CO2 adsorption on NaOH and acid modified montmorillonite: Response surface methodology and machine learning modeling

This study investigates the use of modified montmorillonite (MMT) for CO₂ adsorption through an integrated approach combining Machine Learning (ML) modeling and Response Surface Methodology (RSM). MMT was modified with controlled treatments of sodium hydroxide (NaOH) and hydrochloric acid to enhance its CO₂ adsorption capacity under varying operational conditions, including temperature, pressure, time, and modification parameters, such as acid molarity and NaOH concentration. RSM provided valuable insights into the effects of these parameters, yielding an R² of 0.948, confirming its effectiveness in modeling adsorption trends. However, due to its limitations in capturing complex, nonlinear interactions, advanced ML techniques were employed for improved accuracy. Several ML models were evaluated, including Decision Tree Regressor, Random Forest Regressor, AdaBoost Regressor, Gradient Boosting Regressor, Extreme Gradient Boosting Regressor, Support Vector Machine, Multilayer Perceptron (MLP), Radial Basis Function, and Kernel Ridge Regressor. Among the models tested, the MLP performed the best, achieving an R² of 0.9982 across the entire dataset, outperforming the RSM model. The MLP’s superior predictive accuracy is attributed to its ability to capture intricate, nonlinear patterns, making it the most reliable method for predicting CO₂ adsorption capacity in this study. This research underscores the critical role of machine learning in optimizing CO₂ adsorption models, emphasizing its potential to address the complex interactions between operational and modification parameters that traditional methods struggle to capture. The integration of ML significantly enhances predictive accuracy and provides a powerful tool for optimizing MMT-based adsorbents in environmental applications aimed at mitigating climate change. • Montmorillonite is modified using acid and NaOH to enhance CO₂ adsorption. • RSM provided insights but failed to precisely predict CO₂ adsorption capacity. • Several ML algorithms were applied to predict CO₂ adsorption performance. • The MLP model achieved an R² of 0.9982 and predicted trends in key variables. • The models revealed critical parameter effects, boosting MMT adsorbent performance.