Journal of Membrane Science 246 (2005) 235–247 Ultrafiltration of BSA in pulsating conditions: an artificial neural networks approach Stefano Curcio, Germana Scilingo, Vincenza Calabr` o, Gabriele Iorio ∗ Department of Chemical Engineering and Materials, University of Calabria, Via P. Bucci-Cubo 45/a, 87030 Arcavacata di Rende (CS), Italy Received 19 February 2004; received in revised form 1 July 2004; accepted 8 September 2004 Available online 28 October 2004 Abstract The aim of the present paper is to analyze membrane systems behavior, operating in pulsating conditions, by means of artificial neural networks (ANNs). Different ANNs have been developed, by means of Matlab ® Neural Network Toolbox, to model the ultrafiltration process of aqueous BSA solutions through poly-ethersulfone membranes. A specific neural network architecture, constituted by one input layer, two hidden layers and one output layer, has been finally identified by a trial-and-error procedure. The network has been trained through a selected set of experimental data obtained for a lab-scale flat sheet membrane module, equipped with a device capable of producing periodic pulses of the applied trans-membrane pressure (TMP) and feed flow rate. It has been found that the developed neural network is capable of offering very accurate predictions of actual system behavior either when it is tested within the range used for training or when the inputs combination has been never exploited during learning phase. The observed reliability of neural networks predictions of membrane performances has suggested to use them for searching an optimal pulsation frequency profile able to maximize permeate flux. The utilization of such a pulse frequency profile allows obtaining, on the basis of theoretical evaluations only, significant improvements of membrane performances with respect to UF experiments performed at fixed and constant pulsation frequencies. © 2004 Elsevier B.V. All rights reserved. Keywords: Neural network; Pulsating conditions; Membrane; Ultrafiltration 1. Introduction Membrane ultrafiltration (UF) is a pressure driven pro- cess used for separation, concentration and purification of macromolecules from solutions [1,2]. UF offers plenty of advantages; nevertheless, one of the main drawbacks of this process is the rapid permeate flux decay, due to the mem- brane fouling. In order to improve membrane performances, several different methods, based either on the chemical mod- ifications of the surface or on the variation of system fluid dynamics, have been proposed [3–5]. One of the most suc- cessful fluid dynamics techniques consists in periodically changing both the velocity profiles in the membrane mod- ∗ Corresponding author. Tel.: +39 0984 49 6711; fax: +39 0984 49 6655. E-mail addresses: stefano.curcio@unical.it (S. Curcio), germanascilingo@libero.it (G. Scilingo), vincenza.calabro@unical.it (V. Calabr` o), gabriele.iorio@unical.it (G. Iorio). ule and the applied trans-membrane pressure (TMP), thus promoting a periodic relaxation of the deposited cake and the disruption of concentration polarization layer. The ap- plication of pulsating conditions has shown that the perme- ate flux decay is not so steep as it is found in traditional membrane processes [6], thus promising very interesting per- spectives for the improvement of membrane module pro- ductivity by changing feed flow rate and TMP during the experiment. The individuation of operating parameters values that maximize membrane performances can be accomplished if a dynamic model is available. This has to be able to predict the permeate flux decay as a function of all the phenomena involved in the process. The formulation of such a model is complicated, however, by several aspects that, in some cases, have not yet been fully understood. At present, the avail- able models are based either on some restrictive hypothesis that limit their applications to a few of simple cases, or on 0376-7388/$ – see front matter © 2004 Elsevier B.V. All rights reserved. doi:10.1016/j.memsci.2004.09.004