Modelagem fuzzy-cinética e monitoramento espectrofotométrico da hidrólise enzimática do bagaço de cana-de açúcar em sistema de reatores com separação de fases

Abstract

The decrease in fossil fuel reserves combined with the growing global energy demand fosters the search for alternatives. Within this scope, a possible solution may arise from the use of biofuels, such as ethanol, produced based on renewable sources. In Brazil, in alignment with such efforts, this alcohol is already produced on a large scale from the fermentation of sugarcane juice. This process can have its yield increased by using the remaining sugarcane bagasse itself as a substrate in alcoholic fermentation. The ethanol produced in this way is called second-generation ethanol (2G). To enable this process, the bagasse must be pretreated, exposing its fibers and allowing its hydrolysis. Hydrolysis is responsible for breaking the cellulose molecules into monosaccharides that can be consumed by yeast during fermentation. Enzymatic hydrolysis occurs under mild conditions and releases fewer compounds that inhibit fermentation, which, despite the high cost of the enzyme, makes it more attractive compared to chemical hydrolysis. Based on this context, the present project aimed to study a non-conventional reaction system built in the research group, using experimental data from liquid chromatography and infrared absorption spectroscopy obtained previously. Due to the complexity of the reaction system, the application of phenomenological models is impractical, and semi-mechanistic approaches are presented as an alternative. Therefore, a fuzzy-kinetic model based on Michaelis-Menten kinetics was developed aiming to obtain an equation system capable of describing the process for different feed conditions. This study presents significant results, highlighting the creation of an innovative kinetic model for the enzymatic hydrolysis of sugarcane bagasse in reactors with phase separation. The identifiability analysis of the model parameters revealed its robustness in describing the reaction kinetics. Anticipating possible deviations between the model and the real reaction, it was chosen to develop a monitoring system using near-infrared spectroscopy. In this work, a predictive concentration model based on spectra was developed through a combination of variable selection by genetic algorithm and partial least squares regression, resulting in a model with cross-validation errors of ±0, 012 g/L for cellobiose, ±0, 650 g/L for glucose, and ±0, 072 g/L for xylose.