Deep Study on Fouling Modelling of Ultrafiltration Membranes Used for OMW Treatment: Comparison Between Semi-empirical Models, Response Surface, and Artificial Neural Networks

Abstract Olive oil production generates a large amount of wastewater called olive mill wastewater. This paper presents the study effect transmembrane pressure and cross flow velocity on decrease in permeate flux different ultrafiltration membranes (material pore size) when treating two-phase (olive washing wastewater). Both semi-empirical models (Hermia adapted to tangential filtration, combined model, series resistance model), as well statistical machine learning methods (response surface methodology artificial neural networks), were studied. Regarding Hermia despite good fit, main drawback is that it does not consider possibility these mechanisms occur simultaneously same process. According accuracy fit models, terms R 2 SD, both model able represent experimental data well. indicates cake layer formation blockage contributed membrane fouling. The inorganic showed greater tendency irreversible fouling, with higher values /R T (adsorption/total resistance) ratio. Response ANOVA are significant variables respect for all networks, tansig function presented better results than selu function, presenting high , ranging from 0.96 0.99. However, comparison analyzed depending membrane, one fits others. Finally, through this work, was possible provide understanding modelling used treatment