基于因果关系的模型诊断  

New method of using causal relations for model-based fault diagnosis

作　　者：张立明[1,2] 赵剑[1,2] 赵相福[1,2] 欧阳丹彤[1,2] 白岩[3] 

机构地区：[1]吉林大学计算机科学与技术学院,长春130012 [2]吉林大学符号计算与知识工程教育部重点实验室,长春130012 [3]吉林大学第一医院,长春130021

出　　处：《吉林大学学报（工学版）》2009年第4期1052-1056,共5页Journal of Jilin University:Engineering and Technology Edition

基　　金：国家自然科学基金项目(60773097;60873148);新世纪优秀人才支持计划项目(20060532);'973'国家高技术研究发展规划项目(2003AA118020);吉林省科技发展计划项目(20080107)

摘　　要：提出一种基于因果关系的思想求解所有极小诊断的新方法。此方法与Reiter的模型诊断方法有着本质的不同,不用求解冲突集及冲突集的碰集,一次直接求出所有的极小诊断,避免了由于调用ATMS时的组合爆炸引起的NP-完全问题。并根据因果关系的思想提出了极小诊断组成原理和极小诊断元件个数原理,可直接求出所有的极小诊断,避免了非极小诊断的产生,提高了诊断效率,并且不会丢失正确的解。该算法程序容易编制,且效率较好,可以满足复杂被诊断对象的实时性要求。A novel method of model-based diagnosis by using causal relations was proposed for computation of all minimal diagnoses. This approach is substantially different from the classical model-based diagnosis methods. Unlike the classical methods, it can directly obtain all the minimal diagnoses without computing all the conflict sets and therefore the hitting sets of the collection of the corresponding conflict sets. The NP-complete problems induced by combinatorial explosion due to calling ATMS can be avoided. Based on the principle of causality the composing mechanism of minimal diagnosis and the theory of the number of elements in a minimal diagnosis were put forward. As a result, the non-minimal diagnoses can never be produced, and the true solution can not be missed. The algorithm is easy to program, and the proposed method can significantly improve the diagnosis efficiency to satisfy real-time requirement even for a complex system.

关 键 词：人工智能 基于模型的诊断 极小诊断 因果关系 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]