The removal of heavy metals from industrial wastewater remains a major environmental challenge, particularly for nickel-contaminated effluents generated by electroplating, mining, and metal-finishing industries. In this study, the dynamic adsorption behaviour of Ni²⁺ ions in a fixed-bed column packed with modified nanocellulosic adsorbents was investigated, and a machine learning-nbased predictive framework was developed to enhance process modelling. Breakthrough data obtained under varying operating conditions namely bed height, influent concentration, and flow rate were analysed using three conventional kinetic models: Thomas, Weibull, and Wolborska. Among these models, the Thomas model provided the best representation of the adsorption process, exhibiting the highest agreement with experimental data (R² > 0.97), while the Weibull and Wolborska models showed comparatively lower predictive capability. To overcome limitations associated with traditional kinetic modelling, a Support Vector Machine (SVM) regression approach was implemented to predict Ni²⁺ removal efficiency. The Cubic SVM model demonstrated strong predictive performance with a coefficient of determination (R²) of 0.90 and low prediction errors (RMSE = 4.78, MSE = 22.85, MAE = 3.66). Diagnostic evaluation using predicted–actual plots and residual analysis confirmed the robustness and reliability of the developed model. The results demonstrate that integrating experimental adsorption studies with machine learning algorithms significantly improves the predictive capability and optimization of fixed-bed adsorption systems. This hybrid modelling framework offers a promising strategy for designing efficient and sustainable adsorption processes for heavy metal removal from wastewater.
Etuk et al. (Mon,) studied this question.