铁道科学与工程学报2025，Vol.22Issue(4)：1802-1814,13.DOI:10.19713/j.cnki.43-1423/u.T20240928
基于数据驱动贝叶斯网络的地铁施工事故致因差异化分析
Differential analysis of causes of subway construction accidents based on data-driven Bayesian networks
摘要
Abstract
This study collected 213 subway construction accident investigation reports as the data source to conduct differentiated analysis of the causes of different types of subway construction accidents and explore the root causes.First,based on the improved Human Factor Analysis and Classification System(HFACS)model,38 accident causes including personnel,equipment,management,and environment were identified from the accident investigation reports.Subsequently,the Apriori algorithm was used as a data-driven approach to explore the correlation between causes,and the Expectation-Maximization algorithm(EM)was applied for parameter learning.Then,a directed Bayesian Network(BN)model for subway construction accidents was developed accordingly.Finally,based on the reverse reasoning and sensitivity analysis of Bayesian network,the causal paths and key causes of various types of accidents were determined,thereby determining the root causes of subway construction accidents.The results indicate that inadequate construction monitoring and inadequate technical skills of workers are the root causes of collapse accidents.The lack of strict review of construction schemes and inadequate implementation of safety regulations are the root causes of high-altitude falling accidents.Inadequate resource management and lack of on-site command are the root causes of vehicle injury accidents.The root causes of object beating accidents are equipment or materials in an unsafe state and inadequate investigation and prevention of hidden dangers.Insufficient safety personnel and untimely communication among workers are the root causes of mechanical injury accidents.The root causes of electric shock accidents are inadequate safety education and resource management.The research results can provide effective references for the safety management of subway construction enterprises,and the"chain breaking"control measures proposed based on specific accident types can reverse the trend of accident occurrence in time.关键词
地铁施工事故/数据驱动/关联规则挖掘/贝叶斯网络(BN)模型/人为因素分析与分类系统(HFACS)Key words
subway construction accidents/data-driven/association rule mining/Bayesian network(BN)model/human factors analysis and classification system(HFACS)分类
交通工程引用本文复制引用
霍小森,杜爽,谭琪麟,焦柳丹,曹欢..基于数据驱动贝叶斯网络的地铁施工事故致因差异化分析[J].铁道科学与工程学报,2025,22(4):1802-1814,13.基金项目
教育部人文社科项目(21YJCZH048) （21YJCZH048）
重庆市自然科学基金资助项目(CSTB2023NSCQ-MSX0724) （CSTB2023NSCQ-MSX0724）
重庆市教委科技项目(KJQN202300732) （KJQN202300732）
