Applying of ANFIS and GMDH network for modeling of Ternary Mixtures LLE (water, phosphoric acid, Isoamyl acetate)

ANFIS and GMDH type Neural Networks (NN) employing the Pareto design were used to predict ternary systems (water, phosphoric acid, Isoamyl acetate) LLE data, for which the experimental LLE data and application of some thermodynamic models such as UNIQUAC were reported previously by some of the authors. Experimental data were divided to two set: Training and Testing sets. To measure the goodness of present model, four statistical parameters were introduced as objective functions and multi-objective evolutionary optimization algorithm was employed. The performance and reliability of this type of networks was successfully approved for considered systems based on the obtained results and good agreements were found between experimental and predicated data.