Rolling Element Bearing Diagnosis Based on Signal Reconstruction from Edited Cepstrum

Abstract

A feature extraction scheme for rolling element bearing (REB) fault diagnosis by editing the cepstrum of the original vibration is introduced in this paper. In the presented approach, the order analysis technology is utilized to convert an even-time-spaced scaled signal to an even-angle-spaced signal by resampling the acquired signal. The discrete lines belonging to gears are removed by editing the cepstrum. Then, the signal is reconstructed from the edited cepstrum. Lastly, clear characteristic frequencies related with the faulty REB can be obtained by the envelope spectrum analysis. Simulation studies indicate that this method can be employed to detect the REB faults in a gearbox with a high accuracy by removing the strong gear vibration disturbances.