Flight Performance Evaluation During Takeoff And Approach Based On Machinelearning

As the highest incidence phase of flight accidents,takeoff and approach have been paid increasingly attention in the field of flight safety.As the important controller of the aircraft in the above phases,the pilot's operational risk has an important effect on the safety of the passengers.Accurately identify the operational risk of the pilots and evaluate their flight performance,is of great significance to the flight safety of civil aviation.Through the theoretical study of the flight operation characteristics,an identification model of operational risk is built based on the flight data,and the evaluation of flight performance based on the identification of operational risk is eventually realized.The main contributions are as follows:1)By comparing the current methods of related fields to this subject,the general idea of operational risk identification model is proposed during the takeoff and approach.The concept and basic knowledge of flight data are introduced simultaneously.2)The feature extraction algorithm and feature dimensionality reduction algorithm of flight data is built.Feature extraction algorithm can extract the data of the specified phase from the Quick Access Recorder(QAR),which also has the ability to eliminate the singular values.Feature dimensionality reduction algorithms are mainly used for the high-dimensional feature,which can optimize the accuracy and the efficiency of the identification model.Then the risk category is used as the evaluation basis of flight performance.3)An identification model of the operational risk is constructed.Based on the characteristics of flight operations,the model uses machine learning algorithm to identify and mine the potential risks of different types of risk operations.The experimental results show that the proposed method has achieved excellent identification performance.It can effectively evaluate the flight performance of the pilot during the takeoff and approach.