Effect of Aluminum Loading on Mesoporous Silica-Supported Catalysts for Enhanced Solketal Synthesis from Glycerol via Acetalization: An Integrated Response Surface Methodology and Artificial Neural Network Framework for Parameter Optimization

Abstract

The conversion of surplus glycerol from biodiesel production into solketal represents an effective strategy to enhance biofuel sustainability; however, the development of efficient and reusable solid acid catalysts with optimized acidity remains challenging. In this study, sulfated mesoporous SiO2/Al2O3 catalysts with varying Si/Al ratios were synthesized and evaluated for glycerol acetalization with acetone. Process parameters were optimized using response surface methodology (RSM), considering catalyst loading, temperature, reaction time, and acetone: glycerol molar ratio. An artificial neural network (ANN) model was further developed to capture nonlinear relationships and validate predictive performance within the experimental domain. The Si/Al (50:1) catalyst exhibited excellent performance, achieving 70.0% glycerol conversion, 64.9% solketal yield, and 92.8% selectivity under the optimized conditions (7 wt % catalyst, 60 °C, 120 min, and 6:1 acetone: glycerol molar ratio). NH3-TPD analysis revealed that this catalyst possesses a high density of moderate-strength acid sites, which are primarily responsible for its superior catalytic performance. The RSM model demonstrated high predictive accuracy, with a deviation of less than 2.2% between predicted and experimental values. The ANN model also showed strong predictive capability (overall R = 0.9962), confirming its suitability as a complementary nonlinear modeling tool within the studied range. Reusability tests indicated a gradual decline in activity while maintaining high selectivity (>87%) after four cycles. This deactivation behavior is associated with a reduction in accessible acid sites, as supported by NH3-TPD analysis. These results demonstrate that controlled Si/Al composition effectively tunes catalyst acidity and performance, while the combined use of RSM and ANN provides a complementary modeling approach for reliable prediction within the defined experimental space, offering a practical strategy for sustainable glycerol valorization under mild conditions.