Remaining Useful Life Prediction Method of Aero-Engines Based on LSTM

Fault diagnosis and remaining useful life prediction of aero-engines are important parts of its health management, which is of great significance for reducing maintenance costs and effectively preventing the occurrence of unexpected accidents. In order to improve the accuracy of engine fault prediction, this paper proposes a remaining useful life (RUL) prediction method of aero-engines, which is based on long short-term memory (LSTM) network. Firstly, in order to reduce noise and eliminate the influence caused by singular samples, the time-series in the dataset of aero-engines are processed by wavelet transform and normalization. Secondly, a LSTM prediction network is constructed, and the network is trained by clipping responses (i.e. RUL tag values) and adjusting network parameters. Finally, the RUL prediction is performed. A data set of aero-engines from practical engineering applications is used to validate the effectiveness of the proposed method. Compared with several other prediction algorithms, the proposed method in this paper effectively improves the prediction accuracy of RUL, and provides a decision-making basis for the maintenance of aero-engines.