Application Investigation On Wavelet Analysis To Feature Extraction Of Reciprocating Machinery

Machinery monitoring and fault diagnosis is a comprehensive technology, which essentially is pattern recognition of machine operating conditions. The key issue is feature extraction and classification. Research on fault diagnosis for reciprocating machinery is of particular interest due to the wide application of reciprocating machinery in industry manufactures.This paper mainly studies the application of the wavelet method in the feature extraction of the non-stable signals in the fault diagnosis of reciprocating machinery. The main works are as follows:(1) The recent advances of the time-frequency analyzing method in the fault diagnosis of reciprocating machinery and their application in the fault diagnosis.(2) Study the wavelet analysis and wavelet-packet decomposition theory. By studying the performance of the wavelet basis, indicate that only selecting right wavelet basis corresponding signal feature to decompose and extract feature, can effectively recognize diagnosis information and apply the wavelet analysis to engineering field. Then, bring forward the conception of setting up the clinic of fault diagnosis.(3) Study the application of the wavelet denoising method in the fault diagnosis of reciprocating machinery. By simulative experimental and engineering application analysis, prove the wavelet denoising method can effectively restrain the noise contained in non-stable signals, and effectively reserve the fault information.(4) Using wavelet-packet decomposition in the fault diagnosis of reciprocating machinery, decompose the fault signal with wavelet packet decomposition, study the fault feature in the wavelet packet decomposition.(5) Developing the reciprocating machinery's feature analysis system based on virtual instrument technology, which can work well for the reciprocating machinery fault diagnosis.This paper studies the method mentioned with experiment, the results say: using the wavelet method in the denoising of signals, can filtrate the signal in different frequency scale, solved the problem that the lowpass filter can not separate the signal and the noise when their frequency scale are superposed. It can be used as a pre-processing method in the fault diagnosis of the reciprocating machinery. The wavelet-packet decomposition can decompose signal to different frequency segments, and reconstruct the signal in these segments, then effectively extract the fault information contained in non-stable signals by analyzing relative segments.