Aircraft Hard Landing Prediction Using LSTM Neural Network

Abstract

Hard landing is a severe accident during the flight landing phase, which threats the aircraft architecture and passengers' safety. This study proposed a model named LSTM for aircraft hard landing prediction, which provides advanced warning to take proper measures. The unique structure of LSTM model makes it have the superior capability to capture the long temporal dependency of time series QAR data for hard landing forecasting. Experiments were conducted using the A320 QAR dataset consisting of 853 hard landing flights and 1082 normal landing flights. Comparing the performance of the proposed LSTM model to other tradition prediction models, the results suggest that LSTM model is effective and achieves high prediction accuracy of hard landing.