The aim of this work was to validate the dynamic model for the average permeate flux that considers membrane fouling as a dynamic process from non-equilibrium to equilibrium to predict permeate flux decline with time in the ultrafiltration of macromolecules. This is one of the most accepted ultrafiltration dynamic models and it considers cake formation as the main fouling mechanism. For this purpose, several experiments were performed at pilot plant scale and the results were compared with the ones predicted by the model. A ceramic TiO2-Al2O3 monotubular membrane (Tami, S.A., France) of 5 kDa molecular weight cut-off was used. Aqueous solutions with different concentrations of polyethylene glycol were used as feed. The flow rate and the transmembrane pressure were varied while the temperature was maintained constant at 25 degrees C. As the model does not consider cake compaction, the best results were obtained for low transmembrane pressures, as expected. Crossflow velocity was observed to be one of the main factors affecting fouling. The best agreement between the experimental results and those predicted by the model was obtained for low crossflow velocities as cake formation can be considered as the main fouling mechanism in this case.