Rolling bearing fault diagnosis method based on improved morphology VMD-TCN

In order to fully utilize the vibration signals for fault identification, a rolling bearing fault diagnosis method based on improved morphological VMD-TCN is proposed. Firstly, the rolling bearing vibration signals are preprocessed by the improved morphology filtering method Average Combined Difference Morphological Filtering (ACDMF) to retain the useful signals; secondly, the signals are decomposed by the Variational Mode Decomposition (VMD), and the optimal number of decomposition of IMF components is determined according to whether or not the frequency center of IMF components is over-decomposed, and the optimal number of decomposition of IMF components is determined by the crag evaluation index. The IMF component with rich fault information is selected by the crag evaluation index, and nine features of mean, variance, peak, crag, RMS, peak factor, pulse factor, waveform factor, and margin factor are calculated; finally, the TCN model processes these attributes for training and prediction, and ultimately yields the accurate fault diagnosis rate. Using the rolling bearing vibration signals from Western Reserve University to verify the described method, the method in this paper is able to extract effective fault features and can accurately identify rolling bearing fault states.