Machine learning-assisted prediction of heavy metal adsorption from tannery effluent wastewater using Delonix-regia with random forest, isotherms, and kinetics modelling

This study explores the effectiveness of Delonix regia as a low-cost bio-adsorbent for removing heavy metals from tannery effluents collected from 42 locations in Vaniyambadi, Vellore district. Key operational parameters—pH, adsorbent dosage, metal ion concentration, contact time, and temperature—were optimised using a Random Forest (RF) machine learning model. Heavy metals such as Nickel (Ni), Lead (Pb), and Cadmium (Cd) were found in high concentrations, indicating severe contamination from tannery discharges. Batch adsorption experiments demonstrated significant removal efficiencies, with actual and RF-predicted values closely aligned (e.g., 78.8 % vs. 77.9 % at 60 min), confirming the model’s reliability. Kinetic analysis showed strong agreement with the pseudo-second-order model (R² > 0.99), indicating that chemisorption governs the adsorption process. Isotherm studies confirmed the Langmuir model fit, suggesting monolayer adsorption on homogeneous surfaces. The integration of machine learning and experimental data proved effective in optimising adsorption conditions and understanding mechanisms. Overall, Delonix regia is validated as a sustainable and efficient solution for heavy metal removal from industrial wastewater, contributing to environmentally friendly and cost-effective treatment strategies.