| China's rapid economic development makes people's demand for travel increase,so China's rail transit construction has also been developed rapidly.But the problem of train vibration caused by the development of high-speed,heavy and lightweight is becoming more and more prominent.The vehicle body vibration is a key parameter reflecting the train structure health status,the train vibration state and wheel-rail contact performance.Meanwhile it is an important basis for train body vibration reduction and evaluation of ride comfort.Therefore,in-depth study of the relevant parameters affecting the vehicle body vibration and accurate acquisition of the vehicle body vibration acceleration is the premise to improve the safety of the train operation,the stability of the vehicle equipment and the reliability of the ride comfort level.In this context,the following researches are carried out in this thesis:(1)Construction of rail vehicle dynamics model.Multi-body dynamics model of rail vehicle is established by SIMPACK software to study the coupling mechanism between the structures,and the vibration response of the vehicle body under the excitation of track irregularities was simulated.The purpose is to provide a basis for the subsequent research ideas of vehicle body vibration coupling factor analysis and modeling.(2)Using appropriate data preprocessing methods to improve data quality.There are some problems in the data set collected by GJ-5,such as missing value,wide difference in the value range of different parameters and large sample size.Multiple interpolation methods are adopted to deal with the problem of missing value,data normalization method is used to eliminate the influence caused by dimensional difference among variables,and irrelevant parameters in sample variables are removed according to prior knowledge to achieve data dimension reduction.(3)Analysis and screening of coupling factors of vehicle body vibration.Pearson correlation coefficient,Spearman correlation coefficient,Kendall correlation coefficient and maximum information coefficient(MIC)are used to quantificationally explore the correlation between vehicle body vibration acceleration and other detection parameters.According to the correlation coefficient,the low-correlation parameters are removed to achieve the dimension reduction of the sample,and 12 coupling factors affecting the horizontal vibration of the vehicle body,10 coupling factors affecting the vertical vibration of the vehicle body,and 12 coupling factors affecting the lateral vibration of the vehicle body are screened out.(4)Establish vehicle body vibration acceleration prediction models.Using single regression tree algorithm,time-series regression tree algorithm and bagged regression tree algorithm to establish the vehicle body horizontal,vertical and lateral vibration acceleration prediction models,respectively.R-squared,mean absolute error(MAE),mean square error(MSE)and root mean square error(RMSE)are used to evaluate the performance of the model,and also as the basis for adjusting the model parameters.(5)Experimental verification.Proof of validity of vehicle body vibration coupling factor analysis: the bagged regression tree model trained with a reduced amount of sample data is comparable to the bagged regression tree model trained under all detection parameters.Verify the advantages of the bagged regression tree algorithm: compare the training results of the bagged regression tree model with the predict vehicle body vibration acceleration established by the multiple linear regression model,support vector machine,and BP neural network training model.Verify the prediction effect of the bagged regression tree model: use three new data sets as the verification set of the model to obtain the predicted value of vibration acceleration.By plotting and calculating the fitting degree and error between the actual values and predicted values,it is verified that the bagged regression tree model has a good prediction effect with a prediction accuracy of 94%. |
PDF Full Text Request