基于贝叶斯网络城市燃气管道失效事故后果评价研究

发布时间：2018-11-09 10:35

【摘要】：在城市燃气管道运行过程中,存在着爆炸、燃烧和中毒等危险性,管道失效发生泄漏会伴随着人员伤亡、财产损失等事故后果,已经成为燃气输送的主要隐患。因此,评估与预测燃气管道事故后果严重度是十分有必要的,为燃气管道的风险管理者提供科学的决策依据。将事故后果严重与事故发生概率结合,有利于管理者提出阻止事故发生与扩大的具体措施。本文尝试建立一个燃气管道失效事故后果评价模型,其主要目的是在信息不确定和数据缺失的情况下,预测评估事故后果严重度。首先,在前人研究的基础上对燃气事故后果进行综合分析,结合系统论的思想探讨了燃气事故后果的演变规律。根据对燃气事故常见的火灾和爆炸两种事故类型的研究,分析其带来的事故后果,其包括人员伤亡、事故损失、社会影响和环境影响,将这四个方面作为衡量事故后果严重度的指标。从事故发展的全过程出发,提出影响事故后果严重度指标的主要因素。该研究为后文将贝叶斯网络应用到管道失效事故后果评价中提供理论基础。其次,构建贝叶斯网络结构。第一步是选取网络节点,第二步是确定网络节点之间的关系。其中最为重要的是节点之间关系的确定,本来采取机器学习和咨询专家相结合的方法来确定节点之间的关系,构建贝叶斯网络结构。再次,确定贝叶斯网络参数。根据查阅的25起燃气管道失效事故调查报告,对应各个节点变量,根据每个节点变量的取值范围,获取网络节点样本数据,再应用样本统计法计算网络节点的条件概率,得出各节点变量条件概率表,从而构建贝叶斯网络燃气管道事故后果评价模型。最后,结合实例介绍和验证模型。通过具体分析,应用构建的贝叶斯网络的推理功能对事故后果进行评价,将模型评估的事故后果与实际发生的事故后果进行对比分析,验证模型的有效性,得出贝叶斯网络可以应用到事故后果评价中。
[Abstract]:In the process of urban gas pipeline operation, there are risks such as explosion, combustion and poisoning. The leakage of pipeline failure will be accompanied by casualties, property losses and other accident consequences, which has become the main hidden danger of gas transmission. Therefore, it is necessary to evaluate and predict the severity of gas pipeline accidents, and provide scientific decision basis for gas pipeline risk managers. The combination of serious accident consequence and accident occurrence probability is helpful for managers to put forward concrete measures to prevent the accident from occurring and expanding. This paper attempts to establish a model for evaluating the consequences of gas pipeline failure accidents. The main purpose of this model is to predict and evaluate the severity of accident consequences under the condition of uncertain information and missing data. Firstly, on the basis of the previous studies, the author makes a comprehensive analysis of the consequences of gas accidents, and discusses the evolution law of the consequences of gas accidents with the thought of system theory. Based on the study of two common types of fire and explosion in gas accidents, the consequences of the accidents are analyzed, including casualties, accident losses, social impacts and environmental impacts. These four aspects are used as indicators to measure the severity of the consequences of the accident. Starting from the whole process of accident development, the main factors influencing the severity of accident consequence are put forward. This study provides a theoretical basis for the later application of Bayesian network to the evaluation of pipeline failure consequences. Secondly, the Bayesian network structure is constructed. The first step is to select the network nodes, the second step is to determine the relationship between the network nodes. The most important thing is to determine the relationship between nodes. The method of machine learning and consulting experts is used to determine the relationship between nodes and to construct Bayesian network structure. Thirdly, the Bayesian network parameters are determined. According to the investigation report of 25 gas pipeline failure accidents, corresponding to each node variable, according to the value range of each node variable, the network node sample data is obtained, and then the conditional probability of the network node is calculated by using the sample statistics method. The conditional probability table of each node variable is obtained, and the Bayesian network accident consequence evaluation model of gas pipeline is constructed. Finally, the model is introduced and validated with examples. Through the concrete analysis, the inference function of the constructed Bayesian network is applied to evaluate the accident consequence, and the accident consequence of the model evaluation and the actual accident consequence are compared and analyzed to verify the validity of the model. It is concluded that Bayesian network can be applied to the accident consequence evaluation.
【学位授予单位】：首都经济贸易大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TU996

【参考文献】