Study On Wheel Degradation And Re-profiling Strategy Of Railway Wagons With Multi-covariates

The increase in cargo volume and operating speed of railway wagons,as well as complex factors such as the various operating environment,have brought challenges to railway wagons maintenance.As the final force component of rail freight cars,wheels determine whether railway wagons can run safely and efficiently.The monitoring and detection technology of railway wagons in China is becoming more and more advanced,which improves the economy and operating efficiency of railway wagons.It is of great significance to analyze the degradation status of different wheels and to propose accurate and appropriate re-profiling strategies.Thus,this thesis collects railway wagons wheel maintenance data and various operating environment data,and conducts research on wheel degradation process and maintenance strategies.Firstly,covariates that affect the wheel degradation of railway wagons are seleted.After the normality test and the homogeneity of variance test,the Kruskal-Wallis test and the Mann-Whitney U test are used to screen out some covariates that are significantly related to wheel degradation from the all covariates.Further,stepwise regression is used to select the best combination of mileage variables and filtered covariates that can predict wheel degradation.For different wheel individuals,multilevel model and moving average method are employed to predict the covariate in the best combination,which is useful for degradation modeling.Secondly,degradation model that considers multiple covariates is established.Based on the support vector regression model and input variable data of the best combination,the wheel degradation is predicted,and the grid search and cross-validation are combined to optimize the parameters.The optimal model was compared with other algorithm models.In addition,the effects of different covariables on degradation modeling are analyzed.The actual covariate and mileage variable,the predicted covariate and the mileage variable are put into the support vector regression model respectively,and the predicted value of tread wear and the actual tread wear value are obtained.Through comparison,the validity of the multi-level model and moving average method to predict covariates and support vector regression to predict tread wear is verified.Finally,a re-profiling strategy that considers multiple covariates is established.In order to obtain the wheel life and take the wheel diameter difference as the maintenance index,the corresponding relationship between the amount of diameter degradation and the amount of tread wear is fitted and analyzed,and the BP neural network is used to model wheel repair rules and predict the diameter of the left and right wheels after reprofiling.Considering the tread wear and wheel diameter difference,the yellow line and red line warning mechanism of wheel health status is established.The whole train wheel re-profiling process considering the actual engineering limits is modeled to minimum unit mileage cost.Sensitivity analysis of model parameter is conducted to verify the stability of the proposed strategy.A comparative analysis of whether considering the covariate in the modeling process shows the economics and superiority of the proposed method in this thesis.