Quantitative decoupling of the key factors and mechanisms of Fenton-pretreated membrane fouling mitigation via machine learning-assisted causal inference.

Fenton pretreatment effectively mitigates organic membrane fouling via synergistic oxidation and coagulation. However, its key drivers and intrinsic mitigation mechanisms remain unknown. Here, a progressive data-driven framework integrating machine learning (ML), Shapley additive explanations (SHAP), and causal inference methods across three extensive datasets (>6,700 samples) was developed to quantitatively elucidate Fenton-induced fouling mitigation from macro- to microscopic levels. The developed ML models exhibited exceptional prediction performance across all datasets (R2adj = 0.804-0.976). The results of SHAP analysis revealed that Fenton pretreatment largely governed fouling mitigation (44.1%). Specifically, Fe2+ was 1.8 times more critical than H2O2 was, and both were significantly nonlinearly correlated with fouling mitigation. By capturing these nonlinear dose-response relationships, we identified the optimal trace Fenton dosage levels (Fe2+ = 0.1 mM; H2O2 = 0.075 mM) tailored for specific organic foulants (sodium alginate). Microscopically, causal inference revealed the underlying mitigation mechanisms: notably, oxidation-driven electrostatic repulsion (51.7%) contributed four times more than the coagulation-driven size exclusion effect did (12.8%). This enhanced electrostatic repulsion primarily weakened foulant-foulant intermolecular cohesion (57.6%), which far outweighed the reduction in foulant-membrane adhesion (29.6%). Ultimately, architectural restructuring of the fouling layer (72.7%) driven by these attenuated interfacial interactions effectively mitigated membrane fouling, decisively outweighing the contribution from the reduction in the deposited foulant mass (27.3%). In this study, we not only quantitatively decouple the mechanistic pathways underlying Fenton-mitigated membrane fouling but also provide a versatile and transferable data-driven paradigm for elucidating complex membrane fouling phenomena.