Life Prediction for Machinery Components Based on CNN-BiLSTM Network and Attention Model
Mengyong Wang, Jian Cheng, Hongyu Zhai
Abstract
Predictive maintenance is the focus of industrial Internet applications. The key to achieving it is to effectively predict the Remaining Useful Life (RUL) of the core components of the device. With the development of integrated circuit and sensor technique, data-driven approaches show good potential on RUL prediction. This paper proposes a new data-driven approach with CNN-BiLSTM network for RUL prediction, Using Convolutional Neural Network (CNN) model to extract local features and Bidirectional Long Short-Term Memory (BiLSTM) to make full use of the sensor date sequence in bidirection. After that, the attention layer is used to focus our attention on the most critical features. The comparison experiments on public data sets show that the model proposed in this paper effectively improves the accuracy of life prediction.