Evaluation of bearing performance degradation based on MMFE and extensible k-medoids clustering algorithm

Traditional bearing performance degradation evaluation is often qualitative analysis, and vertical vibration signals are mostly taken as the study object to ignore the correlation among vibration information in different directions. Here, the multivariate multi-scale fuzzy entropy (MMFE) for evaluating the complexity of multi-channel time series was introduced into feature extraction of bearing operation states to construct MMFE features for considering the correlation among bearing vibration information in different directions. Combining the k-medoids algorithm and the extenics theory, a quantitative evaluation model of bearing performance degradation was established. The clustering center was obtained with k-medoids clustering of bearing normal state samples. Boundary of extensible set was determined according to Euclidean distances between sample points and clustering center. Furthermore, extensible correlation function was used to construct the bearing performance degradation evaluation model, it was verified by using bearing life-cycle fatigue tests. The test results showed that the proposed method can effectively identify bearing early performance degradation and quantitatively evaluate bearing performance degradation degree. © 2022 Chinese Vibration Engineering Society. All rights reserved.