Machine learning-based modeling and analysis of perfluoroalkyl and polyfluoroalkyl substances controlling systems in protecting water resources

Perfluoroalkyl and polyfluoroalkyl substances (PFAS) are extensively distributed, highly persistent, and hazardous compounds in water resources threating human health and ecosystems, therefore requiring effective controlling and management systems. Machine learning (ML)-based procedures are novel approaches through which the PFAS-controlling systems can be improved cost-effectively and rapidly from different aspects. The few accomplished ML -based studies in PFAS-controlling systems showed considerable performance, with > 80% prediction strength in outputs, for example, treatment performance, identification of the susceptible groundwater resources, and PFAS defluorination energy in > 70% of the studies. Despite such a great performance, there is no systematic study of various aspects of PFAS-controlling systems, for example, modeling and analysis of PFAS degradation and distribution mechanisms, optimization, alarm management, troubleshooting, and appropriate operation and maintenance of these systems. Therefore, this study reviews key aspects and parameters that can take advantage of ML procedures in achieving cost-effective PFAS control in water resources.

Automated summary

Perfluoroalkyl and polyfluoroalkyl substances (PFAS) are hazardous compounds that are extensively distributed in water resources, posing a threat to human health and ecosystems. Machine learning (ML) techniques can be used to improve PFAS control systems and achieve cost-effectiveness and rapidity. Previous studies have shown that ML-based PFAS control systems perform well, with prediction accuracy exceeding 80% in areas such as treatment performance prediction, identification of groundwater resources, and PFAS defluorination energy prediction.