Fault Feature Extraction Method of Rolling Bearing Based on Attemptable Variational Mode Decomposition

Aiming at the problem of difficult parameter selection in variational mode decomposition (VMD) when extracting weak fault features of bearings, an attemptable variational mode decomposition (AVMD) method is proposed. This method ascertains the penalty coefficient based on the center frequency of each mode, and subsequently employs the pulse kurtosis index as the standard to measure the information content of mode features. The number of modes is determined through an iterative process. Finally, the decomposed main mode envelope is demodulated to complete the extraction of fault feature. By applying this method and empirical mode decomposition (EMD), ensemble empirical mode decomposition (EEMD) and VMD, a comparative analysis concering the fault signals of the outer ring, inner ring and rolling element of the bearing is conducted. The research results indicate that this method can adaptively determine the number of modes and penalty coefficients, and achieve the extraction of weak fault features of the bearing. Compared with EMD and EEMD, the amplitude of the characteristic frequency spectral lines is doubled, and the computational efficiency is also doubled. The proposed method is sensitive to the weak features of bearing with faults, thereby enhancing the reliability and adaptive of VMD in decomposing signals.