Weak Fault Feature Extraction Method of Rolling Bearings Based on MVO-MOMEDA Under Strong Noise Interference

Abstract

Aiming at the problem that the weak information of rolling bearing fault features in a strong background noise environment, and the filter length and fault period of important parameters in multipoint optimal minimum entropy deconvolution algorithm (MOMEDA) depend on human experience selection. This article proposes a rolling bearing weak fault feature extraction method based on multiverse optimization algorithm (MVO) optimized MOMEDA under strong noise interference. First, establish a new index of multiobjective optimization, the peak factor of envelope spectrum is taken as the fitness value, and use the powerful global search ability of MVO to select the best parameter combination of the MOMEDA method adaptively. Second, the weak fault signal is enhanced by the MOMEDA algorithm. Finally, the enhanced signal is decomposed using the ensemble empirical modal decomposition (EEMD), and the fuzzy entropy feature set is constructed, which is input to the support vector machine (SVM) for classification and identification. To verify the feasibility of the method in this article, the rolling bearing data from Case Western Reserve University and the drivetrain dynamics simulator (DDS) testbed were selected for comparison experiments. The experimental results show that compared with minimum entropy deconvolution (MED), maximum correlation kurtosis deconvolution (MCKD), and MOMEDA, the classification accuracy of the proposed method increased by 5.36%, 16.82%, and 13.45%, respectively. Compared with particle swarm optimization algorithm (PSO) and fireworks algorithm (FWA), the MVO algorithm has faster convergence speed and stronger stability when optimizing MOMEDA problems. Even under strong background noise, it still has high accuracy.