Anticipate UF membrane fouling.

Ultrafiltration (UF) is as an effective alternative to using alcohol precipitation to recover mostly shear-thinning exopolysaccharide (EPS)-a carbohydrate polymer-from a dilute fermentation broth. If a membrane fouls, however, it can significantly alter the flux and retention characteristics of the UF process.

Published data on membrane fouling are limited to using proteinaceous feed solutions, such as bovine serum albumin. A dynamic model simulating membrane fouling is essential to provide insights into the nature of membrane fouling during the UF of shear-thinning fermentation broth.

Researchers at the University of Delaware (Department of Animal and Food Sciences, Townsend Hall, 531 S. College Ave., Newark, DE 19717) set out to develop a model that predicts the fouling of UF membranes when they filter highly viscous fluids with shear-thinning properties. The scientists also wanted to verify the prediction model using experimental data.

Results from this research will benefit the users of commercial membrane processes involving fluids whose rheological properties are shear-thinning. These processes include the concentration by UF of juices, skim milk and other EPS materials.

Xanthan broth produced from a fermentation was used as the model shear-thinning fluid. A two-layer exponential model was developed based on the dynamics and rheology of fouling cakes formed on the membrane surface during UF. Researchers obtained experimental data using a tubular polysulfone hollow-fiber UF cartridge and a plate-type module.

The two-layer model simulates membrane fouling caused by concentration and rheology gradients during the UF of shear-thinning fluids. This is difficult to explain using a concentration polarization or resistance-in-series model. The thickness of deposits reached a steady state when the solute advection toward the membrane balanced the solute mass carried away at the trailing edge by the creeping EPS.

The experimental measurements of membrane fouling were in agreement (maximum error 5%) with the prediction model. These results can be correlated with operating parameters, allowing investigators to obtain an understanding as to how to control the fouling process.

Further information. Y.M. Lo; phone: [phone]; fax: [phone]; email:[e-mail].