Application of neural network system in prediction of solvent activity in polymer solutions

Aqueous two-phase systems (ATPS) containing two different types of polymer molecules have been extensively used in biotechnology and polymer processing. Recently, smartpolymers such as polyethylene glycol (PEG) have been used in separation and purification of biomolecules such as aminoacids and proteins. ATPS are suitable systems for partitioning of biomolecules in two immiscible phases.1,2 Therefore, determining the thermodynamic properties and study of thephase behavior of polymer solutions are indispensable in modeling and simulation of biomolecules extractionIn this study, the activity of polymer solutions was modeled using the artificial neural network in wide range of temperature and weight fraction of polymers. The resultsobtained from the ANN model have been compared with those obtained from the UNIQUAC-FV model. The results obtainedin this work indicate that ANN is the effective method for prediction of activity of polymer solutions and have better accuracy and simplicity compared with the UNIQUAC-FV methods