Loading...
AN INTEGRATED EXPERIMENTAL, TECHNO-ECONOMIC, AND MACHINE LEARNING FRAMEWORK FOR FORWARD OSMOSIS USING SODIUM METASILICATE SOL-GEL DRAW SOLUTION AND TREATED SEWAGE EFFLUENT FEED SOLUTION
Citations
Altmetric:
Video URL
Date
2026-06
Collections
Files
Firas Alkadour_ OGS Approved Dissertation.pdf
Adobe PDF, 4.63 MB
- Embargoed until 2028-07-01
Abstract
This dissertation addresses the practical limitation of sodium metasilicate (SMS) use in arid regions, namely the need for water to dilute the solution before soil application. To overcome this challenge, forward osmosis (FO) approach was proposed in which SMS sol-gel was used as novel draw solution and treated sewage effluent (TSE) as the feed solution, enabling water recovery from TSE and dilution of SMS without regeneration. The study combined experimental investigation, techno-economic assessment, machine-learning (ML) prediction, and surrogate optimization. Experimentally, a hollow fiber (HF) FO system equipped with a thin-film composite (TFC) membrane was evaluated under different feed and draw flow rates, transmembrane pressures (TMP), draw-solution molarities, and membrane orientations. Performance was assessed using water flux, reverse solute flux (RSF), specific reverse solute flux (SRSF), membrane selectivity, and specific solution cost (SSC). The system achieved a maximum water flux of 17.82 LMH, ion rejection of 97.81%, and SSC ranging from 0.16 to 0.37 $/m3. To generalize beyond the experimental conditions, a dual-output ML framework was developed using 651 literature-derived data points covering HF and flat-sheet membranes (FSMs) and 17 input variables to simultaneously predict flux and RSF. Among eight models, CatBoost achieved the highest accuracy, with test R2 values of 0.962 for flux and 0.928 for RSF, while interpretability analysis identified variables for membrane selection and process optimization. In the final stage, the verified ML model was applied as a surrogate optimization tool to overcome experimental limitations. Whereas the HF-TFC experiments covered 32 scenarios, the ML framework expanded the assessment to 512 operating cases for HF-TFC and further enabled comparison with FSM. Within the evaluated cases, HF-TFC under AL-FS orientation provided favorable balance between cost and reverse solute leakage at 40 L/h feed flow rate, 50 L/h draw flow rate, 0.3-1.0 bar TMP, and 0.3-0.4 M draw-solution molarity, achieving SSC of 0.19-0.22 $/m3 and SRSF of 0.40-0.51 kg/m3, whereas FSM-TFC under AL-DS orientation achieved the minimum SSC at 40 L/h feed flow rate, 25 L/h draw flow rate, 0.3 bar TMP, and 0.3 M draw-solution molarity, achieving 0.129 $/m3 with SRSF of 0.532 kg/m3.