基于故障树贝叶斯网的山区高速公路事故成因分析

An Analysis of Crash Factors for Freeways in Mountain Areas Based on Fault Tree and Bayesian Network

为分析山区高速公路事故致因因素,采用故障树及贝叶斯网方法,分析各因素对事故的贡献率。建立以事故为顶事件,驾驶员、车辆、道路及环境为次顶事件,事故因素为基本事件的故障树模型;构建基于故障树的贝叶斯网模型,对人、车、路及环境各方面单独分析,定义证据节点的诊断方式计算4个方面的后验概率,找出各方面中最敏感的事件。结果表明,对事故影响较大的基本因素为“未保持安全距离”“超限超载”“长大下坡坡底”和“夜间无照明”;驾驶员、车辆、道路及环境对事故的贡献率分别为54.4%,32.2%,48.1%和4.5%;对人、车、路及环境最敏感的事件分别为“操作不当”、“超限超载”、“长大下坡坡底”和“能见度不足”;基于故障树的贝叶斯网模型不仅可以更高效和简便地对基本因素进行概率推理,而且还能得出更可靠的推论。

To analyze crash factors for freeways in mountain areas,the contribution rate of each factor to traffic accidents is analyzed through fault tree analysis and Bayesian network.A model based on fault tree analysis is developed,which regards crashes as the top event;four factors of driver,vehicle,road,and environment are regarded as intermediate events;and crash factors are regarded as basic events.A Bayesian network model based on the fault tree is developed.The driver,vehicle,road,and environment factors are analyzed separately.A diagnostic method of evidence nodes is defined to calculate the posterior probability of four aspects,and the most sensitive event of each aspect is identified.The results show that the basic factors which have major impacts on traffic accidents are“not keeping safe distance”,“overrun or overload”,“end of long and steep downhill segments”,and“no illumination at night”.The contribution rates of driver,vehicle,road,and environment to traffic accidents are 54.4%,32.2%,48.1%,and 4.5%,respectively.The most sensitive factors to driver,vehicle,road,and environment are“improper operation”,“overrun or overload”,and“end of long and steep downhill segment”and“low visibility”.The Bayesian network model based on the fault tree not only can make more efficiently and conveniently probabilistic reasoning on basic factors,but also can draw more reliable inferences.