Amino-functionalized novel biosorbent for effective removal of fluoride from water: process optimization using artificial neural network and mechanistic insights

Abstract

Aqueous fluoride (${{\text{F}}}^{-}$) pollution is a global threat to potable water security. The present research envisions the development of novel adsorbents from indigenous Limonia acidissima L. (fruit pericarp) for effective aqueous defluoridation. The adsorbents were characterized using instrumental analysis, e.g., TGA-DTA, ATR-FTIR, SEM–EDS, and XRD. The batch-mode study was performed to investigate the influence of experimental variables. The artificial neural network (ANN) model was employed to validate the adsorption. The dataset was fed to a backpropagation learning algorithm of the ANN (BPNN) architecture. The four-ten-one neural network model was considered to be functioning correctly with an absolute-relative-percentage error of 0.633 throughout the learning period. The results easily fit the linearly transformed Langmuir isotherm model with a correlation coefficient ${(R}^{2})$ > 0.997. The maximum ${{\text{F}}}^{-}$ removal efficiency was found to be 80.8 mg/g at the optimum experimental condition of pH 7 and a dosage of 6 g/L at 30 min. The ANN model and experimental data provided a high degree of correlation (${R}^{2}$ = 0.9964), signifying the accuracy of the model in validating the adsorption experiments. The effects of interfering ions were studied with real ${{\text{F}}}^{-}$ water. The pseudo-second-order kinetic model showed a good fit to the equilibrium dataset. The performance of the adsorbent was also found satisfactory with field samples and can be considered a potential adsorbent for aqueous defluoridation.