Train body vibration state prediction and analysis method and system based on improved PCA-SVM network

By improving the PCA-SVM network and K-means algorithm, the state clustering of vehicle body vibration sample data is solved, and combined with the data feature enhancement strategy, the problem of high computational complexity and dimensionality disaster in multi-category classification is solved, and the problem of SVM is achieved more Accurate prediction of vehicle body vibration status.

CN118643398BActive Publication Date: 2024-11-05EAST CHINA JIAOTONG UNIVERSITY
View PDF 4 Cites 0 Cited by

Patent Information

Application Number
CN202411117683.9
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-08-15
Publication Date
2024-11-05
Estimated Expiration
2044-08-15

AI Technical Summary

Technical Problem

The existing SVM has problems such as high computational complexity, disaster in the vibration state of the train body and lack of global optimization in the multi-category classification, resulting in inaccurate prediction and analysis results.

Method used

The improved PCA-SVM network is adopted to cluster the vehicle body vibration sample data through the K-means algorithm, and combine the data feature enhancement strategy and sensitivity adjustment strategy to generate the vehicle body vibration state prediction model to improve the prediction accuracy.

Benefits of technology

It improves the calculation accuracy of the vibration state of the vehicle body, can quickly predict the vibration state of the railway rolling stock and vehicle on uneven tracks, and reduces the calculation complexity and storage requirements.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118643398B_ABST
    Figure CN118643398B_ABST
Patent Text Reader

Abstract

The present invention discloses a train body vibration state prediction and analysis method and system based on an improved PCA-SVM network. The method comprises: using a K-means algorithm to perform state clustering processing on at least one target body vibration sample data to obtain at least one state data set, inputting at least one state data set and a body vibration state standard indicator set corresponding to the at least one state data set into an SVM multi-class classifier, training and generating a body vibration state prediction model, thereby improving the accuracy of calculating the body vibration state and facilitating the rapid prediction of the vibration state of railway locomotives and vehicles on uneven tracks.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention belongs to the technical field of train body vibration analysis, and in particular relates to a train body vibration state prediction and analysis method and system based on an improved PCA-SVM network. Background Art

[0002] At present, the express logistics industry is developing rapidly, among which the demand for high-value and time-sensitive cargo transportation is growing rapidly, resulting in higher requirements for the vehicle body condition during operation. However, incorrect track geometry, uneven track surface or unbalanced wheels may cause vehicle body vibration, or inaccurate manufacturing or unbalanced installation of wheels may cause vehicle body vibration.

[0003] In order to correctly understand the relationship between track unevenness and vehicle body vibration characteristics, many scholars use dynamic models to predict the vehicle body vibration state. However, this requires a lot of professional knowledge and a long period of model simulation, and is difficult to apply in practice. Traditional manual detection and data analysis are time-consuming and laborious, and manual analysis often has some errors. In order to avoid errors and improve work efficiency, the SVM multi-class classifier in machine learning is used to predict and analyze the vibration state of the train body.

[0004] Existing SVMs are very powerful in binary classification problems, but they have some limitations when dealing with multi-category classification, such as high computational complexity: for each category pair, it trains two independent SVM models. As the number of categories increases, the amount of computation and storage requirements increase significantly; feature space expansion: each new category is equivalent to generating a new dimension in the original feature space, which may lead to dimensionality disaster and affect model performance; lack of global optimization: one-to-one and one-to-many methods do not consider the relationship between all categories, which may lead to local optimal solutions instead of global optimal solutions. Summary of the invention

[0005] The present invention provides a train body vibration state prediction and analysis method and system based on an improved PCA-SVM network, which are used to solve the technical problem of inaccurate train body vibration state prediction and analysis results.

[0006] In a first aspect, the present invention provides a train body vibration state prediction and analysis method based on an improved PCA-SVM network, comprising:

[0007] Acquire at least one vehicle body vibration sample data, and perform structural abstraction processing on the at least one vehicle body vibration sample data according to a preset structural abstraction strategy to obtain at least one target vehicle body vibration sample data;

[0008] Performing state clustering processing on the at least one target vehicle body vibration sample data using a K-means algorithm to obtain at least one state data set, wherein a certain state data set contains various vehicle body vibration sample data of the same state, and the certain state data set is associated with a certain vehicle body vibration state standard indicator set;

[0009] The at least one state data set and the vehicle body vibration state standard indicator set corresponding to the at least one state data set are input into the SVM multi-class classifier to train and generate a vehicle body vibration state prediction model, wherein the SVM multi-class classifier includes a data feature enhancement strategy, and the expression of the data feature enhancement strategy is:

[0010] ,

[0011] In the formula, To add functions, To weaken the function, is the state function, is the error function, is the change threshold, is the characteristic parameter index, is the characteristic state, is the feature dimension, is the feature weight, is the characteristic kernel function parameter, is the modal amplitude, is the structural wave number, is the excitation frequency, is the shape function coefficient, is the reduction factor, To reduce the number of waves, is the state weight, To reconstruct the weights, is the target domain state, is the marginal probability function, To represent the state, is the state wave number, is the maximum error value, is the error penalty coefficient, is the error kernelization mapping coefficient, is the error wave number, is the activation function, is the weight coefficient, is the network weight;

[0012] ,

[0013] In the formula, is the weight threshold, is a strong weight, is the upper limit weight, is the bias weight;

[0014] ,

[0015] In the formula, is the optimal weight, The worst feature, For ideal characteristics, To search for features, is the upper limit feature, is the characteristic bias, is the characteristic position, is the number of features, is the optimal number of features;

[0016] Acquire the real-time vibration signal of rail transit equipment according to the preset sensitivity adjustment strategy to obtain the real-time vehicle body vibration data;

[0017] After the structure abstraction processing and the state clustering processing, the real-time vehicle body vibration data is input into the vehicle body vibration state prediction model, and the vehicle body vibration state prediction model outputs a vehicle body vibration state prediction result.

[0018] In a second aspect, the present invention provides a train body vibration state prediction and analysis system based on an improved PCA-SVM network, comprising:

[0019] A processing module is configured to obtain at least one vehicle body vibration sample data, perform structural abstraction processing on the at least one vehicle body vibration sample data according to a preset structural abstraction strategy, and obtain at least one target vehicle body vibration sample data;

[0020] A clustering module is configured to perform state clustering processing on the at least one target vehicle body vibration sample data using a K-means algorithm to obtain at least one state data set, wherein a state data set contains various vehicle body vibration sample data of the same state, and the state data set is associated with a vehicle body vibration state standard indicator set;

[0021] The training module is configured to input the at least one state data set and the vehicle body vibration state standard indicator set corresponding to the at least one state data set into the SVM multi-class classifier, train and generate a vehicle body vibration state prediction model, wherein the SVM multi-class classifier includes a data feature enhancement strategy, and the expression of the data feature enhancement strategy is:

[0022] ,

[0023] In the formula, To add functions, To weaken the function, is the state function, is the error function, is the change threshold, is the characteristic parameter index, is the characteristic state, is the feature dimension, is the feature weight, is the characteristic kernel function parameter, is the modal amplitude, is the structural wave number, is the excitation frequency, is the shape function coefficient, is the reduction factor, To reduce the number of waves, is the state weight, To reconstruct the weights, is the target domain state, is the marginal probability function, To represent the state, is the state wave number, is the maximum error value, is the error penalty coefficient, is the error kernelization mapping coefficient, is the error wave number, is the activation function, is the weight coefficient, is the network weight;

[0024] ,

[0025] In the formula, is the weight threshold, is a strong weight, is the upper limit weight, is the bias weight;

[0026] ,

[0027] In the formula, is the optimal weight, The worst feature, For ideal characteristics, To search for features, is the upper limit feature, is the characteristic bias, is the characteristic position, is the number of features, is the optimal number of features;

[0028] An acquisition module is configured to acquire a real-time vibration signal of rail transit equipment according to a preset sensitivity adjustment strategy to obtain real-time vehicle body vibration data;

[0029] The output module is configured to input the real-time vehicle body vibration data into the vehicle body vibration state prediction model after the structure abstraction processing and the state clustering processing, and the vehicle body vibration state prediction model outputs the vehicle body vibration state prediction result.

[0030] According to a third aspect, an electronic device is provided, comprising: at least one processor, and a memory communicatively connected to the at least one processor, wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can perform the steps of the train body vibration state prediction and analysis method based on the improved PCA-SVM network according to any embodiment of the present invention.

[0031] In a fourth aspect, the present invention further provides a computer-readable storage medium having a computer program stored thereon. When the program instructions are executed by a processor, the processor executes the steps of the train body vibration state prediction and analysis method based on the improved PCA-SVM network according to any embodiment of the present invention.

[0032] The train body vibration state prediction and analysis method and system based on the improved PCA-SVM network of the present application adopts the K-means algorithm to perform state clustering processing on at least one target body vibration sample data to obtain at least one state data set, and inputs the at least one state data set and the body vibration state standard indicator set corresponding to the at least one state data set into the SVM multi-class classifier to train and generate a body vibration state prediction model, thereby improving the accuracy of the measurement of the body vibration state and facilitating the rapid prediction of the vibration state of railway locomotives and vehicles on uneven tracks. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0034] Figure 1 A flow chart of a train body vibration state prediction and analysis method based on an improved PCA-SVM network provided by an embodiment of the present invention;

[0035] Figure 2 A structural block diagram of a train body vibration state prediction and analysis system based on an improved PCA-SVM network provided by an embodiment of the present invention;

[0036] Figure 3 It is a schematic diagram of the structure of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0037] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0038] See also Figure 1 , which shows a flow chart of a train body vibration state prediction and analysis method based on an improved PCA-SVM network of the present application.

[0039] like Figure 1 As shown, the train body vibration state prediction and analysis method based on the improved PCA-SVM network specifically includes the following steps:

[0040] Step S101 : acquiring at least one vehicle body vibration sample data, and performing structural abstraction processing on the at least one vehicle body vibration sample data according to a preset structural abstraction strategy to obtain at least one target vehicle body vibration sample data.

[0041] In this step, the expression of the structure abstraction strategy is:

[0042] ,

[0043] ,

[0044] In the formula, is the dimension feature, is the dimension feature, is a sample set of training data points, is the cost function, for The keynote, is the addition symbol, is the multiplication symbol, for The homomorphism of for The linked list, is a co-algebra, is an algebraic structure, is a double algebra, is the initial algebra, for The keynote, for The homomorphism of for 's linked list.

[0045] Step S102, using the K-means algorithm to perform state clustering processing on the at least one target vehicle body vibration sample data to obtain at least one state data set, wherein a certain state data set contains various vehicle body vibration sample data of the same state, and the certain state data set is associated with a certain vehicle body vibration state standard indicator set.

[0046] In this step, Euler's formula is used to define The characteristic relationship in the set is then used to divide the cluster boundaries according to the different cluster centers of the data, and the distribution function of the clustering model is calculated. , and calculate the reliability of each clustering result to obtain at least one state data set, wherein the expression for calculating the reliability of each clustering result is:

[0047] ,

[0048] ,

[0049] ,

[0050] ,

[0051] ,

[0052] In the formula, is the reliability expression of clustering results, is the weight coefficient of the feature data at the tth iteration, To express the constraint relationship between features, To express the degree of deviation between features, is the number of clusters in the data set, is the number of constraint relations, is the distribution sequence, is the distribution function weight, is the cumulative variance function, is the maximum value of the hidden layer score in the data point, is the iteration reliability, is the reliability characteristic value.

[0053] Step S103: input the at least one state data set and the vehicle body vibration state standard indicator set corresponding to the at least one state data set into a SVM multi-class classifier to train and generate a vehicle body vibration state prediction model.

[0054] In this step, the time domain and frequency domain statistical parameters of track irregularity are used as indicators of different vehicle body vibration states (absolute peak value, standard deviation, absolute average value), and then a vehicle body vibration state standard indicator set corresponding to at least one state data set is obtained. The regular inspection of the line by the track inspection vehicle can provide the required inspection data to meet the needs. It is obtained by collecting on-site data such as track inspection data, train formation, and line maintenance and repair.

[0055] It should be noted that the decision function of the SVM multi-class classifier is transformed into a boundary optimization problem, specifically:

[0056] ,

[0057] ,

[0058] ,

[0059] In the formula, is the pattern vector size, is a predefined function, is an adjustable parameter, is the mode vector oscillation function, is the adjustable parameter weight, is the penalty factor, is the slack variable, is the category label vector and T is the transpose.

[0060] The kernel function of the SVM multi-class classifier is a radial basis kernel function, and the expression of the radial basis kernel function is:

[0061] ,

[0062] ,

[0063] In the formula, is the number of training set samples, is the kernel function weight, is the kernel function, is the characteristic main mode, is the kth training sample, is the radial basis kernel function, is the sample mean, is the i-th training sample, is the jth training sample, is the random perturbation mode, is the kernel function width parameter.

[0064] In order to improve the classification efficiency of the classifier, a data feature enhancement strategy for the SVM multi-class classifier is constructed. The expression of the data feature enhancement strategy is:

[0065] ,

[0066] In the formula, To add functions, To weaken the function, is the state function, is the error function, is the change threshold, is the characteristic parameter index, is the characteristic state, is the feature dimension, is the feature weight, is the characteristic kernel function parameter, is the modal amplitude, is the structural wave number, is the excitation frequency, is the shape function coefficient, is the reduction factor, To reduce the number of waves, is the state weight, To reconstruct the weights, is the target domain state, is the marginal probability function, To represent the state, is the state wave number, is the maximum error value, is the error penalty coefficient, is the error kernelization mapping coefficient, is the error wave number, is the activation function, is the weight coefficient, is the network weight;

[0067] ,

[0068] In the formula, is the weight threshold, is a strong weight, is the upper limit weight, is the bias weight;

[0069] ,

[0070] In the formula, is the optimal weight, The worst feature, For ideal characteristics, To search for features, is the upper limit feature, is the characteristic bias, is the characteristic position, is the number of features, is the optimal number of features;

[0071] In a specific embodiment, before inputting the at least one state data set and the vehicle body vibration state standard indicator set corresponding to the at least one state data set into the SVM multi-class classifier, PCA parameter dimension reduction and information optimization processing are performed on each target vehicle body vibration sample data in the at least one state data set.

[0072] Specifically, the data is centered and the sample mean is calculated, where the expression for calculating the sample mean is:

[0073] ,

[0074] ,

[0075] ,

[0076] ,

[0077] In the formula, is the sample set, For the first sample, For the second sample, is the nth sample, is the first observation value of the i-th sample, is the i-th sample, is the second observation value of the i-th sample, is the pth observation value of the ith sample, is the mean expression, is the number of samples, is the sample after centering;

[0078] Calculate the covariance matrix of the original state data set to obtain the state data set after dimensionality reduction and original centering, where the state data set after dimensionality reduction and original centering is characterized as follows:

[0079] ,

[0080] ,

[0081] ,

[0082] In the formula, is the covariance matrix, is the number of samples in the dataset, is the sample matrix, is the weighted eigenvalue, is the coefficient corresponding to the eigenvector, For the modal, is the eigenvalue of the covariance matrix, is the feature dimension, For transposition, The main mode number.

[0083] In this step, the SVM multi-class classifier includes a data feature enhancement strategy, and the expression of the data feature enhancement strategy is:

[0084] ,

[0085] In the formula, To add functions, To weaken the function, is the state function, is the error function, is the change threshold, is the characteristic parameter index, is the characteristic state, is the feature dimension, is the feature weight, is the characteristic kernel function parameter, is the modal amplitude, is the structural wave number, is the excitation frequency, is the shape function coefficient, is the reduction factor, To reduce the number of waves, is the state weight, To reconstruct the weights, is the target domain state, is the marginal probability function, To represent the state, is the state wave number, is the maximum error value, is the error penalty coefficient, is the error kernelization mapping coefficient, is the error wave number, is the activation function, is the weight coefficient, is the network weight;

[0086] ,

[0087] In the formula, is the weight threshold, is a strong weight, is the upper limit weight, is the bias weight;

[0088] ,

[0089] In the formula, is the optimal weight, The worst feature, For ideal characteristics, To search for features, is the upper limit feature, is the characteristic bias, is the characteristic position, is the number of features, is the optimal number of features;

[0090] Step S104, acquiring the real-time vibration signal of the rail transit equipment according to the preset sensitivity adjustment strategy to obtain the real-time vehicle body vibration data.

[0091] In this step, the CA-YD-3193 piezoelectric vibration acceleration sensor is selected to collect vibration signals of rail transit equipment. The collected vibration signals are transmitted to the monitoring service terminal using the rail transit vehicle monitoring network. The sensitivity of traditional vibration sensors does not meet the requirements, so an Alpha sensitivity enhancement strategy is proposed to improve its sensitivity. Specifically:

[0092] The expression of sensitivity adjustment strategy is:

[0093] ,

[0094] ,

[0095] In the formula, is the sensor sensitivity, is the structural wave number, is the angle function, is the template thickness parameter, is the shape function, is the orthogonal weight, is the derivative of r, is the dimensionless frequency, is the lower surface structural mode, is the upper surface structural mode, is the structural modal threshold, is the sampling index, is the gradient index, is the weighting coefficient.

[0096] Step S105, inputting the real-time vehicle body vibration data into the vehicle body vibration state prediction model after the structure abstraction processing and the state clustering processing, and the vehicle body vibration state prediction model outputs a vehicle body vibration state prediction result.

[0097] In summary, the method of the present application uses the K-means algorithm to perform state clustering processing on at least one target vehicle body vibration sample data to obtain at least one state data set, inputs at least one state data set and a vehicle body vibration state standard indicator set corresponding to at least one state data set into a SVM multi-class classifier, trains and generates a vehicle body vibration state prediction model, improves the accuracy of measuring the vehicle body vibration state, and is conducive to quickly predicting the vibration state of railway locomotives and vehicles on uneven tracks.

[0098] See also Figure 2 , which shows a structural block diagram of a train body vibration state prediction and analysis system based on an improved PCA-SVM network of the present application.

[0099] like Figure 2 As shown, the train body vibration state prediction and analysis system 200 includes a processing module 210 , a clustering module 220 , a training module 230 , an acquisition module 240 and an output module 250 .

[0100] The processing module 210 is configured to obtain at least one vehicle body vibration sample data, perform structural abstraction processing on the at least one vehicle body vibration sample data according to a preset structural abstraction strategy, and obtain at least one target vehicle body vibration sample data;

[0101] The clustering module 220 is configured to perform state clustering processing on the at least one target vehicle body vibration sample data using a K-means algorithm to obtain at least one state data set, wherein a state data set contains various vehicle body vibration sample data of the same state, and the state data set is associated with a vehicle body vibration state standard indicator set;

[0102] The training module 230 is configured to input the at least one state data set and the vehicle body vibration state standard indicator set corresponding to the at least one state data set into the SVM multi-class classifier, train and generate a vehicle body vibration state prediction model, wherein the SVM multi-class classifier includes a data feature enhancement strategy, and the expression of the data feature enhancement strategy is:

[0103] ,

[0104] In the formula, To add functions, To weaken the function, is the state function, is the error function, is the change threshold, is the characteristic parameter index, is the characteristic state, is the feature dimension, is the feature weight, is the characteristic kernel function parameter, is the modal amplitude, is the structural wave number, is the excitation frequency, is the shape function coefficient, is the reduction factor, To reduce the number of waves, is the state weight, To reconstruct the weights, is the target domain state, is the marginal probability function, To represent the state, is the state wave number, is the maximum error value, is the error penalty coefficient, is the error kernelization mapping coefficient, is the error wave number, is the activation function, is the weight coefficient, is the network weight;

[0105] ,

[0106] In the formula, is the weight threshold, is a strong weight, is the upper limit weight, is the bias weight;

[0107] ,

[0108] In the formula, is the optimal weight, The worst feature, For ideal characteristics, To search for features, is the upper limit feature, is the characteristic bias, is the characteristic position, is the number of features, is the optimal number of features;

[0109] The acquisition module 240 is configured to acquire the real-time vibration signal of the rail transit equipment according to a preset sensitivity adjustment strategy to obtain real-time vehicle body vibration data;

[0110] The output module 250 is configured to input the real-time vehicle body vibration data into the vehicle body vibration state prediction model after the structure abstraction processing and the state clustering processing, and the vehicle body vibration state prediction model outputs a vehicle body vibration state prediction result.

[0111] It should be understood that Figure 2 Modules and references documented in Figure 1 Therefore, the operations and features described above for the method and the corresponding technical effects are also applicable to Figure 2 The modules in it will not be described in detail here.

[0112] In other embodiments, the embodiments of the present invention further provide a computer-readable storage medium having a computer program stored thereon, wherein when the program instructions are executed by a processor, the processor is caused to execute the train body vibration state prediction and analysis method based on the improved PCA-SVM network in any of the above method embodiments;

[0113] As an implementation mode, the computer-readable storage medium of the present invention stores computer-executable instructions, and the computer-executable instructions are configured as follows:

[0114] Acquire at least one vehicle body vibration sample data, and perform structural abstraction processing on the at least one vehicle body vibration sample data according to a preset structural abstraction strategy to obtain at least one target vehicle body vibration sample data;

[0115] Performing state clustering processing on the at least one target vehicle body vibration sample data using a K-means algorithm to obtain at least one state data set, wherein a certain state data set contains various vehicle body vibration sample data of the same state, and the certain state data set is associated with a certain vehicle body vibration state standard indicator set;

[0116] The at least one state data set and the vehicle body vibration state standard indicator set corresponding to the at least one state data set are input into the SVM multi-class classifier to train and generate a vehicle body vibration state prediction model, wherein the SVM multi-class classifier includes a data feature enhancement strategy, and the expression of the data feature enhancement strategy is:

[0117] ,

[0118] In the formula, To add functions, To weaken the function, is the state function, is the error function, is the change threshold, is the characteristic parameter index, is the characteristic state, is the feature dimension, is the feature weight, is the characteristic kernel function parameter, is the modal amplitude, is the structural wave number, is the excitation frequency, is the shape function coefficient, is the reduction factor, To reduce the number of waves, is the state weight, To reconstruct the weights, is the target domain state, is the marginal probability function, To represent the state, is the state wave number, is the maximum error value, is the error penalty coefficient, is the error kernelization mapping coefficient, is the error wave number, is the activation function, is the weight coefficient, is the network weight;

[0119] ,

[0120] In the formula, is the weight threshold, is a strong weight, is the upper limit weight, is the bias weight;

[0121] ,

[0122] In the formula, is the optimal weight, The worst feature, For ideal characteristics, To search for features, is the upper limit feature, is the characteristic bias, is the characteristic position, is the number of features, is the optimal number of features;

[0123] Acquire the real-time vibration signal of rail transit equipment according to the preset sensitivity adjustment strategy to obtain the real-time vehicle body vibration data;

[0124] After the structure abstraction processing and the state clustering processing, the real-time vehicle body vibration data is input into the vehicle body vibration state prediction model, and the vehicle body vibration state prediction model outputs a vehicle body vibration state prediction result.

[0125] The computer-readable storage medium may include a program storage area and a data storage area, wherein the program storage area may store an operating system and application programs required for at least one function; the data storage area may store data created based on the use of the train body vibration state prediction and analysis system based on the improved PCA-SVM network, etc. In addition, the computer-readable storage medium may include a high-speed random access memory, and may also include a memory, such as at least one disk storage device, a flash memory device, or other non-volatile solid-state storage device. In some embodiments, the computer-readable storage medium may optionally include a memory remotely arranged relative to the processor, and these remote memories may be connected to the train body vibration state prediction and analysis system based on the improved PCA-SVM network via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0126] Figure 3 is a schematic diagram of the structure of an electronic device provided by an embodiment of the present invention, such as Figure 3 As shown, the device includes: a processor 310 and a memory 320. The electronic device may also include: an input device 330 and an output device 340. The processor 310, the memory 320, the input device 330 and the output device 340 may be connected via a bus or other means. Figure 3 The example of connecting through a bus is taken. The memory 320 is the above-mentioned computer-readable storage medium. The processor 310 executes various functional applications and data processing of the server by running the non-volatile software programs, instructions and modules stored in the memory 320, that is, the train body vibration state prediction and analysis method based on the improved PCA-SVM network in the above-mentioned method embodiment is realized. The input device 330 can receive input digital or character information, and generate key signal input related to user settings and function controls of the train body vibration state prediction and analysis system based on the improved PCA-SVM network. The output device 340 may include display devices such as display screens.

[0127] The electronic device can execute the method provided by the embodiment of the present invention, and has the functional modules and beneficial effects corresponding to the execution method. For technical details not described in detail in this embodiment, please refer to the method provided by the embodiment of the present invention.

[0128] As an implementation mode, the electronic device is applied to a train body vibration state prediction and analysis system based on an improved PCA-SVM network, and is used for a client, and includes: at least one processor; and a memory connected to the at least one processor in communication; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can:

[0129] Acquire at least one vehicle body vibration sample data, and perform structural abstraction processing on the at least one vehicle body vibration sample data according to a preset structural abstraction strategy to obtain at least one target vehicle body vibration sample data;

[0130] Performing state clustering processing on the at least one target vehicle body vibration sample data using a K-means algorithm to obtain at least one state data set, wherein a certain state data set contains various vehicle body vibration sample data of the same state, and the certain state data set is associated with a certain vehicle body vibration state standard indicator set;

[0131] The at least one state data set and the vehicle body vibration state standard indicator set corresponding to the at least one state data set are input into the SVM multi-class classifier to train and generate a vehicle body vibration state prediction model, wherein the SVM multi-class classifier includes a data feature enhancement strategy, and the expression of the data feature enhancement strategy is:

[0132] ,

[0133] In the formula, To add functions, To weaken the function, is the state function, is the error function, is the change threshold, is the characteristic parameter index, is the characteristic state, is the feature dimension, is the feature weight, is the characteristic kernel function parameter, is the modal amplitude, is the structural wave number, is the excitation frequency, is the shape function coefficient, is the reduction factor, To reduce the number of waves, is the state weight, To reconstruct the weights, is the target domain state, is the marginal probability function, To represent the state, is the state wave number, is the maximum error value, is the error penalty coefficient, is the error kernelization mapping coefficient, is the error wave number, is the activation function, is the weight coefficient, is the network weight;

[0134] ,

[0135] In the formula, is the weight threshold, is a strong weight, is the upper limit weight, is the bias weight;

[0136] ,

[0137] In the formula, is the optimal weight, The worst feature, For ideal characteristics, To search for features, is the upper limit feature, is the characteristic bias, is the characteristic position, is the number of features, is the optimal number of features;

[0138] Acquire the real-time vibration signal of rail transit equipment according to the preset sensitivity adjustment strategy to obtain the real-time vehicle body vibration data;

[0139] After the structure abstraction processing and the state clustering processing, the real-time vehicle body vibration data is input into the vehicle body vibration state prediction model, and the vehicle body vibration state prediction model outputs a vehicle body vibration state prediction result.

[0140] Through the description of the above implementation modes, those skilled in the art can clearly understand that each implementation mode can be implemented by means of software plus a necessary general hardware platform, or of course by hardware. Based on such an understanding, the above technical solution can essentially or in other words be embodied in the form of a software product that contributes to the prior art. The computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a magnetic disk, an optical disk, etc., and includes a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods of each embodiment or some parts of the embodiment.

[0141] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A train body vibration state prediction and analysis method based on an improved PCA-SVM network, characterized in that: include: Acquire at least one vehicle body vibration sample data, and perform structural abstraction processing on the at least one vehicle body vibration sample data according to a preset structural abstraction strategy to obtain at least one target vehicle body vibration sample data; Performing state clustering processing on the at least one target vehicle body vibration sample data using a K-means algorithm to obtain at least one state data set, wherein a certain state data set contains various vehicle body vibration sample data of the same state, and the certain state data set is associated with a certain vehicle body vibration state standard indicator set; The at least one state data set and the vehicle body vibration state standard indicator set corresponding to the at least one state data set are input into the SVM multi-class classifier to train and generate a vehicle body vibration state prediction model, wherein the SVM multi-class classifier includes a data feature enhancement strategy, and the expression of the data feature enhancement strategy is: , In the formula, To add functions, To weaken the function, is the state function, is the error function, is the change threshold, is the characteristic parameter index, is the characteristic state, is the feature dimension, is the feature weight, is the characteristic kernel function parameter, is the modal amplitude, is the structural wave number, is the excitation frequency, is the shape function coefficient, is the reduction factor, To reduce the number of waves, is the state weight, To reconstruct the weights, is the target domain state, is the marginal probability function, To represent the state, is the state wave number, is the maximum error value, is the error penalty coefficient, is the error kernelization mapping coefficient, is the error wave number, is the activation function, is the weight coefficient, is the network weight; , In the formula, is the weight threshold, is a strong weight, is the upper limit weight, is the bias weight; , In the formula, is the optimal weight, The worst feature, For ideal characteristics, To search for features, is the upper limit feature, is the characteristic bias, is the characteristic position, is the number of features, is the optimal number of features; Acquire the real-time vibration signal of rail transit equipment according to the preset sensitivity adjustment strategy to obtain the real-time vehicle body vibration data; After the structure abstraction processing and the state clustering processing, the real-time vehicle body vibration data is input into the vehicle body vibration state prediction model, and the vehicle body vibration state prediction model outputs a vehicle body vibration state prediction result.

2. The train body vibration state prediction and analysis method based on the improved PCA-SVM network according to claim 1 is characterized in that: The expression of the structural abstraction strategy is: , , In the formula, is the dimension feature, is the dimension feature, is a sample set of training data points, is the cost function, for The keynote, is the addition symbol, is the multiplication symbol, for The homomorphism of for The linked list, is a co-algebra, is an algebraic structure, is a double algebra, is the initial algebra, for The keynote, for The homomorphism of for 's linked list.

3. The train body vibration state prediction and analysis method based on the improved PCA-SVM network according to claim 1 is characterized in that: The K-means algorithm is used to perform state clustering processing on the at least one target vehicle body vibration sample data to obtain at least one state data set, including: Defined by Euler's formula The characteristic relationship in the set is then used to divide the cluster boundaries according to the different cluster centers of the data, and the distribution function of the clustering model is calculated. , and calculate the reliability of each clustering result to obtain at least one state data set, wherein the expression for calculating the reliability of each clustering result is: , , , , , In the formula, is the reliability expression of clustering results, is the weight coefficient of the feature data at the tth iteration, To express the constraint relationship between features, To express the degree of deviation between features, is the number of clusters in the data set, is the number of constraint relations, is the distribution sequence, is the distribution function weight, is the cumulative variance function, is the maximum value of the hidden layer score in the data point, is the iteration reliability, is the reliability characteristic value.

4. The train body vibration state prediction and analysis method based on the improved PCA-SVM network according to claim 1 is characterized in that: Before inputting the at least one state data set and the vehicle body vibration state standard indicator set corresponding to the at least one state data set into the SVM multi-class classifier, the method further includes: Perform PCA parameter dimension reduction and information optimization processing on each target vehicle body vibration sample data in the at least one state data set.

5. The train body vibration state prediction and analysis method based on improved PCA-SVM network according to claim 4 is characterized in that: The performing PCA parameter dimension reduction and information optimization processing on each target vehicle body vibration sample data in the at least one state data set comprises: The data is centralized and the sample mean is calculated. The expression for calculating the sample mean is: , , , , In the formula, is the sample set, For the first sample, For the second sample, is the nth sample, is the first observation value of the i-th sample, is the i-th sample, is the second observation value of the i-th sample, is the pth observation value of the ith sample, is the mean expression, is the number of samples, is the sample after centering; Calculate the covariance matrix of the original state data set to obtain the state data set after dimensionality reduction and original centering, where the state data set after dimensionality reduction and original centering is characterized as follows: , , , In the formula, is the covariance matrix, is the number of samples in the dataset, is the sample matrix, is the weighted eigenvalue, is the coefficient corresponding to the eigenvector, For the modal, is the eigenvalue of the covariance matrix, is the feature dimension, For transposition, The main mode number.

6. The train body vibration state prediction and analysis method based on improved PCA-SVM network according to claim 1 is characterized in that: in, The decision function of the SVM multi-class classifier is transformed into a boundary optimization problem, specifically: , , , In the formula, is the pattern vector size, is a predefined function, is an adjustable parameter, is the mode vector oscillation function, is the adjustable parameter weight, is the penalty factor, is the slack variable, is the category label vector and T is the transpose.

7. The train body vibration state prediction and analysis method based on improved PCA-SVM network according to claim 1 is characterized in that: The kernel function of the SVM multi-class classifier is a radial basis kernel function, and the expression of the radial basis kernel function is: , , In the formula, is the number of training set samples, is the kernel function weight, is the kernel function, is the characteristic main mode, is the kth training sample, is the radial basis kernel function, is the sample mean, is the i-th training sample, is the jth training sample, is the random perturbation mode, is the kernel function width parameter.

8. The train body vibration state prediction and analysis method based on improved PCA-SVM network according to claim 1 is characterized in that: The expression of the sensitivity adjustment strategy is: , , In the formula, is the sensor sensitivity, is the structural wave number, is the angle function, is the template thickness parameter, is the shape function, is the orthogonal weight, is the derivative of r, is the dimensionless frequency, is the lower surface structural mode, is the upper surface structural mode, is the structural modal threshold, is the sampling index, is the gradient index, is the weighting coefficient.

9. A train body vibration state prediction and analysis system based on an improved PCA-SVM network, characterized in that: include: A processing module is configured to obtain at least one vehicle body vibration sample data, perform structural abstraction processing on the at least one vehicle body vibration sample data according to a preset structural abstraction strategy, and obtain at least one target vehicle body vibration sample data; A clustering module is configured to perform state clustering processing on the at least one target vehicle body vibration sample data using a K-means algorithm to obtain at least one state data set, wherein a state data set contains various vehicle body vibration sample data of the same state, and the state data set is associated with a vehicle body vibration state standard indicator set; The training module is configured to input the at least one state data set and the vehicle body vibration state standard indicator set corresponding to the at least one state data set into the SVM multi-class classifier, train and generate a vehicle body vibration state prediction model, wherein the SVM multi-class classifier includes a data feature enhancement strategy, and the expression of the data feature enhancement strategy is: , In the formula, To add functions, To weaken the function, is the state function, is the error function, is the change threshold, is the characteristic parameter index, is the characteristic state, is the feature dimension, is the feature weight, is the characteristic kernel function parameter, is the modal amplitude, is the structural wave number, is the excitation frequency, is the shape function coefficient, is the reduction factor, To reduce the number of waves, is the state weight, To reconstruct the weights, is the target domain state, is the marginal probability function, To represent the state, is the state wave number, is the maximum error value, is the error penalty coefficient, is the error kernelization mapping coefficient, is the error wave number, is the activation function, is the weight coefficient, is the network weight; , In the formula, is the weight threshold, is a strong weight, is the upper limit weight, is the bias weight; , In the formula, is the optimal weight, The worst feature, For ideal characteristics, To search for features, is the upper limit feature, is the characteristic bias, is the characteristic position, is the number of features, is the optimal number of features; An acquisition module is configured to acquire a real-time vibration signal of rail transit equipment according to a preset sensitivity adjustment strategy to obtain real-time vehicle body vibration data; The output module is configured to input the real-time vehicle body vibration data into the vehicle body vibration state prediction model after the structure abstraction processing and the state clustering processing, and the vehicle body vibration state prediction model outputs the vehicle body vibration state prediction result.

Citation Information

Patent Citations

  • A prediction method of a track-body vibration state based on depth learning

    CN109034192A

  • Bearing vibration operation state prediction model training method and prediction method and device

    CN110334562A

  • Bearing fault prediction method and system based on convolution noise reduction auto-encoder

    CN116467652A

  • Train compartment vibration monitoring method and vibration signal feature library establishment and application methods

    US20240044695A1