A remaining useful life prediction method with long-short term feature processing for aircraft engines

Abstract As one of the key components of aircraft, any failure of the engine can lead to serious accidents. The reliability and safety can be guaranteed by predicting the remaining useful life of the aircraft engine. The data-driven approaches are suitable for predicting the remaining useful life of the aircraft engine, but they generally suffer from the following challenges: (i) how to capture the real degradation trend of the engine; (ii) how to efficiently and fully utilize the temporal correlation between the sensor data; (iii) how to handle highly nonlinear data. In order to address these challenges, an effective data-driven remaining useful life prediction method is proposed in this paper. Firstly, a long-term differential technique is proposed to extract forward differential features, which fully reflects the actual degradation trend in the entire lifetime. Then, the Fibonacci window is proposed for short-term feature extension, which makes full use of the temporal correlation of the historical data and effectively reduces the extra computational load. Finally, the CatBoost algorithm is used to predict the remaining useful life on the highly nonlinear data, and superior prediction performance is obtained. In order to verify the effectiveness of the proposed method, the experiments are carried out on the aircraft engine dataset provided by NASA.