Use of the correlated EEMD and time-spectral kurtosis for bearing defect detection under large speed variation

Abstract

Vibration-based diagnosis is in common use for the health monitoring of rolling element bearing (REB). This paper is concerned with a new defect detection method of the REB under large speed variation by using the correlated ensemble empirical mode decomposition (EEMD) and time-spectral kurtosis (TSK), which are accomplished in two phases. During the first phase, vibration signals are decomposed into intrinsic mode functions (IMFs) with the EEMD, which are then correlated with the original signal in order to determine the shaft IMFs, and hence the distinct instantaneous rotation frequencies (IRFs). During the second phase, the TSK is adopted to determine the fault IMFs, which are further used to reconstruct the fault signal. As a result, the non-stationary signal in the time domain is transformed into the cyclostationary signal in the angular domain with respect to the IRFs by resampling with equal angle increments. Simulations and experiments are carried out to validate the feasibility of the proposed method. It is shown that proposed method offers a potential improvement over the conventional short time Fourier transform and order tracking-based method.