Bioprocess optimization: evaluation of performance using deterministic and genetic algorithm approaches

Author(s)

Elmer Alberto Ccopa Rivera

• Document type: Doctoral Thesis
• Press: Campinas, SP.
• Institution: Universidade Estadual de Campinas (UNICAMP). Faculdade de Engenharia Química
• Defense date: 2006-12-13

Examining board members

Rubens Maciel Filho; Beatriz Vahan Kilikian; Francisco Maugeri Filho; Carlos Eduardo Vaz Rossell; Daniel Ibraim Pires Atala

• Advisor: Aline Carvalho da Costa; Rubens Maciel Filho

Abstract

This thesis addresses the issues of modeling, systematic and reliable optimization methodologies, analysis and computational tools. These fields can capture essential quantitative features of the processes and have a great impact on the modern practices of petrochemical and mechanical process performance improvement. It is evident that the same should occur with bioprocesses, a field in which very little studies and developments has been made. However, it is expected that the bioprocesses built upon methods that consider the problem analysis, codified into a mathematical tool appropriate for the use of computer aided process engineering tools, will transform numerical results into useful engineering solution to finally produce a better understanding of the bioprocess. Different types of models, suitable for computer-aided applications, have been studied for this purpose. The proposed methods are applied to the development and optimization of bioethanol production. The main part of the thesis consists of journal papers (Chapters 3 - 8). Chapter 3 focuses on the optimization techniques to estimate the kinetic model parameters of batch fermentation process for ethanol production using Saccharomyces cerevisiae. The potential of Quasi-Newton (QN) and Real-Coded Genetic Algorithm (RGA) to solve the estimation problem is considered to find out the optimal solution. Chapter 4 provides an adaptive hybrid model that considers the effect of temperature on the kinetics of alcoholic fermentation, represented by neural networks. The potential of a RGA and a binary-coded genetic algorithm (BGA) to re-estimate the parameters of the neural network is evaluated. These algorithms were compared to the quasi-newton algorithm (QN) in terms of performance of the hybrid model. In the Chapter 5 a procedure for the development of a robust mathematical model for an industrial alcoholic fermentation process was evaluated. The proposed model is a hybrid neural model, which combines mass and energy balance equations with Functional Link Networks to describe the kinetics. These networks have been shown to have a good non linear approximation capability, although the estimation of its weights is linear. Chapter 6 proposes a new methodology to estimation of kinetic parameters. The first step is to obtain initial values for all parameters in the model and then a RGA is used to calculate the parameters in the model. The third step is to identify the most significant of the parameters using Plackett and Burman design (PB) and finally the most significant are optimized using a QN algorithm, which converges much more quickly than RGA to the optimal. In the Chapter 7 the modeling of biotechnological processes is studied with focus on developing methodologies that can be used always that a re-estimation of parameters is necessary. The performance of a hybrid neural model and a first-principles model, both considering the effect of temperature on the kinetics, are evaluated not only by their accuracy in describing experimental data, but mainly by the difficulties involved in the adaptation of their parameters. In the Chapter 8 the non-linear model of an extractive alcoholic fermentation process, represented by neural networks, is optimized using RGA and BGA to determine the optimal operational conditions. In order to check the validity of the computational modeling, the results were compared to the optimization of a deterministic model, whose parameters were experimentally determined as functions of the temperature. Furthermore, Appendix A provides additional information regarding the verification of the non-linear behavior of the variables of the extractive alcoholic fermentation process. With the purpose to compare the optimization performances of this process by genetic algorithms, Appendix B presents the results of optimization using successive quadratic programming (AU)