基于图模型的储罐区火灾事故多米诺效应评定

Evaluation of domino effect in fire accidents in storage tank area based on graph model

为量化原油储罐区域火灾事故中由多米诺效应引发的事故风险,基于图模型提出一种同时考虑初级火灾事故及其扩展概率的多米诺效应评定算法。首先,采用事故树计算储罐的年度初级火灾事故起数的期望值;其次,以产生初级火灾事故的储罐为节点,以火灾扩展方向为边构建图模型;然后,基于图模型构建各个节点发生事故的贝叶斯网络;最后,借助贝叶斯网络计算各节点在多米诺效应下发生事故起数的总期望值,并将提出的算法应用到某储罐区火灾事故多米诺效应的评定中。结果表明:相对于仅考虑单一储罐,考虑周围储罐发生事故的多米诺效应后,部分储罐的事故起数的总期望值有明显提高;在事故风险评定中不应忽视多米诺效应这一因素。

In order to quantify the accident risk caused by domino effect of fire accidents in the crude oil storage tank area,a graph-based domino effect evaluation algorithm was developed which considers both the probabilities of primary fire and its probability of subsequent events.Firstly,the accident tree method was used to calculate the expected number of primary fire accidents in the tank.Then the graph model was constructed with the storage tanks in which primary fire occurs as nodes and the probable fire expansion directions as edges.Afterwards,the Bayesian networks presenting the fire expansion of each node were build based on the abovementioned graph model.Finally,the Bayesian network was used to calculate the total expectation of each node’s accidents under the domino effect,and the proposed algorithm was applied to evaluate the domino effect in a crude oil tank farm.The results show that compared with the expected value of accidents considering only a single storage tank,the total expected value of accidents in some storage tanks is increasing significantly after considering the domino effect of accidents in surrounding storage tanks,and that the domino effect should not be ignored in accident risk assessment.