Projecto e Aplicação de Controladores Difusos em Tempo-Real
Authors
Abstract
This dissertation intends to give a contribution to the understanding of the potentialities of fuzzy logic and fuzzy sets in automatic control. In this context, fuzzy logic and fuzzy sets constitute an efficient way to deal with the fuzziness and imprecision related to the knowledge of the processes to be controlled. Although relatively recent, fuzzy logic has come to affirm itself in some areas of knowledge, existing currently a high number of applications using it.The theoretical foundations of fuzzy control are studied based on concepts from fuzzy logic and fuzzy set theory.
The structure of the fuzzy controller is presented, the functions and the parameters associated to its constituent modules are described, namely, the fuzzification module, the knowledge base, the inference mechanism and the defuzzification module.
In this work, some types of fuzzy controllers are studied, particularly, Mamdani controllers, the fuzzy sliding mode controller and the Sugeno controller. Questions related to the analysis, design and tuning of fuzzy controllers are discussed.
The thematic of adaptive fuzzy control is also studied, being presented the structure of the adaptive fuzzy controller. The adaptive characteristics of the fuzzy controller are developed accordingly to three perspectives: scaling factors adaptation, membership functions adaptation and rules adaptation. In the particular case of the scaling factors and the membership functions, new adaptation techniques are proposed and tested.
The controllers being studied have been applied to the control of a laboratorial thermal process, PT 326, of Feedback Instruments. This process present some features found in several industrial processes, and for this reason, the achieved results are relevant to the implantation of this control methodology in industrial environments.
The performance of the fuzzy controllers has been evaluated and quantified in some situations. The performance of the fuzzy controllers was also compared to other control methodologies, such as the conventional control and the adaptive neural control.
In order to implement all the studied controllers, a computational tool was developed to assist in the different stages of the design, particularly of fuzzy controllers, and in the simulation and real time process control.