Adaptive range selection for parameter optimization of VMD algorithm in rolling bearing fault diagnosis under strong background noise

The optimized variational modal decomposition (VMD) algorithm is widely used in the diagnosis of rolling bearing faults. However, the subjectivity of the optimization range can compromise the effect of fault feature extraction under strong background noise (SBN). To enhance the fault diagnosis accuracy of rolling bearings under SBN, an adaptive range selection for parameter optimization of the VMD algorithm was developed. The proposed algorithm utilizes a method based on peak spectral clustering and center frequency to determine the optimal range of mode number and penalty factor. An optimization process based on the weighted kurtosis spectrum L2/L1 norm is then employed as the fitness function to obtain the optimal values of modes and penalty factor. Experimental results have demonstrated that the proposed method achieves a 20.02 % increase in fault diagnosis accuracy compared to the classical adaptive variational mode decomposition (AVMD) method.