ADSORPTION OF SINGLE AND TERNARY METAL SOLUTIONS ON THE BIOCHAR-NANOMATERIAL COMPOSITE: A COMBINED BATCH ADSORPTION STUDY AND ADSORPTION PREDICTION USING MACHINE LEARNING TECHNIQUES
Date of Award
Master of Science
Accumulation of heavy metals in different environmental compartments and their toxicity even at trace level concentration necessitates the study of their efficient removal. Furthermore, metals could co-exist in the environment which is a complex scenario as there would be competition among the metals in terms of removal efficiency. This study presents the effective removal of trace level toxic metals (Hg2+, Cd2+, Pb2+) in both single and ternary metal solutions through adsorption on the successfully synthesized composite (SC) of pinewood-derived biochar (BC) and graphene oxide (GO) nanomaterials. Moreover, different linear regression tools (Gaussian Process (GP), Random Forest (RF), and Feed Forward Back Propagation (FFBP)) from the machine learning (ML) toolbox were used to make the comparison between actual and their predicted adsorption. The structural and morphological analysis of the SC showed that GO was successfully coated on the surface of the BC. GO coating increased the surface area, porosity, functional groups, and adsorption efficiency of these toxic metals on the SC as compared to the unmodified BC. The factors affecting adsorption efficiency were metal concentration, pH, and the ratio of BC and GO in the SC. The adsorption efficiency in single metal solution was found 94-98% for Hg2+, 92-94% for Cd2+, and 96-99% for Pb2+ and for ternary metal solutions 94-96% for Hg2+, 95-97% for Cd2+, and 97-99% for Pb2+ at pH 6 and SC with BC/GO (w/w) ratio as 1:10. However, for unmodified BC, the adsorption efficiency was less in both single and ternary solutions. Thus, results indicate that modification of BC with GO increases adsorption efficiency as compared to unmodified BC. Furthermore, for all three metals, Freundlich's adsorption isotherm was followed in both single and ternary solutions. Regeneration of the SC was also attained by adsorbate desorption, producing a competent and cost-effective adsorbent for the removal of toxic metals from our environment. Furthermore, from the ML toolbox mean squared error (MSE) values between the actual efficiency and predicted efficiency were calculated which was negligible in the case of GP, with regression coefficient (R2) equal to 1. This implied that GP was the most suitable linear regression model among other models (RF, FFBP) for the available data sets. These predicted values through different ML models could significantly reduce the experimental workload for various parameters in predicting the removal efficiency of the synthesized composite for the target toxic metals. Thus, these models help in reducing experimental time and predicting the most appropriate combination for the best result in the future.
This thesis is only available for download to the SIUC community. Current SIUC affiliates may also access this paper off campus by searching Dissertations & Theses @ Southern Illinois University Carbondale from ProQuest. Others should contact the interlibrary loan department of your local library or contact ProQuest's Dissertation Express service.