QAR数据飞行级异常检测研究  

Research on flight level anomaly detection of QAR data

作　　者：宫淑丽[1] 马广佑 邓华佳 曹力[1] Gong Shuli;Ma Guangyou;Deng Huajia;Cao Li(College of Civil Aviation,Nanjing University of Aeronautics&Astronautics,Jiangsu Nanjing,210016,China)

机构地区：[1]南京航空航天大学民航学院,江苏南京210016

出　　处：《机械设计与制造工程》2021年第10期68-73,共6页Machine Design and Manufacturing Engineering

基　　金：国家自然科学基金资助项目(U2033201)。

摘　　要：快速存取记录器(QAR)记录了表征飞机系统状态以及飞行状态的参数,QAR数据异常检测对飞机状态监控、飞行品质评估和保障飞行安全等具有重大意义。基于QAR数据的时间相关性,采用滑窗与聚类相结合的方法实现了单维参数的异常检测;基于QAR数据的空间相关性,采用多元线性回归方法实现了多维参数的异常分析。以某一航班的QAR数据为例进行仿真分析,结果表明,所采用的算法可以有效检测出异常数据,并能确定异常产生的原因,为QAR数据在飞行品质评估与飞机状态监控中的应用奠定了理论基础。Quick access recorder(QAR)is used to record the parameters that are related to the state of the aircraft systems and operation of flight.The QAR data anomaly detection is of great significance to aircraft state monitoring,flight quality evaluation and flight safety.Based on the time correlation of QAR data,the method which combines the sliding window and clustering is adopted to realize the anomaly detection of one-dimensional parameters.Owing to the spatial correlation of QAR data,the multidimensional parameter anomaly analysis is realized by using the correlation analysis method to determine the cause of the anomaly.A certain flight data with anomaly is selected to verify the methods proposed.The results show that the method proposed not only can effectively detect the abnormal data,but also can determine the causes of abnormal.This research lay a theoretical basis for QAR data in the application of flight quality assessment and aircraft condition monitoring.

关 键 词：QAR数据 时间序列 异常数据检测 滑动窗口 聚类 

分 类 号：V328[航空宇航科学与技术—人机与环境工程]