomposition layers, that is, the number of intrinsic mode function components; About time The partial derivative of , which represents the gradient; is the unit impulse function, which is used to define the Hilbert transform; j is an imaginary unit, satisfying , used for complex number operations; For the The eigenmode function components at time The value of is a complex exponential modulation term, which is used to modulate the analytical signal to a position where the fundamental frequency is zero. For the The center frequency of the eigenmode function components; For a given signal.

[0081] In order to solve the constrained variation problem described in equation (9), the alternating direction multiplier method is used to solve it, and the augmented Lagrangian function is introduced. By using the Lagrangian operator and the quadratic penalty operator, the constrained variation problem is transformed into an unconstrained variation problem. The augmented Lagrangian function as follows:

[0082] (10);

[0083] In formula (10), is the Lagrange multiplier, used to enforce constraints; is a regularization parameter that controls the strength of the bandwidth constraint.

[0084] Continuously update the intrinsic mode function components and center frequency To minimize the spectral overlap between modes and to satisfy the iteration termination condition by alternating iterations through the Lagrangian operator.

[0085] After obtaining the most suitable VMD decomposition layer number, the vibration signal and current signal are decomposed by VMD. The time domain and frequency domain diagrams after decomposition are as follows: Figure 5 and Figure 6 shown.

[0086] At this time, in this embodiment, feature extraction is performed on the vibration signal and the current signal respectively to obtain the first time domain feature, the first frequency domain feature and the first energy feature of the vibration signal and the second time domain feature, the second frequency domain feature and the second energy feature of the current signal.

[0087] The first and second time domain features both include mean, kurtosis, effective value, and crest factor. The first and second frequency domain features both include dominant frequency, harmonic energy ratio, power spectrum density, and total harmonic distortion. The first and second energy features both include energy entropy.

[0088] The method for determining the first energy characteristic includes: performing variational modal decomposition on the vibration signal to obtain multiple first intrinsic mode function components, calculating the energy of each first intrinsic mode function component, and obtaining the first energy characteristic based on the energy calculation of all first intrinsic mode function components, specifically calculating through formula (1). The method for determining the second energy characteristic includes: performing variational modal decomposition on the current signal to obtain multiple second intrinsic mode function components, calculating the energy of each second intrinsic mode function component, and obtaining the second energy characteristic based on the energy calculation of all second intrinsic mode function components, specifically calculating through formula (1).

[0089] Among them, the vibration signal is subjected to variational modal decomposition to obtain multiple first intrinsic modal function components, specifically including: using a dual-index comprehensive evaluation method based on reconstruction accuracy and modal independence to determine the first decomposition level number used when performing variational modal decomposition on the vibration signal, performing variational modal decomposition on the vibration signal according to the first decomposition level number, and obtaining multiple first intrinsic modal function components.

[0090] Among them, variational modal decomposition is performed on the current signal to obtain multiple second intrinsic mode function components, specifically including: using a dual-index comprehensive evaluation method based on reconstruction accuracy and modal independence to determine the second decomposition layer number used when performing variational modal decomposition on the current signal, performing variational modal decomposition on the current signal according to the second decomposition layer number, and obtaining multiple second intrinsic mode function components.

[0091] Among them, a dual-index comprehensive evaluation method based on reconstruction accuracy and modal independence is used to determine the first decomposition layer number used for variational modal decomposition of the vibration signal, specifically including: designing multiple first initial decomposition layers; for each first initial decomposition layer number, performing variational modal decomposition on the vibration signal according to the first initial decomposition layer number to obtain multiple first initial intrinsic modal function components, calculating the sum of all first initial intrinsic modal function components to obtain a reconstructed vibration signal, calculating the first root mean square error under the first initial decomposition layer number based on the vibration signal and the reconstructed vibration signal, calculating the first average mutual correlation coefficient under the first initial decomposition layer number based on all first initial intrinsic modal function components, and calculating the first comprehensive score under the first initial decomposition layer number based on the first root mean square error and the first average mutual correlation coefficient; selecting the first initial decomposition layer number with the smallest first comprehensive score as the first decomposition layer number used for variational modal decomposition of the vibration signal.

[0092] Among them, a dual-index comprehensive evaluation method based on reconstruction accuracy and modal independence is used to determine the second decomposition layer number used for variational modal decomposition of the current signal, specifically including: designing multiple second initial decomposition layers; for each second initial decomposition layer number, performing variational modal decomposition on the current signal according to the second initial decomposition layer number to obtain multiple second initial intrinsic modal function components, calculating the sum of all second initial intrinsic modal function components to obtain the reconstructed current signal, calculating the second root mean square error under the second initial decomposition layer number based on the current signal and the reconstructed current signal, calculating the second average mutual correlation coefficient under the second initial decomposition layer number based on all second initial intrinsic modal function components, and calculating the second comprehensive score under the second initial decomposition layer number based on the second root mean square error and the second average mutual correlation coefficient; selecting the second initial decomposition layer number with the smallest second comprehensive score as the second decomposition layer number used for variational modal decomposition of the current signal.

[0093] Step 3: fuse and reduce the dimension of the obtained time-frequency domain dataset and energy entropy dataset to output a multimodal dataset.

[0094] This embodiment is achieved by ( , ), ( , The constructed time-frequency domain dataset and energy entropy dataset were fused using the Canonical Correlation Analysis (CCA) algorithm, pairwise. To select appropriate feature dimensions to improve model performance, reduce the risk of overfitting, and increase computational efficiency, the Principal Component Analysis (PCA) algorithm was used to reduce the feature dimensions of the fused dataset and construct a multimodal dataset.

[0095] This embodiment fuses the features of the vibration signal and the current signal through canonical correlation analysis, effectively extracts the correlation information between the two, and improves the accuracy of fault diagnosis.

[0096] Canonical correlation analysis aims to find the linear combination of two sets of variables so that the correlation between the two sets of linear combinations is maximized. The specific principle and formula of canonical correlation analysis are as follows: For two data sets X and Y , find two projection vectors and ,Will X and Y The linear combination of is expressed as:

[0097] (11);

[0098] In formula (11), For the dataset X In the projection vector The linear combination results under ; For the dataset Y In the projection vector The linear combination results are as follows.

[0099] for and Two projection vectors, so that the correlation coefficient The maximum, its correlation coefficient is defined as:

[0100] (12);

[0101] To ensure that the results are interpretable and numerically stable, the constraints are:

[0102] (13);

[0103] In the above formula, for and covariance of for variance; for The superscript T stands for transpose.

[0104] This embodiment uses principal component analysis to reduce the dimensionality of the fused dataset, reduce the risk of overfitting, improve computational efficiency, and obtain a multimodal dataset.

[0105] Principal component analysis is a statistical method widely used in data dimensionality reduction and feature extraction. Its core idea is to project the original high-dimensional data into a low-dimensional space through linear transformation while retaining the variance information of the original high-dimensional data as much as possible.

[0106] Assume that the original data matrix is , is the number of samples, The purpose of principal component analysis is to find a transformation matrix , is the number of features after dimensionality reduction, so that the new feature matrix Preserve as much variance information of the original data matrix as possible, through eigenvalue decomposition, covariance matrix It can be expressed as:

[0107] (14);

[0108] In formula (14), U is the eigenvector matrix of the covariance matrix Σ.

[0109] The feature extraction, feature fusion and feature dimensionality reduction of this embodiment have the following advantages.

[0110] (1) Improving information integrity through multi-dimensional feature fusion: The energy entropy obtained by combining the time-frequency domain features with the VMD decomposition can not only reflect the basic statistical characteristics and frequency distribution of the signal, but also capture its nonlinear dynamic characteristics, thereby achieving a comprehensive description of the signal characteristics.

[0111] (2) Enhanced ability to analyze complex signals: The time-frequency domain features are suitable for characterizing the overall trend of the signal, and the energy entropy obtained by VMD decomposition can quantify the local nonlinear modal complexity. The combination of the two can accurately characterize the global and local characteristics of the signal and significantly improve the accuracy of analyzing complex signals.

[0112] (3) Improve system robustness: Multi-dimensional feature fusion effectively reduces the sensitivity of a single method to noise and reduces performance degradation caused by signal variation or interference, thereby enhancing the system's noise resistance and stability.

[0113] (4) Optimize model performance: Improve the model's expressiveness and capture subtle changes in the signal; enhance the ability to distinguish between different types of signals and reduce the risk of misclassification and missed classification; maintain high accuracy in the presence of noise interference or signal anomalies to ensure reliable analysis results.

[0114] In summary, this embodiment achieves a more comprehensive and in-depth signal description through multi-dimensional feature fusion, significantly improving the model's analytical ability, distinguishing ability, and robustness.

[0115] At this time, in this embodiment, feature fusion and feature dimensionality reduction are performed on the first time domain feature, the first frequency domain feature, the first energy feature, the second time domain feature, the second frequency domain feature, and the second energy feature to obtain fused and reduced dimensionality features.

[0116] Among them, feature fusion and feature dimensionality reduction are performed on the first time domain feature, the first frequency domain feature, the first energy feature, the second time domain feature, the second frequency domain feature and the second energy feature to obtain the fused and reduced dimension feature, specifically including: using the typical correlation analysis algorithm to perform feature fusion on the first time domain feature, the first frequency domain feature, the second time domain feature and the second frequency domain feature to obtain the first fused feature; using the typical correlation analysis algorithm to perform feature fusion on the first energy feature and the second energy feature to obtain the second fused feature; using the principal component analysis algorithm to perform feature dimensionality reduction on the first fused feature and the second fused feature to obtain the fused and reduced dimension feature.

[0117] Step 4: Improve the Dung Beetle Optimizer (DBO) algorithm and use the improved Dung Beetle Optimizer (IDBO) algorithm to optimize the hyperparameters of the Support Vector Machine (SVM) and build the IDBO-SVM fault diagnosis model.

[0118] SVMs, with their superior performance in processing small, nonlinear, and high-dimensional data, have become an important tool for fault diagnosis. However, SVM classification performance is highly dependent on its hyperparameters. Traditional parameter optimization methods, such as grid search and genetic algorithms, have high computational complexity when processing high-dimensional feature data and are prone to falling into local optima. The DBO algorithm, with its excellent global search capabilities and convergence speed, has become a powerful tool for SVM parameter optimization. However, the DBO algorithm still suffers from problems such as insufficient initial population diversity and a tendency to fall into local optima. Therefore, this embodiment improves the DBO algorithm by combining a dynamic weighting strategy with an efficient search mechanism and multimodal feature fusion technology to provide a new and efficient solution for motor bearing fault diagnosis.

[0119] The dung beetle optimization algorithm mathematically models the dung beetle's ball rolling, reproduction, foraging and stealing behaviors to find the global optimal solution.

[0120] Dung beetles use celestial clues to navigate and achieve the linear rolling behavior of the dung ball. Assuming that the dung beetle moves in a specific direction in the entire search space, the position of the dung beetle will be continuously updated during this process. The position update formula is:

[0121] (15);

[0122] In formula (15), For the A dung beetle in the +1 position information at iteration time; For the A dung beetle in the Position information at the iteration; is a natural coefficient assigned a value of -1 or 1, Indicates no deviation. Indicates deviation from the original direction; is a constant used to control The weighting coefficient is usually between 0 and 1, and can be 0.1 in this embodiment; For the A dung beetle in the - Position information at 1 iteration; is a constant used to control The weighting coefficient is usually between 0 and 1, and can be 0.3 in this embodiment; is a parameter used to simulate the intensity change of light; is the global worst position.

[0123] The dung beetle optimization algorithm proposes a boundary selection strategy to simulate the area where dung beetles lay eggs in a safe environment, which is defined as follows:

[0124] (16);

[0125] In formula (16), is the lower limit of the spawning area; is the local optimal position; R is a dynamic parameter that changes with the number of iterations; is the lower bound of the optimization problem; It is the upper limit of the spawning area; is the upper bound of the optimization problem; is the maximum number of iterations.

[0126] The female dung beetle lays only one egg in each iteration. The dung beetle hides the egg in a brood ball. The location of the brood ball is defined as follows:

[0127] (17);

[0128] In formula (17), For the The brooding ball is in the +1 position information at iteration time; and are two independent random vectors of size 1×Dim, where Dim is the dimension of the optimization problem; For the The brooding ball is in the The position information at the iteration.

[0129] The boundaries of the optimal foraging area are defined as follows:

[0130] (18);

[0131] In formula (18), is the lower limit of the optimal foraging area; is the global optimal position; is the upper limit of the optimal foraging area.

[0132] The formula for updating the position of the dung beetle foraging is as follows:

[0133] (19);

[0134] In formula (19), For the A small dung beetle in the +1 position information at iteration time; For the A small dung beetle in the Position information at the iteration; is a random number that follows a normal distribution; is a random vector belonging to (0, 1).

[0135] The position update formula of the thief dung beetle is as follows:

[0136] (20);

[0137] In formula (20), For the A thief dung beetle +1 position information at iteration time; S is a constant value; g is a random vector of size 1×Dim, following a normal distribution; For the A thief dung beetle The position information at the iteration.

[0138] The specific process of the dung beetle optimization algorithm is as follows.

[0139] (1) Initialize the population and parameters: set parameters such as the initial population size and the maximum number of iterations, and randomly generate the initial positions of dung beetle individuals.

[0140] (2) Calculating the fitness value: Calculate the fitness value of each dung beetle individual according to the objective function. In this embodiment, the objective function is defined as the cross-validation accuracy of the SVM.

[0141] (3) Simulating dung beetle behavior: Ball rolling behavior: Use equation (15) to update the position of the dung beetle to simulate the ball rolling process. Reproduction behavior: Use equation (17) to define the position of the brood ball to simulate the reproduction process. Foraging behavior: Use equation (19) to update the position of the young dung beetle to simulate the foraging process. Stealing behavior: Use equation (20) to update the position of the thief dung beetle to simulate theft behavior.

[0142] (4) Boundary selection strategy: Use formula (16) to determine the upper and lower limits of the egg-laying area to ensure that dung beetles move within a safe area.

[0143] (5) Update the best position: record the current global best position and the global worst position for the next iteration.

[0144] (6) Determine the termination condition: Check whether the maximum number of iterations has been reached or other termination conditions have been met. If not, return to (2) and continue iterating; otherwise, output the final solution.

[0145] The improved dung beetle optimization algorithm uses Figure 7 The idea shown in the paper is to improve the dung beetle optimization algorithm by using Logistic chaos mapping, golden sine strategy, multi-objective selection strategy, adaptive step size, differential evolution strategy and dynamic weight strategy, and use the improved dung beetle optimization algorithm to find the hyperparameters of the support vector machine and construct the IDBO-SVM fault diagnosis model.

[0146] (1) Using the population initialization method of Logistic chaotic mapping, the generated sequence has good uniformity and randomness, which can make the initial population uniformly distributed in the search space. The expression of Logistic chaotic mapping is:

[0147] (twenty one);

[0148] In formula (21), For the The chaotic sequence value of the iteration; It is a control parameter, usually takes a value of 4; For the The chaotic sequence value of the iteration is used to initialize the position of individuals in the population to ensure that the population is evenly distributed in the search space.

[0149] (2) The golden sine strategy is used in the position update of the dung beetle's rolling ball behavior to narrow the search space and conduct sufficient search in the local area, so that the algorithm's global search ability and local development ability are well balanced. The position update formula using the golden sine strategy is as follows:

[0150] (twenty two);

[0151] In formula (22), and is the golden ratio coefficient; The golden ratio, ; For the A dung beetle in the +1 position information at iteration time; For the A dung beetle in the Position information at the iteration; for[ ]A random number in the range; for[ ]A random number in the range; For the A dung beetle in the -1 position information at iteration.

[0152] (3) A multi-objective selection strategy is introduced into the reproductive behavior of dung beetles. By combining the influence of the local optimal solution and the global elite solution, the selection of the breeding area is made more flexible. The position update formula using the multi-objective selection strategy is as follows:

[0153] (twenty three);

[0154] In formula (23), For the The brooding ball is in the +1 position information at iteration time; is the weight coefficient, which is used to balance the influence of local optimal solution and global elite solution. The value range of is [0, 1]; To be a local optimal solution, we need to ensure that and between; To provide a global elite solution, we need to ensure and between; is the disturbance intensity coefficient, which is used to control the magnitude of random disturbance; represents a random disturbance term that follows a normal distribution with mean 0 and variance 1.

[0155] If the calculated result exceeds the bounds, a bounds correction is performed:

[0156] (twenty four);

[0157] In formula (24), the min function is calculated using formula (23) .

[0158] In a limited area, the multi-objective selection strategy can better utilize existing information and avoid blind search. By weighing the influence of local optimal solutions and global elite solutions, it can achieve a better balance between exploration and development in a limited space, explore potential fault characteristics more comprehensively, and improve diagnostic accuracy.

[0159] (4) Adaptive step length and differential evolution strategy are introduced into the foraging behavior of dung beetles. The adaptive step length dynamically adjusts the step length according to the distance between the current solution and the global optimal solution, balancing exploration and development. The differential evolution strategy uses multiple individuals in the group to generate new solutions, enhancing population diversity and global search capabilities. The improved position update formula is:

[0160] (25);

[0161] In formula (25), For the A small dung beetle in the +1 position information at iteration time; are three different individuals randomly selected from the current population; is the first scaling factor, controlling the intensity of differential evolution; is the second scaling factor, used to control the maximum value of the step size; is a tuning parameter used to control the sensitivity of step length to distance changes; is the global optimal solution; For the A small dung beetle in the Position information at the iteration; It is the upper boundary of the spawning area; It is the lower bound of the spawning area.

[0162] By improving the foraging behavior of the dung beetle optimization algorithm and introducing adaptive step size and differential evolution strategies, the algorithm's global search capability and population diversity can be significantly improved without destroying the basic framework of the original algorithm. These improvements not only enhance the performance of the algorithm, but also improve its application effect in fault diagnosis.

[0163] (5) A dynamic weight strategy is introduced into the stealing behavior of dung beetles. The position update formula of the dynamic weight strategy is:

[0164] (26);

[0165] In formula (26), For the A thief dung beetle +1 position information at iteration time; and Are weight parameters, in the early stage of iteration Set it to a larger value to make the dung beetle explore a better area near the location of the best food, increase the global optimization ability of the algorithm, and in the later stages of the iteration Gradually increase, so that the dung beetle has the ability to jump out of the local optimum.

[0166] To improve the DBO algorithm, in this embodiment, in the population initialization phase, a logistic chaotic map is used to generate the initial population. This method has good uniformity and randomness, making the distribution of population resources more balanced in the search space, thereby improving the initial exploration capability of the algorithm. In the position update phase of the dung beetle rolling behavior, a golden sine strategy is introduced to achieve a good balance between global search capability and local development capability by narrowing the search space and conducting sufficient search in local areas. In the position update phase of the dung beetle breeding behavior, a multi-objective selection strategy is adopted, combining the influence of local optimal solutions and global elite solutions to make the selection of breeding areas more flexible. At the same time, the rationality of the search range is ensured through boundary correction, further improving the exploration and development effects of the algorithm. In the position update phase of the dung beetle foraging behavior, an adaptive step size and differential evolution strategy are introduced to dynamically adjust the step size to balance exploration and development, and use multiple individuals in the population to generate new solutions, thereby enhancing population diversity and global search capability. In the position update phase of the dung beetle stealing behavior, a dynamic weight strategy is adopted to increase the algorithm's global optimization capability in the early iteration stage and gradually enhance the local search capability in the later iteration stage, enabling the algorithm to escape the local optimum. The DBO algorithm is improved by the above method, and the improved DBO algorithm is used to find the hyperparameters of SVM to construct the IDBO-SVM fault diagnosis model.

[0167] Using the improved DBO algorithm to find the hyperparameters of SVM includes: defining the optimization goal and initialization, setting the hyperparameters that need to be optimized in SVM, setting the parameters of the improved DBO algorithm, setting the algorithm population size, initializing the population position and fitness value, the maximum number of iterations, etc.; executing the improved DBO algorithm, using the cross-validation accuracy of SVM as the fitness value, calculating and updating the position and fitness value of the dung beetle, and continuously updating the fitness value of each individual, the individual optimal position and the global optimal position; using the two global optimal positions as penalty parameters and kernel function parameters The IDBO-SVM fault diagnosis model is constructed based on the value of

[0168] Support vector machine is a machine learning model widely used in classification tasks. Its core idea is to find an optimal hyperplane in the feature space that can not only correctly classify all training samples, but also maximize the distance between positive and negative samples and the decision boundary.

[0169] The optimal hyperplane is given by the weight vector w and the bias vector Common Definitions:

[0170] (27);

[0171] In formula (27), is the classification result, For the input vector, for the two-class data point , For the input features, which are , For the The category label corresponding to the input feature is .

[0172] For linearly separable datasets, the basic optimization problem (hard margin case) can be formulated as:

[0173] (28);

[0174] In formula (28), is the number of samples.

[0175] For nonlinearly separable data sets, SVM uses the kernel trick to map the data points in the original feature space to a high-dimensional space. In this high-dimensional space, the data may become linearly separable, thereby finding an effective separation hyperplane. The kernel function K The definition is as follows:

[0176] (29);

[0177] In formula (29), For the input features; It is a mapping function, whose main function is to map data from low dimension to high dimension and perform calculations in high-dimensional space.

[0178] In practical applications, data is often not completely linearly separable. Therefore, the concept of soft interval is introduced to allow some sample points to slightly violate the interval condition. The optimization problem then becomes:

[0179] (30);

[0180] In formula (30), w is the weight vector; is the bias vector; It is a slack variable that allows sample points to slightly violate the interval condition; C is the penalty parameter, which is a regularization parameter that controls the degree of penalty; is the number of samples; For the slack variables; For the The category labels corresponding to the input features; For the input features.

[0181] Step 5: Based on the obtained multimodal data set, the IDBO-SVM fault diagnosis model is used to diagnose the motor bearing fault and determine the fault type of the motor bearing.

[0182] The fault type label is numerically encoded, the fault features (i.e., the features after fusion dimensionality reduction) are aligned with the corresponding fault type label, the training set and the test set are divided into a ratio of 7:3, the IDBO-SVM fault diagnosis model is trained using the training set, and the test set is input into the trained IDBO-SVM fault diagnosis model to predict the fault type of the motor bearing. The obtained diagnosis results are compared with the actual classification results, the performance indicators of the fault diagnosis model are calculated, and the fault diagnosis model is evaluated, such as Figure 8 and Figure 9 As shown, Figure 8 The 10 fault types and corresponding labels are: pitting inner ring, pitting outer ring, engraving machine inner ring, engraving machine outer ring, indentation outer ring, drilling outer ring, pitting inner ring + outer ring, indentation inner ring + outer ring, engraving machine inner ring + outer ring, drilling inner ring + outer ring, and the labels are 1, 2, 3, 4, 5, 6, 7, 8, 9, and 10 respectively.

[0183] At this time, in this embodiment, the fused dimensionality reduction features are used as input, and the fault diagnosis model is used to determine whether the motor bearing has a fault and the fault type if a fault exists.

[0184] Among them, the fault diagnosis model is support vector machine.

[0185] The hyperparameters of the support vector machine are determined by the improved dung beetle optimization algorithm. The improved dung beetle optimization algorithm is an algorithm obtained by improving the dung beetle optimization algorithm. The improvements include: using Logistic chaotic mapping to initialize the population, that is, formula (21); using the golden sine strategy to update the position update formula of the dung beetle rolling behavior, and obtaining the first update formula, that is, formula (22); using the multi-objective selection strategy to update the position update formula of the dung beetle breeding behavior, and obtaining the second update formula, that is, formula (23); using the adaptive step size and differential evolution strategy to update the position update formula of the dung beetle foraging behavior, and obtaining the third update formula, that is, formula (25); using the dynamic weight strategy to update the position update formula of the dung beetle stealing behavior, and obtaining the fourth update formula, that is, formula (26).

[0186] The hyperparameters of the support vector machine include penalty parameters and kernel function parameters.

[0187] This embodiment can effectively fuse multi-source heterogeneous data, overcome the problem of incomplete representation of single modal feature information, and significantly improve the discrimination and robustness of fault features. A dual-index comprehensive evaluation method based on reconstruction accuracy and modal independence is used to dynamically determine the number of VMD decomposition layers, which solves the limitation of the traditional method that the number of decomposition layers depends on empirical settings and avoids feature distortion caused by over-decomposition or under-decomposition. By improving the DBO algorithm through Logistic chaos mapping, golden sine strategy, multi-objective selection strategy, adaptive step size, differential evolution strategy and dynamic weight strategy, it effectively balances global search and local development capabilities, solves the problem that traditional optimization algorithms are prone to falling into local optimality and slow convergence, and makes the hyperparameters of SVM adaptively match the distribution characteristics of fault data, significantly improving the accuracy of the fault diagnosis model.

[0188] This embodiment discloses a motor bearing fault diagnosis method based on an improved DBO-SVM, comprising: collecting vibration and current signals from the motor bearing; decomposing the original vibration and current signals using VMD to obtain a certain number of intrinsic mode function components and calculate energy characteristics; calculating the typical characteristics of the original vibration and current signals in the time-frequency domain; performing dimensionality reduction and fusion of the extracted features through CCA and PCA to form a multimodal dataset; improving the DBO algorithm to enhance robustness and search efficiency; optimizing the parameters of the SVM using the improved DBO algorithm; and inputting the multimodal dataset into the improved DBO-SVM model to diagnose the motor bearing fault type. This embodiment improves diagnostic accuracy and provides a new solution for motor bearing fault diagnosis.

[0189] Example 2.

[0190] This embodiment provides a motor bearing fault diagnosis device, which includes a vibration sensor, a current sensor, and a processor.

[0191] The vibration sensor is installed on the motor bearing and is used to collect the vibration signal of the motor bearing.

[0192] The current sensor is installed on the motor bearing and is used to collect the current signal of the motor bearing.

[0193] The processor is communicatively connected to the vibration sensor and the current sensor respectively, and the processor is used to execute the motor bearing fault diagnosis method described in Example 1.

[0194] Example 3.

[0195] In an exemplary embodiment, a computer device is provided. The computer device may be a server or a terminal. The internal structure diagram thereof may be as follows: Figure 10As shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, abbreviated as I / O) and a communication interface. The processor, memory and input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and computer program in the non-volatile storage medium. The database of the computer device is used to store data. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a motor bearing fault diagnosis method is implemented.

[0196] Those skilled in the art will understand that Figure 10 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than shown in the figure, or combine certain components, or have a different component arrangement.

[0197] In an exemplary embodiment, a computer device is provided, including a memory and a processor. The memory stores a computer program, and the processor implements the motor bearing fault diagnosis method in Example 1 when executing the computer program.

[0198] Example 4.

[0199] In an exemplary embodiment, a computer-readable storage medium is provided, storing a computer program. When the computer program is executed by a processor, the motor bearing fault diagnosis method in embodiment 1 is implemented.

[0200] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0201] The technical features of the above embodiments can be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0202] This document uses specific examples to illustrate the principles and implementation methods of this application. The description of the above examples is only intended to help understand the method and core concept of this application. At the same time, for those skilled in the art, based on the concept of this application, there may be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting this application.

Claims

1. A motor bearing fault diagnosis method, characterized in that: The motor bearing fault diagnosis method comprises: Obtain vibration signals and current signals of motor bearings; performing feature extraction on the vibration signal and the current signal respectively to obtain a first time domain feature, a first frequency domain feature, and a first energy feature of the vibration signal, and a second time domain feature, a second frequency domain feature, and a second energy feature of the current signal; the first energy feature and the second energy feature are both energy entropy; Performing feature fusion and feature dimensionality reduction on the first time domain feature, the first frequency domain feature, the first energy feature, the second time domain feature, the second frequency domain feature, and the second energy feature to obtain a fused and reduced dimensionality feature; The fused dimensionality-reduced features are used as input, and a fault diagnosis model is used to determine whether a motor bearing has a fault and the fault type when a fault exists. The fault diagnosis model is a support vector machine, and the hyperparameters of the support vector machine are determined by an improved dung beetle optimization algorithm. The improved dung beetle optimization algorithm is an algorithm obtained by improving the dung beetle optimization algorithm. The improvement includes: using Logistic chaotic mapping to initialize the population, using the golden sine strategy to update the position update formula of the dung beetle rolling behavior to obtain a first update formula, using a multi-objective selection strategy to update the position update formula of the dung beetle breeding behavior to obtain a second update formula, using an adaptive step size and differential evolution strategy to update the position update formula of the dung beetle foraging behavior to obtain a third update formula, and using a dynamic weight strategy to update the position update formula of the dung beetle stealing behavior to obtain a fourth update formula.

2. The motor bearing fault diagnosis method according to claim 1, characterized in that: The first time domain feature and the second time domain feature both include: mean, kurtosis, effective value and peak factor; The first frequency domain feature and the second frequency domain feature both include: spectrum main frequency, harmonic energy ratio, power spectrum density and total harmonic distortion; The method for determining the first energy feature includes: performing variational modal decomposition on the vibration signal to obtain a plurality of first intrinsic mode function components, calculating the energy of each of the first intrinsic mode function components, and obtaining a first energy feature based on the energy calculation of all the first intrinsic mode function components; The method for determining the second energy characteristic includes: performing variational modal decomposition on the current signal to obtain multiple second intrinsic mode function components, calculating the energy of each of the second intrinsic mode function components, and obtaining the second energy characteristic based on the energy calculation of all the second intrinsic mode function components.

3. The motor bearing fault diagnosis method according to claim 2, characterized in that: performing variational modal decomposition on the vibration signal to obtain a plurality of first intrinsic mode function components, specifically comprising: determining a first decomposition level number used in performing variational modal decomposition on the vibration signal using a dual-index comprehensive evaluation method based on reconstruction accuracy and modal independence, and performing variational modal decomposition on the vibration signal according to the first decomposition level number to obtain the plurality of first intrinsic mode function components; The current signal is subjected to variational modal decomposition to obtain a plurality of second intrinsic mode function components, specifically comprising: determining the second decomposition layer number used when performing variational modal decomposition on the current signal using a dual-index comprehensive evaluation method based on reconstruction accuracy and modal independence, and performing variational modal decomposition on the current signal according to the second decomposition layer number to obtain a plurality of second intrinsic mode function components.

4. The motor bearing fault diagnosis method according to claim 3, characterized in that: A dual-index comprehensive evaluation method based on reconstruction accuracy and modal independence is used to determine the first decomposition level used when performing variational modal decomposition on the vibration signal, specifically comprising: designing multiple first initial decomposition levels; for each of the first initial decomposition levels, performing variational modal decomposition on the vibration signal according to the first initial decomposition level to obtain multiple first initial intrinsic mode function components, calculating the sum of all the first initial intrinsic mode function components to obtain a reconstructed vibration signal, calculating a first root mean square error under the first initial decomposition level based on the vibration signal and the reconstructed vibration signal, calculating a first average cross-correlation coefficient under the first initial decomposition level based on all the first initial intrinsic mode function components, and calculating a first comprehensive score under the first initial decomposition level based on the first root mean square error and the first average cross-correlation coefficient; selecting the first initial decomposition level with the smallest first comprehensive score as the first decomposition level used when performing variational modal decomposition on the vibration signal; A dual-index comprehensive evaluation method based on reconstruction accuracy and modal independence is used to determine the second decomposition layer number used when performing variational modal decomposition on the current signal, specifically including: designing multiple second initial decomposition layers; for each second initial decomposition layer number, performing variational modal decomposition on the current signal according to the second initial decomposition layer number to obtain multiple second initial intrinsic modal function components, calculating the sum of all the second initial intrinsic modal function components to obtain a reconstructed current signal, calculating the second root mean square error under the second initial decomposition layer number based on the current signal and the reconstructed current signal, calculating the second average mutual correlation coefficient under the second initial decomposition layer number based on all the second initial intrinsic modal function components, and calculating the second comprehensive score under the second initial decomposition layer number based on the second root mean square error and the second average mutual correlation coefficient; selecting the second initial decomposition layer number with the smallest second comprehensive score as the second decomposition layer number used when performing variational modal decomposition on the current signal.

5. The motor bearing fault diagnosis method according to claim 1, characterized in that: Performing feature fusion and feature dimensionality reduction on the first time domain feature, the first frequency domain feature, the first energy feature, the second time domain feature, the second frequency domain feature, and the second energy feature to obtain a fused and reduced dimensionality feature, specifically including: Performing feature fusion on the first time domain feature, the first frequency domain feature, the second time domain feature, and the second frequency domain feature using a canonical correlation analysis algorithm to obtain a first fused feature; Performing feature fusion on the first energy feature and the second energy feature using a canonical correlation analysis algorithm to obtain a second fused feature; A principal component analysis algorithm is used to perform feature dimensionality reduction on the first fused features and the second fused features to obtain fused dimensionality reduction features.

6. The motor bearing fault diagnosis method according to claim 1, characterized in that: The hyperparameters of the support vector machine include penalty parameters and kernel function parameters.

7. A motor bearing fault diagnosis device, characterized in that: The motor bearing fault diagnosis device includes: a vibration sensor, a current sensor and a processor; The vibration sensor is installed on the motor bearing, and the vibration sensor is used to collect the vibration signal of the motor bearing; The current sensor is installed on the motor bearing, and is used to collect the current signal of the motor bearing; The processor is communicatively connected to the vibration sensor and the current sensor respectively, and the processor is used to execute the motor bearing fault diagnosis method according to any one of claims 1 to 6.

8. A computer device comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the motor bearing fault diagnosis method according to any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the motor bearing fault diagnosis method according to any one of claims 1 to 6 is implemented.

Images