Prediction of Permeate Flux in Ultrafilt

In the realm of membrane filtration, predicting permeate flux is essential for determining the membrane surface area required, a key parameter for scaling-up, equipment sizing, and cost assessment. Various models, both phenomenological (including gel-polarization, osmotic pressure, resistance-in-series, and fouling models) and non-phenomenological, have been developed to elucidate limiting phenomena and forecast permeate flux effectively. Typically, these models are tailored to specific synthetic solutions and membrane systems that demonstrate reliable predictive capabilities. However, their performance might diminish in complex matrices such as fruit juices. This study presents a comprehensive review of various model approaches for permeate flux prediction in ultrafiltration (UF) processes, accompanied by a comparative analysis of selected models’ predictive accuracy. The models were evaluated using experimental data from previously reported studies on three fruit juices—bergamot, kiwi, and pomegranate—processed in a cross-flow system over a duration of 10 hours. A robust statistical examination, including residual analysis, was employed to validate each model's predictive capacity. The findings indicated that the validated phenomenological models exhibited considerable variability in prediction, with R-squared values ranging from 75.91% to 99.78%.