Neural Network-based High-speed Train Wheel Set Size Prediction And Repair Strategy Research

With the rapid development of high-speed railways,high-speed rail is an important link for urban rail transportation,and any minor faults on train components may affect the operation safety of high-speed trains.The wheelset is a key part of the train operation,and it is very easy to wear during the operation process.The maintenance and repair work such as timely detection,repair or replacement of the state accounts for a large proportion of the cost.Healthy wheelset status is an important condition to ensure the safe and stable operation of the train,and can maintain the passenger’s riding comfort while extending the service life of the wheel rail.Therefore,predicting the value of the wheelset hidden danger index to ensure that the size information of the wheelset is within the safe use range is conducive to the railway management department to formulate effective maintenance measures in a timely manner and improve the safety,comfort and economy of the train.The main research object of this thesis is EMU wheelsets,which are analyzed and researched on a large number of on-site measured data.On this basis,an adaptive differential evolution least squares back propagation neural network(ADE-LMBP)prediction model for wheel size data is established,and a wheelset wear model(including wheel flange thickness wear model and wheels)is established.Diameter wear model),based on the wear model,a single-wheel repair strategy optimization goal is proposed,and a better repair strategy proposal is given through the Monte Carlo simulation method.The specific work is as follows:First of all,this paper conducts preliminary data analysis on the change of the shape parameters of all wheels of CRH2 A EMU with the train operation.Taking the wheel diameter value and the rim thickness value as prediction objects,an ADE-LMBP optimization algorithm is proposed,in which the adaptive differential evolution algorithm(ADE)is used to optimize the initial weights and thresholds of the LMBP model.In order to test the prediction performance of the ADE-LMBP model and combine the requirements of the size prediction of the moving wheels,a simulation experiment of a univariate nonlinear function was designed.By comparing with the prediction results of BP neural network,LMBP neural network and DE-LMBP model,the rationality of applying ADE-LMBP prediction model to the wheel set size data is verified.In addition,based on the numerical analysis of a large number of measured data of CRH2 A EMUs,a rim thickness wear model and a wheel diameter wear model were established respectively.In the optimization study of single-wheel repair strategy,the optimization target is set to the longest service life of the wheel and the minimum number of repairs,and the constraint condition is set to the wear limit of the wheel diameter and rim thickness.The Monte Carlo simulation method is given five kinds of best repairing schemes effectively extend the service life of the wheelset.