Risk warning for long landing during aircraft plateau test flights based on transfer learning

Ensuring flight safety is a fundamental priority in the aviation industry, and mitigating the risk of long landing accidents during the aircraft landing phase holds the utmost significance. Nevertheless, when dealing with certain novel specialized aircraft engaged in test flight missions at plateau airports, the limited availability of landing data presents challenges in directly constructing effective long landing risk warning models for these high-altitude airports. To tackle this issue, this study proposes a transfer learning-based approach for long landing risk warning, intending to utilize the Tradaboost method, which relies on sample weight transfer. By harnessing abundant historical flight data from landings at low-altitude airports, the method supplements the scarcity of data from plateau airport landings, thereby constructing a machine learning model capable of effectively warning about long landing risks during aircraft landings at plateau airports. The proposed method is validated using historical flight data from the aforementioned novel specialized aircraft. The case analysis reveals that the model achieves a recall rate of 0.83 for long landing risk warnings at plateau airports, effectively providing timely warnings when the aircraft’s landing altitude reaches 50ft during plateau landings.