Interest of neural networks for the optimization of the crossflow filtration process
Résumé
In order to build up a model representing the effect of transmembrane pressure and crossflow velocity on crossflow filtration results at quasi-steady state, an approach based on neural networks is proposed. For filtralions of various products (raw cane sugar remelt, natural gum solution) on different membranes (micro- and ultrafiltralion) with or without co-current permeate flow, the modelling of both permeate flux and retention rate could be obtained after only five experimental trials. Compared to more classical modelling techniques, the neural networks were showed to be sometimes better suited and are useful when the effects of hydrodynamical conditions on filtration results are strongly nonlinear. Thanks to established models, it was possible to determine, with a good safety margin, an optimum region in every case studied.