An osmotic membrane bioreactor–clarifier system with a deep learning model for simultaneous reduction of salt accumulation and membrane fouling

• Published in: Chemosphere (Oxford) 2021-06, Vol.272, p.129872, Article 129872
• Main Authors: Viet, Nguyen Duc, Im, Sung-Ju, Kim, Chang-Min, Jang, Am
• Format: Article
• Language: English
• Subjects: Artificial Intelligence; Bioreactors; Clarifier; Deep Learning; Hybrid configuration; Membranes, Artificial; Osmosis; Osmotic membrane bioreactor; Salinity reduction; Waste Water; Water Purification
• ISSN: 0045-6535; 1879-1298
• DOI: 10.1016/j.chemosphere.2021.129872

This study investigates a novel hybrid configuration of an osmotic membrane bioreactor–clarifier (OMBRC) to achieve the simultaneous reduction of salt accumulation and membrane fouling. Compared with the conventional OMBR, the OMBRC demonstrated 14 times lower conductivity after 40 d of operation, achieving maximum values exceeding 25,000 and 1800 μS/cm, respectively. The average water flux and flux recovery were approximately 3 and 6 times higher in the OMBRC than in the OMBR, respectively. The ammonium, total organic carbon, and total nitrogen removals of the combined system were measured to be 15%, 11% and 7% higher in the hybrid OMBRC process than in the OMBR, respectively. The hybrid process also reduced the foulant layer thickness in this system to only 15 μm compared with 28 μm in the OMBR. An artificial intelligence-based model was successfully developed for long-term prediction, indicating that the advantages afforded by the hybrid OMBRC can be maintained over long periods of operation with 22 times lower conductivity and 5 times higher water flux compared with the OMBR. A longer lifespan of FO membrane is also predicted in the OMBRC compared to that in the OMBR with the replacements are recommended at 100th and 40th day, respectively. [Display omitted] •A novel hybrid system of osmotic membrane bioreactor-clarifier was introduced.•14 times lower conductivity was observed in novel OMBRC than in conventional OMBR.•Hybrid system showed 3 times higher water flux and 7–15% higher contaminant removal.•2 times thinner and more reversible foulant layer was observed in the hybrid OMBRC.•Deep learning RNN model was firstly developed for prediction of system performance.