Untitled

Crossflow microfiltration experiments were performed on aqueous dispersions of titanium dioxide through a 0.1 µm pore size ceramic membrane at various operating parameters. The initial transient flux decline follows dead-end filtration theory, with the membrane resistance determined from the initial flux and the cake resistance determined from the rate of flux decline due to cake build-up. For long times, the observed fluxes reached steady- or nearly steady-state values, presumably as a result of the cake growth being arrested by the shear exerted at its surface. The steady-state fluxes increased with increasing inlet crossflow velocity and decreasing feed concentration. Rheological work has shown that the titania dispersions exhibit shear-thinning behaviour. Extreme sensitivity with pH was observed, whereby the dispersion viscosity can be changed by as much as an order of magnitude with pH variation at constant volume fraction. The steady-state permeate flux values were determined from the steady-state model based on the Kozeny–Carman equation for cake resistance and Darcy’s law applied over the filter area to relate filtration rate to average pressure difference between the feed and permeate sides of the filter. The model includes a cake resistance of the cake layer, which was determined for the titanium dioxide dispersions by fitting the experimental flux data to the model. The resulting fluxes obtained from the model using simple values for the membrane resistance, cake resistance, and rheological parameters for each data set are in good agreement with the measured fluxes.

Fig. 5. Comparison between predicted and experimentally observed flux–time curves at different operating variables. respective values of the membrane resistance and the cake resistance cited above were used. The symbols are the measured fluxes, and the solid lines are the fluxes predicted by the dead-end theory and by the steady-state model based on the Kozeny–Carman equation for cake resistance and Darcy’s law applied over the filter area to relate filtration rate to average pressure difference between the feed and permeate sides of the filter. Tube constriction due to formation of thick cake layers was accounted for in the theoretical predictions. Predicted fluxes show a notably good correspondence with experimental values. This is remarkable in view of the fact that correlations employed in the model were developed for systems vastly different from the crossflow microfiltration system used in this work. The filtration flux correlated well with wall shear stress, reducing to a unique relation in the absence of pore plugging, i.e., when the filter resistance is dominated by the properties of the deposit and flow conditions. Thus, the filter tube diameter and rheological properties of the dispersions will influence the wall shear stress and have a consequent effect on the flux.