Detection Of Rail Corrugation Based On Time-Frequency Analysis And Data Mining

In recent years,high-speed railways have developed rapidly in my country,and various diseases in railway rails have gradually appeared.Rail corrugation is one of the railway diseases,which can cause non-smooth operation of high-speed trains and even affect the safety of trains.At present,the elimination of rail corrugation in our country is polishing.In severe cases,the rail needs to be replaced,therefore,timely and effective detection of rail corrugation is of certain significance to the safety of train operation.Based on the vehicle dynamic response data set,the method of combining time-frequency analysis and data mining for intelligent diagnosis is proposed in this paper.This article mainly focuses on vehicle dynamic response data set,through wavelet packet filtering processing,EEMD time-frequency decomposition,rough set attribute reduction and support vector machine for intelligent diagnosis.Firstly,the wavelet packet analysis method is used to filter out other information and noise mixed in the vehicle dynamic response signal.According to the frequency range where the corrugated signal is mainly concentrated,the wavelet basis function and the number of decomposition layers are determined.Signals in different frequency bands are obtained through wavelet packet decomposition,and the frequency bands of the main corrugated signal are reconstructed to obtain the filtered signal.Secondly,the ensemble empirical mode decomposition is performed on the signal after wavelet packet filtering to obtain a series of IMF components.In order to select the IMF components containing the main corrugation information,the correlation coefficient between the original signal and each IMF component is calculated,and finally select the component that contains corrugated information.Finally,the effective value and kurtosis value of the main information components are calculated as the characteristic indexes describing the rail corrugation,and a total of 8 characteristic indexes are obtained in combination with the vehicle dynamic response data set.Rough set attribute reduction is used to simplify the index,so as to further improve the classification accuracy,reduce the running time of the model,and combine with the support vector machine for intelligent diagnosis.The experimental results show that the classification accuracy of the method combining time-frequency analysis and data mining reaches 99.03%,which reduces the running time of the model,and the detection effect is better than before reduction.In order to verify the method more effectively,the classification detection method based on wavelet packet and the classification detection method based on ensemble empirical mode decomposition are used to classify and diagnose the rail corrugation,and compared with the method in this paper.The results show that the accuracy,recall and Fl_score of this method are higher than the other two methods,which further proves the effectiveness and feasibility of this method.