Modelagem e simulação da produção de biossurfactante em biorrefinaria integrada biodiesel-bioetanol

Abstract

Biosurfactants are amphipathic molecules produced by enzymatic or microbiological routes. Biosynthesis can occur by esterifying sugars with fatty acids, with lipase as the main enzyme that catalyzes the reaction. However, such products are not yet available on the market, which is dominated by synthetic surfactants, derived from non-renewable sources. In view of this, there is a need to propose new production routes with greater economic viability than those already existing. Thus, the work in question proposes to model and simulate a bioprocess for biosurfactant synthesis through enzimatic route. The simulation was performed using equation-oriented software (EMSO).Two possible routes for the production of biosurtants were evaluated: the first employs the esterification, by immobilized lipase, of Free Fatty Acids (FFA) with xylose, followed by recovery and reuse of the enzymes, and the product separation / purification process using liquid-liquid extraction; a second consideration to the selection / purification of the product by use processes. Regarding the first proposal, in order to simulate the liquid-liquid balance, it was necessary to know the thermodynamic parameters. A solution commonly used to predict thermodynamic phase equilibrium when two liquid phases might be present is the non-random model of two liquids (NRTL). Thus, a specific objective of this work was to train a neural network to predict equilibrium parameters of the NRTL model, and then to obtain estimates of activity coefficients for compounds of interest in this work. An optimal neural configuration was obtained, adequately minimizing errors in training, validation and testing. The simulation of the first route was then carried out at EMSO, using the extractor developed for the separation process. However, there was no convergence since there is still a need for a greater amount of experimental information on the phases formed, which can be used as good initial guesses to the code implemented in the EMSO software. Despite the non-convergence, NRTL-neural modeling is an important methodological contribution to the study of separation of the esterification product via liquid-liquid extraction, which may be better applied in future works. As the separation by extraction is still not concise enough, we proceeded to simulate the second route to be evaluated, based on a proposed precipitation sequence. Considering such a process 4 alternative, the modeling and simulation were carried out successfully. The percentage of solids in the product was about 14% AGL and 86% biosurfactant. Regarding the study of energy expenditure, in general, there was a greater absolute amount of heat associated with cooling (coolers that precede the first precipitator and the second precipitator, - 216.5162 kW) than heating (esterification reactor, 20.1748 kW). With the simulations carried out it was possible to verify that the degree of purity found in previous work available in the literature could be reached for the production of biosurfactants from residues of the biorefinery environment via enzymatic catalysis, becoming an alternative in obtaining sugar esters.