Estimativa de parâmetros de modelos não-lineares usando o EMSO, um programa de código aberto para modelagem, simulação, controle e otimização de processos

Abstract

Simulators are useful tools for applications ranging from project validation, plant operability to increased production and cost reduction. These characteristics, and others, have increased industrial interest in software for modeling, simulation, control, and optimization of processes. However, there is still great dissatisfaction among users, mainly regarding use/learning and high cost. In the present work, non-linear parameter estimation routines were developed on the EMSO (Environment for Modelling, Simulation and Optimization) platform, a free and open-source program, aiming at the simulation and discrimination of the most suitable models for describing the kinetics of cell growth of the yeast Saccharomyces cerevisiae in a batch bioreactor for alcohol production. The experimental data used in this work were adopted from works available in the literature. The kinetic models of Monod, Moser, Contois and Andrews were evaluated. The criteria used to discriminate the models were the mean absolute deviation, the mean absolute percentage error and the square of the correlation coefficient between predicted and observed values. Based on the results obtained, it was possible to conclude that the Contois model presented the best results with an absolute mean deviation of 4.1683 g/L and 0.2451 g/L for the substrate and cell growth curve, respectively, and an error average absolute percentage of 9.82% and 17.67% for the substrate and cell growth curve, respectively. Furthermore, it presented correlation coefficient values above 0.98 for both curves.