Comprehensive review on machine learning methodologies for modeling dye removal processes in wastewater. (20th January 2023)

• Record Type: Journal Article
• Authors: Bhagat, Suraj Kumar; Pilario, Karl Ezra; Babalola, Olusola Emmanuel; Tiyasha, Tiyasha; Yaqub, Muhammad; Onu, Chijioke Elijah; Pyrgaki, Konstantina; Falah, Mayadah W.; Jawad, Ali H.; Yaseen, Dina Ali; Barka, Noureddine; Yaseen, Zaher Mundher
• Is Part Of: Journal of cleaner production. Volume 385(2023)
• Publication Date: 2023-01-20
• Subjects: Machine learning models -- Cost of the media and process -- Data availability -- Decolorization process optimization -- Dye removal prediction -- Dye removal mechanism -- Future road map; Factory and trade waste -- Management -- Periodicals; Manufactures -- Environmental aspects -- Periodicals; Déchets industriels -- Gestion -- Périodiques; Usines -- Aspect de l'environnement -- Périodiques; 628.5
• DOI: 10.1016/j.jclepro.2022.135522
• Languages: English
• ISSNs: 0959-6526

Abstract

Abstract: A wide range of dyes are being disposed in water bodies from several industrial runoff and the quantity is rapidly increasing over the years. From an environmental safety point of view, it is urgent to improve the removal process of dyes. It is important to understand the stochastic and highly redundant behavior of the process of dye removal (DR ) in wastewater treatment. This leads to better utilization of Machine Learning (ML) models for both optimization as well as prediction of the DR process efficiency. In this review, 200 papers (Years, 2006–2021) have been systematically reviewed from the Web of Science and Scopus-indexed journals, covering a total of 84 journals. All applied ML models have been thoroughly studied in the review and analyzed in terms of their architecture setup, hyper-parameters selection, performance, advantages, and disadvantages. A wide range of optimization methods for hyper-parameters tuning were analyzed and discussed scientifically. Explicit information about the data sizes, splitting structure for training-validation-testing along with input and output selection approaches have been logically reviewed and discussed. Data availability, transparency, and reusability have been reported adequately. Various software for data-driven modeling have been discussed with their pros and cons. Trends in statistical evaluators (among about 60 types) have been discussed with their pros and cons including their sensitivity with the data fluctuations. Moreover, the most popular performance metrics have reported. In addition, the DR mechanism has reviewed and discussed inclusively. Extensive media used for the decolorization were discussed thoroughly, including their physical and chemical characteristics, along with feasibility and equilibrium data based on Langmuir model. The cost of the applied media in the decolorization process reported adequately. Finally, the research gap and future road map of the next 5 years, which bridge the gap of the domain are scientifically drafted along with the limitations. This critical review not only provides the appraisal of growth of DR research integrated with ML in the last couple of decades but also scouts the potential studies where all experimental, chemical and modeling processes should be taken under consideration. Graphical abstract: Image 1 Highlights: An extensive review on dye removal modeling established over 2 decades. All machine learning predictive models and optimizers are exhibited and assessed. Data availability, application of hyper-parameter optimization are analyzed. Removal media, dye color, and economical global assessment are appraised. 5 years of future prospective road map are suggested to be explored.