Neural network model prediction of chromium separation in polyelectrolyte-enhanced ultrafiltration

Abstract

Effectiveness of polyelectrolyte-enhanced ultrafiltration in chromium recovery from its aqueous solutions was tested experimentally. Two chromium species, Cr(III) and Cr(VI) ions, were the subject of ultrafiltration processes enhanced with two water-soluble, ion-exchanging polyelectrolytes. These were: poly(sodium 4-styrenesulfonate) - PSSS (for Cr(III) ions recovery) and poly(diallyldimethylammonium chloride) - PDDAC (for Cr(VI) ions recovery). Experimental ultrafiltration tests with two different membranes and model solutions of appropriate chromium ions (5 and 50 mg dm -3 ), at different pH and with various polyelectrolyte doses, provided numerical data for the artificial neural networks training procedure. Numerical neural network models made prediction of chromium retention coefficient (R) under different process conditions (pH, polymer dose, concentration of selected Cr form) possible. Strongly nonlinear dependences of retention coefficient (R) on pH and polymer : metal concentration ratio for both chromium species, represented by experimental data, were identified and modeled by neural networks correctly. Good compatibility between experimental data and neural network predictions was observed.