Hybrid modeling of adsorption process using mass transfer and machine learning techniques for concentration prediction

This study presents a comprehensive hybrid modeling framework that integrates computational fluid dynamics (CFD) with machine learning (ML) techniques to predict chemical concentration distributions during the adsorption of organic compounds onto porous materials. The primary goal is to improve the understanding and prediction of mass transfer behavior in adsorption processes through a data-driven approach. Data points, including x and y coordinates and corresponding solute concentrations (C), were generated through CFD simulations by solving mass transfer equations under varying conditions. Three supervised regression models of Kernel Ridge Regression (KRR), Decision Tree Regression (DT), and Radial Basis Function Support Vector Machine (RBF-SVM) were developed to map spatial coordinates to solute concentrations. Prior to model training, the dataset underwent rigorous preprocessing including outlier removal using the z-score method and normalization. To improve model performance, hyperparameters were optimized using the bio-inspired Barnacles Mating Optimizer (BMO) algorithm. Model evaluation based on R2, root mean square error (RMSE), and mean absolute error (MAE) demonstrated that RBF-SVM outperformed the other models, achieving an R2 of 0.9537, RMSE of 3.5136, and MAE of 1.5326. DT and KRR also turned out strong performance. These findings confirm the effectiveness of ML, particularly RBF-SVM, in analyzing complex spatial dependencies in solute transport processes.