Rolling bearing fault diagnosis based on linear prediction - time shift filtering and deconvolution

Aiming at the problems of the early-stage periodic transient impact of rolling bearings in asynchronous motors being not obvious and the low signal-to-noise ratio of fault signals, a fault diagnosis method based on the combination of linear prediction (LP) and time shifting (TS) filtering, as well as the extraction of multipoint kurtosis (MK) from the output signal of multipoint optimal minimum entropy deconvolution adjusted (MOMEDA) is proposed. Firstly, the instantaneous power signal is subjected to linear prediction and time shifting filtering to eliminate the noise of the power supply fundamental frequency and suppress the noise of eccentric harmonics. Then, deconvolution operation is performed on the fault signal to extract the fault period and enhance the fault. Finally, the multipoint kurtosis diagram is used to identify the fault frequency band of the deconvolution output signal, and the square envelope analysis is conducted on the frequency band to extract the fault characteristic frequency.