Automating process design by coupling genetic algorithms with commercial simulators: a case study for hybrid MABR processes. (1st August 2022)

• Record Type: Journal Article
• Authors: Yang, Cheng; Belia, Evangelia; Daigger, Glen T.
• Is Part Of: Water science and technology. Volume 86:Number 4(2022)
• Page Start: 672
• Page End: 689
• Publication Date: 2022-08-01
• Subjects: genetic algorithms -- hybrid MABR -- process design -- SUMO; Water -- Pollution; Sewage -- Purification; Water quality management; Periodicals; 628.168
• DOI: 10.2166/wst.2022.234
• Languages: English
• ISSNs: 0273-1223

Abstract

Abstract: The development of commercial software and simulators has progressed to assist engineers to optimize design, operation, and control of wastewater treatment processes. Commonly, manual trial-and-error approaches combined with engineering experience or exhaustive searches are used to find candidate solutions. These approaches are becoming less favorable because of the increasingly elaborate process models, especially for new and innovative processes whose process knowledge is not fully established. This study coupled genetic algorithms (GAs), a subfield of artificial intelligence (AI), with a commercial simulator (SUMO) to automatically complete a design task. The design objective was the upgrade of a conventional Modified Ludzack-Ettinger (MLE) process to a hybrid membrane aerated biofilm reactor (hybrid MABR). Results demonstrated that GAs can (1) accurately estimate five influent wastewater fractions using eleven typical measurements – 3 out of 5 estimated fractions were nearly the same and the other two were within 7% relative errors and (2) propose reasonable designs for the hybrid MABR process that reduce footprint by 17%, aeration by 57%, and pumping by 57% with significantly improved effluent nitrogen quality (TN<3 mg-N/L). This study demonstrated that tools from AI promote efficiency in wastewater treatment process design, optimization and control by searching candidate solutions both smartly and automatically in replacement of manual trial-and-error methods. The methodology in this study contributes to accumulating process knowledge, understanding trade-offs between decisions, and finally accelerates the learning pace for new processes. HIGHLIGHTS: GAs are coupled with SUMO for automatic process design and optimization. GAs can accurately calibrate influent characteristics with typical measurements. GAs can optimize designs and operations automatically with multiple objectives. The coupling approach advances simulators use by reducing manual trials and errors. Process knowledge is extremely important in guiding the usage of AI tools. Graphical Abstract