A genetic-based fuzzy membership set for a fuzzy logic control system

number: 
2218
إنجليزية
Degree: 
Author: 
Saad Ghassan Ibrahim Zako
Supervisor: 
Dr. Ausama A. Al-Ani
year: 
2009

Abstract: Fuzzy logic and fuzzy control are incorporated in many forms to different applications. Whether it is used in design or modeling of systems, large number of parameters are needed to be specified in the design process. Among these parameters the selection of the membership types does not gained much attention as other parameters. This thesis will focus on this parameter and presents this approach to the design of a fuzzy logic controller based on a selection of membership function with different types for good performance of control systems. Changing the type of membership function on the universe of discourse (UOD) for the input variables is automatically managed by the genetic algorithm (GA) to select best shapes and led to the possibility of designing simple fuzzy logic controller capable to provide for good performance control strategy. The PID-like fuzzy logic controller is designed to be in a simple form except the shape of membership functions which will be optimized using the genetic algorithm. The output of the FLC is taken to be singletons switched to a predefined levels calculated based on the number of the input membership functions with symmetric distribution along the universe of discourse. In other words, the distribution along the universe of discourse is taken to be symmetric and uniformly equidistant to improve the performance of the design systems, scaling factor are also added to the chromosome by the genetic algorithm to be optimized. This thesis presents the design of a fuzzy logic controller with variable types of membership function for good performance control system. Two examples to examine the applicability of the design methodology were calculated, type 1 3rd order differential equations as a linear system and the inverted pendulum as a non-linear system. The obtained results show the applicability of the presented methodology. Comparison to other design approaches with respect to some well known dynamic performance characteristics like overshoot, rise time, settling time and steady state error (S.S.E) are also presented. The result shows that the shape of the membership functions is very important to be taken into consideration, and mixing between the different types of memberships need to be given more and more attention, and need not to be neglected as many designers do.