Spatial correlation and temporal attention-based LSTM for remaining useful life prediction of turbofan engine

Abstract

Remaining useful life (RUL) prediction has always been a core task of prognostics and health management technology, which is crucial to the reliable and safe operation of mechanical equipment. In recent years, data-driven methods have played an increasingly important role in RUL prediction. However, most methods have no effective mechanism to measure the importance of different variables at different times, and lack information extraction in the temporal dimension, which seriously affects the prediction accuracy of RUL. To solve the negative impact of these problems, this paper proposes a long short-term memory framework based on spatial correlation and temporal attention mechanism (SCTA-LSTM). Firstly, spatial correlation attention fully considers the relationship between variables, and adaptively measures the importance of different variables in input data at different times. Then, temporal attention enhances further the information extraction ability of LSTM in the temporal dimension, and finally, the fully connected network is used to predict RUL. To verify the effectiveness of the proposed SCTA-LSTM, two different turbofan engine simulation datasets from the Prognostics Center of Excellence at NASA Ams Research Center are used for modeling and testing. The experimental results show that this method outperforms other existing methods.