Bearing Remaining Useful Life Prediction Based on an Improved Back Propagation Neural Network
Volume 10, Number 6, September 2014 - SC 46 - pp. 653-657
XINGHUI ZHANG1, LEI XIAO2, and JIANSHE KANG11 Mechanical Engineering College, Shijiazhuang, CHINA
2 The State Key Lab of Mechanical Transmission, Chongqing University, Chongqing, CHINA
(Received on March 29, 2014; revised on June 7, 2014)
Bearings are the key components in most of rotating machineries. Their failures can lead to catastrophic disasters. The accuracy of remaining useful life (RUL) prediction has a great influence on the preventive maintenance activity. RUL prediction based on standard back propagation neural network (BPNN) already exists. However, training standard BPNN needs more time and sometimes it may converge to local optima which can have contrary influence on the accuracy. Existing BPNN improving works used dynamic learning rate, momentum item and utilized genetic algorithms or other random researching algorithm to optimize the adjustment of connect weights in the network. In this paper, an improved BPNN based on Levenberg-Marquardt algorithm and momentum item is proposed. It can predict the bearing’s RUL with a good performance. Finally, the bearing simulation life data sets are used to validate the proposed method. The results show that the prediction accuracy of the proposed method is superior to other existing BPNNs.
Click here to download the paper.
Please note : You will need Adobe Acrobat viewer to view the full articles.