基于数据驱动贝叶斯网络的化工事故风险分析  

Risk analysis of chemical accidents based on data-driven Bayesian network

作　　者：林其彪[1] 李鑫[2] 葛樊亮 阳富强[2] LIN Qibiao;LI Xin;GE Fanliang;YANG Fuqiang(Zhicheng College,Fuzhou University,Fuzhou Fujian 350108,China;School of Environmental and Safety Engineering,Fuzhou University,Fuzhou Fujian 350108,China)

机构地区：[1]福州大学至诚学院,福建福州350108 [2]福州大学环境与安全工程学院,福建福州350108

出　　处：《中国安全生产科学技术》2024年第4期180-185,共6页Journal of Safety Science and Technology

基　　金：国家自然科学基金项目(52274181);福建省科技计划区域发展项目(2023Y3001)。

摘　　要：为减少化工厂风险分析中的主观干预,基于关联规则和贝叶斯网络构建1种数据驱动风险分析模型。该模型涵盖3个任务项,分别为数据集项、关联规则驱动项和贝叶斯网络风险评估项。首先,收集事故报告和事故因素构建事故数据库;其次,将事故数据导入Apriori算法,并根据关联规则的因素相关性确定贝叶斯网络和条件概率表结构;然后,基于事故因素出现频率计算先验概率和条件概率,并采用Fussel-Vesely计算事故因素的敏感度;最后,收集94起危险化学品中毒窒息事故实例,运用数据驱动风险分析模型评估事故因素的影响大小。研究结果可为减少和避免化工事故提供一定参考,有助于提高相关企业的整体安全水平。To reduce the subjective intervention in risk analysis of chemical plants,a data-driven risk analysis model based on the association rules and Bayesian network was constructed,which covered three task items,including data set item,association rule driven item and Bayesian network risk assessment item.Firstly,the accident reports and accident factors were collected to construct an accident database.Secondly,the accident data were imported into the Apriori algorithm,and the structure of Bayesian network and conditional probability table were determined based on the factor correlation of association rules.Then,the prior probability and conditional probability were calculated according to the occurrence frequency of accident factors,and the sensitivity of factors was calculated by Fussel-Vesely.Finally,a chemical poisoning and asphyxiation accident was used as an example,and the influence of the accident factors was assessed using the data-driven risk analysis model.The research results can provide new ideas to reduce and avoid chemical accidents,and help improve the overall safety level of related enterprises.

关 键 词：风险分析 数据驱动 事故数据 关联规则 贝叶斯网络 

分 类 号：X937[环境科学与工程—安全科学]