Risk assessment method for civil aircraft approach and landing at high plateau based on QAR data

The high plateau approach and landing of civil aircraft is a high-risk stage of high plateau flight. To effectively identify the risk and its grade of this approach and landing, a long short term memory-deep neural network (LSTM-DNN) deep learning risk assessment method is proposed based on the variable fuzzy identification of entropy weights. This method utilizes high-altitude flight data recorded by the quick access recorder (QAR), referencing the advisory notices from the flight operations quality assurance (FOQA) of civil aviation as well as the industry QAR monitoring standards. The method combines indicator importance analysis with Delphi expert surveys to extract five key monitoring items for civil aviation high-altitude approach and landing risk assessment, including significant changes in heading during landing, low trajectory, large descent rate during the 610−305 m approach, touchdown vertical acceleration during landing, and high descent rate during the 153−15 m approach. To overcome the subjective bias of the evaluation index weight, the entropy weight method is then used to determine the evaluation index weight, with the risk level membership function constructed based on the variable fuzzy identification method. Finally, the LSTM-DNN risk assessment model for civil aircraft approach and landing at high plateau is established. Taking the Chengdu−Lhasa approach and landing segment as an example, this study extracted the QAR data to train and test the risk assessment model, and compared the results with those of the evaluation methods such as Logistic multiple regression and support vector machines (SVM). The results show that the recognition rate of the proposed method reaches 94.18% on average with the highest being 94.79%, verifying the effectiveness of the method. © 2024 Beijing University of Aeronautics and Astronautics (BUAA). All rights reserved.