Machine learning-based modeling of malachite green adsorption on hydrochar derived from hydrothermal fulvification of wheat straw.

This study investigated the efficiency of hydrochar derived from hydrothermal fulvification of wheat straw in adsorbing malachite green (MG) dye. The characterizations of the hydrochar samples were determined using various analytical techniques like SEM, EDX, FTIR, X-ray spectroscopy, BET surface area analysis, ICP-OES for the determination of inorganic elements, elemental analysis through ultimate analysis, and HPLC for the content of sugars, organic acids, and aromatics. Adsorption experiments demonstrated that hydrochar exhibited superior removal efficiency compared to feedstock. The removal efficiency of 91 % was obtained when a hydrochar dosage of 2 g L was used for 20 mg L of dye concentration in a period of 90 min. The results showed that the study data followed the Freundlich isotherms as well as the pseudo-second order kinetic model. Moreover, the determined activation energy of 7.9 kJ mol indicated that the MG adsorption was a physical and endothermic process that increased at elevated temperatures. The study also employed an artificial neural network (ANN), a machine learning approach that achieved remarkable R (0.98 and 0.99) for training and validation dataset, indicating high accuracy in simulating MG adsorption by hydrochar. The model's sensitivity analysis demonstrated that the adsorbent dosage exerted the most substantial influence on the adsorption process, with MG concentration, pH, and time following in decreasing order of impact.