电子学报 ›› 2018, Vol. 46 ›› Issue (3): 589-594.DOI: 10.3969/j.issn.0372-2112.2018.03.011

• 学术论文 • 上一篇    下一篇

结合问题特征的分组式诊断方法

刘梦1,2, 欧阳丹彤1,2, 刘伯文1,2, 张立明1,2, 张永刚1,2   

  1. 1. 吉林大学计算机科学与技术学院, 吉林长春 130012;
    2. 吉林大学符号计算与知识工程教育部重点实验室, 吉林长春 130012
  • 收稿日期:2016-10-12 修回日期:2017-01-12 出版日期:2018-03-25
    • 通讯作者:
    • 张永刚
    • 作者简介:
    • 刘梦,女,1993年生于河南洛阳,吉林大学硕士研究生,研究方向为基于模型诊断、SAT问题.E-mail:2238356051@qq.com;欧阳丹彤,女,1968年生于吉林长春,吉林大学教授,主要研究方向为基于模型的诊断、自动推理和模型检测.E-mail:ouyangdantong@163.com;刘伯文,男,1993年生于吉林延边,吉林大学硕士研究生,研究方向为基于模型诊断.E-mail:1591365445@qq.com;张立明,男1980年生于吉林长春,吉林大学博士,主要研究方向为基于模型诊断.E-mail:limingzhang@jlu.edu.cn
    • 基金资助:
    • 国家自然科学基金 (No.61272208,No.61402196,No.61672261); 浙江省自然科学基金 (No.LY16F020004)

Grouped Diagnosis Approach Using the Feature of Problem

LIU Meng1,2, OUYANG Dan-tong1,2, LIU Bo-wen1,2, ZHANG Li-ming1,2, ZHANG Yong-gang1,2   

  1. 1. College of Computer Science and Technology, Jilin University, Changchun, Jilin 130012, China;
    2. Key Laboratory of Symbolic Computation and Knowledge Engineering(Jilin University), Ministry of Education, Changchun, Jilin 130012, China
  • Received:2016-10-12 Revised:2017-01-12 Online:2018-03-25 Published:2018-03-25
    • Corresponding author:
    • ZHANG Yong-gang

摘要: 模型诊断方法是人工智能领域重要的系统故障自动检测方法,被广泛应用于软件故障检测和硬件诊断.近年来由于电路规模和复杂度不断增大,其诊断难度也不断增大.本文通过对电路模型特征的研究,结合LLBRS-tree(Last-Level Based on Reverse Search-tree)诊断算法提出分组式诊断方法GD(Grouped Diagnosis):首先结合电路特征确定组件的故障相关性并对电路组件进行分组,可缩减电路中需检测的规模;其次,利用分组后电路并结合非诊断解定理和SAT(SATisfiability)求解特征定位部分非诊断解,从而避免该部分的一致性检测来加速求解.本文算法可应用于电子电路故障诊断领域,并且实验结果表明该算法与LLBRS-tree算法相比求解效率平均提高了1.5倍,最多提高了3倍.

关键词: 基于模型的诊断, 问题特征, 分组, SAT求解器, 集合枚举树

Abstract: Model-based diagnosis is an automatic fault detection approach in artificial intelligence. It is used in software fault detection and hardware diagnosis. Recently, the difficulty of circuit diagnosis is increasing with the increasing size and complexity of the circuit. After studying the characteristics of the circuit model, this paper proposes the grouped diagnosis (GD) approach based on the LLBRS-tree (Last-Level Based on Reverse Search-tree) algorithm. Firstly, the component grouped method is used to identify the component's faulty correlation and group the components. And the scale of the circuit to be detected can be reduced. Secondly, through the grouped circuit, the non-diagnostic solution theorem is given to locate some non-diagnostic solutions with the feature of satisfiability. It can help us avoid checking consistency on these non-diagnostic solutions so as to accelerate the processing. Our algorithm can be used in electronic circuit fault diagnosis. And the experimental results show that it improves the efficiency by 1.5 times and even 3 times compared with the LLBRS-tree algorithm.

Key words: model-based diagnosis, the feature of problem, grouped, SAT solver, set-enumeration-tree

中图分类号: