Application of GMDH-type neural network to prediction of liquid-liquid equilibrium for system water + ethanol + trans-decalin

In recent years, the traditional neural networks are used to model LLE, but the performance of them related to their topologies. This paper presents the Group Method of Data Handling (GMDH) technique to estimate LLE data of a ternary system (water + ethanol + trans-decalin) at 300.2-315.2 K. The GMDH-type neural networks are selforganizing networks that the number of nodes and layers of them are adjusted during training process. The phase diagrams for the ternary mixtures including both the experimental and estimated tie-lines are presented. The results compared in terms of root mean square deviations between experimental and calculated values which were obtained using the GMDH and UNIFAC models.

کلیدواژه ها:

Liquid–Liquid Equilibrium ، Ternary System ، Artificial Neural Network ، Group Method of Data Handling