Automatic Fault Diagnosis for Rolling Element Bearings

Abstract

Vibration-based condition monitoring plays an important role in rolling element bearing maintenance. Based on features in bearing vibration signals, envelope analysis is very popular because of its effectiveness in bearing fault diagnostics. However, its effectiveness heavily relies on selection of the frequency band which has been accomplished manually. In this research, we develop an automated signal analysis procedure including frequency band selection and fault signature identification. Band selection is based on wavelet packet transform and signal energy decomposition. Wavelet packet transform decomposes the spectrum of a bearing vibration signal into finite frequency bands. Then Root Mean Square is applied to locate the band with the highest energy suitable for envelope analysis. Further, cepstrum analysis is employed to identify repetitive nature in the enveloped signal which is associated to bearing fault signature. The techniques developed are verified using experimental data from Bearing Data Center of Case Western Reserve University.