Revisiting microplastic removal and release by point-of-use ultrafiltration membranes: 1-year monitoring and interpretable machine learning

Membrane filtration has emerged as a promising technology for removing microplastics (MPs) and reducing associated health risks, particularly in point-of-use (POU) systems for drinking water. However, concerns about MP release from polymeric membranes have raised questions about their overall benefits. This study revisits MP removal and release by simulating the use of four commercial POU ultrafiltration (UF) membranes under a one-year filtration scheme. Results indicate that effective MP removal began from approximately 3-6 months of operation onward, as evidenced by consistent positive removal trends. Cake filtration modeling indicated that the formation of fouling layers might play an important role in rejecting MPs and alleviating MP release by protecting the polymer surface. Besides, correlation analysis showed that membrane characteristics had a significant influence on MP release only at the initial stage. In addition, interpretable machine learning, using SHapley Additive exPlanations (SHAP), highlighted that filtration parameters, i.e., transmembrane pressure (TMP), filtration volume, permeability, and total resistance, accounted for 57.6-70.6 % of the feature importance in predicting the MP concentration in the membrane permeate. In this case, extending the membranes' end-of-life (EOL) could enhance MP removal by leveraging fouling while reducing plastic waste and costs from membrane replacement. However, extending EOL comes with trade-offs, including declining water quality, increased TMP, higher energy consumption, and greater carbon emissions. These challenges underscore the need for a balanced approach that maximizes sustainability while ensuring operational efficiency.