| Summary: | 博士 === 國立臺灣科技大學 === 電機工程系 === 99 === As compared with conventional controllers (e.g., PID control, linear quadratic control … ), the fuzzy controllers present excess robustness and performance in plants with imprecise, unknown, or nonlinear characteristics. Large-scale complex fuzzy systems such as multi-input-multi-output interconnected systems, lead to a large computation and a significant amount of rules. Furthermore, a severe drawback is the existence of steady-state error. This thesis presents a single input PID-type fuzzy logic controller for simultaneously reducing the steady-state error and the required number of fuzzy rules.
The factors affecting a fuzzy controller include its linguistic variables of input and output, fuzzification design, membership function design, fuzzy rules base, inference engine and defuzzification design. The rule base and the membership functions are considered as two most important factors. To optimize the control performances, this thesis proposed a new fitness function and applied genetic algorithms which incorporate「Elite method」to choose the shapes of the membership functions and the scaling factors.
Finally, several simulations using Matlab/Simulink were investigated. The results showed that it can be successfully performed in two-rotor-helicopter system and permanent magnet linear synchronous motor (PMLSM) system. In this thesis, a new framework was also proposed to implement the direct thrust control method for a real-word PMLSM more important precision position control system. The steady-state error of experimental results was about 0.045mm, and the proposed control systems require only three rules.
|
|---|
