The prediction of VLE data by conventional thermodynamic methods is tedious and requires determination of “constants”
which is arbitrary in many ways hence there is need to adopt Adaptive Neuro Fuzzy Inference System with its associative
property and its ability to learn and recognize highly nonlinear and complex relationships. In this work to create a model
for prediction of Vapor Liquid Equilibrium data for ethanol water system, the UniSim Process Design Software was used
to simulate VLE data for three different ethanol water molar mixtures. An ANFIS model having 5 layers with 9 hidden
neurons representing the fuzzy logic rules was thereafter developed to predict the vapor phase fraction of the system. The
inputs to the model were the temperature and pressure values simulated on the UNISIM software and the model was
trained using input triangular membership function, output linear membership function and the hybrid optimizer to
predict the mole fraction and vapor phase fraction of the system taken as outputs. The performance of the ANFIS model
gave R-square values of 0.9927, 0.9999 and 0.9999 for the three molar mixtures studied. A comparative plot for the
prediction showed a perfect fit for all the three systems for the ANFIS predicted values. It can therefore be concluded that
ANFIS model gives a superior predictive capability than conventional models.
Browse additional articles authored by one or more contributors to this paper.