Optimal hybrid artificial intelligence models for predicting the adsorptive removal of dyes and phenols from aqueous solutions using an amine-functionalized graphene oxide/layered triple hydroxide nanocomposite

Abstract

Water pollution is a serious environmental problem with a significant negative impact on human health and the ecosystem. Adsorption is an attractive process for water decontamination. Developing artificial intelligence models capable of predicting the adsorption performance of newly developed materials could save huge amounts of resources and efforts. In the present study, a novel polyethyleneimine/graphene oxide/layered triple hydroxide (PEI/GO/LTH) nanocomposite was synthesized, characterized, and applied to adoptively remove harmful phenolic (bisphenol A) and azo dye (acid red 1) from wastewater samples. The results revealed that the PEI/GO/LTH nanocomposite is highly effective in removing these two pollutants. The adsorption isotherms and kinetics of these two pollutants are best fitted by the Langmuir isotherm and pseudo-second-order models, respectively. Additionally, the synthesized nanocomposite is easily and highly regenerable with an insignificant loss in performance when repeatedly used. Besides the above investigations, the present study also employs support vector machines (SVM) and Bayesian optimization as tools for predicting the adsorptive removal of acid red 1 (AR1) dye and bisphenol A (BPA) from contaminated water samples by the synthesized PEI/GO/LTH nanocomposite. The models achieved promising results as the correlation coefficients, during the testing phase, reached 97.3 and 96.6 % for the AR1 and BPA data using BO-SVM models, respectively.