Modelling of the ultrafiltration process is crucial for effective optimization of full-scale modules, and considerable effort has been devoted to describing the permeate flux during ultrafiltration. Many models use a detailed description of the mass transport and are thus valuable for the understanding of the fundamental mechanisms of ultrafiltration. However, for the prediction of the flux in multi-component solutions, the complexity of these models is a drawback. In industrial applications, the process streams are often complex mixtures of several components. This calls for a combination of experiments and calculations instead of a strictly theoretical approach. In the work presented here, a method combining parametric studies and simulations has been used. The method is general due to its empirical nature, and its versatility has been demonstrated by applying it to three feed solutions with different filtration characteristics: bleach plant effluent with flux dependent on both transmembrane pressure and cross-flow velocity, concentrated bleach plant effluent with flux dependent on transmembrane pressure, but independent of cross-flow velocity and a colloidal silica solution with flux independent of transmembrane pressure but dependent on cross-flow velocity. Tubular polymeric membranes have been used, and both laminar and turbulent flow conditions have been investigated. It was shown that it exist an optimal cross-flow velocity when treating solutions with flux dependent on transmembrane pressure. The optimal velocity depends on the frictional pressure drop, which is a function of solution viscosity and module configuration. (C) 2009 The Institution of Chemical Engineers. Published by Elsevier BY. All rights reserved.