| Summary: | 碩士 === 大同工學院 === 資訊工程研究所 === 87 === Fuzzy set theory was first invented by Lotfi Zadeh in 1965 and has been developed over more than thirty years. Using fuzzy logic, we can build a fuzzy model including both the properly identified structure and parameters. After a rough fuzzy model is created, how to modify the structure and how to adjust the system parameters so that the fuzzy model can achieve high performance are important problems.
In this thesis, a new fast tuning algorithm is proposed to expedite converging process. In order to improve the disadvantages of the time-consuming gradient descent method, the principle of this new algorithm is only to tune the consequent parts of the fuzzy rules and the membership functions of the fuzzy model are always fixed. The fast tuning method includes two main phases. Using the average of the output data to initialize the consequent part values for each fuzzy rule is the first phase. Second, the importance of this phase is to add the system error outputs to the consequent part of the fuzzy rules as the updated values. We will describe the proposed algorithm in detail in this thesis.
Besides, in order to simplify the complexity of designing the fuzzy model, we use the technique of the grey relational analysis to obtain the relationship between the system input and output variables. According to the grey relational degrees, the more important input variables are selected to establish the fuzzy model.
Finally, some simulation results are presented to verify that the proposed method can converge speedily and has better inference capability than conventional methods.
|
|---|
