关键词: 滚动轴承, 多尺度样本熵, 故障诊断, 贝叶斯网络, 算术优化算法

Abstract:

To address the issues of traditional MSE， such as the loss of feature information during the coarse-graining processes and the difficulty in extracting feature information from fault signals at large scale factors， a method for rolling bearing fault diagnosis was proposed based on improved refined composite multiscale sample entropy（IRCMSE） combined with an AOA optimized Bayesian network. The traditional coarse-graining processes， which involved averaging， were replaced with a cross-sampling method to obtain time series at each scale， and the method of calculating entropy values at different scales was changed to extract feature information from the time series. IRCMSE was used to extract the fault feature information of rolling bearings， forming fault feature samples. These fault feature samples were then input into the Bayesian network model optimized by AOA for fault recognition. The improved method is experimentally compared with fault diagnosis methods based on MSE， multiscale dispersion entropy， and refined composite multiscale sample entropy （RCMSE）， verifying the feasibility of the proposed method and demonstrating a higher fault recognition rate.

Key words: rolling bearing, multiscale sample entropy （MSE）, fault diagnosis, Bayesian network, arithmetic optimization algorithm（AOA）