论文部分内容阅读
针对产生式专家系统中当规则数和事实数较大时,推理所产生的组合爆炸问题,提出了一种基于状态空间正交划分的模式匹配算法。 这里,状态空间是事实集合, 事实是术语、关系和术语值的三元组。 根据术语在规则中的不同作用, 首先将状态空间正交划分为输入子空间, 中间子空间和输出子空间。 然后, 基于状态空间正交划分, 提出了产生式模糊推理的模式匹配算法。 并将该算法运用于铝电解槽模糊专家系统AEGFES。试验表明该算法可显著提高模糊推理的匹配速度, 特别是可减少推理轮数, 和第2轮以后的匹配次数。
Aimed at the problem of combinatorial explosion generated by reasoning when the number of rules and the factual number are large in production expert system, a pattern matching algorithm based on orthogonal partition of state space is proposed. Here, the state space is a fact set, and the fact is a triple of terms, relations, and term values. According to the different roles of terms in rules, the state space is first divided into input subspace, intermediate subspace and output subspace orthogonally. Then, based on the orthogonal partition of state space, a pattern matching algorithm of production fuzzy inference is proposed. The algorithm is applied to AEGFES, an aluminum cell fuzzy expert system. Experiments show that the algorithm can significantly improve the matching speed of fuzzy reasoning, especially reducing the number of inference rounds and the number of matches after the second round.