基于贝叶斯网络的复杂系统多态可靠性分析  

Reliability analysis of multi-state system based on Bayesian networks

作　　者：戚珩[1] 李光[1] 姜晨[1] 阚树林[2] 

机构地区：[1]中船重工第704研究所,上海200072 [2]上海大学机电工程与自动化学院,上海200031

出　　处：《现代制造工程》2014年第1期92-96,共5页Modern Manufacturing Engineering

摘　　要：为了评估多态复杂系统的可靠性,识别系统中的薄弱环节,提出了一种基于贝叶斯网络的多态系统可靠性评估方法。在运用故障树进行分析的基础上进一步将故障树映射为贝叶斯网络。考虑到部件故障状态以及部件间故障逻辑关系的不确定性,该方法运用贝叶斯网络的条件概率描述部件间的不确定联系,采用三状态节点来描述系统和部件不同的故障状态,融合各验前信息计算各底事件和顶事件的发生概率,分析各个底事件对系统整体可靠性的影响,并计算各事件的重要度指标,通过分析查找系统的薄弱环节,以应用实例对提出方法的可靠性进行了验证。Studies a reliability analysis method using Bayesian networks to evaluate the reliability of multi-state system and determine the system weakness. The mapping method is presented to establish the Bayesian networks based on the fault tree analysis. Considering both the fuzziness of fauh state and the uncertainty of fault logical relationship between components, the fault states of system and components are described by fuzzy numbers, and the relationship between components described as conditional probability table of Bayesian networks. The occurrence probability of the basic events and the top event can be calculated by fusing various kinds of prior information by analyzing the effects of each component on the system reliability and calculating the importance measures of each event. The vulnerable parts of the system are found out by the above analysis. An application example is illustrated to verify the effectiveness and feasibility of the proposed method.

关 键 词：贝叶斯网络 多态系统 可靠性 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]