ROLLING ELEMENT BEARING FAULT DIAGNOSIS USING COMPLEX GAUSSIAN WAVELET

This paper is focused on fault diagnosis of bearings due to localized defects i.e. spall on the bearing components, which is essential to the design of high performance rotor bearing system. The methodology proposed in this paper for fault diagnosis of rolling element bearings, utilizes autocorrelation of raw vibration signals to reduce the dimension of vibration signals with minimal loss of significant frequency content. Dimension of vibration signal is reduced to 10% with negligible loss of information. To extract most appropriate features from auto-correlated vibration signals and for effective classification of faults, vibration signals are decomposed using complex Gaussian wavelet. Total 150 signals of healthy and defective bearings at rotor speeds 250, 500, 1000, 1500 and 2000 rpm with three loading conditions are considered. 1-D continuous wavelet coefficients of these samples are calculated at the seventh level of decomposition (27 scales for each sample). Maximum Energy to Shannon Entropy ration criterion is used to determine scale corresponding to characteristic defect frequency. Statistical features are extracted from the wavelet coefficients corresponding to selected scales. Finally, bearing faults are classified using Support Vector Machine (SVM) method. The test results show that the SVM can be used efficiently for bearing fault classification. It is also observed that classification accuracy is improved by using autocorrelation.