| Summary: | 碩士 === 國立中央大學 === 電機工程研究所 === 94 === This thesis investigates the technique of the modifications of triangular membership functions for performance improvement of fuzzy systems by
means of genetic algorithms.
First, we propose a simple method to modify the traditional triangular membership functions. It can to tune the membership function of error by
modify method. Then, output of the fuzzy controller is transmitted into a PID controller to control plant. The controllers can stabilize a process with a minimal amount of prior knowledge and improve rise time, settling time and maximum overshoot effectively. Some illustrative examples demonstrate that satisfactory performance is achieved. Simulation results show that a better
performance is obtained, when modified membership functions are utilized.
By means of genetic algorithms, the parameters of the aforementioned controllers are determined to have a better performance. These experiment results show technique’s power and its advantages:
1. Systematic search.
2. This technique can save much more time and effort than that of
conventional trial-and-error design method.
3. This technique does not need extra professional knowledge or mathematical analysis about system dynamics.
4. The performance specifications can be achieved by means of designing a proper fitness function.
|
|---|
