Study On The Automatic Detection Of Bearing Fault Based On Time-frequency Analysis

With development of the science and technology, machines are developing in the direction of larger size, higher speed and more complicated structure, and machine fault will bring more serious effect, so fault diagnosis is becoming more and more important for industry, and has got great development in both theory and practice. The development of science and technology, especially the fast growth of processing technology of non-stationary signal provides potential method for the machinery fault diagnosis.This thesis mainly concerns the application of the fault diagnosis method based on time-frequency analysis in mechanical part fault diagnosis. The concept and property of STFT,WVD and CWT were expatiated via theoretical analysis and simulation and the CWT is found to be better for mechanical localized fault diagnosis. Based on the CWT representation of vibration signal, an automatic detection algorithm for signal transients is proposed, through which the periodical feature about the localized fault is overlapped and thus enhanced in the polar diagram and the feature about signal noise and about non-periodical components is de-noise oppositely. The application of the proposed algorithm in cone bearing localized fault shows that this method is rather effective in detecting the periodical transients for mechanical fault diagnosis. Based on the CWT representation, a mechanical fault diagnosis method through the reassignment of CWT coefficients, and the application of this method in cone bearing show effectiveness in representing the transient feature about the local fault in the time-scale plane.