Study On The Fault Diagnosis Methods Of Rolling Bearing Based On Wavelet Transform And Empirical Mode Decomposition

The Rolling bearing is an universal part of all electromechanical equipment. Therefore, fault monitoring and diagnosis has been deeply researched in fault diagnosis. This paper explores its diagnosis algorithms of time-frequency analysis and pattern recognition based on wavelet transformation, empirical mode decomposition and AR model.1. Fault diagnosis of rolling bearing based on Morlet wavelet is studied. Firstly, the shape characteristic of Morlet wavelet is analysized. Secondly, the smoothness index (SI) method is given for rolling bearing fault diagnosis. After the Shannon entropy, kurtosis and smoothness index are comparative analysized in detail, resulting in a new algorithm seeking the best wavelet scale. Lastly, the algorithm is applied in detected signal.2. Wavelet packet method is applied in rolling bearing fault diagnosis. A new algorithm seeking the best node is proposed which has ideal effects after detected signal is analysized. It is based on the best base searching algorithm of Coifman and Wickerhauser and replaces the Shannon entropy by wavelet coefficients energy.3. The idea of using wavelet packet and Hilbert-Huang transform is put out in rolling bearing fault diagnosis. Signal after wavelet packet transform is decomposed by empirical mode decomposition method. Then the IMFs separations are reserved which have a larger correlation coefficient. When the spectrum analysis is finished, the fault feature frequency and multi-frequency are obtained. It can detects the fault feature frequency in all kinds of fault types and normal signals of rolling bearing.4. Fault pattern recognition of rolling bearing is studied. Wavelet packet transform, empirical mode decomposition and AR model method is used in pattern recognition. Wavelet packet de-noise must be executed before it. Then the signal is decomposed by empirical mode decomposition algorithm. Lastly the fault of sample signal is detected by the M-distance between them after the AR model is constructed. A new method named mutual correlation coefficient about weight coefficient allocation is proposed. It is proved that this algorithm can recognize the pattern of rolling bearing correctly.Finally, the obtained results are summarized and future work is addressed.