| Summary: | 博士 === 國立臺灣科技大學 === 電機工程系 === 99 === This dissertation proposes a novel control method for identification of a class of uncertain systems by using on-line adaptive T-S fuzzy-neural modeling. And the robust controller is designed to compensator modeling errors and external disturbances. This dissertation uses the mean value theorem to transform the nonlinear system dynamic into a virtual linear system because the most systems are nonlinear. Then the T-S fuzzy-neural model can identify the dynamic model of the linearized system. Although T-S fuzzy-neural modeling is an efficient identification method for uncertain systems, it encounters serious problem of fuzzy rules explosion in processing a high dimensional system. Furthermore, this problem leads to large computing time. Therefore, we propose a kind of hierarchical structure through which the complex structure of fuzzy-neural networks can be modeled by using a family of subsystems with fewer dimensions. By this hierarchical structure, the fuzzy rules and the computation time will decrease. Finally, this dissertation gives some examples for affine nonlinear systems, and the simulation results illustrate that the proposed controller design presents good performances and effectiveness.
|
|---|
