摘要
基于模型的诊断问题在人工智能领域内一直备受关注,将诊断问题转换成SAT(Satisfiable)问题成为解决基于模型诊断问题的一个重要方法.基于目前高效诊断方法 LLBRS-Tree(Last-Level Based on Reverse Search-Tree)的研究,本文提出电路分块诊断方法 ACDIAG(Abstract Circuit Diagnosis)方法,对电路进行分块来缩减电路规模,利用LLBRS-Tree方法对分块后抽象电路求得极小块诊断解;提出诊断解拓展方法,结合分块后电路结构特征对每个极小块诊断解进行直接扩展得到极小诊断解,避免对抽象电路还原后才能得到所有解的问题.
Model-based diagnosis problem has been attracting much attention in the field of artificial intelligence. It is an important technique for solving model-based diagnosis problem by converting diagnosis problem to SAT. Based on the research of LLBRS-Tree,this paper proposes an ACDIAG method. Firstly,the circuit blocking method is used to block the circuit via circuit structure so as to downscale circuit. Then,minimal block diagnoses are acquired on the abstract circuit after blocking via LLBRS-Tree. Secondly,diagnosis extending method is given to extend minimal block diagnoses to obtain other diagnoses directly via circuit structure. It avoids the draw back that extending diagnoses need to restore the abstract circuit.
作者
欧阳丹彤
刘伯文
刘梦
张立明
张永刚
OUYANG Dan-tong;LIU Bo-wen;LIU Meng;ZHANG Li-ming;ZHANG Yong-gang(College of Computer Science and Technology,Jilin University,Changchun,Jilin 130012,China;Ministry Education Key Laboratory of Symbolic Computation and Knowledge Engineering,Jilin University,Changchun,Jilin 130012,China)
出处
《电子学报》
EI
CAS
CSCD
北大核心
2018年第7期1571-1577,共7页
Acta Electronica Sinica
基金
国家自然科学基金(No.61672261
No.61502199
No.61402196
No.61373052)
浙江省自然科学基金(No.LY16F020004)
关键词
基于模型诊断
SAT问题
枚举树
抽象
model-based diagnosis
SAT problem
enumeration tree
abstract
