Research On Unstable Approach Risk Analysis And Early Warning Technology Based On QAR Flight Big Data

Flight quality monitoring is an important task to ensure the stable and sustainable safety development of the civil aviation industry,and has an important role in preventing flight safety accidents in advance.Quick Access Recorder（QAR）provides comprehensive and complete data for flight risk study by recording parameters such as flight position,flight attitude and flight operations during flight.The process of aligning the aircraft with the runway during the landing phase is known as approach,and the occurrence of unstable approach can easily lead to serious events such as heavy landing,tail wipe,and runway deviation during landing.Therefore,the risk analysis and precise warning of unstable approach events can effectively guarantee the safety of aircraft landing.In this paper,we use QAR data provided by the Civil Aviation Research Institute of China for the whole year of 2019,and conduct a comprehensive study on the unstable approach risk and space-time characteristics of civil aviation aircraft nationwide for the whole year of 2019,and provide quantitative early warning on the unstable approach risk.The main research work of this paper is as follows:（1）Approach risk analysis based on Bayesian networksIn order to explore the correlation between unstable approach events and related parameters such as pilot operation,weather conditions,and aircraft attitude for risk analysis.In this paper,we first analyze the impact of the pilot’s maneuvers and the external environment on the flight during landing,such as the impact of throttle decomposer angle,wind,and wind direction on the aircraft attitude,based on QAR data using association rules.Then Bayesian network is used to model the unstable approach event,and the Bayesian network is constructed by combining the results of association rule analysis,and the parameters of Bayesian network are learned by using the QAR data of the approach phase,and the conditional probability of Bayesian network is calculated,and finally the constructed model is used to analyze the event influencing factors of the approach phase risk.（2）Spatio-temporal characteristics of the unstable approach and analysis of the elementsMost of the unstable approach events occur during the landing phase of the flight,i.e.,they occur in the area near the airport.Therefore,the events are correlated with airports to explore the association between the occurrence of unstable approach events and airport attributes and meteorological attributes,such as airport elevation,weather,and season.In this chapter,an exploratory spatial and temporal analysis of the unstable approach is conducted,and the correlation between the unstable approach and the elements is analyzed using Pearson correlation coefficients and geographically weighted correlation coefficients.For elemental analysis,this chapter uses geographically weighted regression analysis,multi-scale geographically weighted regression analysis and spatio-temporal geographically weighted regression analysis to analyze different temporal and spatial scales of the unstable approach events,and analyze the spatio-temporal elemental correlations of the unstable approach events.（3）Unstable Approach Risk Early WarningIn order to reduce the hazards caused by unstable approach events,this chapter provides quantitative early warning for unstable approach risks,including pitch angle exceedance,cross-roll angle exceedance and descent rate exceedance,so that pilots can sense them in advance.Firstly,the features with higher correlation are selected by feature engineering to improve the accuracy of the warning model and reduce the difficulty of the warning model.Finally,the XGBoost model in the tree model and the Long Short Time Memory（LSTM）model in the neural network model are used to build the unstable approach warning model,and the QAR data are used to train and optimize it,so as to provide accurate warning of unstable approach risk.