Modelo de predicción de la concentración de cloroformo durante el proceso de destilación de una mezcla metanol-cloroformo

Abstract

The goal of this work is to model the pressure swing distillation system to separate the methanol-chloroform mixture and predict the concentration of chloroform by artificial intelligence. The methanol-chloroform mixture generates a minimum boiling azeotrope with approximately 64 mol% chloroform at 327 K under atmospheric pressure. The simulations of the distillation system are carried out with the DWSIM software. Related experimental data from the literature have been used to construct the model. The prediction model uses one hidden layer and 100 neurons in the hidden layer. The temperature and the molar fraction of chloroform in the feed, the reflux ratio, and reboiler temperature in the low and high-pressure column have been selected as input variables and the molar fraction of chloroform and flow rate in the distillate and residue of the columns as output variables. Pearson's correlation coefficient of 0.99999 and mean squared error of 1.52 E-14 for a set with 100 training and test data from the network; and a statistical p-value greater than 0.05 in the network validation with a new set of 25 data, confirm that there is reasonable compliance between the predicted values and the actual data. The results indicate that the artificial neural network model proved to be efficient in predicting the chloroform concentration obtained by distilling a methanol-chloroform mixture with a constant feed of 100 kmol / h in a pressure swing distillation system that operates with two columns maintained at 1 and 10 atm. It is recommended to use the prediction model to calculate the composition of the products obtained distilling by pressure oscillation other binary or multi-component mixtures that exhibit minimum boiling azeotropes.