An investigation into using fuzzy logic techniques to control a real-world application

Abstract

In this thesis fuzzy control is analyzed and applied to two complex processes. A process is deemed to be complex if it possesses characteristics that make it non-linear. Examples of such characteristics are backlash, hysteresis, saturation and dead zones, which are common in industrial processes. These characteristics do not allow for the easy implementation of controllers and often there would be a requirement to employ some non-linear form of control. Often with complex processes linearization techniques would be employed to enable the application of linear controllers. However, these controllers will only operate over a limited range and will have degradation in performance when subjected to unpredictable changes in parameters. Fuzzy controllers can handle nonlinear characteristics in complex processes very well and also provides some transparency between the human machine interface. Fuzzy control is based upon the fuzzy methodology that was introduced by Lotfi Zadeh (Zadeh, 1965) in his seminal paper on fuzzy sets. The first practical application (Mamdani and Assilian, 1975) paved the way for fuzzy control and although this alternate paradigm of control came up against much criticism it managed to capture the interest of many researchers. Although research into fuzzy control saw a slump in the late 70's and early 80's it rapidly progressed in the 90's with literally thousands of research papers being published. In this study this alternate paradigm of control is investigated and applied to two processes. An initial study was done on the fundamental concepts of fuzzy logic and how this methodology can be applied to develop a controller that is fuzzy logic based. A plethora of information exist on fuzzy theory however only the fundamental concepts are dealt with here. This is sufficient for the realization of a rudimentary controller. A detailed investigation was also done on the development of the rules of the knowledge base of the controller with the emphasis on the development of a fuzzy controller that has characteristics akin to a proportional-integral (PI) controller.