Advanced search
Start date

Development and analysis of predictive hybrid control by fuzzy logic for polymerization processes

Nádson Murilo Nascimento Lima
Total Authors: 1
Document type: Doctoral Thesis
Institution: Universidade Estadual de Campinas (UNICAMP). Faculdade de Engenharia Química
Defense date:
Examining board members:
Reinaldo Giudici; Alexandre Tresmondi; Jorge Vicente Lopes da Silva; Fernando José Von Zuben
Advisor: Rubens Maciel Filho

Controllers design has currently a great importance in the field of açademic and industrial research. A well-projected controller may mean the success regarding the aims of production, and it also provides the production of materials with desirable specifications and it allows that the system operates under specific restrictions, considering aspects related to operability, safety and minimization of residue. Then, it is known that modeling stages are fundamental for delineation of controller strategies. However, the obtaining of precise and applicable mathematical representations to the control of most of relevant process in chemical engineering is a challenging task, because of the presence of nonlinear dynamic behaviors and space-time mutable. Therefore, an effort is done to obtain the simpler models, but provided with the essential representativity inherent to production systems. Thus, suitable control structures could be designed for each specific necessity. This work focus on the development of four multivariable nonlinear predictive hybrid advanced controllers, based on multivariable nonlinear functional fuzzy models, to polymerization processes. Such systems present high complex dynamic and hard mathematical modeling, making difficult to apply classic controllers or advanced control methodologies based on conventional models. Two study cases for the performance analysis of the proposed controllers were considered: the copolymerization process of methyl methacrylate and vinyl acetate, and industrial copolymerization of ethene/1 butene with Ziegler-Natta catalysis. The phenomenologic models of the two processes already are described in relevant literature, and they are considered as virtual plants to create dynamic data and to implement the suggested developments. Based on computational simulations, multivariable nonlinear functional dynamic fuzzy models were made – they demonstrated excellent capacity for outputs prediction from input dynamic data – and they were, subsequently, inserted in the MPC (Model-based Predicitve Control) control structure. The choice of MPC methodology to develop the proposal structures is because of its wellknown industry applicability to multivariable chemical processes, beyond it makes possible the incorporation of restrictions in controlled and manipulated variables. Finally, the performance among outlined hybrid controllers and two predictive control strategies fairly widespread in the literature were compared. The regulatory and servo problems were analyzed and satisfactory results were observed in both conditions. This demonstrates the high potential of proposed algorithms to control multivariable nonlinear systems. (AU)