| Summary: | 碩士 === 國立交通大學 === 控制工程系 === 82 === In this thesis, we study the neural-network-based fuzzy
inference systems. To realize the rule reasoning of fuzzy
inference systems, two fuzzy neural networks, the FNN and
NFNN, are presented in this thesis. The proposed fuzzy
neural networks can acquire the fuzzy logical rules by
employing the learning capability of neural networks.
Moreover, for simplifying the structures of the proposed
fuzzy neural networks, the redundant rules and linguistic
terms should be removed from the FNN and the NFNN. With this
problem, we utilize the fuzzy similarity measure in the FNN to
combine the similar rules and linguistic terms. While for the
NFNN, the fuzzy rules are reduced by means of a rule
combination procedure. At last, the fuzzy modeling of
nonlinear systems are applied to illustrate the proposed
fuzzy neural networks. By means of the simulations, both the
FNN and the NFNN can be successfully used on the fuzzy
modeling of nonlinear systems. Moreover, the fuzzy
similarity measure and the rule combination procedure are
effective to reduce the structures of the FNN and the NFNN.
|
|---|
