The Research And Applicaton Of Time-Frequency Analysis Method In Fault Feature Extraction Of Gearbox

The purpose of this article is comparing the effect of time-frequency analysis methods in gear and rolling bearing fault feature extraction,finding appropriate feature extraction methods.To this end,establishing a gearbox failure experiment device,simulating different gear failures and rolling bearing failures,introducing the characteristics of the frequency distribution of gear and rolling bearing and calculate methods.First through the time domain and frequency domain to compare the parameters of the normal and fault signals.And through the envelope demodulation analysis and cepstrum analysis to extract the feature of the fault signals,through analysis obtain that envelope demodulation analysis by using hilbert transform can extract the fault signal from the high-frequency modulation signal,thereby avoiding the interference of low-frequency signal,obtaining a good result to the modulation signal demodulator.Cepstrum can.simplify the original spectrum to the single line of cepstrum through the fourier transform,so that the spectrum of the complex cycle components become clear.for analysing the complex signals which has multi-frequency components, can find the problems which are difficult to identify in the power spectrum, achieving better results.In the practical applications,a large number of signals are nonlinear and stochastic signals.Because the fault signal generating from the research gearbox is based on non-random,the artical will against the signal of gearbox easily arise broken teeth fault in the course.In the location of changing speed,load,test points,using short-time fourier transform and Choi-williams distribution to extract feature.Comparing the effect of various factors'influence,obtaining the effect using this methods to extract the characteristics of the gear fault information is visual,but the results were not satisfactory.This artical also choose broken teeth, tooth wear,rolling bearing outer ring and inner ring scratches faults,using wavelet transform and EMD decomposition methods to break it down,and through the envelope spectrum and the marginal spectrum to extract feature,then comparing the effect of two methods.For the gear failure,using wavelet decomposition and envelope spectrum for feature extraction achieved good results, using the EMD decomposition and marginal spectrum not achieved good effect.But for rolling bearing outer ring and inner ring scratches fault,using the above methods for feature extraction, results are not satisfactory. Against the cross-term issue of Wigner-ville distribution existing for multi-component signal,using the method of based on EMD'sWigner distribution to verify the digital simulation vibration signal,obtaining for inhibiting the cross-term this method having good effect.At the same time,based on using the local hilbert marginal spectrum method for roller bearing inner and outer ring scratch fault signal feature extraction,achieveing fairly good results.