Research On Flight Risk In Landing Using QAR Data

In recent years,the air transport industry has developed rapidly,and the total volume of air traffic has been increasing.Aviation safety is one of the most important contents of civil aviation.Ensuring aviation safety is the eternal theme of aviation industry development.Flight risk refers to the trigger of one or more factors under certain conditions during the entire flight of an aircraft,resulting in the occurrence of flight exceedance events,resulting in the consequences of different severity levels.There is a strong correlation between flight risk factors and target unsafe events,therefore,the analysis of flight risk has become the focus of aviation industry.Flight data refers to data describing aircraft movement state and operation index of aircraft system.Aircraft has a variety of acquisition equipment,in which Quick Access Recorder(QAR)data is often used in the analysis of flight exceedance events detection,flight quality analysis,maintenance fault detection and other fields because of its own easy storage,high sampling frequency,and many recording parameters;at present,there is no quantitative analysis of flight risk based on QAR data.So,this paper,aiming at the landing stage of aircraft flight,starting with QAR data application and flight risk analysis,analyzes the QAR data of landing by clustering analysis,extracts the key flight parameters of landing stage,and realizes the visualization of key flight parameters.In order to extract the key risk points,we hackle the risk scenario in landing by Situational awareness.The key risk points are quantized with QAR data as the evaluation index,and the Bayesian network is introduced.The key risk points are used as nodes of Bayesian network for setting up the risk analysis model for landing stage.A large number of historical QAR data are used to determine the prior probability of the network.Finally,a well-established network structure is used for forward reasoning analysis and reverse diagnosis analysis.The main work of this paper is as follows:First,in view of the problem that the flight risk is difficult to quantify,a quantitative method of time point + parameter value is proposed.By extracting the value of the parameters at the critical time and judging whether the value is in the normal range,the risk is determined,and the determination of the normal interval should be based on the industry experience.Secondly,there are too many indicators of QAR data.How to extract key parameters from QAR data is a problem that must be solved.This paper proposes a key parameter extraction method based on clustering algorithm and parameter reduction.The original QAR data are preprocessed to unify the dimension and parameter collection frequency.Then the hierarchical clustering algorithm is used to cluster the QAR data,and the redundant parameters are eliminated by the clustering results combined with the industry experience and flight characteristics,and the key parameters are finally obtained by iteration.For the convenience of analysis,the key parameters are visualized.Finally,in order to find out the relationship between flight risk and unsafe events,a flight risk analysis model based on Bayesian network is proposed.The situational awareness and Bowtie model are applied to the qualitative analysis of the landing stage risk scenario,and the nodes and structures of the Bayesian network are determined by the key parameters extracted from the QAR data,and the Bayesian network is studied by a large number of QAR data to determine the prior probability of the network;using the established model to do reasoning and diagnosis analysis for flight risk.500 QAR data of A320 type provided by an airline company are used for experiment.The analysis results show that the method of key risk quantification and risk analysis through Bayesian network is effective,and the relationship between risk factors and unsafe events can be objectively and accurately analyzed.