| In recent years,the operating mileage of my country’s high-speed railways has increased year by year,and the operating speed has also increased steadily.High-frequency driving leads to various diseases of the rails,and rail grinding is an effective means to solve the diseases,but whether the rail grinding is reasonable will directly affect the safety and stability of the train operation,so more refined rail grinding standards are particularly important.Based on the analysis of domestic and foreign rail grinding standards and related research,this paper establishes the dynamic model of high-speed EMU,and uses software such as Python and SIMPACK to conduct co-simulation.Based on the measured wheel-rail data of high-speed railway,parameters such as wheel-rail contact state,contact point distribution,and vehicle dynamic performance are obtained through simulation,so as to study the influence of rail profiles with different deviation values in the rail head area on running vehicles.The main results and conclusions achieved are as follows:(1)Firstly,the phenomenon that the profile data of rails collected on the spot will be deviated due to interference is analyzed.Taking the rail profile with bump data as an example,three commonly used filtering methods,namely Gaussian filtering,cubic spline interpolation,and moving average,are used to process the data.And the root mean square error(RMSE)is used to select the processing method with better smoothing effect on the rail profile data.After data comparison and image observation,it is concluded that the Gaussian filtering method has a better smoothing effect on the rail profile data.(2)In view of the current situation of rail profile data collection in my country,in the process of rail alignment,the rail head is used as the basis,and according to the structural characteristics of the rail head,the oblique lines on both sides of the rail head and the rail apex are selected as alignment standards.And use the minimum area method to solve the angle that needs to be rotated during the alignment process,this alignment method can show a better alignment effect.(3)In order to obtain the rail profiles with different grinding deviations at the same lateral position of the rail head,the currently widely used data clustering methods K-Means clustering,BIRCH clustering,Gaussian mixture model clustering,K-Medoids clustering to cluster the grinding deviation curves.The elbow rule and the silhouette coefficient method can well solve the problem that the clustering algorithm needs to specify the number of classification clusters.With the help of the commonly used clustering algorithm evaluation indicators CH index,DB index,Dunn index,and XB index,the K-Means cluster is finally obtained.This kind of algorithm has the best effect when dealing with the rail grinding deviation curve.(4)After the clustering algorithm obtains the clustering result,there is a problem that the result contains multiple qualified grinding deviation curves.Based on the distribution position of the wheel-rail contact points when each deviation curve corresponds to the rail profile and the wheel matching,with the help of Kendall correlation coefficient,Spearman rank correlation coefficient,and Pearson correlation coefficient,in the least squares distance method,arithmetic mean method,and cosine similarity method three The optimal algorithm is selected from among the representative curve selection methods.(5)In the form of MATLAB and SIMPACK co-simulation,from the aspects of vehicle operation safety,stability,stability,contact state,etc.,analyze the influence of the rail profile with grinding deviation in different areas on the vehicle operation state,and finally according to the impact The horizontal position of the rail head is roughly divided into three areas:-30mm~-24 mm,-24mm~+26mm,and +26mm~+32mm.And from the analysis of derailment coefficient,wheel load reduction rate,stability index,etc.,it is concluded that when the dynamic performance index exceeds the limit by 10%,20%,30%,40%,and 50%,the corresponding deviation limit of each area is obtained.value. |