In this work an experimental procedure was established to assess the main engineering parameters characterizing the ultrafiltration (UF) recovery of a commercial sample of sodium alginate from model solutions, using both a laboratory-scale and a pilot-scale plant equipped with ceramic tubular UF membrane modules. Several total recycle tests were performed in the laboratory-scale plant so as to assess the effects of transmembrane pressure difference (Delta P), feed superficial velocity (vs) and solute concentration (C-BR) in the ranges of 0.5-4.5 bar,4-10 m/s, and 3-22 kg m(-3), respectively, on the permeation flux under a constant process temperature of 50 degrees C. As C-BR increased from approximately 3 to 7 kg m(-3), the limiting permeation flux (J(P infinity)) decreased almost linearly from 140 to 320 dm(3) m(-2) h(-1) (depending on vs) to about 40 dm(3) m(-1) h(-1), the latter value being independent of vs. The change in slope of the plot J(P infinity)-vs-log(C-BR)(CBR) was shown to be due to the transition from turbulent to laminar flow. Two empirical dimensionless correlations relating the modified Sherwood, Reynolds and Schmidt numbers, valid in the laminar or turbulent flow regime, enabled prediction of the permeation flux of two independent batch mode validation tests at high and low initial feed solute concentrations to within < 10%. The correlations were also used to estimate the permeation flux for other UF tests, performed previously in a pilot-scale plant equipped with a commercial ceramic tubular membrane module operating in either batch or total recycle mode at 60 degrees C and high or low transmembrane pressure, with an accuracy of 5.6-15.2%. The two empirical correlations should be suitable as design tools for further scaling-up of the UF process under study. (c) 2008 Elsevier B.V. All rights reserved.