摘要
面向航天器测控管理,研究了一种基于专家系统(ES)、案例推理(CBR)以及故障树(FT)的混合智能诊断技术.文中,故障树双向混合推理机制被用于实现航天器故障定位和预测.同时案例推理的k最近邻检索策略(KNN)采用了简单实用、易收敛特性的多感官群集算法(MSA).基于案例推理和故障树的航天器专家系统(SESCF)采用了2种融合模式.案例推理和故障树采用独立运行模式,专家系统与案例推理和故障树之间则采用了松耦合运行模式.出于改善推理效率的目的,文中提出了一种将遥测信息转化为语义信息的结合特定推理方法的非线性转换方法.某卫星供配电分系统的测试证实了SESCF系统诊断的有效性.测试结果表明,相对于专家系统,SESCF系统具有更高的诊断准确度和可靠性.SESCF系统采用的非线性转换方法在航天器故障诊断过程中简单实用且容错性较好.
A hybrid intelligent algorithm with expert system （ES） combining case-based reasoning （CBR） and fault tree （FT） for the diagnosis of spacecraft measurement and control management is studied. FT with bi-direction combination reasoning mechanism is set up to realize spacecraft fault location and prediction. CBR optimized by k-nearest neighbor （KNN） method using multi-sense swarm algorithm （MSA）, which is applicable and easy to converge, is constructed. Spacecraft ES combining CBR and FT （SESCF） runs under two kinds of models. Stand-alone model combines CBR with FT, while Loose-Coupling model combines ES with CBR and FT. Considering the improvement of reasoning efficiency, a non-linear transformation merging with special reasoning for changing remote sensing information to semantic data is realized. The fault diagnosis generated by SESCF has been verified using the experiment results of a satellite＇s electric supply and distri- bution sub-systems. It is also proved that SESCF is more accurate and reliable comparing with traditional ES. Application of the non-linear transformation to SESCF shows feasibility and good redundancy in spacecraft fault diagnosis
出处
《信息与控制》
CSCD
北大核心
2010年第1期106-113,共8页
Information and Control
基金
国家863计划资助项目(2007AA04Z218)
国家自然科学基金资助项目(60573059)
北京市重点学科建设项目(XK100080537)
关键词
航天器
混合智能
专家系统
故障树
案例推理
spacecraft
hybrid intelligence
expert system
fault tree
case-based reasoning.
