Density and viscosity calculations for polar solutions via neural networks

Abstract

Density and viscosity were measured for solutions of butylamine + ethanol, butylamine + ethanol + water, and butylamine + benzylamine + water over a temperature range of 303.15 - 323.15 K at atmospheric pressure. Three artificial neural networks were also developed for predicting the density and viscosity of multicomponent polar liquid mixtures with emphasis on aqueous solutions. The connection weights of the networks were determined from binary data. Those trained networks were capable of calculating density to within ± 1 % for both binary and ternary systems and calculating viscosity generally as good as that from the corresponding-states model of Lee and Wei for aqueous systems. © 1994, The Society of Chemical Engineers, Japan. All rights reserved.