系统可靠性评估的超椭球贝叶斯网络及其灵敏度方法  

System Reliability Assessment Method Based on Hyper-ellipsoid Bayesian Networks and Their Sensitivities

作　　者：陈东宁[1,2] 姚成玉[3] 

机构地区：[1]燕山大学河北省重型机械流体动力传输与控制重点实验室,秦皇岛066004 [2]先进锻压成形技术与科学教育部重点实验室(燕山大学),秦皇岛066004 [3]燕山大学河北省工业计算机控制工程重点实验室,秦皇岛066004

出　　处：《中国机械工程》2015年第4期529-535,552,共8页China Mechanical Engineering

基　　金：国家自然科学基金资助项目(51405426);河北省自然科学基金资助项目(E2012203015);河北省教育厅资助科研项目(ZH2012062)

摘　　要：利用区间模型描述根节点的失效可能性,解决根节点的失效可能性不易精确获取的问题;通过引入超椭球模型来界定不确定性参量的取值范围,解决区间贝叶斯网络在求取可靠性指标时计算结果相对保守的问题;定义超椭球贝叶斯网络的灵敏度指标,为找到系统的关键环节提供依据;结合贝叶斯网络双向推理求解出在根节点失效可能性已知的条件下,叶节点的失效可能性、根节点状态的后验可能性;给出了可靠性评估实例。Failure probabilities of root nodes were described by interval model to solve the difficulty to obtain the failure probabilities accurately.Hyper-ellipsoid model was utilized to define the ranges of the uncertain parameters to improve the relatively conservative reliability indices calculated by the interval Bayesian networks method.Sensitivity indices of hyper-ellipsoid Bayesian networks were proposed to provide basis for finding the key link of the system.The leaf node＇s failure probability and the root nodes＇ state posterior possibilities were solved by combining with Bayesian networks bidirectional inference under the condition of the root nodes＇ failure probabilities as known.Reliability assessment example was given at last.

关 键 词：可靠性评估 贝叶斯网络 超椭球 区间模型 灵敏度 

分 类 号：TB114.3[理学—概率论与数理统计]