A highly effective, recyclable, and novel host-guest nanocomposite for Triclosan removal: A comprehensive modeling and optimization-based adsorption study.

In this research paper, response surface methodology (RSM), generalized regression neural network (GRNN), and Adaptive Neuro-Fuzzy Inference System (ANFIS) were employed to develop prediction models for Triclosan (TCS) removal by a novel inclusion complex (host-guest complex). Hence, β-cyclodextrin (β-CD) and poly(ethylene glycol) (PEG) host-guest complex loaded on the multi walled carbon nanotube (MWCNT/PEG/β-CD) was prepared and characterized by Raman, NMR, TGA, XRD, SEM, TEM, and point of zero charge (pHpzc) technique. The effects of MWCNT/PEG/β-CD dose (g), temperature (°C), antibiotic concentration (mg L-1), and sonication time (min), each at five levels were investigated as independent factors. Central composite design (CCD) of RSM setup was applied in combination with ANFIS and GRNN training dataset for evaluation purposes. Moreover, the kinetic, isotherm equilibrium, and thermodynamic parameters of adsorption of TCS on MWNT-PEG/β-CD nanocomposite was examined. To assess the accuracy of results, several statistics such as R2, RMSE (root mean square error), mean squared error (MSE), MAE (mean absolute error), sum of the absolute error (SAE), %AAD (absolute average deviation), average relative error (ARE), hybrid fractional error function (HYBRID), Marquart's percentage standard deviation (MPSD), and Pearson's Chi-square measure (χ) were checked. The results of ANFIS approach were found to be more trustworthy than GRNN model since better statistical analysis were attained. However, it was known that the GRNN is easier and take a little time for modeling than the ANFIS approach.