LSTM-Based Risk Identification for Civil Aircraft Tire Speed Limit Exceedance During Takeoff

Abstract This study proposes a data-driven approach for predicting the risk of tire speed limit exceedance (TSLE) during aircraft takeoff. Using flight data from the Quick Access Recorder (QAR), key factors such as aircraft weight, wind components, ambient temperature, and pressure altitude are analyzed. A Long Short-Term Memory (LSTM) neural network is developed to predict takeoff tire speed and evaluate exceedance risks within a quantile-based statistical framework enhanced by bootstrap confidence intervals. The model demonstrates high predictive accuracy and strong generalization capability under diverse flight conditions. By integrating time-series learning with probabilistic risk assessment, the proposed method offers an objective and interpretable solution for the early detection of TSLE events. The findings indicate that this framework can effectively support data-driven safety management and improve operational risk mitigation during aircraft takeoff in civil aviation.