建筑物地震次生火灾的贝叶斯网络推理模型与研究.docVIP

建筑物地震次生火灾的贝叶斯网络推理模型研究 颜峻 讲师 (中国劳动关系学院 安全工程系，北京 100048) 收稿日期：201-07-02；修稿日期：201-00-00 学科分类与代码：6201030（学） 中图分类号： X 文献标志码：A 资助项目：中国劳动关系学院院级科研项目“”（编号：1YY002） 【摘要】 【关键词】；；Study on Inference Model of Seismic Secondary Fire Based on Bayesian Networks YAN Jun (China Institute of Industrial Relations，Department of Safety Engineering，Beijing，100048，China) Abstract: In order to research the evolutionary path of seismic secondary fire induced by combustible gas leak in building, Bayesian networks was built to character the dependent relationship between the intermediate cause events in development stages of urban earthquake gas leak induced fire by analyzing earthquake fire spread process and the status of node variables and their ranges were determined. By analyzing earthquake records and expert empirical judgments conclusions, the conditional probability of node variables in Bayesian network structure were given. posterior probabilities were compared in stages of earthquake secondary fire occurrence, spread and out of control based on the same conditions of seismic characteristics and architectural attributes, but node variable values of fire fighting, rescue, and water supply network damage degree were different. Inference results showed, the timeliness and effectiveness of firefighting actions have significantly affections on fire occurrence probability, and ambient wind speed and the distance between buildings have great impact on the probability of fire spread. Key words: earthquake; combustible gas leak; secondary fire; Bayesian network; inference 0 引言 1　研究现状 2　 2.1　城市地震次生燃气泄漏火灾演化过程可以分为建筑物受地震作用破坏、管道燃气泄漏起火、消防力量扑救、建筑物间火灾蔓延等4个阶段。建筑物地震次生火灾发展过程的贝叶斯网络拓扑结构，如图1所示。贝叶斯网络结构由节点及有向弧线组成，是一个有向无环图。其中，每个节点分别对应着离散型变量，网络结构表示了离散型随机变量集合之间的独立和条件独立关系。图中的每条弧线代表了变量之间的存在函数依赖关系，如果有一条弧线从到，则是的双亲或直接前驱。如果节点之间没有弧线连接，表示它们条件独立[10,11]。 灾难经验表明，在破坏性地震中建筑抗震性能是决定建筑物遭受破坏程度的最主要因素，因此论文将地震烈度和建筑物抗震性能2个因素作为决定建筑遭受破坏的两个父节点。当建筑物一旦遭受严重破坏其内部燃气管网必将破坏。考虑到若建筑内部未铺设燃气管道则不会出现“燃气管道破坏”这一中间事件，因