Research On Fault Feature Extraction Method Of Locomotive Bearing Based On Empirical Wavelet Transform

With the continuous development of the railway,the safety performance of locomotives becomes more and more important.As the key moving parts of locomotives,the bearings health status will directly affect the operation of locomotive.Therefore,it is of great significance to study the fault diagnosis of locomotives bearing.In order to extract the fault information of locomotive bearing effectively,this paper studies the bearing fault diagnosis method based on empirical wavelet transform.The main research work is as follows:Wavelet transform,wavelet packet transform and empirical mode decomposition(EMD)are studied,and those three methods are analyzed by numerical experiments.The frequency band division in the wavelet transform and wavelet packet transform process is independent of the signal itself,and it is related to the sampling frequency and the number of decomposition layers of the signal.Their decomposition results are related to the selected wavelet function,and the construction of the wavelet function is independent of the signal itself.At the same time,their wavelet basis functions need a priori assumption,which makes them lack self-adaptability.The EMD method decomposes signals based on the time domain characteristics of the signal itself,and does not require any basic functions to be defined in advance.The basis function and frequency band division of this method are entirely determined by the characteristic scale of the signal itself,and its decomposition is adaptive,but it lacks a complete theoretical basis,and there are also some defects such as over-decomposition.The fault diagnosis method of locomotive bearing based on Empirical Wavelet Transform(EWT)is studied.The boundary of the divided frequency band is calculated by the frequency corresponding to the local maximum value of the Fourier spectrum of the signal,thereby the analysis frequency bands of the signal are divided.Then the orthogonal wavelet filter banks are constructed based on the sub-bands to decompose the signal adaptively to obtain the fault characteristic components.However,when EWT is applied to decompose signals,the division of frequency bands is easily disturbed by noise components,which results in errors in the division results.In order to solve the problem that EWT band division is susceptible to noise interference,an improved EWT method based on signal spectrum envelope is studied.Firstly,the cubic spectrum is used to calculate the envelope of the signal spectrum in the background of strong noise to weaken the interference of the noise frequency domain peak to the frequency band division,and then select the frequency corresponding to the local maximum of the spectrum envelope to calculate the frequency band.Based on the boundary of the frequency band division,an orthogonal wavelet filter bank is constructed to decompose the signal.The improved EWT method based on spectrum envelope is applied to extract fault features of roller bearings,and the bearing fault features are effectively extracted.The improved EWT method based on the signal spectrum envelope is prone to overshoot,undershoot and incomplete envelope when interpolating the signal spectrum,which results in local extremum or no extremum.Aiming at the above deficiencies,an improved EWT method based on time-frequency kurtosis spectrum is studied.The time-frequency kurtosis spectrum of the signal is obtained by time-frequency transform,and the frequency corresponding to the local minimum value of the kurtosis spectrum is taken as the boundary of the sub-band,and then the orthogonal wavelet filter bank is constructed to decompose the signal by EWT.This method has been applied to the fault extraction of locomotive running bearing and achieved ideal results already.