Artificial Neural Network Modeling of Equilibrium Relationship for Partially Miscible Liquid-Liquid Ternary System

Abstract

The equilibrium relationship for a ternary mixture containing one pair of partially miscible components can be expressed in the form of a ternary diagram depicting a binodal curve. Depending upon the location of the point representing the composition of the mixture, the ternary diagram may be divided into three parts 1, 2 & 3, for, whether the point is on the Binodal curve, single phase region or two phase region, respectively. Because of this unique nature of equilibrium relationship modeling of such partially miscible ternary system becomes very complex. Artificial neural network (ANN) is an upcoming modeling tool & has high accuracy levels even for processes involving multivariable non-linear relationships. The objective of the present work is to develop ANN models for the system of acetic acid-water-benzene for prediction of type of resulting mixture whether single homogeneous liquid phase, two immiscible liquid phases or a single equilibrium liquid phase of the pair of partially miscible system. The equilibrium data generated experimentally has been used. Different topology of ANN architecture has been tried for model 1 & 2. Selection of ANN model is based on the comparison of % relative error for predicted output values for ANN model-1 & 2. The highlight of the present work is the successful incorporation of the linguistics variables in a model with coded values.