K-means multi-working-condition clustering vehicle fault diagnosis method based on correlation analysis

Abstract

The invention discloses a k-means multi-working-condition clustering vehicle fault diagnosis method based on correlation analysis. The method comprises the following steps: obtaining vehicle offline working condition parameters; obtaining an optimal clustering center number of k-means clustering based on an elbow method and an average contour coefficient method; performing k-means clustering analysis on the vehicle offline working condition parameters to obtain threshold ranges of vehicle component signal parameters corresponding to different working conditions; and acquiring real-time vehicle part signal parameters, and comparing and analyzing the real-time vehicle part signal parameters with the threshold ranges of the vehicle part signal parameters corresponding to different working conditions to obtain a vehicle fault diagnosis result. According to the method, the defect that fault diagnosis of an artificial intelligence algorithm is poor in effect under small samples and variable working conditions at present can be overcome, the state of the vehicle part in the complex working condition environment can be diagnosed, and a new method is provided for fault diagnosis of the vehicle running in the complex working condition environment.

Description

Technical Field

[0001] The invention belongs to the technical field of vehicle fault diagnosis, and in particular relates to a k-means multi-operating condition clustering vehicle fault diagnosis method based on correlation analysis. Background Art

[0002] As the traffic environment becomes increasingly complex, vehicle fault diagnosis technology is essential to improving safety, extending service life and reducing maintenance costs. The current mainstream vehicle fault diagnosis methods are as follows:

[0003] Sensor signal diagnosis: Various sensors such as accelerometers, acoustic emission sensors, temperature sensors, etc. collect vibration signals, acoustic signals, temperature signals, etc. For example, Qiu et al. proposed a fault diagnosis method based on vibration signal adjustment spectrum, which is suitable for bearing fault diagnosis under different working conditions and limited sample conditions. Osman developed a new normalized Hilbert-Huang transform (NHHT) technology based on vibration signals for bearing fault detection and proved its robustness under different experimental settings.

[0004] Feature extraction and intelligent diagnosis: Traditional signal analysis and processing methods rely on manual feature extraction, require a lot of expert experience, and have great application limitations. With the application of neural networks, feature extraction has become intelligent, and the original vibration signal can be directly used as a model input to complete fault diagnosis. However, when dealing with variable operating condition bearing fault diagnosis, the simple structure and limited nonlinear computing power of traditional neural networks limit their application. The concept of deep learning came into being, which improves nonlinear fitting capabilities and high-dimensional data processing capabilities through more complex network structures and deep feature extraction.

[0005] Domain Adaptation and Transfer Learning: It is challenging to collect enough rolling bearing fault samples under different working conditions. Domain adaptation-based methods, such as one-dimensional multi-resolution deep domain adaptation neural network, achieve accurate diagnosis of rolling bearing faults under different working conditions by reducing the distribution difference between the source domain and the target domain, demonstrating its effectiveness in diagnosis under different working conditions.

[0006] Fault diagnosis based on classification: Common fault diagnosis methods based on classification, such as Bayesian network (BN), K nearest neighbor (KNN), support vector machine (SVM), etc., are used to identify different working conditions and fault types. Among them, SVM is widely used in fault diagnosis due to its good generalization performance and accuracy, especially the selection of its parameter optimization method is crucial to the model classification effect.

[0007] Existing diagnostic methods based on sensor equipment require a lot of capital and maintenance costs. Diagnostic methods based on artificial intelligence perform well in big data environments, but the diagnostic effect is poor in small sample and complex working conditions, which is not conducive to engineering implementation.

[0008] In view of the above problems, it is urgent to propose a k-means multi-operating condition clustering vehicle fault diagnosis method based on correlation analysis to improve the fault diagnosis effect of vehicle components under small samples and complex operating conditions. Summary of the invention

[0009] In order to solve the above technical problems, the present invention proposes a k-means multi-operating condition clustering vehicle fault diagnosis method based on correlation analysis to solve the problems existing in the above prior art.

[0010] To achieve the above object, the present invention provides a k-means multi-operating condition clustering vehicle fault diagnosis method based on correlation analysis, comprising the following steps:

[0011] Obtain vehicle offline operating parameters;

[0012] Obtain the optimal number of cluster centers for k-means clustering based on the elbow method and the average silhouette coefficient method;

[0013] Performing k-means cluster analysis on the vehicle offline operating condition parameters to obtain threshold ranges of vehicle component signal parameters corresponding to different operating conditions;

[0014] Real-time vehicle component signal parameters are obtained, and the real-time vehicle component signal parameters are compared and analyzed with threshold ranges of vehicle component signal parameters corresponding to different working conditions to obtain vehicle fault diagnosis results.

[0015] Optionally, before performing k-means cluster analysis on the vehicle offline operating condition parameters, the method includes:

[0016] A correlation analysis is performed on the vehicle offline operating condition parameters based on the Spearman correlation coefficient, and typical operating condition parameters are selected.

[0017] Optionally, the process of obtaining the optimal number of cluster centers for k-means clustering based on the elbow method and the average silhouette coefficient method includes:

[0018] The first optimal number of cluster centers of k-means clustering is obtained based on the elbow method, and the second optimal number of cluster centers of k-means clustering is obtained based on the average silhouette coefficient method. Based on engineering practice, a comparative analysis is conducted on the first optimal number of cluster centers and the second optimal number of cluster centers to determine the final optimal number of cluster centers of k-means clustering.

[0019] Optionally, the process of obtaining the first optimal number of cluster centers of k-means clustering based on the elbow method includes:

[0020] Based on the elbow method, the intra-group sum of squares under different numbers of cluster centers is obtained, and the corresponding curve graph is drawn. The first optimal number of cluster centers is determined based on the point where the curve graph begins to show a straight line appearance.

[0021] Optionally, the process of obtaining the second best number of cluster centers of k-means clustering based on the average silhouette coefficient method includes:

[0022] Based on the average silhouette coefficient method, the silhouette coefficients under different numbers of cluster centers are obtained, and the corresponding curve graph is drawn. By analyzing the silhouette coefficients in the curve graph, the second optimal number of cluster centers is determined.

[0023] Optionally, the process of performing k-means cluster analysis on the vehicle offline operating condition parameters to obtain threshold ranges of vehicle component signal parameters corresponding to different operating conditions includes:

[0024] A k-means clustering analysis is performed on the offline working condition parameters of the vehicle to determine the number of working conditions as the optimal number of cluster centers; after clustering is completed, the signal parameters of each component of the vehicle are analyzed to obtain the threshold range of the signal parameters of the vehicle components corresponding to different working conditions.

[0025] Optionally, the process of comparing and analyzing the real-time vehicle component signal parameter with the threshold range of the vehicle component signal parameter corresponding to different working conditions includes:

[0026] The distances between the real-time vehicle component signal parameters and different cluster centers are obtained and sorted, and the working conditions corresponding to the real-time vehicle component signal parameters are obtained based on the cluster center with the shortest distance to the real-time vehicle component signal parameters; when the real-time vehicle component signal parameters exceed the threshold range of the vehicle component signal parameters under the corresponding working conditions, the vehicle fault diagnosis results are obtained.

[0027] The present invention also provides an electronic device, comprising: a memory and a processor; the memory is used to store a program; the processor is used to execute the program to implement each step of the k-means multi-operating condition clustering vehicle fault diagnosis method based on correlation analysis.

[0028] The present invention also provides a readable storage medium on which a computer program is stored. When the computer program is executed by a processor, each step of the k-means multi-operating condition clustering vehicle fault diagnosis method based on correlation analysis is implemented.

[0029] Compared with the prior art, the present invention has the following advantages and technical effects:

[0030] The method proposed in the present invention first obtains the vehicle's offline operating condition parameters, performs k-means cluster analysis on the vehicle's offline operating condition parameters, and obtains the threshold range of the vehicle component signal parameters corresponding to different operating conditions; then obtains the real-time vehicle component signal parameters, compares and analyzes the real-time vehicle component signal parameters with the threshold range of the vehicle component signal parameters corresponding to different operating conditions, and obtains the vehicle fault diagnosis results. It can avoid the shortcomings of the current artificial intelligence algorithm fault diagnosis that is not effective under small samples and variable operating conditions, and can diagnose the status of vehicle components under complex operating conditions, providing a new method for fault diagnosis of vehicles traveling under complex operating conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] The drawings constituting a part of the present application are used to provide a further understanding of the present application. The illustrative embodiments and descriptions of the present application are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0032] Figure 1 A schematic diagram of a method flow of an embodiment of the present invention;

[0033] Figure 2 This is a schematic diagram of the thermal matrix of correlation coefficients of four operating parameters of an embodiment of the present invention;

[0034] Figure 3 A schematic diagram of a k-value curve determined by the elbow method according to an embodiment of the present invention;

[0035] Figure 4 A schematic diagram of a k-value curve determined by the average silhouette coefficient method according to an embodiment of the present invention;

[0036] Figure 5 Schematic diagram of clustering visualization results according to an embodiment of the present invention. DETAILED DESCRIPTION

[0037] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0038] It should be noted that the steps shown in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and that, although a logical order is shown in the flowcharts, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0039] Embodiment 1

[0040] like Figure 1 As shown, this embodiment provides a k-means multi-operating condition clustering vehicle fault diagnosis method based on correlation analysis, including the following steps:

[0041] Obtain vehicle offline operating parameters;

[0042] Obtain the optimal number of cluster centers for k-means clustering based on the elbow method and the average silhouette coefficient method;

[0043] Performing k-means cluster analysis on the vehicle offline operating condition parameters to obtain threshold ranges of vehicle component signal parameters corresponding to different operating conditions;

[0044] Real-time vehicle component signal parameters are obtained, and the real-time vehicle component signal parameters are compared and analyzed with threshold ranges of vehicle component signal parameters corresponding to different working conditions to obtain vehicle fault diagnosis results.

[0045] It is feasible that before performing k-means cluster analysis on the vehicle offline working condition parameters, the following steps are included: performing correlation analysis on the vehicle offline working condition parameters, and the Spearman correlation coefficient calculation formula is as follows:

[0046]

[0047] In the above formula, r i is a single time point operating parameter data. is the average value over a period of time; i For a single time point data of another working condition, is its average value in the same period of time; n is the number of time points in that period of time, and ρ is the Spearman correlation coefficient.

[0048] Calculate the correlation coefficient of each operating condition parameter, eliminate the operating condition parameters with large correlation coefficients, and select typical operating condition parameters according to the actual engineering environment requirements to solve the redundant effects caused by too many operating conditions.

[0049] The process of obtaining the optimal number of cluster centers (k value) of k-means clustering based on the elbow method and the average silhouette coefficient method can be implemented as follows:

[0050] Use the elbow method: by changing the size of k and calculating the within-group sum of squares (WCSS) under different k values, the optimal number of clusters k, that is, the number of the first best cluster centers, is determined by finding the point where the graph begins to appear as a straight line, that is, the point where the rate of WCSS reduction begins to slow down.

[0051] Silhouette coefficient method: The silhouette coefficient method is also used to confirm the optimal number of cluster centers (k value) of k-means clustering. By calculating the silhouette coefficient under different k values ​​(the silhouette coefficient range is [-1, 1]), if the silhouette coefficient of the current clustering is small, then the current clustering method is not suitable; if the silhouette coefficient is large, then the current clustering method is appropriate, and then the second optimal number of cluster centers is obtained.

[0052] By analyzing the first optimal number of cluster centers and the second optimal number of cluster centers, the selection of the k value in the subsequent k-means clustering is comprehensively determined.

[0053] It is feasible to perform k-means cluster analysis on the vehicle offline working condition parameters in Matlab to obtain the threshold range of the vehicle component signal parameters corresponding to different working conditions, including:

[0054] The vehicle offline operating parameters are directly clustered using the kmeans function in Matlab, where the k value is set to the k value selected based on the elbow method and the average silhouette coefficient method to obtain k clusters.

[0055] After clustering is completed, the signal data of each component of the vehicle is analyzed to obtain the threshold range of the component signal data corresponding to each cluster.

[0056] The above clusters are visualized in Matlab to obtain k cluster centers, and the threshold ranges of vehicle component signal parameters corresponding to different clusters are obtained (different clusters here refer to different working conditions), and the cluster centers and the obtained component signal thresholds are saved in the .mat folder.

[0057] The process of performing fault diagnosis on real-time vehicle signal data that can be implemented includes:

[0058] The distances between the real-time vehicle component signal parameters and different cluster centers are obtained and sorted, and the working conditions corresponding to the real-time vehicle component signal parameters are obtained based on the cluster center with the shortest distance to the real-time vehicle component signal parameters. Specifically, according to the real-time selected working condition data, the working condition at that moment is determined to be the working condition corresponding to the offline state according to which cluster the real-time selected working condition data falls into.

[0059] According to the threshold range of the vehicle component signal parameters under different working conditions, it is determined whether the vehicle component signal parameter under the working condition at that moment exceeds the corresponding offline threshold range. If exceeded, it is determined that the component signal data under the working condition at that moment is abnormal.

[0060] Embodiment 2

[0061] This embodiment performs fault diagnosis on real-time vehicle components based on a k-means clustering study of offline data of a certain vehicle operating condition.

[0062] Extract the four types of working condition data of a vehicle offline, including engine speed, accelerator pedal opening, vehicle speed, and gear position, and perform cross-correlation analysis on the four types of working condition data in Matlab. Figure 2This is the correlation coefficient matrix heat map of the four operating condition parameters. The correlation between the four operating condition parameters is relatively high. Considering the actual driver operation, the accelerator pedal and steering wheel angle signal parameters are selected for k-means operating condition clustering.

[0063] Since the working environment of this part of offline data is uncertain, the k value of k-means clustering is also uncertain. Therefore, the elbow method and average silhouette coefficient method are used to determine the final k value of k-means clustering:

[0064] Elbow method: The elbow method is used to determine the optimal number of cluster centers (k value) for k-means clustering. It determines the optimal number of clusters k by changing the size of k and calculating the within-group sum of squares (WCSS) under different k values. It determines the optimal number of clusters k by finding the point where the graph begins to appear as a straight line, that is, the point where the rate of reduction of WCSS begins to slow down.

[0065] Average silhouette coefficient method: The silhouette coefficient method is also used to confirm the optimal number of cluster centers (k value) of k-means clustering. By calculating the silhouette coefficient under different k values ​​(the silhouette coefficient range is [-1, 1]), if the silhouette coefficient of the current clustering is small, the current clustering method is not suitable; if the silhouette coefficient is large, the current clustering method is appropriate.

[0066] Figure 3 is the k value curve determined by the elbow method, Figure 4 The k value curve is determined by the average silhouette coefficient method. Combined with the actual project, it is more practical to finally determine that k is 9. Therefore, the number of working conditions for the offline data research of the vehicle is determined to be 9, and k=9 in the k-means clustering.

[0067] In Matlab, k-means clustering is performed on the offline accelerator pedal opening and steering wheel angle, k=9, the cluster centers are printed and the clustering results are visualized. Figure 5 To visualize the clustering results, Table 1 shows the specific cluster center points (i.e., different working conditions).

[0068] Table 1

[0069] cluster Cluster center value Working conditions Cluster1 (0.0913，0.1274) Condition 1 Cluster2 (36.7068，53.4096) Condition 2 Cluster3 (30.0004，3.4425) Condition 3 Cluster4 (25.8997，24.4463) Condition 4 Cluster5 (5.3595，24.1276) Condition 5 Cluster6 (7.7712，74.5556) Condition 6 Cluster7 (81.7836，11.0680) Condition 7 Cluster8 (8.0127，46.4623) Condition 8 Cluster9 (12.8437，2.3641) Condition 9

[0070] The range of vehicle component parameter signals corresponding to different working conditions is obtained through offline data analysis.

[0071] The following is a fault diagnosis of the real-time signal data:

[0072] First, the working condition is determined based on the steering wheel angle and accelerator pedal opening at a certain moment in real time. The working condition closest to the cluster center obtained offline is determined. Then, it is identified whether the signal data of the vehicle components under this working condition exceeds the threshold range obtained offline. If so, it is determined to be abnormal. Table 2 shows the recognition results of some real-time fault diagnosis.

[0073] Table 2

[0074]

[0075] Embodiment 3

[0076] This embodiment provides an electronic device, including: a memory and a processor; the memory is used to store a program; the processor is used to execute the program to implement each step of the k-means multi-operating condition clustering vehicle fault diagnosis method based on correlation analysis.

[0077] Embodiment 4

[0078] This embodiment provides a readable storage medium on which a computer program is stored. When the computer program is executed by a processor, each step of the k-means multi-operating condition clustering vehicle fault diagnosis method based on correlation analysis is implemented.

[0079] The above are only preferred specific implementations of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed in the present application 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.

Claims

1. A k-means multi-operating condition clustering vehicle fault diagnosis method based on correlation analysis, characterized in that: The following steps are involved: Obtain vehicle offline operating parameters; Obtain the optimal number of cluster centers for k-means clustering based on the elbow method and the average silhouette coefficient method; Performing k-means cluster analysis on the vehicle offline operating condition parameters to obtain threshold ranges of vehicle component signal parameters corresponding to different operating conditions; Real-time vehicle component signal parameters are obtained, and the real-time vehicle component signal parameters are compared and analyzed with threshold ranges of vehicle component signal parameters corresponding to different working conditions to obtain vehicle fault diagnosis results.

2. The method according to claim 1, characterized in that Before performing k-means cluster analysis on the vehicle offline operating condition parameters, the method includes: A correlation analysis is performed on the vehicle offline operating condition parameters based on the Spearman correlation coefficient, and typical operating condition parameters are selected.

3. The method according to claim 1, characterized in that The process of obtaining the optimal number of cluster centers for k-means clustering based on the elbow method and the average silhouette coefficient method includes: The first optimal number of cluster centers of k-means clustering is obtained based on the elbow method, and the second optimal number of cluster centers of k-means clustering is obtained based on the average silhouette coefficient method. Based on engineering practice, a comparative analysis is conducted on the first optimal number of cluster centers and the second optimal number of cluster centers to determine the final optimal number of cluster centers of k-means clustering.

4. The method according to claim 3, characterized in that The process of obtaining the first optimal number of cluster centers for k-means clustering based on the elbow method includes: Based on the elbow method, the intra-group sum of squares under different numbers of cluster centers is obtained, and the corresponding curve graph is drawn. The first optimal number of cluster centers is determined based on the point where the curve graph begins to show a straight line appearance.

5. The method according to claim 3, characterized in that: The process of obtaining the second best number of cluster centers for k-means clustering based on the average silhouette coefficient method includes: Based on the average silhouette coefficient method, the silhouette coefficients under different numbers of cluster centers are obtained, and the corresponding curve graph is drawn. By analyzing the silhouette coefficients in the curve graph, the second optimal number of cluster centers is determined.

6. The method according to claim 1, characterized in that The process of performing k-means cluster analysis on the vehicle offline operating condition parameters to obtain the threshold ranges of the vehicle component signal parameters corresponding to different operating conditions includes: A k-means clustering analysis is performed on the offline working condition parameters of the vehicle to determine the number of working conditions as the optimal number of cluster centers; after clustering is completed, the signal parameters of each component of the vehicle are analyzed to obtain the threshold range of the signal parameters of the vehicle components corresponding to different working conditions.

7. The method according to claim 6, characterized in that The process of comparing and analyzing the real-time vehicle component signal parameters with the threshold ranges of the vehicle component signal parameters corresponding to different working conditions includes: The distances between the real-time vehicle component signal parameters and different cluster centers are obtained and sorted, and the working conditions corresponding to the real-time vehicle component signal parameters are obtained based on the cluster center with the shortest distance to the real-time vehicle component signal parameters; when the real-time vehicle component signal parameters exceed the threshold range of the vehicle component signal parameters under the corresponding working conditions, the vehicle fault diagnosis results are obtained.

8. An electronic device, characterized in that: include: Memory and processor; The memory is used to store programs; The processor is used to execute the program to implement the various steps of the k-means multi-operating condition clustering vehicle fault diagnosis method based on correlation analysis as described in any one of claims 1 to 7.

9. A readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, each step of the k-means multi-operating condition clustering vehicle fault diagnosis method based on correlation analysis as described in any one of claims 1 to 7 is implemented.

Images