Ionic liquids: syntheses, experimental applications, thermodynamic modeling and simulation with Machine Learning

• Autor(a) principal: Garcia, Luis Alberto Gallo
• Data de Publicação: 2024
• Tipo de documento: Tese
• Idioma: eng
• Título da fonte: Biblioteca Digital de Teses e Dissertações da USP
• Texto Completo: https://www.teses.usp.br/teses/disponiveis/97/97139/tde-27062025-093621/
• topic: machine learning; equilíbrio líquido-líquido; FAME; FAME; ionic liquids; liquid-liquid equilibrium; líquidos iônicos; machine learning; modelagem termodinâmica; thermodynamic modeling; transesterificação; transesterification

This PhD thesis had as its general objective the use of ionic liquids as extracting solvents in separation processes, as formers of aqueous two-phase systems and as catalysts for the transesterification reaction; as well as simulating with machine learning models the prediction of fatty acid methyl esters (FAME) using molecular information from ionic liquids. Four main studies were developed. i) the use of alternative solvents to form new biphasic aqueous systems and to be able to study the liquid-liquid equilibrium of the system composed of the ionic liquid 1-butyl-3-methylimidazolium bromide ([BMIM][Br]); 1-butyl3-methylimidazolium chloride ([BMIM][Cl]); and 1-butyl-3-methylimidazolium hydroxide ([BMIM][OH]) + sodium citrate + water at 298.15 K. The experimental data obtained were modeled with the NRTL thermodynamic model. The use of this model generated deviations of less than 2%. These ionic liquids were able to form biphasic aqueous systems in the presence of sodium citrate. ii) the use of ionic liquids in separation processes of the azeotropic mixture n-butyl acetate + n-butanol and to study the phase stability of the system n-butyl acetate + n-butanol + IL ([BMIM][Cl], [EMIM][EtSO4], [EMIM][BF4]) at 298.15 K and 101.3 kPa. The experimental data were modeled and the distribution coefficient and separation factor were determined to study the viability of these ionic liquids as solvents. All the systems studied presented separation values greater than 1, indicating the possibility of being used at an industrial level. iii) to optimize the FAME synthesis process through the ionic liquid tetrabutylammonium L-argininate [N4444][L-arg] used as catalyst. A response surface design was applied through a rotatable composite core design. The influence of the variables (temperature, reaction time, methanol/oil molar ratio, catalyst loading) were studied in the FAME synthesis process. The maximum FAME yield of 95.2% was obtained under the following optimal operating conditions: temperature of 56.7 ºC, reaction time of 89.7 min, molar ratio of methanol and oil of 6.5:1, and 5.5 wt.% catalyst load. The transesterification reaction catalyst demonstrated high catalytic activity in the production of FAME. iv) perform a simulation with the Python programming language in Jupyter Notebook using machine learning models (K-nearest neighbor, Random Forest Regressor, Decision Tree Regressor model, Gradient Boosting Regressor and Multilayer Perceptrons) to predict FAME yields using information on the molecular interaction of ionic liquids obtained through the mathematical thermodynamic model conductor-like screening model for realistic solvation (COSMO-RS). According to statistical criteria such as [coefficient of determination (R2), the mean absolute error (MAE) and the root mean square error (RMSE)], the Gradient Boosting Regressor model was the one that obtained the best statistical performance to predict FAME from ionic liquids with catalytic activity. The COSMO-RS thermodynamic model and machine learning models can be used to predict fatty acid methyl esters with molecular information from ionic liquids.