Modelagem de processos enzimáticos e fermentativos usando otimização por enxame de partículas

Abstract

The process modeling and simulation is a greatly important procedure for many chemical and biotechnological processes. The process simulation allows to predict elementary behavior of the state variables of the process, leading to many economical and process advantages, such as the avoidance of losses of time and materials for not knowing the process particularities, the safety guarantee, the product quality, and, mostly the process optimization, permitting to study and to reach the best conditions of a process, which shall yield in more products with quality produced with less effort and expenditures. In this work, the modeling and simulation of two biological processes enzymatic hydrolysis and solid state fermentation were performed in order to develop models and estimate parameters that enable an engineer to predict the process behavior and to make decisions about the process. The modeling procedure also involves the computing of differential equations, and algebraic-differential equations; in this manner, the engineer must be able to use different numerical integration methods. Mainly two parameters estimation procedures were used Particle-Swarm Optimization and Levenberg-Marquardt -, and two numerical integration methods were also resorted Runge-Kutta and Dormand-Prince. Experimental data from previous works were used to perform several tests in order to assure that the models were predicting correctly the state variables of the process and, in this manner, were reliable and useful. In the first paper, several mechanistic and empirical models are tested to fit the enzymatic hydrolysis experimental data; statistical tests were performed to verify which of those models would best describe the process, which was found to be and entirely empirical non-autonomous model. The following papers are about the modeling of the solid-state fermentation process. The model was found to be very accurate and adequate to be used for predictions, mainly for the bed temperature of the packed-bed bioreactor, since it could predict the temperature gradients along the time and height of the bed. Also, some numerical procedures such as parameters identifiability, to realize which were the most important parameters to be estimated, and model reparametrization, to reduce the total number of parameters to be estimated and avoid magnitude problems of the model, were successfully performed. This work has shown that the modeling and simulation of processes holds an enormous importance for industry, and different techniques can be applied with more or less effort and success. Further, hopefully, this work has contributed to the state of the art of modeling, in a general way, for biological processes.