In the operation of urban rail transit,rail corrugation occurs,which is particularly severe in sections with egg fastener.To predict the development of rail corrugation in shear-type fastener sec-tions,a 16-month tracking test was conducted on a certain subway line,and the third-octave frequency spectrum of rail corrugation,the peak-to-peak moving average value of wave depth(PPR),and its development pattern were analyzed.The test results indicate that PPR has good repeatability in the development pattern of corrugation and can be used to predict the trend of rail corrugation development.However,due to the complex mechanisms behind corrugation formation,it is difficult to obtain a uni-versal prediction method based on traditional mechanical mechanism characteristics.Based on the theory of machine learning,a rail wear prediction method using an improved polynomial expansion linear regression model is proposed.The results of the study show that the proposed prediction method has a root mean square error of 4.246 μm in predicting the moving peak-to-peak value average(PPR)and an over-limit rate error of 5.42％,indicating that the improved polynomial expansion linear regression model has high accuracy and predictive capability in rail corrugation forecasting.
维普
万方数据