Nova metodologia para análise de identificabilidade e estimação de parâmetros de modelos fenomenológicos

AUTOR(ES)
FONTE

IBICT - Instituto Brasileiro de Informação em Ciência e Tecnologia

DATA DE PUBLICAÇÃO

2012

RESUMO

Model identification is the activity of constructing representative mathematical models of real systems. Several techniques are available for its completion, which depend on the level of information available about the studied system. In this work techniques for gray-box identification, which use both phenomenological models and experimental data, were studied by focusing on two main steps: identifiability analysis and parameter estimation. Identifiability analysis is a fundamental tool in the identification process and must precede the parameter estimation. Your goal is to verify, from a model structure and experimental data which parameters of a model can be identified. In this work the most important methodologies available in the literature were analyzed. Subsequently, a new methodology was developed, based on concepts of control theory, which aims to overcome the limitations inherent in other methods. The parameters estimation was made considering two distinct dynamic optimization methods: the single-shooting method and simultaneous method. Furthermore, in this same step, several formulations of objective function were tested, checking the effect of minimizing the square of the derivative of the error when compared to the traditional approach, which does not make use of the derivative. To elucidate the methodology, two case studies were proposed: Continuous Alcoholic Fermentation Reactor and Williams-Otto Semi-Batch Reactor. In both cases, the experimental data were generated from the simulation of the model under know parameters, operational conditions and inference in order to verify the effect of each of these peculiarities. To consolidate the studies developed a real application was also tested, and this identifies the model of a polymerization reactor HDPE using data from an industrial plant.

ASSUNTO(S)

processos químicos : modelagem controle de processos químicos modelos matemáticos

Documentos Relacionados