Machine learning approaches to predict adsorption performance of sugarcane derived-carbon dot -based composite in the removal of dyes

This study presents a sustainable and low-cost approach to uniformly blend nanofillers (TiO2) and CDs) into chitosan polymer matrices, aiming to enhance the adsorption capacity and regeneration efficiency of the composite. The effect of degree of deacetylation on the composite was examined using crosslinkers EDA and ECH, achieving a DD of 78.8 % and demonstrating improved chemical and thermal stability compared to chitosan alone. Furthermore, SEM, TEM, and EDX mapping confirmed successful dispersion of nanofillers within the crosslinked chitosan matrix, with excellent adsorption performance towards cationic and anionic dyes (1666.67 mg/g and 1630 mg/g, respectively). The study also presents a novel approach that employs machine learning techniques to accurately predict optimization results. This study combines experiments and AI techniques using Python and machine learning algorithms on a dataset of over 600 experiments, achieving high forecast accuracy with the Random Forest model, evidenced by an R2 value of 1 and the lowest mean squared error for both dyes. The adsorption process followed Freundlich and the pseudo-second-order model. The statistical physical model explained a mixed orientation of anionic dye molecules and a multi-molecular mechanism of cationic dye molecules over nanocomposite, with FTIR results indicating electrostatic attraction, 7C-7C interaction, and hydrogen bonding as key factors in the adsorption process. The composite material used in the study showed a consistent ability to adsorb dyes over 5 regeneration cycles in the column adsorption study, indicating its reusability. The results of this comprehensive investigation contribute to advancing sustainable wastewater treatment through the development of innovative nanocomposites and the use of sophisticated modelling techniques.