Aero-engine Life Prediction Based on ARIMA and LSTM with Multi-Head Attention Mechanism  
Author Ning Shen


Co-Author(s) Tiantian Niu; Rui Peng; Kaiye Gao


Abstract The performance variations of aero-engine is a key concern for airlines, At present, the urgent need are failure prediction and health management technology of aero-engine, the remaining useful life (RUL) prediction is one of the core technologies of health management technology. In this paper, we propose a combined model for aero-engine life prediction which based on Long Short-Term Memory network based on multi-headed attention mechanism and Autoregressive Integrated Moving Average Model. Using the multi-headed attention mechanism to focus on the information of the input sequence related to the current time step, can improve the level of attention of the LSTM neural network to the key information of the sequence data, By extracting important time features from different moments, and using ARIMA time series model to extract linear features advantage will be more accurate. The model experiments based on actual flight data indicate that the prediction result of this model is better than the traditional statistical methods and machine learning methods, so it is effective and it has advantages in improving the accuracy of remaining useful life prediction of aero-engine.


Keywords aero-engine; remaining useful life; Multi- Head Attention Mechanism; LSTM; ARIMA
    Article #:  RQD28-311

Proceedings of 28th ISSAT International Conference on Reliability & Quality in Design
August 3-5, 2023