Green synthesis of bimetallic Ni-Fe nanocatalysts for hydrogen production: a comparative study of linear and ensemble machine learning models

Abstract

In this study, NiFe bimetallic nanoparticles were prepared by an environmentally friendly green synthesis method using rose petal extract. XRD analysis revealed an average crystallite size of 4 nm, while SEM images showed rod-shaped and partially spherical morphologies. EDX analysis revealed 31.90 % Ni, 12.04 % Fe and 56.06 % C content, while Raman spectroscopy confirmed structural integrity and the presence of metal-oxide bonds. The catalytic performance was studied at different temperature, amount of catalyst and solution concentrations in the NaBH4 hydrolysis reaction. In 0.5 wt% NaOH medium, the maximum hydrogen production rate was 445.64 mlmin(-1)-g(-1). Kinetic analyses calculated the activation energy of the NiFe catalyst to be 63.5 kJ/mol. In the second stage of the study, XGBoost, Gradient Boosting and various linear regression models were used to predict the hydrogen production rate. The data set was enriched with polynomial, logarithmic and interaction terms, and the risk of over-learning was reduced by k-fold cross-validation. The results show that the XGBoost (R-2: 0.980) and Gradient Boosting (R-2: 0.963) models provide the highest accuracy, while the error metrics demonstrate the superiority of ensemble methods over linear models. The findings reveal that NiFe nanoparticles produced by green synthesis offer high catalytic efficiency in hydrogen production and advanced machine learning techniques are effective in predicting and optimizing the production rate. This integrated approach contributes significantly to the advancement of sustainable hydrogen production technologies by accelerating catalyst development processes in the field of clean energy.