危险化学品道路运输风险预测模型研究

Risk forecast and prediction model with the hazardous chemical road transportation

当前,我国危险化学品道路运输事故频发,事故后果严重使之不同于一般交通事故。为了减少危险化学品道路运输事故,合理控制事故风险,提出了一种基于贝叶斯网络的危险化学品道路运输事故风险预测模型。利用事故致因理论和D-S证据理论融合专家意见建立贝叶斯网络结构,以2016—2018年发生的447起数据作为训练样本,通过EM算法进行参数学习。对2020年1月发生的15起事故案例进行了情景模拟,事故后果类型、事故等级的预测准确率达到80%。经因果推理发现,环境条件为最常见的情况下,事故后果中发生泄漏的可能性最大;进一步推理发现,当司机状态为"疲劳驾驶"而其他节点状态为最常见时,易发生"油箱泄漏"和"爆炸"事故。

The purpose of this paper is to study the risk-involving factors that may lead to the hazardous chemical road transportation accidents and the interactive disasters among the risk factors. To achieve the above research goals,we have chosen Bayesian network to express the causative consequential relationship and their mutually dependable relationship among the risk factors. And,then,we have managed to analyze some uncertain events while reasoning the function of Bayesian network. Thus,in so doing,we have first of all collected large amounts of sampling data to analyze the process of the taking place of the accidents,and,then,manage to explore the accident appearance mechanism from the following 4 aspects,i. e. the environment,the people involved,the vehicle’s particular problems and the material to be transported. To access the people involved,the vehicle itself,the goods intended to be transported and their impact on the accident consequences in accordance with the D-S evidence theory,we have managed to integrate the causative relationship among environmental factors and the people involved logically connected with the vehicle,the material in reference to the expert opinions. In the second part of the paper,we have collected 447 data sets from 2016 to 2018 as training samples and EM algorithm,so as to gain some more valuable data via the missing causative theory so as to find the mutual dependable relation between the risk factors by setting up a conditional probability table. And,so far,we have managed to choose 15 accident cases that have ever been taking place in January,2020,for the scenario simulation to verify the model to the prediction and forecast accuracy up to 80% in correspondence with the accident consequential types and the accident level nodes,so as to indicate that the prediction effect of the said model is comparatively nice,which turns out to formulate a relatively complete risk prediction model for the hazardous chemical road transportation.Thus,finally,through the causative reasoning,the maximum probability of leakage can be attained up to 74%,if,or when,the weather is clear and"sunny",with the road type being up to the "high speed one". At the same time,it should be in the clear and lighting daytime condition. Needless to say,the road conditions should be such with "no"obvious abnormal conditions,say,"too dry"or"too wet". In addition,when and if the driver were to be in a state of "over-fatigue"or the amount of goods in transportation is over "30-40 tons". Or,else,the transportation goods belong to the category of "flammable liquids",which are likely to lead to"the tank leakage"and therefore prone to lead to explosion,whose probability may come to39% or 33%,respectively and correspondingly. Thus,it can be seen that the conclusions we have gained indeed have a deductive significance for the road transportation accidents control and the emergency management.