Modeling of reduction kinetics of Cr<sub>2</sub>O<sub>7</sub><sup>-2</sup> in FeSO<sub>4</sub> solution via artificial intelligence methods.

Abstract

This study aims to model the reduction kinetics of potassium dichromate (K₂Cr₂O₇) by ferrous ions (Fe²⁺) in sulfuric acid (H₂SO₄) solutions using artificial intelligence-based regression models. The reaction was monitored potentiometrically under controlled hydrodynamic conditions, and an experimental dataset was generated by varying key parameters including temperature, stirring speed, grain size, and Fe²⁺ and H⁺ concentrations. The dataset contains 263 data points representing the conversion rates at different time intervals and experimental conditions. To explore the predictive capabilities of AI in modeling complex chemical kinetics, we applied and compared several regression models: Gradient Boosting, Random Forest, Decision Tree, K Nearest Neighbors, Linear, Ridge, and Polynomial Regression. Hyperparameter tuning was performed using random search to optimize each model's performance. Among these, the Gradient Boosting Regression model demonstrated the best accuracy with an R² value of 0.975 and RMSE of 0.046. Feature importance analysis revealed that stirring speed and temperature were the most influential parameters. To our knowledge, this is the first study to model the Cr(VI) reduction kinetics using a broad range of AI-based regression methods applied to experimentally derived data. The findings highlight the potential of artificial intelligence in replacing conventional kinetic modeling with faster, cost-effective, and highly accurate approaches.