文献类型：会议论文

英文题名：Bearing Remaining Useful Life Prediction Using FNN-based Feature Principal Component and GRNN

作者：Liao, Zhiqiang[1]; Jia, Baozhu[1]; Kong, Defeng[1]; Ji, Ran[1]; Li, Xiaoyu[1]; Hao, Kang[1]

机构：[1] Naval Architecture and Shipping College, Guangdong Ocean University, Zhanjiang, China

会议论文集：2022 International Conference on Sensing, Measurement and Data Analytics in the Era of Artificial Intelligence, ICSMD 2022 - Proceedings

会议日期：December 22, 2022 - December 24, 2022

会议地点：Harbin, China

语种：英文

外文关键词：Bearing RUL; false nearest neighbor (FNN); feature principal component; GRNN model

外文摘要：In the bearing remaining useful life (RUL) prediction, constructing a health indicator to reflect the running bearings health status is one of the most critical parts, because it determines the performance of the RUL prediction model. This paper proposed a FNN-based feature principal component health indicator construction method, and combined with generalized regression neural network (GRNN) for bearing RUL prediction. The symptom parameters of bearing vibration signal in time domain and frequency domain are extracted at first. Then the dominant features representing bearing degradation characteristics are selected via false nearest neighbor (FNN). The health indicator was constructed with the feature principal component. The selected dominant features and the health indicator are brought into the GRNN model to predict bearing RUL. The proposed approach is verified with data of 2012 IEEE PHM challenge. The experimental results show that the proposed method can effectively improve the prediction accuracy of the bearing RUL, which proves the feasibility and effectiveness of the proposed method. ? 2022 IEEE.