| Abstract
| - A model based on a feedforward back-propagation neuralnetwork was employed to predict thephase equilibrium diagram of the aqueous two-phase systems. ThePEG/potassium phosphate/water system (pH 7) was selected as the model system to demonstrate thepoint of interest. Avariety of molecular weights (MW) of PEG systems including PEG 600,1500, 3400, 8000, and20 000 were considered for training the patterns in order to estimatethe systems with PEGMW of 400 and 1000. After the optimal architecture of the networkwas investigated and finallydetermined, the extrapolated and interpolated simulations by this modelexhibited an excellentagreement with experimental data. The characteristics of the phasediagram such as the binodalcurve and tie lines were illustrated in precision in all trials.The model can associate thedependence of PEG MW with the subtle shift of the corresponding phasediagrams over the testMW range. All the equilibrium data of the PEG/potassium phosphatesystems with continuouslyvariable PEG MW ranging from 20 000 to 400 could be predicted by themodel. The resultsindicated the applicability of the neural network model as adesign-oriented technique foroptimization of extraction condition. The neural network modelshould be a potent means todeal with more complex models such as PEG/dextran systems and partitionof proteins in aqueoustwo-phase systems.
|
