Sintese de um controlador hibrido fuzzy-preditivo : aplicação para processos de polimerização

AUTOR(ES)
DATA DE PUBLICAÇÃO

2003

RESUMO

This work presents the development of a new predictive controller based in functional fuzzy models for polymerizatíon processes. These processes present a highly non-linear dynamic behavior, thus making difficult the development of controllers based on model as conventional predictive controllers. Two case studies were considered to analyse the performance of the proposed control/er, to know: process of copolimerization (Congalidis et al. 1989) and policondensation (Giudici et ai. 1999). The copolimerization process presents a recycle system that leads to some disturbances in the reactor input of the process. This problem is solved through a feedforward controller, developed by Congalidis et al. 1989. In this way, the reactor can be interpreted as an isolated process system for the control system design. For the polycondensation process, a dynamic model was developed through the sequence of ten reactors CSTR. This model satisfactorily represents the dynamic behavior of the main process variables for the development of the control system. Taking this into consideration the success of some applications of predictive controllers in chemical processes, and also its ability to consider the restrictions on the manipulated and control/ed variables. This type of controller was used as a basis for the development of new control algorithm coupling the fuzzy concepts together with the model predictive controllers. Thus, it explicated in this work the development of a methodology for the design the predictive control/er being based on functional dynamic models fuzzy (Takagi e Sugeno, 1985). These models present an excellent capacity to represent dynamic data. Moreover, they allow the inclusion of qualitative or operational information of the process. The fuzzy model determination (rules number and model parameters) is obtained from the process database. The treatment of these data for the fuzzy model determination is carried out by means of mathematical algorithms of clustering and least squares. The modeling by the fuzzy approach showed to have a good potential for representation. The fuzzy internal models were developed based on the functional dynamic fuzzy representation (Takagi e Sugeno, 1985). The proposed fuzzy based controller were compared to the dynamic matrix controller (DMC) and the obtained results showed that the proposed controller is robust and does not require step test as conventional controllers

ASSUNTO(S)

polimerização controle de processos quimicos logica difusa controle preditivo

ACESSO AO ARTIGO

Documentos Relacionados