Reserch Of Adaptive Model Based On Fuzzy Petri Net

With the development of software engineering，new requests have been risen forthe conventional software development. When a system has been accomplished, it canmeet new demands. Or when the environment changes, it can adjust itself to adapt thenew environment. That is to say, the system can adapt the unexpected changing. Theissue has aroused the research to adaptive system and has become important task insoftware engineering. The progress of fuzzy maths provides a method for us todescribe objective world. In our daily life lots of concepts are fuzzy and can’t bequantified. Fuzzy concepts are suit for our reality. So we can utilize fuzzy maths tosolve the issue of adaption.The task of this paper is to construct fuzzy adaptive model. And the core of afuzzy system is fuzzy rules. So the central issue in this paper is to produce fuzzy rules.If the fuzzy rules can be updated according to the changes of demands and externalenvironment, we say that the system has the ability of adaption. In this paper adaptionis divided into static adaption and dynamic adaption, the main difference of which ishow to update fuzzy rules. Static adaption produces fuzzy rules by data pairs andexperts’ experience. After demands and external environment change, the responsesare not accurate. However, dynamic adaption can update itself according to thechanges. Static adaption is mainly based on k-means, c-means, classified-tree, andneural network and so on. In2009, Agus put forward an algorithm for dynamicadaption to construct fuzzy rules. This paper has improved the algorithm. Afterdividing input and output space with fuzzy method, the system can produce fuzzyrules, construct fuzzy rule base and make new rules to response new demands.However, not all new rules can meet them, they should be verified. This paper utilizesthe comparability between Petri net and fuzzy rules. And then fuzzy rules aretranslated into Petri net to construct fuzzy Petri net. We make use of fuzzy Petri netreasoning algorithm to assure the credibility of every rule. At last some rules with high credibility are kept to update fuzzy rule base.This paper has established an adaptive model and validated it. Compared withthe model established by Agus, this model has higher accuracy and the method toconstruct it is simpler. It reduces the complexity to develop a system and validatesfuzzy rules, which keep the purity of fuzzy rule base. This assures the correctness toproduce new rules further.