Fault feature extraction for rolling bearings based on parameter-adaptive variational mode decomposition and multi-point optimal minimum entropy deconvolution

Extracting fault feature is hard to realize because of weak fault impact components and environmental noise interference in vibration signals. Thus, a hybrid fault diagnosis method based on parameter-adaptive variational mode decomposition (VMD) and multi-point optimal minimum entropy deconvolution (MOMEDA) is proposed. Firstly, whale optimization algorithm (WOA) is employed to solve VMD parameter selection problem. Then a series of modes are obtained by parameter-adaptive VMD. Secondly, the effective modes whose index values are greater than the average index value are selected for reconstruction to enhance the impulse related to fault characteristics. Finally, periodic pulse signal is extracted from the reconstructed signal by MOMEDA. Fault characteristic frequencies can be identified from envelope spectra. The proposed method is verified to be effective based on two different experimental datasets. Moreover, the comparisons with fast kurtogram, ensemble empirical mode decomposition (EEMD) and the other latest methods further highlight its superiority of fault feature extraction.