A Study on Modeling and Control of Singularly Perturbed Fuzzy Systems

• Main Authors: Kuo-Jung Lin, 林國榮
• Other Authors: Tzuu-Hseng S. Li
• Format: Others
• Language: en_US
• Published: 2005
• Online Access: http://ndltd.ncl.edu.tw/handle/94390695915828782564

博士 === 國立成功大學 === 電機工程學系碩博士班 === 94 === 　This dissertation presents several fuzzy control designs to stabilize the singularly perturbed fuzzy systems with guaranteed H infinity control performance. Firstly, by using Takagi-Sugeno (T-S) fuzzy model, we transform a nonlinear singular perturbation system into a singularly perturbed fuzzy system, where the corresponding fuzzy slow and fast subsystems of the original singularly perturbed fuzzy system are also obtained. Then we propose a series of fuzzy control designs including state feedback fuzzy control, composite fuzzy control, direct fuzzy control, observer-based fuzzy control, static and dynamic output feedback fuzzy controls. Based on the Lyapunov stability criterion, the stability conditions are reduced to a linear matrix inequality (LMI) problem. By the guarantee Epsilon-bound issue, the allowable perturbation bound Epsilon* can be determined by some simple algebraic inequalities such that the proposed fuzzy control stabilizes the singularly perturbed fuzzy systems for all Epsilon belong to (0, Epsilon*). Moreover, the developed criterion guarantees a minimum disturbance. Furthermore, the singularly perturbed fuzzy system with parametric uncertainties is also examined. For better transient behavior of singularly perturbed fuzzy system, the pole placement problem is also developed. Some practical systems, for example, the nonlinear circuit and the DC motor driver inverted pendulum systems, are given to illustrate the validity of the proposed schemes. All the simulation results demonstrate the proposed fuzzy control designs are feasible and satisfactory. The main contribution of this dissertation is that all the proposed controller gain matrices can be determined by the LMIs and the corresponding Epsilon upper bound can be calculated.