Railway high-voltage circuit breaker fault diagnosis method and device based on improved binary tree classification algorithm
By improving the combination of binary tree classification algorithm and support vector machine, the characteristic vector of the electric vibration acoustic signal of the railway high-voltage circuit breaker is extracted using principal component analysis method, and dimension reduction and classification are performed, which solves the problem of insufficient accuracy and reliability of high-voltage circuit breaker fault diagnosis, achieving more efficient and accurate fault diagnosis.
Patent Information
- Application Number
- CN202411679105.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-11-22
- Publication Date
- 2025-03-25
AI Technical Summary
The electric vibration and acoustic signals generated by railway high-voltage circuit breakers are disturbed by noise, resulting in insufficient accuracy and reliability of fault diagnosis. The fault diagnosis results of the prior art support vector machines are not accurate and have large calculation amounts.
The improved binary tree classification algorithm combined with the support vector machine is used to extract the eigenvector vectors of the electric vibration and acoustic signals through principal component analysis, and dimension unified and covariance matrix dimensionality reduction are carried out to obtain effective components. Mechanical fault diagnosis is performed based on the ratio of the inter-class distribution metric to the intra-class distribution metric and the confidence of the classifier.
It improves the accuracy and speed of fault diagnosis, reduces the number of required second-class classifiers, enhances the robustness of the system, and makes fault diagnosis more rapid and reliable.
Smart Images
Figure CN119689232A_ABST
Abstract
Description
Technical Field
[0001] The invention relates to the field of fault diagnosis of high-voltage circuit breakers, and in particular to a method and device for diagnosing faults of railway high-voltage circuit breakers using an improved binary tree classification algorithm. Background Art
[0002] The electric vibration and sound signals generated by high-voltage circuit breakers during operation are often interfered by various noises. These noises come from the operation of power equipment, environmental interference and the vibration of the equipment itself, which seriously affects the accurate judgment of the working state of the circuit breaker and the reliability of fault diagnosis. At present, a variety of state monitoring and fault diagnosis technologies have been proposed and applied, including methods based on fuzzy theory and radial basis function networks. However, railway high-voltage circuit breakers are mainly composed of actuators, transmissions, insulation supports, breaking mechanisms, etc. The structure is relatively complex, and the operating environment of high-voltage circuit breakers is changeable. Therefore, there are many types of faults caused by the external environment and its own mechanisms. The different properties of the signal characteristics obtained by these methods result in low accuracy of the support vector machine fault diagnosis results and a large amount of calculation. Summary of the invention
[0003] The purpose of the present invention is to overcome the shortcomings and deficiencies of the prior art and to provide a railway high-voltage circuit breaker fault diagnosis device with an improved binary tree classification algorithm, comprising a data feature extraction module, a data processing module and a support vector machine;
[0004] The data feature extraction module collects the feature vector sequence X of the electric vibration and acoustic signal of the railway high-voltage circuit breaker by principal component analysis;
[0005] The data processing module includes a data preprocessing module and a data dimension reduction module, wherein the data preprocessing module is connected to the data feature module and is used to perform dimension unification processing on the feature vector sequence X; the data dimension reduction module is connected to the data feature module and is used to obtain a covariance matrix according to the feature vector sequence X after dimension unification processing, and reduce the dimension of the covariance matrix to obtain effective components;
[0006] The support vector machines are respectively connected to the data dimension reduction modules. The support vector machines obtain the ratio of the inter-class distribution metric to the intra-class distribution metric and the credibility of the classifier as the classification basis according to the effective components and the improved binary tree classification algorithm, and obtain the mechanical failure result according to the ratio of the inter-class distribution metric to the intra-class distribution metric and the credibility of the classifier.
[0007] Preferably, the data dimension reduction module is used to obtain a covariance matrix according to the eigenvector sequence X after dimension unification, and reduce the dimension of the covariance matrix to obtain effective components, specifically comprising the following steps:
[0008] S31, performing decentralized processing on the eigenvector sequence X and obtaining a covariance matrix;
[0009] S32, by performing eigendecomposition on the covariance matrix and obtaining eigenvalues (λ1≥λ2≥…≥λ m ) and the corresponding eigenvector, select the eigenvector a where the first k eigenvalues are located i (i=1,2,…m), forming the projection matrix:
[0010] W=[a1,a2,…,a k ];
[0011] S33. Determine the number of principal components according to the size of the eigenvalue. The eigenvalue value is proportional to the contribution coefficient. The contribution coefficient K t Defined as:
[0012]
[0013] S34. Select the main components with cumulative contribution rate greater than 96% as effective components.
[0014] Preferably, the effective components include training samples and test samples, and the support vector machine obtains the ratio of the inter-class distribution metric to the intra-class distribution metric and the credibility of the classifier used as a classification basis according to the effective components and the improved binary tree classification algorithm, specifically including the following steps:
[0015] S41, the training sample is input into the support vector machine for learning, and the optimal parameters w*, a are obtained. i *, b*, and output model S i ;
[0016] S42, the training samples are input into a support vector machine to train a classifier {c1, c2, ..., c n} corresponding to the weights p1, p2, ..., p n (i=1,…,n), and the weight corresponding to the classifier is used as the credibility of the classifier;
[0017] Among them, the optimal parameters include w*, a i *, b*, w* are weights, a i * is the optimal Lagrangian factor, b* is the optimal threshold; the credibility interval is between 0 and 1, "1" represents complete credibility, that is, it belongs to this category; "0" represents complete unreliability, that is, it does not belong to this category. The closer the credibility is to 1, the higher the accuracy of the classification judgment. Output model S i as follows:
[0018] p(x)=p(y=1 / x)=1 / {1+exp[-f(x)]};
[0019] p(-x)=p(y=-1 / x)=exp[-f(x)] / {1+exp[-f(x)]}.
[0020] Preferably, the ratio of the inter-class distribution metric to the intra-class distribution metric divides the training samples into two categories, and the two categories of training samples respectively identify the training samples belonging to the i-th category according to the credibility of the classifier and mark them with a positive sign, otherwise they are marked with a negative sign.
[0021] Preferably, the test sample is input into a support vector machine to diagnose the mechanical fault category according to the i-th category training sample.
[0022] Preferably, the classifier {c1, c2, ..., c n} is a binary classifier constructed by the binary tree classification algorithm.
[0023] Preferably, the data processing module uses a mean-variance calculation method to perform dimension unification processing on the feature vector sequence X.
[0024] Preferably, it comprises a data storage module, the data storage module comprises a data resource management module, a data storage execution module and a data query service module, the data resource management module;
[0025] The data storage module is used to store the high-voltage circuit breaker data in a mixed manner to form initial data, wherein the high-voltage circuit breaker data includes structured data such as current and voltage and unstructured data such as mechanical data;
[0026] The data resource management module is used to virtualize and standardize the initial data according to the Hadoop cluster technology;
[0027] The data storage execution module is connected to the data resource management module and is used to perform unified dimension processing on the initial data that has been virtualized and standardized;
[0028] The data query service module automatically queries the initial data processed in the same dimension by using the traditional retrieval algorithm and the intelligent retrieval algorithm in parallel.
[0029] Preferably, it comprises a fault diagnosis module, the fault diagnosis query module comprises a signal acquisition module, a signal reading module and a diagnosis result display module;
[0030] The signal acquisition module is used to specify the acquisition mode, sampling frequency and number of channels, and is connected to the data resource management module;
[0031] The signal reading module is used to read the high-voltage circuit breaker data signal and input the high-voltage circuit breaker data signal into the data storage execution module;
[0032] The diagnosis result display module is connected to the data query service module and is used to display the mechanical failure results.
[0033] Preferably, a railway high-voltage circuit breaker fault diagnosis method using an improved binary tree classification algorithm comprises the railway high-voltage circuit breaker fault diagnosis method using an improved binary tree classification algorithm being used to adapt a railway high-voltage circuit breaker fault diagnosis device using an improved binary tree classification algorithm; the method comprises the following steps:
[0034] The data feature extraction module collects the feature vector sequence X of the electric vibration and acoustic signal of the railway high-voltage circuit breaker according to the principal component analysis method;
[0035] The feature vector sequence X is dimensionally unified through the data preprocessing module;
[0036] The data dimension reduction module is used to obtain the covariance matrix of the feature vector sequence X after dimension unification, and the dimension of the covariance matrix is reduced to obtain the effective components;
[0037] The support vector machine obtains the ratio of the inter-class distribution metric to the intra-class distribution metric and the credibility of the classifier as the classification basis according to the effective components and the improved binary tree classification algorithm, and obtains the mechanical failure result according to the ratio of the inter-class distribution metric to the intra-class distribution metric and the credibility of the classifier.
[0038] The beneficial effects of the present invention are as follows: 1. The data dimension reduction module performs dimension reduction processing on the acquired feature vector sequence X by principal component analysis to obtain effective components, thereby ensuring that effective information will not be lost during the dimension reduction process and enhancing the robustness of the system;
[0039] 2. By adopting the improved binary tree algorithm to consider the one-to-one relationship between samples and the credibility of the classifier during the classification process, the number of required two-class classifiers is reduced, the classification accuracy is improved, the diagnosis process is simplified, the classification speed is increased, and fault diagnosis is faster. BRIEF DESCRIPTION OF THE DRAWINGS
[0040] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, without paying creative labor, other drawings obtained based on these drawings still belong to the scope of the present invention.
[0041] Figure 1 is an operation flow chart of the support vector machine of the present invention;
[0042] Figure 2It is a flow chart of the improved classification algorithm of the present invention;
[0043] Figure 3 This is a schematic diagram of the improved binary tree structure of the present invention;
[0044] Figure 4 A block diagram of a data storage module of the present invention;
[0045] Figure 5 It is the user login interface of the fault diagnosis module of the present invention;
[0046] Figure 6 A circuit breaker fault diagnosis interface of the fault diagnosis module of the present invention;
[0047] Figure 7 is the main component contribution rate of the present invention;
[0048] Figure 8 It is a binary tree algorithm fault diagnosis result diagram of the present invention;
[0049] Fig. 9 This is a diagram of the fault diagnosis results of the improved binary tree algorithm of the present invention; DETAILED DESCRIPTION
[0050] In order to make the objectives, technical solutions and advantages of the present invention more clear, the present invention will be further described in detail below with reference to the accompanying drawings.
[0051] It should be noted that all expressions using "first" and "second" in the embodiments of the present invention are for distinguishing two non-identical entities with the same name or non-identical parameters. It can be seen that "first" and "second" are only for the convenience of expression and should not be understood as limitations on the embodiments of the present invention. The subsequent embodiments will not explain this one by one.
[0052] The directions and positions mentioned in the present invention, such as "upper", "lower", "front", "back", "left", "right", "inside", "outside", "top", "bottom", "side", etc., are only for reference to the directions or positions of the drawings. Therefore, the directions and positions used are for explaining and understanding the present invention, but not for limiting the scope of protection of the present invention.
[0053] like Figures 1 to 9 As shown, it is an embodiment of the present invention, which includes a data feature extraction module, a data processing module and a support vector machine;
[0054] The data feature extraction module collects the feature vector sequence X of the electric vibration and acoustic signal of the railway high-voltage circuit breaker by principal component analysis;
[0055] The data processing module includes a data preprocessing module and a data dimension reduction module, wherein the data preprocessing module is connected to the data feature module and is used to perform dimension unification processing on the feature vector sequence X; the data dimension reduction module is connected to the data feature module and is used to obtain a covariance matrix according to the feature vector sequence X after dimension unification processing, and reduce the dimension of the covariance matrix to obtain effective components;
[0056] The support vector machines are respectively connected to the data dimension reduction modules. The support vector machines obtain the ratio of the inter-class distribution metric to the intra-class distribution metric and the credibility of the classifier as the classification basis according to the effective components and the improved binary tree classification algorithm, and obtain the mechanical failure result according to the ratio of the inter-class distribution metric to the intra-class distribution metric and the credibility of the classifier.
[0057] It can be understood that the feature vector sequence X extracted by the data feature extraction module includes feature vectors under four mechanical states.
[0058] The data processing module uses the mean-variance calculation method to perform dimension unification processing on the feature vector X.
[0059] The data dimension reduction module is used to obtain a covariance matrix according to the eigenvector sequence X after dimension unification, and reduce the dimension of the covariance matrix to obtain effective components, which specifically includes the following steps:
[0060] S31, performing decentralized processing on the eigenvector sequence X and obtaining a covariance matrix;
[0061] S32, by performing eigendecomposition on the covariance matrix and obtaining eigenvalues (λ1≥λ2≥…≥λ m ) and the corresponding eigenvector, select the eigenvector a where the first k eigenvalues are located i (i=1,2,…m), forming the projection matrix:
[0062] W=[a1,a2,…,a k ];
[0063] S33. Determine the number of principal components according to the size of the eigenvalue. The eigenvalue value is proportional to the contribution coefficient. The contribution coefficient K t Defined as:
[0064]
[0065] S34. Select the main components with cumulative contribution rate greater than 96% as effective components.
[0066] It can be understood that in order to ensure that effective components are not lost after the principal component analysis method is optimized, we select the four best main components with a cumulative contribution rate greater than 96%, which are called principal component one, principal component two, principal component three, and principal component four respectively.
[0067] The effective components include training samples and test samples, and the support vector machine obtains the ratio of the inter-class distribution metric to the intra-class distribution metric and the credibility of the classifier used as a classification basis according to the effective components and the improved binary tree classification algorithm, specifically including the following steps:
[0068] S41, the training sample is input into the support vector machine for learning, and the optimal parameters w*, a are obtained. i *, b*, and output model S i ;
[0069] S42, the training samples are input into a support vector machine to train a classifier {c1, c2, ..., c n} corresponding to the weights p1, p2, ..., p n (i=1,…,n), and the weight corresponding to the classifier is used as the credibility of the classifier;
[0070] Among them, the optimal parameters include w*, a i *, b*, w* are weights, a i * is the optimal Lagrangian factor, a i * is the optimal threshold; the credibility interval is between 0 and 1, "1" represents complete credibility, that is, it belongs to this category; "0" represents complete untrustworthiness, that is, it does not belong to this category. The closer the credibility is to 1, the higher the accuracy of the classification judgment. Output model S i as follows:
[0071] p(x)=p(y=1 / x)=1 / {1+exp[-f(x)]};
[0072] p(-x)=p(y=-1 / x)=exp[-f(x)] / {1+exp[-f(x)]}.
[0073] The ratio of the inter-class distribution metric to the intra-class distribution metric divides the training samples into two categories. The two categories of training samples respectively identify the training samples belonging to the i-th category according to the credibility of the classifier and mark them with a positive sign, otherwise they are marked with a negative sign.
[0074] The test sample is input into the support vector machine to diagnose the mechanical fault category according to the training sample of the i-th category.
[0075] The classifier {c1, c2, ..., c n} is a binary classifier constructed by the binary tree classification algorithm.
[0076] It includes a data storage module, the data storage module includes a data resource management module, a data storage execution module and a data query service module, the data resource management module;
[0077] The data storage module is used to store the high-voltage circuit breaker data in a mixed manner to form initial data, wherein the high-voltage circuit breaker data includes structured data such as current and voltage and unstructured data such as mechanical data;
[0078] The data resource management module is used to virtualize and standardize the initial data according to the Hadoop cluster technology;
[0079] The data storage execution module is connected to the data resource management module and is used to perform the same dimension processing on the initial data after virtualization and standardization processing;
[0080] The data query service module automatically queries the initial data processed in the same dimension by using the traditional retrieval algorithm and the intelligent retrieval algorithm in parallel.
[0081] It includes a fault diagnosis module, the fault diagnosis query module includes a signal acquisition module, a signal reading module and a diagnosis result display module;
[0082] The signal acquisition module is used to specify the acquisition mode, sampling frequency and number of channels, and is connected to the data resource management module;
[0083] The signal reading module is used to read the high-voltage circuit breaker data signal and input the high-voltage circuit breaker data signal into the data storage execution module;
[0084] The diagnosis result display module is connected to the data query service module and is used to display the mechanical failure results.
[0085] A railway high-voltage circuit breaker fault diagnosis method using an improved binary tree classification algorithm, comprising: the railway high-voltage circuit breaker fault diagnosis method using an improved binary tree classification algorithm is used to adapt a railway high-voltage circuit breaker fault diagnosis device using an improved binary tree classification algorithm; the method comprises the following steps:
[0086] The data feature extraction module collects the feature vector sequence X of the electric vibration and acoustic signal of the railway high-voltage circuit breaker according to the principal component analysis method;
[0087] The feature vector sequence X is dimensionally unified through the data preprocessing module;
[0088] The data dimension reduction module is used to obtain the covariance matrix of the feature vector sequence X after dimension unification, and the dimension of the covariance matrix is reduced to obtain the effective components;
[0089] The support vector machine obtains the ratio of the inter-class distribution metric to the intra-class distribution metric and the credibility of the classifier as the classification basis according to the effective components and the improved binary tree classification algorithm, and obtains the mechanical failure result according to the ratio of the inter-class distribution metric to the intra-class distribution metric and the credibility of the classifier.
[0090] The following are some experimental data of selecting VS1 (VBM7)-12 type high voltage circuit breaker as the research object of the present invention:
[0091] 1. Data dimensionality reduction using principal component analysis. The electrical, vibration and acoustic signals of the circuit breaker are measured using sensors. The contribution rates after dimensionality reduction using principal component analysis are shown in Table 1.
[0092] Table 1 Individual contribution rate and cumulative contribution rate of principal components
[0093] Principal component number Individual contribution rate / % Cumulative contribution rate / % 1 79.80 79.80 2 9.58 89.38 3 4.55 93.93 4 2.48 96.41 5 1.80 98.21 6 1.00 99.21 7 0.37 99.58 8 0.20 99.78 9 0.09 99.87 10 0.08 99.95 11 0.02 99.97 12 0.02 99.99 13 0.01 100
[0094] like Figure 7 As shown in the figure, the individual contribution rates of the principal components are plotted as bar graphs, and the cumulative contribution rates are plotted as curve graphs, which can more intuitively show the changing patterns and characteristics.
[0095] In order to ensure that the effective components are not lost after the principal component analysis is optimized, this patent selects four optimal main components with a cumulative contribution rate greater than 96%, which are respectively called principal component 1, principal component 2, principal component 3, and principal component 4. The characteristic values of a small part of the principal components are shown in Table 2.
[0096] Table 2 Characteristic data of the first four principal components
[0097]
[0098] 3. Improve binary tree fault diagnosis
[0099] The features after dimension reduction by principal component analysis are input into the improved binary tree classification model for classification. There are 90 groups of samples in each state. 60 groups of samples in each state are extracted, totaling 240 groups as training samples, and the remaining 120 groups of samples are used as test samples. The normal state is set as category 1, the loose screw state is set as category 2, the electromagnet stuck state is set as category 3, and the spring aging state is set as category 4. The samples are input into the improved binary tree and original binary tree classification models for classification respectively. The comparison of classification results is shown in Table 3. It can be seen that the classification performance of binary tree and improved binary tree is better, and the training and testing time is shorter. The classification results of binary tree and improved binary tree are shown in the figure below. Figure 8 and Fig. 9 As shown, the number of classification errors after improvement is 1, and the number of classification errors of the original binary tree is 10. The improved binary tree is more superior.
[0100] Table 3 Comparison of classification results
[0101] Classifier Training time / s Test time / s Training classification accuracy / % Test classification accuracy / % 1-v-r 2.013 0.198 100 90.0 1-v-1 1.975 0.296 100 90.0 Binary Tree 1.538 0.235 100 91.7 Improved binary tree 1.602 0.227 100 99.2
[0102] The single coil current signal, vibration signal, sound signal, electric-vibration combined signal, and electric-vibration-sound combined signal feature vector are input into the improved binary tree classification model for fault diagnosis. The diagnosis results are shown in Tables 4, 5, 6, 7, and 8. It can be seen that the fault diagnosis rate of the electric-vibration-sound combined signal feature can reach 99.2%, and the fault diagnosis rate of the electric-vibration-sound combined signal feature can reach 91.7%. The fault diagnosis rates of single current, vibration, and sound signal features are all less than 90%. It can be proved that fault diagnosis using electric-vibration-sound combined features is better than fault diagnosis using single signal features, and inputting the electric-vibration-sound combined feature quantity into the improved binary tree classification model has excellent fault diagnosis performance.
[0103] Table 4 Electro-vibration-acoustic joint feature diagnosis results
[0104] Prediction results Normal state Loose screws Electromagnet stuck Spring aging Total number of samples Correct number / total number 30 / 30 30 / 30 30 / 30 29 / 30 119 / 120 Accuracy 100％ 100％ 100％ 96.7％ 99.2％
[0105] Table 5. Electro-vibration joint feature diagnosis results
[0106] Prediction results Normal state Loose screws Electromagnet stuck Spring aging Total number of samples Correct number / total number 30 / 30 27 / 30 27 / 30 26 / 30 110 / 120 Accuracy 100％ 90％ 90％ 87.7％ 91.7％
[0107] Table 6 Single coil current characteristic diagnosis results
[0108] Prediction results Normal state Loose screws Electromagnet stuck Spring aging Total number of samples Correct number / total number 28 / 30 26 / 30 26 / 30 25 / 30 105 / 120 Accuracy 93.3％ 87.7％ 87.7％ 83.3％ 87.5％
[0109] Table 7 Single vibration signal feature diagnosis results
[0110] Prediction results Normal state Loose screws Electromagnet stuck Spring aging Sample, total, number Correct number / total number 27 / 30 25 / 30 26 / 30 27 / 30 105 / 120 Accuracy 90％ 83.3％ 87.7％ 90％ 87.5％
[0111] Table 8 Single sound signal feature diagnosis results
[0112] Prediction results Normal state Loose screws Electromagnet stuck Spring aging Total number of samples Correct number / total number 28 / 30 24 / 30 24 / 30 24 / 30 100 / 120 Accuracy 93.3％ 80％ 80％ 80％ 83％
[0113] 4. Data storage architecture and fault diagnosis interface design
[0114] Based on the analysis of the characteristics of monitoring data, structured data such as current and voltage and unstructured data such as machinery are mixed and stored, and the database is designed according to the three layers of resources, storage and query. Among them, the resource layer is mainly responsible for virtualizing and standardizing a large amount of data. This layer plans to use Hadoop cluster technology; the storage layer is responsible for storing a large amount of data, and first of all, it is necessary to pre-process the unstructured data. The data is cleaned, sampled, transformed and normalized, and finally the structured data and unstructured data are in the same format; the query layer can realize automatic query of massive data. Its structure is as follows Figure 4 shown.
[0115] According to the diagnostic method, a fault diagnosis interface was designed using GUI, which enables users to connect to the database. The login interface sets the user name, login password and usage permissions, such as Figure 5 The fault diagnosis interface is as shown. Figure 6 As shown, the leftmost column is the acquisition setting, which can specify the acquisition mode, sampling frequency and number of channels, and access the resource layer to virtualize and standardize the data; the middle column is the signal reading interface, which can read the opening and closing data signals, and access the storage layer for storage and preprocessing; the rightmost column is the fault diagnosis interface, which can display the fault diagnosis results and the remaining electrical and mechanical life of the circuit breaker, and this part is connected to the query layer.
[0116] A person skilled in the art can understand that all or part of the steps in the above-mentioned embodiment method can be completed by instructing related hardware through a program, and the program can be stored in a computer-readable storage medium, such as ROM / RAM, disk, CD-ROM, etc.
[0117] The above disclosure is only the preferred embodiment of the present invention, which certainly cannot be used to limit the scope of the present invention. Therefore, equivalent changes made according to the claims of the present invention are still within the scope of the present invention.
[0118] It should be noted that the embodiments of the present invention can be implemented by hardware, software or a combination of software and hardware. The hardware part can be implemented using dedicated logic; the software part can be stored in a memory and executed by an appropriate instruction execution system, such as a microprocessor or dedicated design hardware. Those skilled in the art will appreciate that the above-mentioned devices and methods can be implemented using computer executable instructions and / or contained in processor control codes, such as such codes provided on a programmable memory or a data carrier such as an optical or electronic signal carrier.
[0119] In addition, although the operation of the method of the present invention is described in a particular order in the accompanying drawings, this does not require or imply that these operations must be performed in this particular order, or that all the operations shown must be performed to achieve the desired result. On the contrary, the steps described in the flow chart can change the order of execution. Additionally or alternatively, some steps can be omitted, multiple steps can be combined into one step for execution, and / or one step can be decomposed into multiple steps for execution. It should also be noted that the features and functions of two or more devices according to the present invention can be embodied in one device. On the contrary, the features and functions of a device described above can be further divided into being embodied by multiple devices.
[0120] Although the invention has been described with reference to several specific embodiments, it should be understood that the invention is not limited to the specific embodiments disclosed. The invention is intended to cover various modifications and equivalent arrangements included within the spirit and scope of the appended claims.
Claims
1. A railway high-voltage circuit breaker fault diagnosis device with an improved binary tree classification algorithm, characterized in that: It includes data feature extraction module, data processing module and support vector machine; The data feature extraction module collects the feature vector sequence X of the electric vibration and acoustic signal of the railway high-voltage circuit breaker by principal component analysis; The data processing module includes a data preprocessing module and a data dimension reduction module, wherein the data preprocessing module is connected to the data feature module and is used to perform dimension unification processing on the feature vector sequence X; the data dimension reduction module is connected to the data feature module and is used to obtain a covariance matrix according to the feature vector sequence X after dimension unification processing, and reduce the dimension of the covariance matrix to obtain effective components; The support vector machines are respectively connected to the data dimension reduction modules. The support vector machines obtain the ratio of the inter-class distribution metric to the intra-class distribution metric and the credibility of the classifier as the classification basis according to the effective components and the improved binary tree classification algorithm, and obtain the mechanical failure result according to the ratio of the inter-class distribution metric to the intra-class distribution metric and the credibility of the classifier.
2. The railway high-voltage circuit breaker fault diagnosis method based on the improved binary tree classification algorithm according to claim 1 is characterized in that: The data dimension reduction module is used to obtain a covariance matrix according to the eigenvector sequence X after dimension unification, and reduce the dimension of the covariance matrix to obtain effective components, which specifically includes the following steps: S31, performing decentralized processing on the eigenvector sequence X and obtaining a covariance matrix; S32, by performing eigendecomposition on the covariance matrix and obtaining eigenvalues (λ1≥λ2≥…≥λ m ) and the corresponding eigenvector, select the eigenvector a where the first k eigenroots are located i (i=1,2,…m), forming the projection matrix: <h2 style=";text-align:left;direction:ltr">W=[a1,a2,…,a<h2 style=";text-align:left;direction:ltr"> k <h2 style=";text-align:left;direction:ltr"> ]; S33. Determine the number of principal components according to the size of the eigenvalue. The eigenvalue value is proportional to the contribution coefficient. The contribution coefficient K t Defined as: S34. Select the main components with cumulative contribution rate greater than 96% as effective components.
3. The railway high-voltage circuit breaker fault diagnosis method based on the improved binary tree classification algorithm according to claim 1 is characterized in that: The effective components include training samples and test samples, and the support vector machine obtains the ratio of the inter-class distribution metric to the intra-class distribution metric and the credibility of the classifier used as a classification basis according to the effective components and the improved binary tree classification algorithm, specifically including the following steps: S41, the training sample is input into the support vector machine for learning, and the optimal parameters w*, a are obtained. i *, b*, and output model S i ; S42, the training samples are input into a support vector machine to train a classifier {c1, c2, ..., c n } corresponding to the weights p1, p2, ..., p n (i=1,…,n), and the weight corresponding to the classifier is used as the credibility of the classifier; Among them, the optimal parameters include w*, a i *, b*, w* are weights, a i * is the optimal Lagrangian factor, b* is the optimal threshold; the credibility interval is between 0 and 1, "1" represents complete credibility, that is, it belongs to this category; "0" represents complete unreliability, that is, it does not belong to this category. The closer the credibility is to 1, the higher the accuracy of the classification judgment. Output model S i as follows: p(x)=p(y=1 / x)=1 / {1+exp[-f(x)]}; p(-x)=p(y=-1 / x)=exp[-f(x)] / {1+exp[-f(x)]}.
4. The railway high-voltage circuit breaker fault diagnosis method based on the improved binary tree classification algorithm according to claim 3 is characterized in that: The ratio of the inter-class distribution metric to the intra-class distribution metric divides the training samples into two categories. The two categories of training samples respectively identify the training samples belonging to the i-th category according to the credibility of the classifier and mark them with a positive sign, otherwise they are marked with a negative sign.
5. The railway high-voltage circuit breaker fault diagnosis method based on the improved binary tree classification algorithm according to claim 4 is characterized in that: The test sample is input into the support vector machine to diagnose the mechanical fault category according to the training sample of the i-th category.
6. The railway high-voltage circuit breaker fault diagnosis method based on the improved binary tree classification algorithm according to claim 3 is characterized in that: The classifier {c1, c2, ..., c n } is a binary classifier constructed by the binary tree classification algorithm.
7. The railway high-voltage circuit breaker fault diagnosis method based on the improved binary tree classification algorithm according to claim 1 is characterized in that: The data processing module uses the mean-variance calculation method to perform dimension unification processing on the feature vector sequence X.
8. The railway high-voltage circuit breaker fault diagnosis method based on the improved binary tree classification algorithm according to claim 1 is characterized in that: It includes a data storage module, the data storage module includes a data resource management module, a data storage execution module and a data query service module, the data resource management module; The data storage module is used to store the high-voltage circuit breaker data in a mixed manner to form initial data, wherein the high-voltage circuit breaker data includes structured data such as current and voltage and unstructured data such as mechanical data; The data resource management module is used to virtualize and standardize the initial data according to the Hadoop cluster technology; The data storage execution module is connected to the data resource management module and is used to perform unified dimension processing on the initial data that has been virtualized and standardized; The data query service module automatically queries the initial data processed in the same dimension by using the traditional retrieval algorithm and the intelligent retrieval algorithm in parallel.
9. The railway high-voltage circuit breaker fault diagnosis device with improved binary tree classification algorithm according to claim 8 is characterized in that: It includes a fault diagnosis module, the fault diagnosis query module includes a signal acquisition module, a signal reading module and a diagnosis result display module; The signal acquisition module is used to specify the acquisition mode, sampling frequency and number of channels, and is connected to the data resource management module; The signal reading module is used to read the high-voltage circuit breaker data signal and input the high-voltage circuit breaker data signal into the data storage execution module; The diagnosis result display module is connected to the data query service module and is used to display the mechanical failure results.
10. A railway high-voltage circuit breaker fault diagnosis method based on an improved binary tree classification algorithm, characterized in that: The railway high-voltage circuit breaker fault diagnosis method including the improved binary tree classification algorithm is used to adapt the railway high-voltage circuit breaker fault diagnosis device of the improved binary tree classification algorithm of any one of claims 1 to 9; the method includes the following steps: The data feature extraction module collects the feature vector sequence X of the electric vibration and acoustic signal of the railway high-voltage circuit breaker according to the principal component analysis method; The feature vector sequence X is dimensionally unified through the data preprocessing module; The data dimension reduction module is used to obtain the covariance matrix of the feature vector sequence X after dimension unification, and the dimension of the covariance matrix is reduced to obtain the effective components; The support vector machine obtains the ratio of the inter-class distribution metric to the intra-class distribution metric and the credibility of the classifier as the classification basis according to the effective components and the improved binary tree classification algorithm, and obtains the mechanical failure result according to the ratio of the inter-class distribution metric to the intra-class distribution metric and the credibility of the classifier.