Estruturas de controle de biorreator baseadas em fluxos metabólicos para fermentações micro-aeradas

Abstract

The supply of oxygen as a limiting nutrient in bioreactors is a challenge, as microorganisms may have different metabolic requirements, besides the difficulty of measuring this variable in low levels. Despite the obstacles, micro-aeration has been shown to be essential for bioprocesses such as alcoholic fermentations. Therefore, using yeast catalyzed alcoholic fermentations as a case study, the objective of this thesis is to develop a control system for bioreactors based on metabolic fluxes to control the oxygen metabolic flux under micro-aeration conditions. This work involves two distinct control approaches, which used Saccharomyces cerevisiae micro-aerated cultivations as study case. The execution of the proposal was divided into four stages: (i) Obtaining of the metabolic fluxes matrices by simulations of the metabolic model iND750 and training of artificial neural networks; (ii) Conducting experiments to characterize two industrial cerevisiae strains (FT858L and FERMEL) in mini-reactors cultivations for mathematical modeling of the growth kinetics; (iii) Experimental evaluation of the control strategy with the artificial neural network incorporated; (iv) Mathematical modeling of cell growth, product formation, substrate and oxygen consumption in micro-aerated conditions, followed by operationality analysis and simulation of the control system using a biomimetic algorithm. The experiments were carried out in minimal medium (5.0g.L-1 KH2PO4, 2.0g.L-1 MgSO4.7H2O, 1.5g.L-1 of urea; 3ppm Kamoran), with hexoses as a carbon source, pH 4.5, and temperature of 30oC. In step (ii), the fermentations were carried out in mini-reactors (5 mL) equipped with a CO2 exit, in which the loss of mass due to the production of CO2 was monitored. In step (iii), following the strain selection in terms of performance, the FT858L yeast was used in batch fermentations conducted in 5 L stirred tank bioreactor with a high glucose load. Three different oxygen supply strategies were evaluated and compared: a) simple maintenance of a constant air flowrate; b) use of PID and other controllers to keep the respiratory quotient (RQ) at the desired range; c) use of a control based on metabolic fluxes with neural networks incorporated to maintain the O2 flux. The control based on neural networks was efficient in maintaining the desirable conditions of micro-aeration, leading to high yield values (0.48 gEthanol.gSubstrate-1), productivity (4.2 g.L-1.h-1) and cell viability (95%). This performance was superior to the achieved in other evaluated strategies, which presented a yield between 0.33-0.40 gEthanol.gSubstrate-1 and productivity between 3.4-3.7 g.L-1.h-1. The kinetic model incorporating substrate and ethanol inhibition and based on data generated in simple experiments in mini-reactors was able to describe the growth, product formation, and substrate consumption in conventional and micro-aerated fermentations carried out in bioreactors. The kinetic model was also used in the analysis the process operationality and in the simulations of the biomimetic control algorithm for micro-aeration. The biomimetic controller was able to maintain the O2 and CO2 fluxes at the desired set points, resulting in a productivity of 4.0 g.L-1.h-1. The results obtained demonstrate the importance of the development of precise and robust controllers for the intensification of processes whose performance is favored by micro-aeration.