摘要
报告类事件信息中蕴含大量事发原因及发展脉络,将其进行图形化展示有助于获取其中相关信息,并为基于数据的安全管理决策提供支持。本研究以2017—2021年中国民用航空安全信息系统的8119条中止进近/复飞事件为样本,应用文本挖掘技术,从事件文本数据中提取相关信息,并确定事件原因,结合威胁与差错管理(TEM)模型构建本体模型;利用深度学习和模板匹配方法抽取事件原因及事件演化关系,将收集到的原因和关系转化为图谱形式。使用图数据库Neo4j,将原因、地点、时间等作为节点,关系作为边,共形成了48695个实体节点,64933个关系,构建原因图谱。通过图谱推理和查询,可以进一步发现和验证原因之间的逻辑或演化关系。为了解事件致因逻辑及其演化关系,以及防范此类事件和可能的高风险事件提供参考。
The information of report-type events contains a large number of incident causes and developmental pathways,and its graphical presentation helps to obtain the relevant information therein and provide data-based support for safety management decision-making.This study takes 8119 aborted approach/go-around events of Aviation Safety Information System of CAAC from 2017-2021 as samples,extracts relevant information from event text data by applying text mining technology,determines the causes of the events,and constructs an ontology model by combining with the Threats and Errors Management(TEM)model;extracts the causes of the events and their evolutionary relationships by using deep learning and template matching methods,and converts the collected causes and relationships into the graphical form.Using the graph database Neo4j,the causes,locations,times,etc.are used as nodes and relationships as edges,and a total of 48,695 entity nodes and 64,933 relationships are formed to construct the cause mapping.The logical or evolutionary relationships between causes can be further discovered and verified through graph reasoning and querying.It provides reference for understanding the logic of event causation and its evolutionary relationship,and preventing such events and possible high-risk events.
作者
刘俊杰
于佳楠
衣文正
LIU Junjie;YU Jianan;YI Wenzheng(College of Safety Science and Engineering,Civil Aviation University of China,Tianjin 300300,China)
出处
《综合运输》
2024年第8期110-119,共10页
China Transportation Review
关键词
航空安全
文本分析
致因分析
中止进近/复飞
知识图谱
Aviation safety
Text analytics
Causal analysis
Abort approach/go-around
Knowledge map
