Simulación para la deshidratación y recuperación de bioetanol en un proceso de destilación por oscilación de vapor asistida por bomba de calor

Abstract

The objective of the work was to develop an artificial neural network of the bioethanol recovery and dehydration process in a heat pump assisted steam swing distillation process. Heat pump assisted distillation for bioethanol recovery followed by pressure swing adsorption for bioethanol dehydration is simulated in DWSIM using ChemSep distillation columns and the non-random two-liquid model (NRTL) to thermodynamically model the liquid phase fugaciousness of the azeotropic water-ethanol mixture. A supervised learning artificial neural network was trained with the Bayesian Regularization algorithm and an architecture consisting of 6 input neurons, 40 neurons in the hidden layer and 4 neurons in the output layer. Through the optimization analysis, a base of 201 data was obtained to develop the artificial neural network. A total of 150 data were used for training and 51 data were used to validate the prediction. The prediction capability during the training of the network was evaluated through the optimal mean square error (2.81E-08) and the correlation between the actual and predicted data through Pearson's coefficient (0.99). A P-value above the significance level for all the variables compared allowed us to conclude that there is no statistically significant difference between the means of the actual and predicted results by the artificial neural network.