Adaptive Variational Mode Decomposition for Bearing Fault Diagnosis

Variational Mode Decomposition (VMD) can break down a signal into its intrinsic modes based on the frequency characteristics of the signal, but it suffers from the problem that the decomposition results are affected by preset parameters. This paper proposes an Adaptive Variational Mode Decomposition for Bearing Fault Diagnosis (EO-VMD) to solve the problem. Firstly, to achieve adaptability, composite evaluation index (CEI) is constructed to quantify the performance of different parameter combinations. Secondly, the EO uses the CEI as its fitness function to iteratively find the best solution for the Volume Mass Balance model as the best set of parameters for the VMD. To verify the adaptability and decomposition accuracy, a simulated signal and the Case Western Reserve University bearing fault dataset are tested and compared with Ensemble Empirical Mode Decomposition (EEMD) and Genetic Algorithm-based VMD (GA-VMD). The results suggest that the EO-VMD algorithm can better extract the signal's effective components by adaptively selecting the number of modes and penalty parameters.