Study On Intelligent Diagnosis Technology Of Bearing Based On Wavelet-Envelope Demodulation And Artificial Neural Networks

The rolling bearing is one of the most commonly used part in the mechanical system, so the diagnosis of the rolling bearings is very important. When some parts of the rolling bearings 'surface are damaged, the bearings would collide the other parts periodically with the result that the periodic impulses come into being. The frequency band of the impulse is so broad that it will overcast each connatural frequency of the bearing system, so the impulse will arose each connatural vibration of the bearing system as ideal impulse. So the stable signal turn to be unstable signal. The Fourier transform is localized completely in frequency area, but it can not provide local information c in time area, so it is not suitable for processing stable signal. Having a multi-scale resolution in time and frequency, the Wavelet transform is endowed with the indisputable hegemony in signal processing especially for unstable signal.Artificial neural network(ANN)is widely applied because of its advantages recent years. There are three most remarkable excellence in ANN :Firstly it is nonlinear, secondly it is constructed paralleled,thirdly is its ability of studying and generalizing. In the same time ,ANN is easy to be realized because ANN is consist of a lot of simple neural cells and can solve those problems which are hard to be solved by analytics directly. It is suitable to introduce ANN to rolling bearing fault diagnosis because of its advantages.Aiming at the fault diagnosis of the rolling bearings based on wavelet-envelope demodulation, this paper would devote most of its efforts to the following:1,Introducing systematically the vibration mechanism of the rolling bearings and the vibration character of typical fault;2,Expatiating the essential theories and the applications in signal processing of the Wavelet transform;3,Expatiating the theory of the wavelet-envelop demodulation and wavlet-packet frequency band energy technic, and diagnosing the typical faults of the rolling bearings respectively through the two technics. Then extracting the characteristics vectors.4,Inputting the two kinds of the characteristics vectors to the BP neural network , training the net and diagnosing the modes, the result shows that the characteristics vectors of the wavelet-envelop demodulation technic could reflect the type of the fault more effectively.