测试不可靠条件下多故障诊断方法  

Multiple fault diagnosis method with unreliable test

作　　者：方甲永[1] 肖明清[1] 王学奇[1] 禹航[1] 

机构地区：[1]空军工程大学自动测试系统实验室,西安710038

出　　处：《北京航空航天大学学报》2011年第4期433-438,共6页Journal of Beijing University of Aeronautics and Astronautics

基　　金：国防十一五重点预研资助项目(51317030103)

摘　　要：针对部队复杂系统故障诊断中存在的诊断精度低,虚警率高等问题,提出一种测试不可靠条件下多故障诊断方法.为解决系统诊断贝叶斯网络结构和概率映射表建立困难的问题,通过建立系统的多信号流图模型,从而获得系统诊断贝叶斯网络.将测试不可靠度引入概率映射表,增加了算法工程应用中的鲁棒性.利用后验概率诊断推理将问题归结为不等式约束极值问题,采用0-1规划隐数法对不等式极值问题求解,从而获得最优解.以某型导弹制导系统电子部件为例,验证了该方法对复杂系统多故障诊断的有效性.Aiming at low diagnosis precision and high false alarm of complex systems diagnosis,a multiple fault diagnosis method was proposed.In order to solve the difficulties of building Bayesian networks＇ structure and conditional probabilities table,the system＇s multi-signal flow graphs model was built.By introducing unreliability probability into conditional probabilities table,the proposed method＇s robustness was enhanced.The problem was concluded as the extreme value problem with inequality constraints by using Bayes maximal posterior probability.The optimized result was obtained by the 0-1 programming implicit enumeration.The electronic equipment of guidance system was explored to illustrate the effectiveness of the proposed method.

关 键 词：多信号流图模型 诊断贝叶斯网络 贝叶斯后验概率 0-1规划隐数法 

分 类 号：TP277[自动化与计算机技术—检测技术与自动化装置]