Akademik Veri Yönetim Sistemi
International Journal of Hydrogen Energy, cilt.206, 2026 (SCI-Expanded, Scopus)
This study aimed to i) determine the effects of different functional groups, both individually and synergistically, on catalytic performance in hydrogen production, ii) elucidate the pathways through which the catalytic mechanism occurs according to functional groups, and iii) successfully integrate machine learning algorithms, Adaptive Neuro-Fuzzy Inference System (ANFIS) and ANalysis of Covariance (ANCOVA) methods into catalyst studies. In this respect, this study is a pioneering study as one of the most comprehensive studies on catalysts in the literature. To achieve these objectives, four new hypercross-linked polymers (HCPs) were designed and synthesized with resorcinol, 1-naphthol, and diphenylamine. HCPs were used as catalysts in hydrogen production by methanolysis of NaBH4. Under optimum conditions, a maximum of 11571 mL H2.min−1∙g−1 of hydrogen gas was produced. The functional group effects and the catalytic mechanism were elucidated by XPS, SEM, BET, zeta potential, and DFT analyses. Ridge regression, Artificial Neural Networks (Multilayer Perceptron (MLP), Random Forest Regression, Extreme Gradient Boosting (XGBoost) Regression (R2 = 0.962), and CatBoost Regression (R2 = 0.965) machine learning methods were successfully implemented to predict the H2 production volume (mL) (Output) based on key reaction parameters (Inputs: HCP type, Temperature (°C), NaBH4 amount (mg), Catalyst amount (mg), and Reaction Time (s)). These ML approaches were integrated with ANFIS and ANCOVA statistical methods, completing a comprehensive predictive and mechanistic model study for our hyper-crosslinked polymer catalysts.