Modeling sulfamethoxazole removal by pump-less in-series forward osmosis–ultrafiltration hybrid processes using artificial neural network, adaptive neuro-fuzzy inference system, and support vector machine

Abstract

This study presented an in-series forward osmosis–ultrafiltration (FO–UF) hybrid system for sulfamethoxazole (SMX) removal. Artificial neural network (ANN), adaptive neuro-fuzzy inference system, and support vector machine were employed to predict water flux and SMX removals by FO and FO–UF. This investigation relied on 60 experimental data sets that varied the initial draw solution (DS) concentration (1–5 M), initial SMX concentration (2.5–12.5 mg/L), initial pH (3–11), and natural organic matter (NOM) content (0–18 mg/L as dissolved organic carbon). Experimental results demonstrated that the hybrid system achieved 83%–93% and 91%–99% SMX removals via FO and FO–UF, respectively, while the obtained water flux was 5–14 L/m2h. From the three machine learning models, ANN had the most accurate prediction results, with statistical R2 of 0.96, 0.91 and 0.99 for water flux and SMX removals by FO and FO–UF, respectively. For the best ANN model, relative importance of the input variables to water flux and SMX removals by FO and FO–UF, respectively, was in the following order: DS concentration (41%, 49% and 36% in the aforementioned order), NOM concentration (21%–28%), initial SMX concentration (15%–24%) and initial pH (11%–17%). © 2023 Elsevier B.V.