Research on Predictive Maintenance of Aircraft Based on Long Short-Term Memory Neural Network

Abstract

The proper operation of aircraft systems is of great importance to guarantee flight safety. Aircraft systems are quite complex, especially the surveillance systems in the predictive maintenance model, which incorporates information collection and extraction techniques. Under the premise of ensuring applicability, therefore, reducing the high cost of preventive maintenance and making much accurate estimates or predictions effectively has always been a topic worth studying. In this study, the aircraft system-related data are collected and evaluated by the big data analysis. With LSTM (Long Short-Term Memory) used to process and predict important events of very long intervals and delays in time series. After data cleaning, filtering, and feature engineering, a set of predictive models is finally built. Through the model, replacement time of the aircraft system components can be more accurately predicted. Thereby reducing maintenance costs and optimizing benefits.