基于逻辑树和贝叶斯网络的海洋平台火灾概率分析

Probability analysis on the offshore platform fire based on the logic tree and Bayesian network model

火灾概率分析是海洋平台火灾定量风险评估的重要组成部分,考虑到传统事故树和事件树方法存在一定的局限性,提出了基于逻辑树和贝叶斯网络的火灾概率分析模型。首先采用数理统计方法对墨西哥湾地区2 837起火灾事故进行统计分析,依据事故情况构建逻辑树,然后将逻辑树转化为贝叶斯网络,根据历史数据确定贝叶斯网络各节点的先验概率和条件概率。结果表明:海洋平台火灾事故是设备、人因和组织管理多因素耦合作用的结果;基于贝叶斯网络模型得到海洋平台火灾概率约为1.0×10^(-5),为海洋平台火灾定量风险评估提供了基础数据;由贝叶斯网络模型分析得出,人因操作失误与缺乏作业安全分析的后验概率分别达到0.471和0.119,表明人因组织因素对海洋平台火灾事故具有重要影响。

The present paper is inclined to introduce a probability analysis on the offshore platform fire based on the logic fault tree and the Bayesian network model for determining the probability of the fire and explosion events,which is generally considered to be a critical component of risk-based quantitative assessments. The event tree or the fault tree has commonly been used for predicting the frequency of the fire accidents in oil and gas industry. However,the traditional approaches may suffer from unrealistic assumptions in regard to the deterministic relationship among the root events,middle events and top events. What we would like to introduce is just a hybrid method based on the combination of the logical tree and Bayesian network,which is effective in handling the uncertainties among the eventful accidents or disasters and performing a qualitative and quantitative analysis over the offshore fire accidents. In accordance with the needs,the method of our analysis can just involve the following main steps：（ 1） It is necessary to do extensive investigations of all kinds of the likely offshore fire accidents and sort out all the likely influential factors that may lead to or result in such offshore fire accidents;（ 2） It is necessary for us to establish a logic tree,or an event or fault tree in accordance with the results gained from our aforementioned investigation so as to display the potential accident scenarios and their origins;（ 3） It is necessary to convert Bayesian Network by the logic tree to determine the relation between the likely causes and the accidents concerned by exploring the conditional probability and assess the general frequency of the offshore fires through pursuing the initiating events and the conditional probabilities.The results from the hybrid method are proved that（ a） the offshore fire accidents can usually be caused by the coupling multifactors of the equipment,human and organizational factors;（ b）The Bayesian network can potentially serve as a valuable tool for the offshore fire frequency analysis,which has been worked out at about 1. 0 × 10-5,which helps to provide the basic data for the risk evaluation of the offshore fire and explosion,（ c） The posterior probability of the human error and job safety analysis has been worked out at about 0. 471 and 0. 119,respectively,which indicates that human and organizational factors should account for the essential influence on the offshore fire accidents. Thus,it can be seen that the method we have proposed can be efficiently applied to the said offshore risk assessments,including the assessments of frequency of the offshore fires as well as the quantification of the likely effects of the safety barriers,such as human and organizational factors.