Empirical Wavelet Transform And Its Application In Bearing Fault Feature Extraction

The signal adaptive decomposition method can decompose complex signals into a series of simple components according to the characteristics of the signal itself,and is widely used in fault diagnosis to eliminate noise and extract fault features.In order to improve the ability of the empirical wavelet transform(EWT)to extract bearing fault features,this paper carried out a comparative analysis of adaptive signal decomposition methods,and achieved the following results:The decomposition method,decomposition component and decomposition ability of the commonly used adaptive signal decomposition method are compared and analyzed.It is proved that the empirical wavelet transform is superior to the empirical mode decomposition in terms of anti-modal aliasing and decomposition time,so it is more suitable for bearing vibration signal fault feature extraction.When EWT decomposes a signal,it must achieve a good effect by reasonable division of the Fourier spectrum.However,the actual fault signal is complex and variable,and the classic EWT does not clearly separate the fault features.Therefore,an improved empirical wavelet transform(IEWT)algorithm is proposed.The first,the cyclic envelope technique is proposed,and the "locmaxmin" method is used to segment the signal spectrum to obtain multiple mono-components.The second,the Pearson correlation coefficient is used to measure the correlation degree between adjacent components.And the mono-component with high correlation degree is merged as a new segmentation strategy.Through comparative analysis of fault simulation signals,it is proved that this method can segment spectrum more efficiently and reasonably.Applying IEWT to the actual bearing fault diagnosis,it is found that the fault feature is difficult to locate.In order to improve the efficiency of fault feature extraction,an IMF entropy value evaluation method is proposed to effectively select the IMF component obtained by IEWT decomposition.Through the operation experiments of test-bed,locomotive and wind turbine,the fault feature extraction method is accurately diagnosed in bearing faults in various mechanical systems.Summarizing the above research results,a bearing fault feature extraction method based on IEWT and Intrinsic Mode Function(IMF)entropy value evaluation is proposed and validated by industrial field experiments.