Abstract: Aiming at the non-stationary and nonlinear characteristics of rolling bearing vibration signals and the difficulty in feature extraction of early fault signals, the feature extraction methods commonly used in rolling bearing condition monitoring were studied, and a feature extraction method based on multivariate variational mode decomposition (MVMD) and fractional Fourier transform (FRFT) was proposed and applied to fault diagnosis of rolling bearings. The multi-channel vibration signals collected by multi-sensors were decomposed synchronously by using MVMD algorithm, the ability of multi-channel data fusion was effectively improved, and several intrinsic mode function (IMF) components were obtained at the same time. According to the correlation coefficient method, the component containing the most fault information was selected as the optimal component from the decomposed IMF component, and the optimal component was filtered by FRFT to reduce the interference of noise to the weak fault signal. The filtered signal was demodulated by 1.5-D envelope spectrum, and the fault features were extracted by analyzing the envelope spectrum of the filtered signal. The research results show that the combination of MVMD and FRFT can effectively avoid modal aliasing, and it makes full use of fault feature information, meanwhile it can weaken the interference of low frequency signal and noise, and extract fault feature information effectively.
Key words:  rolling bearing; feature extraction; multivariate variational mode decomposition(MVMD); fractional Fourier transform(FRFT);intrinsic mode function(IMF)

HONG Da, MA Jie, ZHAO Xi-wei.Fault feature extraction of rolling bearing based on MVMD and FRFT[J].Journal of Mechanical & Electrical Engineering,2021,38(10):1284-1291.