Fault Diagnosis Method of Bearings Based on RandWPSO-VMD-MCKD

Abstract The RandWPSO-VMD-MCKD method is introduced to address the challenge of detecting early failures in rolling bearings, which are often masked by noise and difficult to discover. This approach combines Random Weight Particle Swarm Optimisation (RandWPSO) with Variational Mode Decomposition (VMD) and Maximum Correlated Kurtosis Deconvolution (MCKD) to improve the reduction of noise and to emphasise the frequency of faults. By using the RandWPSO algorithm to automatically optimise parameters in both VMD and MCKD, this method aims to effectively diagnose rolling bearing faults. The envelope crest factor Ec is selected as the fitness function to drive the optimisation process. First, the two parameters K and α in the VMD algorithm, which represent the number of decomposed modes and the penalty factor, respectively, are optimised using the RandWPSO algorithm. Next, the parameters L , T and M are optimised in the MCKD algorithm and MCKD is applied to deconvolve the best modal component to improve its impulsive characteristics. Finally, the deconvolved signal components are subjected to envelope spectrum analysis. The findings demonstrate that the methodology is capable of identifying preliminary deficiencies in bearing signals in the presence of background noise.