Performance Degradation Prediction of Rolling Bearing based on KJADE and Holt–Winters

摘要

A performance degradation prediction method is proposed in this paper for condition monitoring and bearings performance degradation prediction. This method is the combination of kernel joint approximate diagonalization of eigen-matrices (KJADE) and Holt–Winters. First, the vibration signals acquired from running bearing are processed through multi-domain features extraction. An optimal feature set was obtained from the multi-domain features through dimensionality reduction and feature fusion using the KJADE algorithm. Then, the between- and within-class scatters were calculated to acquire the performance degradation indicators. Finally, the performance degradation pre- diction model based on Holt–Winters was established to predict the bearing performance degradation. Results show that bearing degradation trend can be effectively identified by the proposed method. Moreover, the prediction accuracy of this method is higher than that of extreme learning machine (ELM).