Neural networks model cross flow microfiltration

Abstract

Cross-flow microfiltration can be the solution to many solid-liquid separation problems. However, its feasibility as a commercial process for use in industry is hampered by low permeation rates that result from membrane fouling. For a given product/membrane combination, it is often difficult to predict fouling rate as a function of operating conditions. This problem arises since it is the consequences of a number of dynamic and simultaneous phenomena, for example, back-diffusion, or electrochemical interactions between the product and the membrane. Compared to ultrafiltration, dynamic-knowledge based models specific to micro filtration are scarce.