Application of Advanced Self-Adaptation Learning and Inference Techniques to Fuzzy Petri Net Expert System

Abstract

Fuzzy production rules are comparatively inefficient to depict vague and modified knowledge in an expert system. Fuzzy Petri nets are more accurate in describing the relative degree of each proposition when exists dynamic knowledge. However, the limited learning ability of fuzzy Petri net constrains its application in dynamic knowledge expert system. In this paper, an advanced self-adaptation learning way based on back-propagation is proposed to train parameters of fuzzy production rules in fuzzy Petri net. In order to reason and learn expediently, fuzzy Petri net without loop is transformed into hierarchy model and continuous functions are built to approximate transition firing and fuzzy reasoning. Simulation results show the adaptive learning techniques can make rule parameters arrive at optimization rapidly. These techniques used in this paper are quite effective and can be applied to most practical Petri net models and fuzzy expert systems