Fault Diagnosis For Gear Based On Wavelet Analysis

Gear drive is one of the common methods of mechanical drive. Fault diagnosis of gear has many important economic and social profits. Traditional methods of fault diagnosis are appropriate for steady signal analysis, but they can not represent time-average characteristic of unsteady signal caused by local defect of gear completely. Theory of wavelet annlysis is appropriate for unsteady signal analysis. So the method of wavelet analysis has been become one of the most conspicuous in new diagnosis methods in recent years.For studying the characteristics of wavelet theory and the vibration signal of gear, This paper has made the following researches on the application of wavelet analysis to fault diagnosis of gear:(1) The math models for the vibration of gear and the vibration of gear fault have been built in this paper. The definition of wavelet and the characteristic of wavelet analysis were introduced also. A method, which can do time-frequency analysis for better analysis of fault signal is needed. Frequency band can be disassembled accurately by the wavelet analysis which has important significance in the field of the weak signal extraction.(2) Wavelet function, threshold value and the levels of disassembling influences the effect of wavelet de-noising. With analysis of simulation, this problem has been approached. After analyzing the experimental signal, we draw the conclusion that we should choose 'rigrsure' threshold value. The levels of disassembling is limited. Two functions named db4 and ciof5 are the more effective wavelet function of de-noising. (3) Applicating continuous wavelet transform and wavelet packet analysis, we extracted the feature of gear and compared with the result of both method. A conclusion that feature extracting base on wavelet packet ranks high resolution ratio was obtained also. And it can reflect the signal power distribution of different frequency band with the power spectrum of signal. So this is a more effective method of feature extraction.The data in the paper is obtained by means of the experiment of gear fault. The results indicate that the methods are appropriate for the local defect diagnosis of gear. A wide application will be reached in the future work.