CN-121997197-A - Train wheel polygon prediction method, system, device, medium and program product

CN121997197ACN 121997197 ACN121997197 ACN 121997197ACN-121997197-A

Abstract

The application provides a train wheel polygon prediction method which comprises the steps of obtaining historical wheel data of a locomotive and train, constructing a data sample corresponding to the historical wheel data, dividing an influence factor set into data samples, carrying out correlation analysis on the influence factor set to determine characteristic indexes of which the correlation coefficients meet a set threshold, dividing the characteristic indexes into a training set and a testing set of a random forest model according to the characteristics and a preset proportion, training the training set based on a random forest algorithm to obtain a train wheel polygon prediction model, obtaining a wheel polygon characteristic vector of the train to be predicted, inputting the train wheel polygon prediction model, and outputting a wheel polygon prediction result. The method can effectively predict the polygonal state of the wheels and provides an important reference for train maintenance. The application also provides a train wheel polygon prediction system, equipment, medium and program product, which have the beneficial effects.

Inventors

ZHU WENLONG
DAI JISHENG
SONG DONGLI
JIANG PING
ZHANG LEI
ZHAN YANHAO
Tang Lizhe
ZHANG XIAORUI
LIAO QISHU

Assignees

株洲中车时代电气股份有限公司

Dates

Publication Date: 20260508
Application Date: 20241107

Claims (10)

1. A method for predicting a polygon of a wheel of a train, comprising: The method comprises the steps of obtaining historical wheel data of a locomotive train and constructing a data sample corresponding to the historical wheel data, wherein the historical wheel data comprises turning data of a turning machine tool, detection data of a locomotive wheel pair fault dynamic detection system and daily mileage data of the train; dividing the data sample into influence factor sets, and carrying out correlation analysis on the influence factor sets to determine characteristic indexes that the correlation coefficient meets a set threshold value: taking the characteristic index as training data, and randomly sampling the training data to generate a training sample set; Training the training set based on a random forest algorithm to obtain a train wheel polygon prediction model; and obtaining a wheel polygon feature vector of the train to be predicted, inputting the wheel polygon feature vector into the train wheel polygon prediction model, and outputting a wheel polygon prediction result of the train to be predicted.
2. The method of claim 1, wherein obtaining historical wheel data for a locomotive train and constructing data samples corresponding to the historical wheel data comprises: acquiring historical wheel data of a locomotive train, and determining continuous characteristics, nominal characteristics and target variables in the historical wheel data; The continuous characteristics comprise the diameter of the wheel, the thickness machining deviation of the rim, the maximum value of the polygon of the front wheel set which is turned in last time, the machining deviation of the diameter of the wheel, the total driving mileage of the vehicle and the accumulated driving mileage after the last turning in last time, the nominal characteristics comprise the type of a carriage, the number of axles and seasons, and the target variable is whether the polygon fault of the wheel exists or not.
3. The method of claim 2, wherein performing a correlation analysis on the set of influencing factors to determine that the correlation coefficient meets a characteristic index of a set threshold comprises: and carrying out correlation analysis on the continuous characteristics in the influence factor set and the target variable by using a Pearson correlation analysis method so as to determine the continuous characteristics of which the correlation coefficient meets a set threshold.
4. The method of claim 1, wherein training the training set based on a random forest algorithm to obtain a train wheel polygon prediction model comprises: randomly extracting characteristic information of the training set to generate a plurality of random sample sets; Generating CART regression trees of a plurality of random sample sets based on a classification regression algorithm, and calculating the coefficient of the basis of the random sample sets; for a plurality of random sample sets where the current node of the classification regression tree is located, returning the classification regression algorithm to the sub-decision tree and stopping recursion when the coefficient of the basis of the plurality of random sample sets is smaller than a coefficient threshold value, or calculating the coefficient of basis of each feature information in the random sample set where the current node is located when the coefficient of the basis of the plurality of random sample sets is not smaller than the coefficient threshold value; Taking the feature information corresponding to the minimum coefficient of the coefficient of each feature information as optimal feature information, dividing a corresponding random sample set into a first feature set and a second feature set in the optimal feature information, and taking the first feature set and the second feature set as a left child node and a right child node of the current node respectively; Calculating the coefficient of the first feature set and the coefficient of the second feature set, returning the classification regression algorithm to the sub-decision tree and stopping recursion when the coefficient of the first feature set and the coefficient of the second feature set are smaller than the threshold value of the coefficient of the first feature set and the coefficient of the second feature set, generating T CART regression trees, and outputting a learning result; and taking the average value of the output values of the T CART regression trees as the output result of the random forest.
5. The method of claim 4, wherein randomly extracting feature information of the training set to generate a plurality of random sample sets comprises: Determining the quantity of feature information in the training set; randomly putting back and extracting a set number of samples from the training set to form a bootstrap sample; for each bootstrap sample, calculating a statistic of interest and obtaining an empirical distribution of the statistic; a training sample set is generated from the empirical distribution.
6. The method for predicting train wheel polygons according to claim 1, wherein training the training set based on a random forest algorithm, after obtaining a train wheel polygon prediction model, further comprises: Randomly sampling the characteristic information of the polygonal characteristic vector of the wheel for a plurality of times to obtain a sampling sample set, wherein the sampled characteristic information is put back into a test set for the next random sampling after each random sampling; constructing a prediction model based on the sampling sample set, and evaluating an accuracy score of the prediction model; adding the feature information corresponding to the prediction model when the accuracy score is highest into an optimal feature subset; And optimizing the train wheel polygon prediction model according to the optimal feature subset.
7. A train wheel polygon prediction system, comprising: The system comprises a data acquisition module, a data analysis module and a data analysis module, wherein the data acquisition module is used for acquiring historical wheel data of a locomotive train and constructing a data sample corresponding to the historical wheel data, and the historical wheel data comprises turning data of a turning machine tool, detection data of a locomotive wheel pair fault dynamic detection system and daily mileage data of the train; the characteristic analysis module is used for dividing the data sample into influence factor sets, and carrying out correlation analysis on the influence factor sets so as to determine that the correlation coefficient meets the characteristic index of a set threshold value: the training data sampling module is used for taking the characteristic index as training data, and randomly sampling the training data to generate a training sample set; The model training module is used for training the training set based on a random forest algorithm to obtain a train wheel polygon prediction model; the wheel prediction module is used for acquiring the wheel polygon feature vector of the train to be predicted, inputting the wheel polygon feature vector into the train wheel polygon prediction model, and outputting the wheel polygon prediction result of the train to be predicted.
8. An electronic device, comprising: A memory for storing a computer program; A processor for implementing the steps of the method according to any one of claims 1 to 6 when said computer program is executed.
9. A computer readable storage medium, characterized in that it has stored thereon a computer program which, when executed, implements the steps of the method according to any of claims 1 to 6.
10. A computer program product comprising a computer program which, when executed, implements the steps of the method according to any one of claims 1 to 6.

Description

Train wheel polygon prediction method, system, device, medium and program product Technical Field The application relates to the technical field of public transportation safety, in particular to a train wheel polygon prediction method, a train wheel polygon prediction system, train wheel polygon prediction equipment, train wheel polygon prediction medium and a train wheel polygon prediction program product. Background With the continuous increase of the running speed of the high-speed train, the interaction between the wheel rails is stronger, and a series of problems such as fatigue stripping, abrasion, polygonal abrasion of the wheels and the like of the tread of the wheels are more easily caused. The problem of wheel polygon is found in various types of high-speed locomotives by each vehicle application unit during maintenance. Meanwhile, the polygonal abrasion of the wheels can also aggravate the interaction of wheel rails, change the dynamic performance of the vehicle, generate high-frequency vibration and noise in the high-speed running process of the vehicle and reduce the comfort of passengers. Currently, the main judging method is mainly focused on working experience of maintenance personnel to verify a future working state of the wheel. However, the subjective performance of the manual experience method is strong, the prediction of high-precision data is basically not met, the current research on the polygon and the formation mechanism of the wheel is still insufficient, the polygon abrasion of the wheel can be caused due to various reasons, and the related maintenance departments can not carry out turning repair operation on the corresponding wheel in time, so that the problem of condition-based maintenance is particularly remarkable, the service life of the wheel is greatly shortened, and the rail traffic operation and maintenance cost is increased. Meanwhile, the train running safety is seriously affected by the polygonal wheels. Disclosure of Invention The application aims to provide a train wheel polygon prediction method, a train wheel polygon prediction system, train wheel polygon prediction equipment, a train wheel polygon prediction medium and a train wheel polygon prediction program product, which can remarkably improve prediction accuracy by constructing a train wheel polygon prediction model for abrasion prediction. In order to solve the technical problems, the application provides a train wheel polygon prediction method, which comprises the following specific technical scheme: The method comprises the steps of obtaining historical wheel data of a locomotive train and constructing a data sample corresponding to the historical wheel data, wherein the historical wheel data comprises turning data of a turning machine tool, detection data of a locomotive wheel pair fault dynamic detection system and daily mileage data of the train; dividing the data sample into influence factor sets, and carrying out correlation analysis on the influence factor sets to determine characteristic indexes that the correlation coefficient meets a set threshold value: Taking the characteristic index as training data, and randomly sampling the training data to generate a training set; Training the training set based on a random forest algorithm to obtain a train wheel polygon prediction model; and obtaining a wheel polygon feature vector of the train to be predicted, inputting the wheel polygon feature vector into the train wheel polygon prediction model, and outputting a wheel polygon prediction result of the train to be predicted. Optionally, obtaining historical wheel data of the locomotive train and constructing a data sample corresponding to the historical wheel data includes: acquiring historical wheel data of a locomotive train, and determining continuous characteristics, nominal characteristics and target variables in the historical wheel data; The continuous characteristics comprise wheel diameter, rim thickness machining deviation, maximum value pre-measurement of a polygon of a front wheel set which is turned in last time, wheel diameter machining deviation, total vehicle mileage and accumulated running mileage after turning in last time, wherein the nominal characteristics comprise carriage type, axle number and seasons; optionally, performing correlation analysis on the influence factor set to determine that the correlation coefficient meets the feature index of the set threshold includes: and carrying out correlation analysis on the continuous characteristics in the influence factor set and the target variable by using a Pearson correlation analysis method so as to determine the continuous characteristics of which the correlation coefficient meets a set threshold. Optionally, training the training set based on a random forest algorithm, and obtaining the train wheel polygon prediction model includes: randomly extracting characteristic information of the training set to generate a plurality of random