Vehicle driving motor gearbox fault diagnosis method and device

Abstract

The invention provides a vehicle driving motor gearbox fault diagnosis method and device, and the method comprises the steps: obtaining a sample data set and a pre-constructed original diagnosis model, and carrying out the model training, and obtaining a trained motor gearbox fault diagnosis model; acquiring a driving motor transmission signal of the to-be-diagnosed automobile; performing multi-granularity scanning on the driving motor transmission signal through a multi-granularity scanning model to obtain a first feature vector; performing feature extraction on the first feature vector through a PCA feature extraction model to obtain a second feature vector; splicing the first feature vector and the second feature vector to obtain a spliced vector; and processing the splicing vector through a deep forest model to obtain a diagnosis result. According to the method and the device, the fault of the gearbox of the vehicle driving motor can be automatically diagnosed, so that the dependence on prior knowledge is reduced, the fault diagnosis process is more universal and interpretable, and meanwhile, the accuracy and the efficiency of fault diagnosis can be improved.

Description

Technical Field

[0001] The present application relates to the technical field of electric vehicles, and in particular to a method and device for diagnosing a gearbox fault of a vehicle drive motor. Background Art

[0002] With the rapid development of the electric vehicle industry and the intensification of market competition, users have increasing requirements for vehicle safety. The gearbox of the drive motor system is a key component of electric vehicles, and its reliability is crucial to the stable operation of electric vehicles. However, under complex working conditions, especially in high temperature, high pressure and high speed environments, the gearbox is susceptible to a variety of complex alternating loads and is prone to failure, resulting in high maintenance costs and potential safety accidents. Therefore, efficient and accurate motor gearbox fault diagnosis technology has become the key to ensuring the safe operation of electric vehicles. Traditional manual time domain and frequency domain analysis methods usually require the use of tools such as waveform diagrams to manually observe and analyze to determine whether there is an abnormality in the transmission signal. In practice, it is found that in the existing methods, feature extraction is limited and it is difficult to fully reflect the fault characteristics of the gearbox. At the same time, testers are required to have a lot of prior knowledge and professional background, and are easily affected by subjectivity, which consumes manpower and material resources and reduces the accuracy and efficiency of fault diagnosis. Summary of the invention

[0003] The purpose of the embodiments of the present application is to provide a vehicle drive motor gearbox fault diagnosis method and device, which can diagnose vehicle drive motor gearbox faults without the need for human intervention, reduce dependence on prior knowledge, make the fault diagnosis process more universal and explainable, reduce manpower and material resources, and thus improve the accuracy and efficiency of fault diagnosis.

[0004] The first aspect of the present application provides a vehicle drive motor gearbox fault diagnosis method, comprising: Obtain a sample data set and a pre-built original diagnostic model; wherein the original diagnostic model includes a multi-granularity scanning model, a PCA feature extraction model and a deep forest model, the deep forest model includes a cascade forest and a random forest layer, and the PCA feature extraction model is connected to each level of the cascade forest; The original diagnostic model is trained by using the sample data set to obtain a trained motor gearbox fault diagnosis model; Acquire the drive motor transmission signal of the vehicle to be diagnosed; Performing multi-granularity scanning on the drive motor transmission signal by using the multi-granularity scanning model to obtain a first eigenvector; Performing feature extraction on the first feature vector by using the PCA feature extraction model to obtain a second feature vector; Concatenate the first feature vector and the second feature vector to obtain a concatenated vector; The splicing vector is processed by the deep forest model to obtain a diagnosis result.

[0005] Based on the above implementation process, this method can diagnose vehicle drive motor gearbox faults without the need for human intervention, reducing dependence on prior knowledge, making the fault diagnosis process more universal and explainable, reducing manpower and material resources, and thus improving the accuracy and efficiency of fault diagnosis.

[0006] Furthermore, the obtaining of the sample data set includes: Acquire the original transmission signal of the drive motor gearbox; wherein the original transmission signal includes the transmission category signal and the transmission load signal collected by the transmission laboratory; Performing multi-dimensional feature extraction on the original transmission signal to obtain a multi-dimensional feature set; A sample data set is constructed according to the multidimensional feature set.

[0007] Furthermore, the multi-dimensional feature extraction is performed on the original transmission signal to obtain a multi-dimensional feature set, including: Calculating permutation entropy according to the original transmission signal; Calculating time domain parameters and frequency domain parameters according to the original transmission signal; Performing EMD decomposition on the original transmission signal to obtain an IMF component signal; Calculate IMF component permutation entropy and IMF component energy according to the IMF component signal; Performing wavelet packet decomposition on the original transmission signal to obtain a decomposed signal; Calculate the reconstruction coefficient energy entropy of the third layer node of the wavelet packet according to the decomposed signal; The permutation entropy, the time domain parameter, the frequency domain parameter, the IMF component permutation entropy, the IMF component energy and the reconstruction coefficient energy entropy are summarized to obtain a multidimensional feature set.

[0008] Furthermore, the multi-granularity scanning model is used to perform multi-granularity scanning on the drive motor transmission signal to obtain a first eigenvector, including: Performing multi-granularity scanning on the drive motor transmission signal through the multi-granularity scanning model to obtain an initial feature vector; wherein the initial feature vector at least includes a time scale feature, a granularity feature, and a fault mode feature; Performing feature transformation on the initial feature vector by using a random forest algorithm to obtain a transformed feature vector; The transformed feature vectors are concatenated to obtain a first feature vector.

[0009] Further, the extracting features from the first feature vector by the PCA feature extraction model to obtain a second feature vector includes: Normalizing the first feature vector to obtain a normalized feature vector; Calculating the average value of each normalized feature vector according to the PCA feature extraction model; Calculating a centralized data set matrix according to the average value and the normalized eigenvector; Calculate the covariance matrix of the data set matrix according to the PCA feature extraction model; Calculating the eigenvalues ​​of the covariance matrix and the eigenvectors corresponding to the eigenvalues; Determine an eigenvector matrix according to the eigenvalues ​​and the eigenvectors corresponding to the eigenvalues; The normalized eigenvector is converted into the space constructed by the eigenvector matrix to obtain a second eigenvector.

[0010] Furthermore, the original diagnostic model is trained by the sample data set to obtain a trained motor gearbox fault diagnosis model, including: Dividing the sample data set into a training data set, a test data set and a validation data set; The original diagnostic model is trained using the training data set to obtain a target model; Performing diagnostic performance verification on the target model using the verification data set to obtain a verification result; Performing a performance test on the target model using the test data set to obtain a test result; Determining the model diagnosis accuracy according to the verification result and the test result; When the model diagnosis accuracy is greater than a preset accuracy threshold, the target model is determined as a motor gearbox fault diagnosis model.

[0011] Furthermore, the training of the original diagnostic model by the training data set to obtain a target model includes: Performing multi-granularity scanning on the training data set by using the multi-granularity scanning model to obtain a third feature vector; Performing feature extraction on the third eigenvector using the PCA feature extraction model to obtain a fourth eigenvector; Concatenate the third eigenvector and the fourth eigenvector to obtain a concatenated vector; Performing K-fold cross validation on the deep forest model according to the splicing vector to obtain a validation result; When it is determined according to the verification result that the deep forest model has not reached a preset convergence threshold, reconstructing the cascade forest; When it is determined according to the verification result that the deep forest model reaches the preset convergence threshold, obtaining a target deep forest model; The multi-granularity scanning model, the PCA feature extraction model and the target deep forest model are aggregated to obtain a target model.

[0012] A second aspect of the present application provides a vehicle drive motor gearbox fault diagnosis device, the vehicle drive motor gearbox fault diagnosis device comprising: A first acquisition unit, used to acquire a sample data set; A second acquisition unit is used for a pre-built original diagnostic model; wherein the original diagnostic model includes a multi-granularity scanning model, a PCA feature extraction model and a deep forest model, the deep forest model includes a cascade forest and a random forest layer, and the PCA feature extraction model is connected to each level of the cascade forest; A training unit, used for training the original diagnosis model through the sample data set to obtain a trained motor gearbox fault diagnosis model; A third acquisition unit is used to acquire a drive motor transmission signal of the vehicle to be diagnosed; A multi-granularity scanning unit, used for performing multi-granularity scanning on the drive motor transmission signal through the multi-granularity scanning model to obtain a first eigenvector; A feature extraction unit, configured to extract features from the first feature vector using the PCA feature extraction model to obtain a second feature vector; A concatenation unit, configured to concatenate the first feature vector and the second feature vector to obtain a concatenated vector; A processing unit is used to process the splicing vector through the deep forest model to obtain a diagnosis result.

[0013] Furthermore, the first acquiring unit includes: An acquisition subunit is used to acquire an original transmission signal of a drive motor gearbox; wherein the original transmission signal includes a transmission category signal and a transmission load signal collected by a transmission laboratory; An extraction subunit, used for performing multi-dimensional feature extraction on the original transmission signal to obtain a multi-dimensional feature set; The construction subunit is used to construct a sample data set according to the multidimensional feature set.

[0014] Furthermore, the extraction subunit includes: A calculation module, used for calculating the permutation entropy according to the original transmission signal; The calculation module is further used to calculate time domain parameters and frequency domain parameters according to the original transmission signal; A decomposition module, used for performing EMD decomposition on the original transmission signal to obtain an IMF component signal; The calculation module is further used to calculate the IMF component permutation entropy and the IMF component energy according to the IMF component signal; The decomposition module is further used to perform wavelet packet decomposition on the original transmission signal to obtain a decomposed signal; The calculation module is further used to calculate the reconstruction coefficient energy entropy of the third layer node of the wavelet packet according to the decomposed signal; The first summarizing module is used to summarize the permutation entropy, the time domain parameter, the frequency domain parameter, the IMF component permutation entropy, the IMF component energy and the reconstruction coefficient energy entropy to obtain a multidimensional feature set.

[0015] Furthermore, the multi-granularity scanning unit comprises: A scanning subunit, configured to perform multi-granularity scanning on the drive motor transmission signal through the multi-granularity scanning model to obtain an initial feature vector; wherein the initial feature vector at least includes a time scale feature, a granularity feature, and a fault mode feature; A transformation subunit, used for performing feature transformation on the initial feature vector by using a random forest algorithm to obtain a transformed feature vector; The concatenation subunit is used to concatenate the transformed feature vectors to obtain a first feature vector.

[0016] Furthermore, the feature extraction unit comprises: A processing subunit, configured to perform normalization processing on the first feature vector to obtain a normalized feature vector; A calculation subunit, used for calculating the average value of each normalized feature vector according to the PCA feature extraction model; The calculation subunit is further used to calculate a centralized data set matrix according to the average value and the normalized eigenvector; The calculation subunit is further used to calculate the covariance matrix of the data set matrix according to the PCA feature extraction model; The calculation subunit is further used to calculate the eigenvalues ​​of the covariance matrix and the eigenvectors corresponding to the eigenvalues; A first determining subunit, configured to determine an eigenvector matrix according to the eigenvalues ​​and the eigenvectors corresponding to the eigenvalues; The conversion subunit is used to convert the normalized eigenvector into the space constructed by the eigenvector matrix to obtain a second eigenvector.

[0017] Furthermore, the training unit comprises: A division subunit, used to divide the sample data set into a training data set, a test data set and a validation data set; A training subunit, used for training the original diagnostic model using the training data set to obtain a target model; A verification subunit, used to verify the diagnostic performance of the target model using the verification data set to obtain a verification result; A testing subunit, used to perform a performance test on the target model using the test data set to obtain a test result; A second determination subunit, configured to determine a model diagnosis accuracy according to the verification result and the test result; The second determination subunit is further configured to determine the target model as a motor gearbox fault diagnosis model when the model diagnosis accuracy is greater than a preset accuracy threshold.

[0018] Furthermore, the training subunit includes: A scanning module, used for performing multi-granularity scanning on the training data set through the multi-granularity scanning model to obtain a third feature vector; An extraction module, used for performing feature extraction on the third eigenvector through the PCA feature extraction model to obtain a fourth eigenvector; A splicing module, used for splicing the third eigenvector and the fourth eigenvector to obtain a splicing vector; A verification module, used to perform K-fold cross-validation on the deep forest model according to the splicing vector to obtain a verification result; A construction module, configured to reconstruct the cascade forest when it is determined according to the verification result that the deep forest model has not reached a preset convergence threshold; An acquisition module, configured to acquire a target deep forest model when it is determined according to the verification result that the deep forest model reaches the preset convergence threshold; The second aggregation module is used to aggregate the multi-granularity scanning model, the PCA feature extraction model and the target deep forest model to obtain a target model.

[0019] A third aspect of the present application provides an electronic device, including a memory and a processor, wherein the memory is used to store a computer program, and the processor runs the computer program to enable the electronic device to execute the vehicle drive motor gearbox fault diagnosis method described in any one of the first aspect of the present application.

[0020] A fourth aspect of the present application provides a computer-readable storage medium storing computer program instructions, which, when read and executed by a processor, executes the vehicle drive motor gearbox fault diagnosis method described in any one of the first aspect of the present application.

[0021] The beneficial effects of the present application are as follows: the method can optimize the algorithm based on the characteristics of electric vehicles as edge computing devices, making the fault diagnosis process more efficient while ensuring accuracy; then, the method can combine multiple feature extraction methods and optimization algorithms to reduce dependence on prior knowledge, making the fault diagnosis process more universal and explainable; On this basis, this method can make the feature extraction more comprehensive by comprehensively considering multiple feature extraction methods such as permutation entropy theory, wavelet packet decomposition technology and EMD decomposition method, thereby more accurately describing the vibration characteristics of the gearbox, which in turn helps to deeply understand its working mechanism and can effectively solve the problem of various mixed faults of the gearbox; at the same time, this multiple feature extraction method can also more effectively utilize existing data, reduce the cost of data acquisition, and improve the efficiency of fault diagnosis. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the embodiments of the present application will be briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present application and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.

[0023] Figure 1 A schematic diagram of a process flow of a vehicle drive motor gearbox fault diagnosis method provided in an embodiment of the present application; Figure 2 A schematic flow chart of another method for diagnosing a fault in a gearbox of a vehicle drive motor provided in an embodiment of the present application; Figure 3 A schematic diagram of a workflow of a deep forest model provided in an embodiment of the present application; Figure 4 A schematic diagram of a training process of a motor gearbox fault diagnosis model provided in an embodiment of the present application; Figure 5 A schematic diagram of the structure of a vehicle drive motor gearbox fault diagnosis device provided in an embodiment of the present application; Figure 6 A schematic structural diagram of another vehicle drive motor gearbox fault diagnosis device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0024] The technical solutions in the embodiments of the present application will be described below in conjunction with the drawings in the embodiments of the present application.

[0025] It should be noted that similar reference numerals and letters represent similar items in the following drawings, so once an item is defined in one drawing, it does not need to be further defined and explained in subsequent drawings. At the same time, in the description of this application, the terms "first", "second", etc. are only used to distinguish the description and cannot be understood as indicating or implying relative importance.

[0026] Example 1 Please see Figure 1 , Figure 1 The present invention provides a flow chart of a method for diagnosing a fault in a gearbox of a vehicle drive motor. The method for diagnosing a fault in a gearbox of a vehicle drive motor includes: S101. Obtain a sample data set and a pre-built original diagnostic model.

[0027] In this embodiment, the original diagnosis model includes a multi-granularity scanning model, a PCA feature extraction model and a deep forest model. The deep forest model includes a cascade forest and a random forest layer. The PCA feature extraction model is connected to each level of the cascade forest.

[0028] S102: Train the original diagnosis model using the sample data set to obtain a trained motor gearbox fault diagnosis model.

[0029] S103: Acquire a drive motor transmission signal of the vehicle to be diagnosed.

[0030] S104 , performing multi-granularity scanning on the drive motor transmission signal through a multi-granularity scanning model to obtain a first eigenvector.

[0031] S105 . Perform feature extraction on the first eigenvector using a PCA feature extraction model to obtain a second eigenvector.

[0032] S106: Concatenate the first eigenvector and the second eigenvector to obtain a concatenated vector.

[0033] S107. Process the concatenated vectors through a deep forest model to obtain a diagnosis result.

[0034] In this embodiment, the method proposes a fault diagnosis method based on a multi-dimensional feature extraction method such as permutation entropy, wavelet packet reconstruction energy entropy and empirical mode decomposition (EMD) energy entropy, and an improved deep forest model based on principal component analysis feature extraction for the multiple fault states of the motor gearbox. Specifically, the method can use a multi-dimensional feature extraction method combined with an improved deep forest model to improve the data transmission and processing capabilities in multi-granularity scanning and cascade forests, while ensuring data diversity, enhancing the representativeness of features within the model, thereby improving the algorithm's operating efficiency and diagnostic performance. In addition, the method does not need to be designed separately for different electric vehicle motor gearbox operating conditions, thereby greatly reducing the computing power requirements and diagnostic costs at the edge of the electric vehicle.

[0035] In this embodiment, the execution subject of the method may be a computing device such as a computer or a server, and no limitation is made in this embodiment.

[0036] In this embodiment, the execution subject of the method may also be a smart device such as a smart phone, a tablet computer, etc., which is not limited in this embodiment.

[0037] It can be seen that the vehicle drive motor gearbox fault diagnosis method described in this embodiment can diagnose the vehicle drive motor gearbox fault without human intervention, reduce dependence on prior knowledge, make the fault diagnosis process more universal and explainable, reduce manpower and material resources, and thus improve the accuracy and efficiency of fault diagnosis.

[0038] Example 2 Please see Figure 2 , Figure 2 The present invention provides a flow chart of a method for diagnosing a fault in a gearbox of a vehicle drive motor. The method for diagnosing a fault in a gearbox of a vehicle drive motor includes: S201, obtaining an original transmission signal of a drive motor gearbox.

[0039] In this embodiment, the original transmission signal includes a transmission category signal and a transmission load signal collected by a transmission laboratory.

[0040] S202. Calculate the permutation entropy according to the original transmission signal.

[0041] S203. Calculate time domain parameters and frequency domain parameters according to the original transmission signal.

[0042] In this embodiment, the characteristic expression of the time domain parameter includes: ; ; ; ; ; Among them, the original signal x(i) is the original signal, N represents the number of samples of signal data or spectrum data, and in the characteristic expression of time domain parameters, N is the total number of sampling points of the time domain signal.

[0043] Among them, TF1 is the mean, which indicates the average value of the signal data, reflecting the DC component or overall offset of the signal. TF2 is the root mean square (RMS), which is used to measure the energy of the signal and is particularly suitable for describing the strength of vibration signals. TF3 is the variance, which reflects the discrete degree of the signal amplitude and is often used to describe signal volatility. TF4 is the crest factor, which measures the relationship between the signal peak and energy and is often used to detect abnormal spikes in the signal. TF5 is the impulse factor, which measures the ratio of the maximum value of the signal to the average value and can highlight the impulse component in the signal.

[0044] In this embodiment, the characteristic expression of the frequency domain parameter includes: ; ; ; ; ; Among them, FF1 is the centroid frequency, which indicates the main frequency distribution position of the signal and can be used to characterize the main frequency components of the signal. FF2 is the frequency variance, which is used to reflect the discreteness of the signal spectrum distribution. FF3 is the square frequency, which is used to reflect the energy concentration of the signal spectrum. FF4 is the spectral standard deviation, which is used to measure the concentration of the spectrum energy distribution. The more dispersed the frequency is, the larger the value is. FF5 is the spectral peak factor, which indicates the peak ratio of the energy distribution in the spectrum.

[0045] Wherein, N represents the number of samples of signal data or spectrum data. In the characteristic expression of frequency domain parameters, N is the total number of frequency components after the signal is transformed into the frequency domain.

[0046] Among them, f i Represents the i-th frequency component in the frequency domain. It is the frequency obtained by Fourier transform, usually at equally spaced frequency points, such as: f1, f2, ..., f N .

[0047] Among them, F(f i ) represents the spectral amplitude (or power spectral density) corresponding to the i-th frequency component. In simple terms: it describes the frequency f i The energy size on the frequency can be understood as the strength of the signal at that frequency; in, It means N times FF4.

[0048] S204, performing EMD decomposition on the original transmission signal to obtain an IMF component signal.

[0049] S205 . Calculate IMF component permutation entropy and IMF component energy according to the IMF component signal.

[0050] S206, performing wavelet packet decomposition on the original transmission signal to obtain a decomposed signal.

[0051] S207. Calculate the reconstruction coefficient energy entropy of the third layer node of the wavelet packet according to the decomposed signal.

[0052] In this embodiment, the characteristic expression of the reconstruction coefficient energy entropy of the third layer node of the wavelet packet includes: S208 , summarizing the permutation entropy, time domain parameters, frequency domain parameters, IMF component permutation entropy, IMF component energy, and reconstruction coefficient energy entropy to obtain a multidimensional feature set.

[0053] In this embodiment, the electric vehicle motor gearbox often contains mixed faults of more than two types of parts, such as mixed faults of bearing inner ring and gear tooth breakage, gear eccentricity and output shaft imbalance, etc. In engineering applications, the vibration signal collected by the acceleration sensor shows strong non-stationarity and nonlinearity, so the fault feature extraction of the motor gearbox will directly affect the accuracy of the later fault mode classification. In addition, changes in the speed, load and diameter of the fault fault of the motor gearbox will also affect the feature extraction of the vibration signal. Therefore, in order to comprehensively consider the fault feature attributes of the bearing and reflect the inherent differences in the fault mode, this method combines time domain, frequency domain, permutation entropy algorithm, wavelet packet decomposition and EMD decomposition technology to extract high-dimensional fault feature space from multiple angles.

[0054] S209: Construct a sample data set according to the multidimensional feature set.

[0055] S210: Obtain a pre-built original diagnosis model.

[0056] S211. Train the original diagnosis model using the sample data set to obtain a trained motor gearbox fault diagnosis model.

[0057] As an optional implementation, the original diagnosis model is trained by using a sample data set to obtain a trained motor gearbox fault diagnosis model, including: Divide the sample data set into training data set, test data set and validation data set; The original diagnostic model is trained using the training data set to obtain a target model; The diagnostic performance of the target model is verified through the verification data set to obtain the verification results; Perform performance test on the target model through the test data set and obtain the test results; Determine the model diagnosis accuracy based on the verification and test results; When the model diagnosis accuracy is greater than a preset accuracy threshold, the target model is determined as the motor gearbox fault diagnosis model.

[0058] In this embodiment, the application of multi-granularity scanning technology can improve the feature diversity of the model, so that it can more comprehensively capture the information hidden in the sample. Among them, this technology uses multiple windows of different sizes to perform multi-scale sampling on the sample, generating a series of feature vectors of different scales, thereby greatly enriching the dimension and depth of the sample information.

[0059] In this embodiment, the method can first convert the characteristic signal obtained in the previous step into a sample data set containing training data and test data. Then, the training data is input into a multi-granularity scanning model. The core part of this model is to perform multi-window scanning on the characteristic signal with n-dimensional fault characteristics. Through this scanning mechanism, the fault mode under different time scales and granularities can be captured, providing a strong basis for subsequent fault identification and classification. During the scanning process, the method uses a random forest algorithm for feature transformation (the random forest can effectively extract more representative and discriminative new features from the original features with its powerful nonlinear feature learning ability). Then, after the random forest transformation, a series of transformed feature vectors can be obtained. Finally, these transformed feature vectors are spliced ​​to form a high-dimensional transformed feature vector. It can be seen that this high-dimensional feature vector not only contains rich information in the original signal, but also enhances the diversity and expression ability of the features through multi-granularity scanning and random forest transformation.

[0060] In this embodiment, the method designs feature windows of sizes d / 16, d / 8, and d / 4 for scanning data with d-dimensional features. Such a design not only ensures the diversity of the generated samples, but also maximizes the extraction and utilization of useful information in the data while ensuring computational efficiency. Based on this, the method can improve model performance while effectively reducing computational costs, thereby providing more convenience and flexibility for practical applications.

[0061] As a further optional implementation, the original diagnostic model is trained by a training data set to obtain a target model, including: Perform multi-granularity scanning on the training data set through a multi-granularity scanning model to obtain a third eigenvector; Perform feature extraction on the third eigenvector using a PCA feature extraction model to obtain a fourth eigenvector; Concatenate the third eigenvector and the fourth eigenvector to obtain a concatenated vector; Perform K-fold cross validation on the deep forest model based on the splicing vector to obtain the validation results; When the deep forest model fails to reach the preset convergence threshold according to the verification results, the cascade forest is rebuilt; When the deep forest model reaches a preset convergence threshold according to the verification result, the target deep forest model is obtained; The multi-granularity scanning model, PCA feature extraction model and target deep forest model are summarized to obtain the target model.

[0062] In this embodiment, the fault diagnosis of the electric vehicle motor gearbox is a key classification problem, and the cascade forest model can provide normal and fault probability prediction. Based on this, the method proposes a method to integrate the PCA model with each level of the cascade forest. This deep forest model based on PCA feature extraction can effectively combine the advantages of random forest (RF) and completely random forest (ERF). On the other hand, PCA has been widely used in various mechanical fault diagnosis studies due to its excellent feature extraction ability, and is often used in conjunction with non-statistical fault diagnosis data-driven methods such as support vector machines and neural networks. Based on this, PCA can effectively solve two major technical problems in gearbox fault diagnosis: one is to reduce the number of fault features in the fault feature pattern, and the other is to overcome the adverse effects of multicollinearity of fault features.

[0063] Please see Figure 3 , Figure 3 A schematic diagram of the workflow of a deep forest model is shown.

[0064] In this embodiment, PCA (Principal Component Analysis) is used to extract the main features of the data. These features extracted by PCA will be spliced ​​with the transformed feature vectors of the original multi-granularity scan output to form a richer feature set. Then, this spliced ​​feature set will be input into the first level of the cascade forest for decision making. Each level of the cascade forest is further learned based on the output results of the previous level. In particular, the output results of each level not only consider the decision results of the previous level, but also splice again with the transformed feature vectors of each granularity extracted by PCA features. This splicing operation ensures that each level can make full use of feature information from different granularities and different transformation methods, thereby improving the expressiveness and generalization ability of the model. Through this cascade method, the cascade forest can gradually deepen the understanding of the data and extract useful information from multiple angles and granularities. The decision of each level is based on the comprehensive consideration of the output of the previous level and the original features, which enables the entire model to show higher accuracy and robustness when dealing with complex problems. In summary, the role of PCA in this step is to extract the main features of the data, which will be spliced ​​with the original features and input into the cascade forest for decision making. The cascade forest makes full use of feature information of different granularities and different transformation methods through step-by-step splicing and decision-making, thereby improving the performance of the model.

[0065] By implementing this implementation method, the method can effectively solve the problems of feature redundancy and low algorithm operation efficiency faced by deep forests when processing single sample data with long characteristics by cascading PCA and cascade forests and extracting corresponding features for each level of cascade forest input. At the same time, this integration strategy can give full play to the advantages of deep forests and make electric vehicle motor gearbox fault diagnosis more accurate and efficient.

[0066] In this embodiment, during the growth of the cascade forest, the construction of each level is accompanied by rigorous k-fold cross-validation to accurately evaluate its diagnostic performance. If the performance fails to reach the expected convergence state, the model will automatically backtrack to the multi-granularity scanning step and restart the construction process of the new cascade forest. This process will be repeated until the diagnostic performance reaches a stable convergence state, and the growth of the cascade forest is determined to be complete. Subsequently, the fully trained model will meet the challenge of test data, make accurate classification predictions, and output detailed diagnostic results.

[0067] Please see Figure 4 , Figure 4A training process diagram of a motor gearbox fault diagnosis model is shown, wherein the method can first perform normalization preprocessing, wavelet packet decomposition and EMD decomposition based on the data collected by the transmission laboratory according to the transmission system category and load, and then input the signal into the multi-granularity scanning model. At this time, the multi-granularity scanning model performs multi-window scanning on the original time series with n-dimensional fault characteristics, and performs feature transformation through random forest, and finally splices the transformed feature vectors to obtain a high-dimensional transformed feature vector. Then, the high-dimensional transformed feature vector is reduced in dimension by principal component analysis (PCA), so that the reduced-dimensional features are spliced ​​with the original transformed feature vectors and input into the first level of the cascade forest for decision-making, and the output results of each level of the cascade forest are spliced ​​with the granularity transformed feature vectors extracted by PCA features, and input into the cascade forest of the next level, and so on, until the diagnostic performance converges and the cascade forest stops growing, and the motor gearbox fault diagnosis model is obtained.

[0068] S212: Acquire a drive motor transmission signal of the vehicle to be diagnosed.

[0069] S213, performing multi-granularity scanning on the drive motor transmission signal through a multi-granularity scanning model to obtain an initial feature vector.

[0070] In this embodiment, the initial feature vector includes at least a time scale feature, a granularity feature, and a fault mode feature.

[0071] S214. Perform feature transformation on the initial feature vector using a random forest algorithm to obtain a transformed feature vector.

[0072] S215. Concatenate the transformed feature vectors to obtain a first feature vector.

[0073] S216: Normalize the first eigenvector to obtain a normalized eigenvector.

[0074] In this embodiment, in order to ensure the consistency and comparability of data, the method adopts data normalization technology to standardize each element in the feature vector so that it conforms to a specific statistical distribution.

[0075] S217. Calculate the average value of each normalized feature vector according to the PCA feature extraction model.

[0076] In this embodiment, in order to extract the most representative principal components, the method uses the principal component analysis (PCA) method and determines the optimal number of principal components through an optimization strategy.

[0077] Among them, PCA aims to reduce the dimension of the original data from n dimensions to k p The key to achieving this goal is to find k pvectors, which are the principal components, are used to project the original data to minimize the projection error.

[0078] Assume that the original feature set is , the original feature set contains m samples and n-dimensional features, Each row of the matrix X represents a sample (or feature vector), for example, j1 , x j2 , …, x jn , represents the eigenvalue of the jth sample. Each column of the matrix X represents a feature (attribute), representing features of different dimensions, for example, x 1i , x 2i , …, x mi , which represents the value of the i-th feature in all samples.

[0079] Among them, M represents other values ​​in the matrix, that is, the eigenvalues ​​of the middle part, which refers to the elements not explicitly listed in the matrix and is used to omit writing all the features. O represents the value of the center part of the matrix. Similar to M, it is a placeholder used to represent the unexpanded part of the center of the matrix. L represents the omitted part, which usually represents the boundary of the matrix, or the part that is not fully listed in the rows and columns. For example: L is used to represent the unexpanded part on the right and bottom of the matrix.

[0080] Among them, x 11 , x 12 , …, x 1n , represents all the eigenvalues ​​of the first sample. x m1 , x m2 , …, x mn , represents all the eigenvalues ​​of the mth sample. m represents the number of samples, and n represents the feature dimension. Each element x ji represents the i-th eigenvalue of the j-th sample.

[0081] At this point, first centralize the data set, that is, find the average value of each feature separately: ; Among them, μ i represents the average value of the i-th eigenvalue, m represents the number of samples, x ji is x in the original feature set X ji represents the i-th eigenvalue of the j-th sample.

[0082] S218. Calculate the centralized data set matrix according to the average value and the normalized eigenvector.

[0083] In this embodiment, the corresponding mean is subtracted from all data to obtain the centralized data set matrix X0: .

[0084] Among them, μ1, μ2, …μ i , μ n , represents the average value of the i-th eigenvalue, m represents the number of samples, x 11 , x 12 , …, x 1n , represents all the eigenvalues ​​of the first sample. x m1 , x m2 , …, x mn , represents all the eigenvalues ​​of the mth sample.

[0085] S219. Calculate the covariance matrix of the data set matrix according to the PCA feature extraction model.

[0086] In this embodiment, the covariance matrix X of the centralized data set matrix X0 is calculated. COV : .

[0087] Among them, X0 represents the data set matrix, n represents the number of samples in the data set matrix X0, is the transposed matrix of the data set matrix X0.

[0088] S220, calculating the eigenvalues ​​of the covariance matrix and the eigenvectors corresponding to the eigenvalues.

[0089] S221. Determine an eigenvector matrix according to the eigenvalues ​​and the eigenvectors corresponding to the eigenvalues.

[0090] In this embodiment, the method can use singular value decomposition (SVD) to find the eigenvalues ​​of the covariance matrix and their corresponding eigenvectors. Then sort the eigenvalues ​​from large to small and select the largest k p Then the corresponding k p The eigenvectors are used as row vectors to form an eigenvector matrix.

[0091] S222. Convert the normalized eigenvector into a space constructed by the eigenvector matrix to obtain a second eigenvector.

[0092] In this embodiment, the method can convert the original feature set into k p In the new space constructed by the eigenvectors, we get k p The dataset X of dimension features new (Second eigenvector): ; Among them, X e is the matrix composed of the principal components (eigenvectors) obtained by PCA, and X is the original feature set.

[0093] This conversion process retains the main information in the data while significantly reducing the dimension of the data, which helps the data analysis process and model training process.

[0094] S223. Concatenate the first eigenvector and the second eigenvector to obtain a concatenated vector.

[0095] S224. Process the concatenated vector through a deep forest model to obtain a diagnosis result.

[0096] In this embodiment, in order to effectively solve the problem of feature masking and redundancy that the original deep forest may encounter when processing high-dimensional transformed feature vectors, the deep forest model proposed by this method can cleverly deeply integrate PCA with each level of the cascade forest, and perform fine feature extraction on the original transformed feature vector input at each level, thereby solving the problem of feature masking and redundancy, and at the same time can greatly improve the diagnostic accuracy of the deep forest.

[0097] In this embodiment, PCA's excellent performance in feature extraction enables it to accurately screen and refine high-dimensional transformation feature vectors. This not only eliminates the adverse effects of multicollinearity that may exist in the original fault features on the random forest, but also further optimizes the performance of the deep forest model itself. Through this improvement, the method can successfully mitigate the negative impact of multicollinearity on gearbox fault diagnosis, thereby significantly improving the accuracy and reliability of gearbox fault diagnosis.

[0098] In this embodiment, the execution subject of the method may be a computing device such as a computer or a server, and no limitation is made in this embodiment.

[0099] In this embodiment, the execution subject of the method may also be a smart device such as a smart phone, a tablet computer, etc., which is not limited in this embodiment.

[0100] It can be seen that the vehicle drive motor gearbox fault diagnosis method described in this embodiment can be used for fault feature extraction of electric vehicle motor gearbox vibration signals by combining permutation entropy theory, wavelet packet decomposition technology and EMD decomposition method on the basis of considering the two traditional feature extraction methods in time domain and frequency domain, and then calculating the permutation entropy value and energy entropy value of the reconstruction coefficient energy entropy of wavelet packet decomposition and the IMF component of EMD decomposition, thereby combining a 31-dimensional high-dimensional feature space. At the same time, the method can also integrate PCA feature extraction with the hierarchical part of the cascade forest to solve the problems of feature redundancy, low operating efficiency and multicollinearity in the original deep forest in gearbox fault diagnosis, and provide a more efficient and accurate method for gearbox fault diagnosis. Finally, the method can effectively reduce the feature dimension and reduce the risk of overfitting through PCA feature extraction technology, so as to maintain good diagnostic performance under small sample conditions. At the same time, as an integrated learning method based on decision trees, deep forest is insensitive to the distribution of data and the correlation between features. Therefore, it is more robust when dealing with small sample problems. It can reduce the operating cost of the entire diagnostic program while ensuring classification accuracy, which is conducive to edge computing on electric vehicles.

[0101] Example 3 Please see Figure 5 , Figure 5 This is a schematic diagram of the structure of a vehicle drive motor gearbox fault diagnosis device provided in this embodiment. Figure 5 As shown, the vehicle drive motor gearbox fault diagnosis device includes: A first acquisition unit 310 is used to acquire a sample data set; A second acquisition unit 320, for pre-constructing an original diagnostic model; wherein the original diagnostic model includes a multi-granularity scanning model, a PCA feature extraction model and a deep forest model, the deep forest model includes a cascade forest and a random forest layer, and the PCA feature extraction model is connected to each level of the cascade forest; A training unit 330 is used to train the original diagnosis model using a sample data set to obtain a trained motor gearbox fault diagnosis model; The third acquisition unit 340 is used to acquire a drive motor transmission signal of the vehicle to be diagnosed; A multi-granularity scanning unit 350 is used to perform multi-granularity scanning on the drive motor transmission signal through a multi-granularity scanning model to obtain a first eigenvector; A feature extraction unit 360 is used to extract features from the first feature vector using a PCA feature extraction model to obtain a second feature vector; A concatenation unit 370, configured to concatenate the first feature vector and the second feature vector to obtain a concatenated vector; The processing unit 380 is used to process the splicing vector through a deep forest model to obtain a diagnosis result.

[0102] In this embodiment, the explanation of the vehicle drive motor gearbox fault diagnosis device can refer to the description in Example 1 or Example 2, and will not be further elaborated in this embodiment.

[0103] It can be seen that the vehicle drive motor gearbox fault diagnosis device described in this embodiment can diagnose vehicle drive motor gearbox faults without the need for human intervention, reducing dependence on prior knowledge, making the fault diagnosis process more universal and explainable, reducing manpower and material resources, and thus improving the accuracy and efficiency of fault diagnosis.

[0104] Example 4 Please see Figure 6 , Figure 6 This is a schematic diagram of the structure of a vehicle drive motor gearbox fault diagnosis device provided in this embodiment. Figure 6 As shown, the vehicle drive motor gearbox fault diagnosis device includes: A first acquisition unit 310 is used to acquire a sample data set; A second acquisition unit 320, for pre-constructing an original diagnostic model; wherein the original diagnostic model includes a multi-granularity scanning model, a PCA feature extraction model and a deep forest model, the deep forest model includes a cascade forest and a random forest layer, and the PCA feature extraction model is connected to each level of the cascade forest; A training unit 330 is used to train the original diagnosis model using a sample data set to obtain a trained motor gearbox fault diagnosis model; The third acquisition unit 340 is used to acquire a drive motor transmission signal of the vehicle to be diagnosed; A multi-granularity scanning unit 350 is used to perform multi-granularity scanning on the drive motor transmission signal through a multi-granularity scanning model to obtain a first eigenvector; A feature extraction unit 360 is used to extract features from the first feature vector using a PCA feature extraction model to obtain a second feature vector; A concatenation unit 370, configured to concatenate the first feature vector and the second feature vector to obtain a concatenated vector; The processing unit 380 is used to process the splicing vector through a deep forest model to obtain a diagnosis result.

[0105] As an optional implementation manner, the first acquiring unit 310 includes: The acquisition subunit 311 is used to acquire the original transmission signal of the drive motor gearbox; wherein the original transmission signal includes the transmission category signal and the transmission load signal collected by the transmission laboratory; The extraction subunit 312 is used to extract multi-dimensional features from the original transmission signal to obtain a multi-dimensional feature set; The construction subunit 313 is used to construct a sample data set according to the multi-dimensional feature set.

[0106] As an optional implementation, the extraction subunit 312 includes: A calculation module, used for calculating the permutation entropy according to the original transmission signal; The calculation module is also used to calculate time domain parameters and frequency domain parameters according to the original transmission signal; A decomposition module is used to perform EMD decomposition on the original transmission signal to obtain an IMF component signal; The calculation module is further used to calculate the IMF component permutation entropy and the IMF component energy according to the IMF component signal; The decomposition module is also used to perform wavelet packet decomposition on the original transmission signal to obtain a decomposed signal; The calculation module is also used to calculate the reconstruction coefficient energy entropy of the third layer node of the wavelet packet according to the decomposed signal; The first summarization module is used to summarize the permutation entropy, time domain parameters, frequency domain parameters, IMF component permutation entropy, IMF component energy and reconstruction coefficient energy entropy to obtain a multi-dimensional feature set.

[0107] As an optional implementation, the multi-granularity scanning unit 350 includes: The scanning subunit 351 is used to perform multi-granularity scanning on the drive motor transmission signal through a multi-granularity scanning model to obtain an initial feature vector; wherein the initial feature vector at least includes a time scale feature, a granularity feature, and a fault mode feature; A transformation subunit 352 is used to perform feature transformation on the initial feature vector by using a random forest algorithm to obtain a transformed feature vector; The concatenation subunit 353 is used to concatenate the transformed feature vectors to obtain a first feature vector.

[0108] As an optional implementation, the feature extraction unit 360 includes: The processing subunit 361 is used to perform normalization processing on the first feature vector to obtain a normalized feature vector; A calculation subunit 362, used for calculating the average value of each normalized feature vector according to the PCA feature extraction model; The calculation subunit 362 is further used to calculate a centered data set matrix according to the average value and the normalized eigenvector; The calculation subunit 362 is further used to calculate the covariance matrix of the data set matrix according to the PCA feature extraction model; The calculation subunit 362 is further used to calculate the eigenvalues ​​of the covariance matrix and the eigenvectors corresponding to the eigenvalues; A first determining subunit 363, configured to determine an eigenvector matrix according to the eigenvalues ​​and the eigenvectors corresponding to the eigenvalues; The conversion subunit 364 is used to convert the normalized eigenvector into the space constructed by the eigenvector matrix to obtain a second eigenvector.

[0109] As an optional implementation, the training unit 330 includes: A division subunit 331 is used to divide the sample data set into a training data set, a test data set and a validation data set; The training subunit 332 is used to train the original diagnosis model through the training data set to obtain a target model; The verification subunit 333 is used to verify the diagnostic performance of the target model through the verification data set to obtain a verification result; The testing subunit 334 is used to perform a performance test on the target model through a test data set to obtain a test result; A second determination subunit 335 is used to determine the model diagnosis accuracy according to the verification result and the test result; The second determination subunit 335 is further configured to determine the target model as a motor gearbox fault diagnosis model when the model diagnosis accuracy is greater than a preset accuracy threshold.

[0110] As an optional implementation, the training subunit 332 includes: A scanning module, used for performing multi-granularity scanning on the training data set through a multi-granularity scanning model to obtain a third eigenvector; An extraction module, used for performing feature extraction on the third eigenvector through a PCA feature extraction model to obtain a fourth eigenvector; A concatenation module, used for concatenating the third eigenvector and the fourth eigenvector to obtain a concatenated vector; The validation module is used to perform K-fold cross validation on the deep forest model according to the splicing vector to obtain the validation result; A construction module, used for reconstructing the cascade forest when it is determined according to the verification result that the deep forest model does not reach a preset convergence threshold; An acquisition module, used to acquire a target deep forest model when it is determined according to the verification result that the deep forest model reaches a preset convergence threshold; The second aggregation module is used to aggregate the multi-granularity scanning model, the PCA feature extraction model and the target deep forest model to obtain the target model.

[0111] In this embodiment, the explanation of the vehicle drive motor gearbox fault diagnosis device can refer to the description in Example 1 or Example 2, and will not be further elaborated in this embodiment.

[0112] It can be seen that the vehicle drive motor gearbox fault diagnosis device described in this embodiment can be used for fault feature extraction of electric vehicle motor gearbox vibration signals by combining permutation entropy theory, wavelet packet decomposition technology and EMD decomposition method on the basis of considering the two traditional feature extraction methods of time domain and frequency domain, and then calculate the permutation entropy value and energy entropy value of the reconstruction coefficient energy entropy of wavelet packet decomposition and the IMF component of EMD decomposition, thereby combining a 31-dimensional high-dimensional feature space. At the same time, the method can also integrate PCA feature extraction with the hierarchical part of the cascade forest to solve the problems of feature redundancy, low operating efficiency and multicollinearity in the original deep forest in gearbox fault diagnosis, and provide a more efficient and accurate method for gearbox fault diagnosis. Finally, the method can effectively reduce the feature dimension and reduce the risk of overfitting through PCA feature extraction technology, so as to maintain good diagnostic performance under small sample conditions. At the same time, as an integrated learning method based on decision trees, deep forest is insensitive to the distribution of data and the correlation between features. Therefore, it is more robust when dealing with small sample problems. It can reduce the operating cost of the entire diagnostic program while ensuring classification accuracy, which is conducive to edge computing on electric vehicles.

[0113] An embodiment of the present application provides an electronic device, including a memory and a processor, wherein the memory is used to store a computer program, and the processor runs the computer program to enable the electronic device to execute the vehicle drive motor gearbox fault diagnosis method in embodiment 1 or embodiment 2 of the present application.

[0114] An embodiment of the present application provides a computer-readable storage medium storing computer program instructions. When the computer program instructions are read and executed by a processor, the vehicle drive motor gearbox fault diagnosis method in embodiment 1 or embodiment 2 of the present application is executed.

[0115] In several embodiments provided in the present application, it should be understood that the disclosed devices and methods can also be implemented in other ways. The device embodiments described above are merely schematic. For example, the flowcharts and block diagrams in the accompanying drawings show the possible architecture, functions and operations of the devices, methods and computer program products according to multiple embodiments of the present application. In this regard, each box in the flowchart or block diagram can represent a module, a program segment or a part of a code, and the module, a program segment or a part of a code contains one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two consecutive boxes can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or flowchart, and the combination of boxes in the block diagram and / or flowchart can be implemented with a dedicated hardware-based system that performs a specified function or action, or can be implemented with a combination of dedicated hardware and computer instructions.

[0116] In addition, the functional modules in the various embodiments of the present application may be integrated together to form an independent part, or each module may exist separately, or two or more modules may be integrated to form an independent part.

[0117] If the function is implemented in the form of a software function module and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application can be essentially or partly embodied in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, and other media that can store program codes.

[0118] The above description is only an embodiment of the present application and is not intended to limit the scope of protection of the present application. For those skilled in the art, the present application may have various changes and variations. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application should be included in the scope of protection of the present application. It should be noted that similar reference numerals and letters represent similar items in the following drawings, so once an item is defined in one drawing, it does not need to be further defined and explained in the subsequent drawings.

[0119] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

[0120] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.

Claims

1. A method for diagnosing a fault in a gearbox of a vehicle drive motor, characterized in that: include: Obtain a sample data set and a pre-built original diagnostic model; wherein the original diagnostic model includes a multi-granularity scanning model, a PCA feature extraction model and a deep forest model, the deep forest model includes a cascade forest and a random forest layer, and the PCA feature extraction model is connected to each level of the cascade forest; The original diagnostic model is trained by using the sample data set to obtain a trained motor gearbox fault diagnosis model; Acquire the drive motor transmission signal of the vehicle to be diagnosed; Performing multi-granularity scanning on the drive motor transmission signal by using the multi-granularity scanning model to obtain a first eigenvector; Performing feature extraction on the first feature vector by using the PCA feature extraction model to obtain a second feature vector; Concatenate the first feature vector and the second feature vector to obtain a concatenated vector; The splicing vector is processed by the deep forest model to obtain a diagnosis result.

2. The vehicle drive motor gearbox fault diagnosis method according to claim 1, characterized in that: The obtaining of the sample data set comprises: Acquire the original transmission signal of the drive motor gearbox; wherein the original transmission signal includes the transmission category signal and the transmission load signal collected by the transmission laboratory; Performing multi-dimensional feature extraction on the original transmission signal to obtain a multi-dimensional feature set; A sample data set is constructed according to the multidimensional feature set.

3. The vehicle drive motor gearbox fault diagnosis method according to claim 2, characterized in that: The multi-dimensional feature extraction is performed on the original transmission signal to obtain a multi-dimensional feature set, including: Calculating permutation entropy according to the original transmission signal; Calculating time domain parameters and frequency domain parameters according to the original transmission signal; Performing EMD decomposition on the original transmission signal to obtain an IMF component signal; Calculate IMF component permutation entropy and IMF component energy according to the IMF component signal; Performing wavelet packet decomposition on the original transmission signal to obtain a decomposed signal; Calculate the reconstruction coefficient energy entropy of the third layer node of the wavelet packet according to the decomposed signal; The permutation entropy, the time domain parameter, the frequency domain parameter, the IMF component permutation entropy, the IMF component energy and the reconstruction coefficient energy entropy are summarized to obtain a multidimensional feature set.

4. The vehicle drive motor gearbox fault diagnosis method according to claim 1, characterized in that: The multi-granularity scanning model is used to perform multi-granularity scanning on the drive motor transmission signal to obtain a first eigenvector, including: Performing multi-granularity scanning on the drive motor transmission signal through the multi-granularity scanning model to obtain an initial feature vector; wherein the initial feature vector at least includes a time scale feature, a granularity feature, and a fault mode feature; Performing feature transformation on the initial feature vector by using a random forest algorithm to obtain a transformed feature vector; The transformed feature vectors are concatenated to obtain a first feature vector.

5. The vehicle drive motor gearbox fault diagnosis method according to claim 1, characterized in that: The step of extracting features from the first feature vector using the PCA feature extraction model to obtain a second feature vector includes: Normalizing the first feature vector to obtain a normalized feature vector; Calculating the average value of each normalized feature vector according to the PCA feature extraction model; Calculating a centralized data set matrix according to the average value and the normalized eigenvector; Calculate the covariance matrix of the data set matrix according to the PCA feature extraction model; Calculating the eigenvalues ​​of the covariance matrix and the eigenvectors corresponding to the eigenvalues; Determine an eigenvector matrix according to the eigenvalues ​​and the eigenvectors corresponding to the eigenvalues; The normalized eigenvector is converted into the space constructed by the eigenvector matrix to obtain a second eigenvector.

6. The vehicle drive motor gearbox fault diagnosis method according to claim 1, characterized in that: The original diagnosis model is trained by the sample data set to obtain a trained motor gearbox fault diagnosis model, including: Dividing the sample data set into a training data set, a test data set and a validation data set; The original diagnostic model is trained using the training data set to obtain a target model; Performing diagnostic performance verification on the target model using the verification data set to obtain a verification result; Performing a performance test on the target model using the test data set to obtain a test result; Determining the model diagnosis accuracy according to the verification result and the test result; When the model diagnosis accuracy is greater than a preset accuracy threshold, the target model is determined as a motor gearbox fault diagnosis model.

7. The vehicle drive motor gearbox fault diagnosis method according to claim 6, characterized in that: The step of training the original diagnostic model using the training data set to obtain a target model includes: Performing multi-granularity scanning on the training data set by using the multi-granularity scanning model to obtain a third feature vector; Performing feature extraction on the third eigenvector using the PCA feature extraction model to obtain a fourth eigenvector; Concatenate the third eigenvector and the fourth eigenvector to obtain a concatenated vector; Performing K-fold cross validation on the deep forest model according to the splicing vector to obtain a validation result; When it is determined according to the verification result that the deep forest model has not reached a preset convergence threshold, reconstructing the cascade forest; When it is determined according to the verification result that the deep forest model reaches the preset convergence threshold, obtaining a target deep forest model; The multi-granularity scanning model, the PCA feature extraction model and the target deep forest model are aggregated to obtain a target model.

8. A vehicle drive motor gearbox fault diagnosis device, characterized in that: The vehicle drive motor gearbox fault diagnosis device comprises: A first acquisition unit, used to acquire a sample data set; A second acquisition unit is used for a pre-built original diagnostic model; wherein the original diagnostic model includes a multi-granularity scanning model, a PCA feature extraction model and a deep forest model, the deep forest model includes a cascade forest and a random forest layer, and the PCA feature extraction model is connected to each level of the cascade forest; A training unit, used for training the original diagnosis model through the sample data set to obtain a trained motor gearbox fault diagnosis model; A third acquisition unit is used to acquire a drive motor transmission signal of the vehicle to be diagnosed; A multi-granularity scanning unit, used for performing multi-granularity scanning on the drive motor transmission signal through the multi-granularity scanning model to obtain a first eigenvector; A feature extraction unit, configured to extract features from the first feature vector using the PCA feature extraction model to obtain a second feature vector; A concatenation unit, configured to concatenate the first feature vector and the second feature vector to obtain a concatenated vector; A processing unit is used to process the splicing vector through the deep forest model to obtain a diagnosis result.

9. An electronic device, characterized in that: The electronic device comprises a memory and a processor, the memory is used to store a computer program, and the processor runs the computer program to enable the electronic device to execute the vehicle drive motor gearbox fault diagnosis method according to any one of claims 1 to 7.

10. A readable storage medium, characterized in that: The readable storage medium stores computer program instructions, and when the computer program instructions are read and executed by a processor, the vehicle drive motor gearbox fault diagnosis method according to any one of claims 1 to 7 is executed.

Images