Integrated experimental and machine learning-driven multi-objective optimization for phenol removal via silica aerogel/ZSM-5 nanocomposite: Isotherm and kinetic analysis, thermodynamic insights and pareto-based process design

Phenol is a highly toxic pollutant frequently detected in industrial wastewater. This study presents the first investigation of a hydrophilic silica aerogel/ZSM-5 nanocomposite for phenol adsorption, integrated with advanced machine-learning modeling and multi-objective optimization. Silica aerogel was synthesized via an ambient-pressure sol-gel route and combined with hydrothermally prepared ZSM-5. The 75:25 aerogel/ZSM-5 composite showed the highest surface area and adsorption performance after hydrophilic conversion. Batch experiments evaluated adsorbent dosage (0.1-0.5 g), initial concentration (5-50 mg L⁻¹), contact time (5-180 min), and temperature (25-55 °C). Under optimized conditions (0.3 g, 10 mg L⁻¹, 90 min, 55 °C), 96.1% removal was achieved. Equilibrium data followed the Freundlich isotherm, indicating multilayer adsorption on a heterogeneous surface, while kinetics obeyed the pseudo-second-order model. Thermodynamic parameters confirmed a spontaneous, endothermic process; the low ΔH° (30.27 kJ mol⁻¹) and activation energy (18.54 kJ mol⁻¹) indicate weak interactions between the adsorbate and adsorbent surface. Ethanol regeneration retained the highest capacity over three cycles. Four machine-learning models trained on 150 data points showed that the NSGA-II-optimized ANN and XGBoost delivered superior accuracy (test R² > 0.995). Multi-objective particle swarm optimization identified a validated Pareto-optimal solution (0.139 g, 11.44 mg L⁻¹, 180 min, 54.2 °C) yielding 90.95% removal and 4.86 mg g⁻¹ capacity. SHAP and partial-dependence analyses ranked contact time and initial concentration as the dominant variables. The integrated experimental-ML framework offers a robust pathway for designing high-performance adsorbents and optimizing phenol-removal processes under conflicting objectives.