中国安全科学学报 ›› 2022, Vol. 32 ›› Issue (3): 174-182.doi: 10.16265/j.cnki.issn1003-3033.2022.03.024

• 公共安全 • 上一篇    下一篇

基于贝叶斯网络的危化品道路运输事故推理模型

鲁义1,2(), 伍江乐1,2, 邵淑珍1, 施式亮1, 周荣义1, 王伟2   

  1. 1湖南科技大学 资源环境与安全工程学院,湖南 湘潭 411201
    2应急管理部 上海消防研究所,上海 徐汇,200032
  • 收稿日期:2021-12-20 修回日期:2022-02-17 出版日期:2022-08-23
  • 作者简介:

    鲁 义 (1986—),男,江西新干人,博士,教授,博士生导师,主要从事火灾科学与技术方面的工作。E-mail:
    鲁义 教授, 施式亮 教授,周荣义 副教授

  • 基金资助:
    湖湘青年英才资助项目(2020RC3047); 国家自然科学基金资助(52074118); 湖南省教育厅优秀青年项目(20B230)

Prediction model for road transport accidents of hazardous chemicals based on Bayesian network

LU Yi1,2(), WU Jiangle1,2, SHAO Shuzhen1, SHI Shiliang1, ZHOU Rongyi1, WANG Wei2   

  1. 1School of Resource & Environment and Safety Engineering, Hunan University of Science and Technology, Xiangtan Hunan 411201, China
    2Shanghai Fire Science and Technology Research Institute of MEM, Shanghai, 200032, China
  • Received:2021-12-20 Revised:2022-02-17 Published:2022-08-23

摘要:

为精准预测危化品道路运输事故风险,首先统计2015—2020年国内1 727例危化品道路运输事故数据,构建以事故影响因素、事故类型、事故应急处理时间及伤亡程度为主要节点的贝叶斯网络(BN)结构;然后在Netica中建立危化品道路运输事故推理模型,根据平均绝对误差(MAE)验证模型的有效性;最后通过正向因果推理和反向诊断推理观察目标节点各变量的后验概率变化,探究在设定条件下的事故发展趋势和事故演变过程。研究表明:该模型可在设定条件下有效进行事故推理预测,通过正向因果推理得出,中午时段,最易发生的事故是因追尾或罐体泄漏而引发的泄漏事故;结合反向诊断推理得出,运载量小于30 t是易燃液体泄漏事故可在0~3 h内处置完成的显著条件。

关键词: 危化品, 道路运输, 贝叶斯网络(BN), 事故推理, Netica

Abstract:

In order to accurately predict road transport accidents of hazardous chemicals, firstly, data of 1,727 such transport accidents in China from 2015 to 2020 were collected, and a Bayesian network (BN) was developed with accident influencing factors, accident types, accident emergency treatment time and the degree of casualties as main nodes. Then, a prediction model for the accidents was established in Netica, and its validity was verified according to the mean absolute error (MAE). Finally, through forward causal reasoning and reverse diagnostic reasoning, the posterior probability changes of each variable of target nodes were observed, and accident development trend and evolution process under set conditions were explored. The results show that the model can effectively predict accidents under set conditions. Through positive causal inference, it is concluded that the most likely form of accident at noon is the leakage accident caused by rear-end collision or tank leakage, while based on reverse diagnostic reasoning, it is found that carrying capacity <30 t is a significant condition for flammable liquid leakage accidents to be successfully disposed of within 0 to 3 hours.

Key words: hazardous chemicals, road transportation, Bayesian network (BN), accident prediction, Netica