Resumen de Use of mathematical methods in the resolution of chemical engineering problems
This thesis consists of a compendium of five works that illustrate the utilization of selected mathematical methods to solve specific chemical engineering problems. Hence, the thesis is intended to cover both, a review of fundamental mathematical procedures for the solution of models raised from chemical phenomena, and a demonstration of their effectiveness to obtain useful novel significant results. The opening paper explores diverse global optimization algorithms to adjust both kinetic constants and the binary interaction parameters (BIPs) for the Peng-Robinson equation of state to the experimental data. Those parameters are essential to determine the model raised from the supercritical transesterification of triolein with methanol to produce biodiesel, with CO2 as cosolvent, consisting of three reversible reaction in series. Here, a novel model merging the ordinary differential equations system raised from kinetic mechanism and the time-dependent thermodynamic state of the complex mixture is presented for diverse operating conditions. Among all results obtained, novel binary interaction coefficients for the intermediate reaction species (dioleins and monooleins) highlight. The second and fourth papers included in this thesis are aimed at the study of lanolin extraction from raw wool, using 5% ethanol in CO2. The former explores solid lanolin extraction under near-critical conditions by means of a mass-transfer model based on the shrinking-core concept, while the latter is addressed at the liquid lanolin supercritical extraction. Both models result in a partial differential equations (PDEs) system determined by the solubility of multiphasic lanolin, Henry-type partition coefficient and the lanolin mass transfer coefficient. Hence, in each paper the raised PDEs system is solved through a different method: in the second paper orthogonal collocation method is employed, while in the fourth paper finite differences method is used combined with the numerical integration of an expression previously obtained by means of the Laplace transform. Finally, an optimization procedure is used in order to fit the extraction parameters to the experimental data, achieving coherent results that agree well with those previously reported. Between the cases exposed, liquid lanolin extraction is significantly complex to model because of the diffusion phenomena that may occur inside the two lanolin fraction mixture added to the diffusion of solvent in the interphase. Therefore, in the third work a nonlinear autoregressive exogenous neural network model is designed to predict the outcoming extracted fraction of lanolin at diverse temperatures, pressures, solvent mass flow rates, wool packing densities and times. The problem with the scarce data available for training of the neural network is overcome by augmenting experimental data using an empirical Weibull function, which correctly predicts the lanolin breakthrough at the extractor exit. This hybrid Weibull - Neural Network algorithm results in a low prediction error and conform a powerful tool for optimizing operating conditions, proved by the fast convergence of genetic algorithm procedure. This thesis closes with Molecular Dynamics simulations for peptide-folding studies, followed by a Principal Component Analysis (PCA) and clustering analysis to understand the Free Energy Landscape of the peptide (FEL). Those methods are aimed at assessing the conformational profile of bombesin, a peptide with interest in drug design as a possible novel agonist and/or antagonist in the fight against cancer. Results suggest that the peptide adopts mainly helical structures at the C-terminus and, to a lesser extent, hairpin turn structures at the N-terminus. Those results agree with those available from NMR in a 2,2,2-trifluoroethanol/water (30% v/v), and point out a suitable a-helix conformation for binding where Trp8 and His12 interaction has a significant role.
