Rolling Bearing Fault Diagnosis Based on Synchroextracting Transform Under Variable Rotational Speed Conditions

Vibration signal processing methods, most of which are appropriate for the bearings under constant rotational speed or speed with little fluctuation, are the fault diagnosis methods of rolling element bearing with the widest use. While under normal operating conditions, bearings often run under variable rotational speed conditions especially when the machine starts and stops or there is a speed fluctuation with bearing fault. Under such circumstances the bearing signal is non-stationary, which makes the traditional bearing fault diagnosis methods inapplicable. Thus, rolling bearing fault diagnosis based on synchroextracting transform is proposed. Spectral kurtosis is used to find the optimal frequency band, based on which a band pass filter is designed. Then the synchroextracting transform is applied to the band-pass filtered signal to get the time-varying frequency demodulated spectrum and extract the fault features. Through the time-varying frequency demodulated spectrum, rolling element bearing fault can be diagnosed without resampling. The method proposed is illustrated by numerical simulated signal analysis, and is further validated via lab experimental rolling bearing vibration signal analyses.