Remaining Useful Life Prediction via Multi-Neural Network Integration
作者
Zhiwen Yu,Hao Huang,Yi Tang,Di Wu,Yiyuan Bao
出处
期刊: [Institute of Electrical and Electronics Engineers] 日期:2021-10-22卷期号:15: 4240-4246被引量:2
标识
DOI:10.1109/cac53003.2021.9728490
摘要
The remaining useful life prediction is an important components of the prognostic and health management system. The traditional remaining useful life prediction for the bearing cannot consider the spatial-temporal features of the degraded signal, and depends the manual extraction of the bearing degra-dation features. The single neural network model is difficult to maintain the best feature extraction in different degradation stages of bearings. In this paper, the remaining useful life prediction framework is developed via the multi-neural network integration. The different neural network can effectively extract the different features of the original signal, adaptive weight integration algorithm is proposed to integrate the multi-neural network. The bearing degradation data is used to verify the effectiveness of the developed framework, which is collected on the PRONOSTIA platform. The experimental results show that the developed framework can effectively reduce the average percentage error for the remaining useful life prediction of bearing.