Knowledge Reasoning And Application For Expert Systems Based On Fuzzy Petri Net

The most important technologies of the development of the expert system, such as knowledge representation, knowledge acquisition, knowledge base maintenance, and reasoning mechanism, are closely related to knowledge. In other words, knowledge plays an important role in the expert system. Fuzzy Petri net (FPN) is an expansion of Petri net, it can describe the fuzzy knowledge efficiently. Many studies on FPN have proved that, FPN is a good method of knowledge representation. It can describe the uncertain and fuzzy knowledge in the real world. Compared with other methods, it supports parallel reasoning and the consistency maintenance of knowledge base. The relation between the representation of fuzzy production system and FPN is discussed and a transformation algorithm is presented. This algorithm changes the isolated fuzzy rules into integrity FPN. The forward reasoning algorithm, backward reasoning algorithm and the two-way reasoning algorithm of the production system are brought forward. Besides, a parallel reasoning algorithm based on FPN is also presented. It avoids unnecessary repetitions of firing of transitions effectively, and improves the efficiency of the reasoning process. Default reasoning is an important method for the real-life reasoning process. A kind of FPN model based on default reasoning is proposed in the paper. In case of the consistent deficiency of the knowledge base, a consistency maintenance algorithm is presented. It can deal with equivalent rules set, redundancy rules set, the conflict rules set, and the cycled rules set. Because the parameters of the FPN, such as weights, thresholds of transitions, degree of true of rules, are difficult to access accurately, the optimization algorithm of parameters based on BP algorithm and genetic algorithm are both discussed in the paper. In order to reduce the complexity of the FPN model, a simplification method of FPN based on knowledge reduction is presented. The simplified model contains the same information as before.