CEEMD And SK Combination Research In Motor Bearing Fault Diagnosis

Motor is the most widely used in the production of modern society a mechanical drive,its running state will directly affect the normal work of the whole production system,the rolling bearing is relatively weak link of the motor,its running state is often directly affect the normal performance of the whole machine.So the rolling bearing fault detection is particularly important.This article in view of the hard threshold wavelet function discontinuities and soft threshold function of the true problem, and then put forward the combination the wavelet threshold with EMD(empirical mode decomposition,EMD)decomposition.Aim at the existing modal alising problem and endpoint effect of EMD decomposition,this paper use of the rolling bearing fault by CEEMD(Complementary Collection Empirical mode decomposition,CEEMD) and the combined endpoint continuation decomposition.CEEMD decomposition and the combined endpoint continuation adding the symmetrical gaussian white noise to the signal,it can effectively solve the EMD decomposition method of modal alising problem and endpoint effect of EMD decomposition.In this paper,combine the wavelet threshold denoising with EMD decomposition or CEEMD decomposition with improved wavelet threshold denoising,then given the combinition implementation steps.First put signal into EMD decomposition or CEEMD decomposition,make the fault signal can be effectively decomposed into different frequency band,use of kurtosis index as the basis of the component selection,to conform the kurtosis index of the IMF(Intrinsic Mode Function,IMF) component refactoring,improved wavelet threshold denoising to reconstruct signal denoising.To test the superiority of the two methods,using Matlab to analyse experiment.Analysis results show that:the futher threshold of wavelet denoising method better than the traditional wavelet in SNR and MSE.To the CEEMD decomposition and the combination of spectral kurtosis diagnosis method is also better than EMD method,as the rolling bearing fault degree aggravating,The EMD decomposition cannot accurate reflection of rolling bearing fault characteristic frequency,but CEEMD decomposition and the method of spectral kurtosis combination can still be good to the bearing fault signal diagnosis.