Spectral DCS-based feature extraction method for rolling element bearing pseudo-fault in rotor-bearing system

The pseudo-fault is a common phenomenon in a system, which is defined as the faulty feature of a healthy component caused by a defective component. Pseudo-fault signals, coupled with signals of both faulty component and healthy component, may lead to false diagnosis for healthy component. It is a challenging task to decouple pseudo-fault feature based on the single-channel pseudo-fault signal of rotating machinery. An issue is presented for feature extraction of the pseudo-fault signal of a rolling element bearing (REB) in a rotor-bearing system. To address the matter, a method is proposed for feature extraction of a REB pseudo-fault in a rotor-bearing system based on spectral degree of cyclostationarity (DCS). Firstly, a dynamic model of the rotor-bearing system is established to produce pseudo-fault signals, considering different deflection exponents and slippage of bearings caused by both outer race and inner race defects. The dynamic and time-frequency characteristics are then analyzed to show the amplitude modulation of pseudo-fault signals. The cyclostationary analysis is implemented for pseudo-fault extraction and verified by numerical simulation. The spectral DCS method is proposed to capture the cyclostationarity of filtered signals ordered by kurtosis, evaluated by the crest factor upon the spectral kurtosis gradient. An experimental system, under four assembly modes and three rotating speeds, is constructed based on a machinery fault simulator (MFS) to verify effectiveness of the proposed spectral DCS method. Experiment results show that quasi-cyclostationarity distribution with respect to kurtosis remains almost constant in the spectral kurtosis and degree of cyclostationary (SK-DCS) spectrogram and that the crest factor upon the spectral kurtosis gradient is closer to zero. Thus, the proposed spectral DCS method can be used as a tool of feature extraction for pseudo-fault in REBs by single-channel signal to avoid false diagnosis.