基于地理加权建模的QAR大数据空间异质性研究

项目名称： 基于地理加权建模的QAR大数据空间异质性研究

项目编号： No.U1533102

项目类型： 联合基金项目

立项/批准年度： 2016

项目学科： 无线电电子学、电信技术

项目作者： 孙华波

作者单位： 中国民航科学技术研究院

项目金额： 23万元

中文摘要： 近年来空中颠簸事件频繁发生，轻则伤人、重则机毁人亡，已严重危害到航空安全。通过飞行品质监控对QAR数据分析可以找到事件原因，但目前QAR数据的分析多局限于对单次事件的表征分析，鲜有结合空间信息和关联事件多方面诱因进行综合分析，尤其是关系的空间异质性。针对此不足，本项目拟采用地理加权建模技术研究QAR大数据空间异质性关系，具体研究内容包括：1）采用地理加权主成分分析技术研究与颠簸相关要素的空间异质分布；2）采用地理加权回归分析技术定量分析颠簸关联要素之间的空间异质关系；3）结合复杂网络和时空分析理论，分析航空网络气象风险时空结构异质性特征；4）构建QAR大数据空间异质性分析原型系统，创建空中颠簸时空分布带，以规避常见的恶劣气象，挖掘不易识别的气象锋面，从而有效降低飞机失控风险，提升飞行安全。本项目将地理加权建模的新方法引入到QAR数据挖掘领域，为FOQA提供一种全新的数据分析模式和解决方案。

中文关键词： QAR大数据；飞行品质监控；空间异质性；地理加权主成分分析；地理加权回归分析

英文摘要： In recent years, aircraft turbulence events (ATE) have occurred frequently and caused bodily injury, even death. The incentives could be found from quick access recorder (QAR) data. The current statistical analysis of the QAR data is still dominated by simply exploring some superficial features of individual events, but rarely analyze them together with spatial information and other relative factors, particularly in exploring spatial heterogeneity in data relationships. Accordingly, this project intends to study spatial heterogeneity in QAR data relations via geographically weighted modelling (GWM) techniques. The contents of this project include: 1) all the ATE related factors will be investigated via geographically weighted principal component analysis (GWPCA); 2) the spatial heterogeneity in data relations within the QAR data and ATE related factors will be studied via geographically weighted regression (GWR); 3) the weather risk within the whole aircraft network will be considered via complex network theories and spatio-temporal analysis; 4) all the methods and techniques will be integrated into a prototype system for exploring spatial heterogeneity in QAR data relationships. This project innovatively use GWM techniques in QAR data mining, and will provide a novel solution and choice for FOQA.

英文关键词： quick access recorder data;fligFlight Operational Quality Assurance;spatial heterogeneity;Geographically Weighted Principal Components Analysis;Geographically Weighted Regression