电子学报 ›› 2017, Vol. 45 ›› Issue (8): 1937-1946.DOI: 10.3969/j.issn.0372-2112.2017.08.019

• 学术论文 • 上一篇    下一篇

基于直觉模糊Petri网的混合推理方法

孟飞翔1, 雷英杰1, 雷阳2, 申晓勇1   

  1. 1. 空军工程大学防空反导学院, 陕西西安 710051;
    2. 武警工程大学, 陕西西安 710086
  • 收稿日期:2016-02-29 修回日期:2016-06-07 出版日期:2017-08-25 发布日期:2017-08-25
  • 作者简介:孟飞翔,男,1986年出生,河南信阳人,现为空军工程大学计算机应用专业博士研究生,主要研究方向为智能信息处理.E-mail:ttimo@163.com;雷英杰,男,1956年出生,陕西华阴人,现为空军工程大学教授,博士生导师,主要研究方向为智能信息处理与智能决策.E-mail:leiyjie@163.com
  • 基金资助:
    国家自然科学基金(No.61272011);国家自然科学青年基金(No.61309022)

Hybrid Reasoning Using Intuitionistic Fuzzy Petri Nets

MENG Fei-xiang1, LEI Ying-jie1, LEI Yang2, SHEN Xiao-yong1   

  1. 1. Air and Missile Defense College, Air Force Engineering University, Xi'an, Shaanxi 710051, China;
    2. Engineering University of Armed Police Force, Xi'an, Shaanxi 710086, China
  • Received:2016-02-29 Revised:2016-06-07 Online:2017-08-25 Published:2017-08-25

摘要: 针对现有的基于模糊Petri网(Fuzzy Petri Nets,FPN)和直觉模糊Petri网(Intuitionistic Fuzzy Petri Nets,IFPN)的推理方法在求解只涉及知识库中部分规则的问题时存在推理过程复杂、效率不高,而且不能对问题产生的原因进行分析等缺陷,提出一种基于IFPN的混合推理方法.该方法将反向推理与正向推理相结合,首先把所要求解的问题转化为目标库所,并引入关联库所、关联变迁和子模型等概念;其次运用反向推理寻找目标库所的关联库所和变迁并构建推理子模型,从而获取问题产生的潜在原因并简化推理模型;最后以子模型作为推理模型,运用正向推理求解目标库所的token值,解决了直接运用原模型进行推理时过程复杂且效率不高的问题.与此同时,通过在模型中引入阈值以及"路径"和"有效路径"等定义,排除无效关联库所,从而找出了问题产生的真正原因.实例验证表明该方法可行且有效,与现有方法的对比分析表明该方法克服了现有方法的缺陷.

关键词: 直觉模糊Petri网, 正向推理, 反向推理, 混合推理

Abstract: Aimed at that existing reasoning methods based on fuzzy Petri nets (FPN) and intuitionistic fuzzy Petri nets (IFPN) have the defects of complicated process and low efficiency in solving the problems only related to part of the knowledge base and they cannot analyze the causes of the problems,a hybrid reasoning method based on IFPN was presented.The method combined the forward reasoning and backward reasoning,firstly,the problems which needed to be solved were converted to goal places; secondly,in order to obtain potential causes of the problems and to simplify the reasoning model,associate places and transitions of goal places were searched by backward reasoning and a submodel was constructed; lastly,the problems of complicated process and low efficiency in using the original model to reason were solved by taking the submodel as the reasoning model and using forward reasoning to compute token values of goal places.Moreover,the real causes of the problems were found out by introducing threshold and the definition of route and active route into the model to remove the invalid associate places.The examples shows that the hybrid reasoning method is feasible and effective,and that compared with the existing methods shows that it overcomes the defects of the existing methods.

Key words: intuitionistic fuzzy Petri nets (IFPN), forward reasoning, backward reasoning, hybrid reasoning

中图分类号: