Study On Detection And Diagnosis Method Of Wheel Polygonization Of High-Speed Trains

With the increase of the speed of high-speed trains,the dynamic interaction between wheel and rail is intensified,and polygonal wear of wheel treads is a common phenomenon.The impact and vibration caused by wheel polygonal wear will cause damage of wheel-rail systems and affect the stability and safety of train operation.Therefore,it is very important to study the detection and recognition method of wheel polygonization.According to the non-stationary characteristics of the wheel vibration signal,a modified ensemble empirical mode decomposition(MEEMD)method is introduced into the analysis of the wheel vibration signal,and the time-frequency characteristics of the components of the signal are researched.The main components are selected and their characteristics of entropy are analyzed and extracted.It is proposed that the wheel polygonization recognition method based on MEEMD entropy and GA-SVM,and the experimental analysis is carried out.The main research work of this paper is as follows:(1)Experimental data acquisition and the analysis and processing of the vibration signal are proposed.The existing detection methods are analyzed.Through field experiments,the vertical vibration acceleration signals of the axle box of two types of wheels with normal wheels and polygonal wheels are collected.According to the non-stationary characteristics of the wheel vibration signal,the MEEMD method having an advantage in the suppression of mode mixing is used to decompose the signal into several IMF signals.In time domain,the vibration of the polygonal wheel has obvious impact,while the vibration of the normal wheel is stable.Envelope spectrum analysis and spectrum analysis are carried out for each IMF component.In the envelope spectrum,the spectral peak of polygonal wheel is the same as the wheel rotation frequency,which is different from the frequency characteristic of the normal wheel.In the spectrum,the first order component of the polygonal wheel includes frequency doubling of the wheel rotation frequency obviously,through which the order number of the dominant wheel polygonization can be obtained.(2)The selection of the main components of the signal and the analysis and extraction of the characteristics of entropy value are proposed.In order to select the components which have greater correlation with the fault from several IMF components,the main components are selected based on the kurtosis value and the correlation coefficient of the component.From the calculation results of the experimental data,this method can find the IMF component that contains the main fault information.According to the characteristics of permutation entropy,which can reflect the complexity and dynamic mutation of signals,the partial mean of multi-scale permutation entropy is used as the characteristic of wheelvibration signals.The calculation and analysis of experimental signals show that the entropy values of two kinds of signals are different,and the feature vectors constructed by their partial mean can reflect the fault feature information.(3)The diagnosis method of the wheel polygonization fault and its model are proposed.The genetic algorithm support vector machine(GA-SVM)is used as the recognition algorithm to diagnose the wheel polygonization fault.The feature vectors of two types of signals are input into GA-SVM.After training and testing,the recognition accuracy is up to97.5%.The effectiveness and superiority of the proposed method is verified by comparison with the recognition results of other recognition methods.Finally,a graphical user interface(GUI)is used to construct the model of the wheel polygonization recognition method,which improves the practicability of the method.