Performance Degradation Assessment of Rolling Bearing Based on Difference of Eigenvalues in Random Matrix Theory

The difference of eigenvalues index based on Random Matrix Theory (RMT) was proposed to overcome the problem that traditional feature extraction methods hardly extract effective feature when the sample data is large and severe noise in signal, which further affects the accuracy of bearing state detection. Firstly, a technique based on matrix randomization is proposed, which can construct a large dimensional random matrix that meeting the requirement of random matrix theory through matrix randomization processing. Secondly, the convergence characteristics of eigenvalues and the difference between matrix eigenvalues in Marchenko-Pastur (M-P) law based on random matrix theory can achieve noise reduction, and a rolling bearing eigenvalue difference index is proposed to reduce noise interference. Finally, rolling bearing full life data collected by Intelligent Maintenance System (IMS) was used for application research. The experimental results show that: the feasibility and effectiveness of M-P law in the field of bearing state monitoring can be verified, and the proposed index not only can detect the abnormal occurrence in advance, but also can accurately describe the degradation process of rolling bearing.