民航管制安全风险主题时空分布规律研究

Spatial and temporal distribution of Air Traffic Control(ATC)safety risk topic

为了探究民航管制安全风险的时空分布规律,基于潜在迪利克雷分布(Latent Dirichlet Allocation,LDA)主题模型识别出民航管制安全风险主题,定义民航管制安全风险主题强度的定量测度指标,运用全局空间自相关分析和冷热点分析对民航管制安全风险主题的时空分布规律进行研究。结果表明:利用LDA主题模型识别出“管制员指令错误风险”等10个管制安全风险主题;“管制员指令错误风险”主题存在较弱的全局空间自相关性,在2018—2021年,全局Moran’s I总体呈现波动增长的趋势;在2018—2021年,“管制员指令错误风险”主题强度高值聚集的区域由西南向东南转移,高值聚集区域数量变少,且不稳定,低值聚集区域发生转移并在2020年后保持稳定。通过全局空间自相关分析和冷热点分析确定了2018—2021年中国民航不同管制区域的管制安全风险的时空分布格局,为局方进行差异化的安全监管提供决策支持。

To explore the temporal and spatial distribution law of civil air traffic control safety risk,this study used the Jieba word segmentation tool to perform Chinese word segmentation,identified the topics and keywords and document-topic distribution of civil air traffic control safety risks based on Latent Dirichlet Allocation(LDA)model,and defined the quantitative measurement index of the topic intensity of civil air traffic control safety risks.Then,the global spatial autocorrelation analysis method based on Moran's I statistic and the cold hot spot analysis based on Getis Ord G∗i local statistic were used to study the spatiotemporal distribution pattern of civil air traffic control safety risk themes.The results are shown as follows:(1)the LDA model can identify 10 control safety risk topics,such as the risk of controller command error,the risk of flight plan error,and the risk of personnel security failure.(2)The global Moran's I statistic for the topic intensity of‘risk of controller command error’is close to 0,indicating a weak global spatial autocorrelation.The global Moran's I index value shifts from negative to positive,and the topic intensity of controller command error risk shows a trend of development from a spatial discrete state to spatial aggregation.(3)During the period from 2018 to 2021,the global Moran's I overall showed a trend of fluctuation and growth.(4)From 2018 to 2021,the areas with a high concentration of topic intensity for‘risk of controller command error’shifted from southwest to southeast,and the number of high-concentration areas decreased and became unstable.(5)Low-value clustering areas have shifted and remain stable after 2020.The spatiotemporal distribution pattern of air traffic control safety risks in different air traffic control areas of China's civil aviation from 2018 to 2021 was determined through global spatial autocorrelation analysis and cold hot spot analysis,providing decision-making support for differentiated safety supervision by the authorities.