Adaptive Neuro-Fuzzy Inference System and Artificial Neural Network Modeling for the Adsorption of Methylene Blue by Novel Adsorbent in a Fixed - Bed Column Method

A column study for the removal of Methylene Blue (MB) by the activated carbon from the leaves of Calotropis Gigantea (CGLAC) was done. The CGLAC was characterized by SEM, FT-IR, Raman, TGA/DTA EDAX, BET, and XPS. TGA/DTA studies of CGLAC showed a high fixed carbon content, which indicated that the activated carbon was highly efficient for adsorption. XPS studies confirmed the presence of the aromatic conjugated pi system, C = O and C = C as the main functional group in CGLAC. The pHpzc studies showed that CGLAC has a negative surface charge density and hence adsorption of the cationic dye would be highly efficient. In the column studies, the effect of the parameters like initial concentration (۱۰۰ – ۵۰۰ mg/L) of dye, bed height (۱, ۱.۵, and ۲ cm), pH (۲, ۶.۵, and ۱۰), flow rate (۳.۵, ۵, and ۶.۵ mL), and temperature (۳۰۳, ۳۱۸ and ۳۳۳ K) for the removal of MB were tested. Dynamic column models were tested for BTC and it was found that BDST model fitted the experimental data most. R۲ for BDST model was found to be greater than ۰.۹۹ for most of the parameters studied. Other models were found to fit BTC model at certain conditions. A higher flow rate showed a better fit towards the Thomas model with R۲ value of ۰.۹۵. The highest % of removal of MB by CGLAC was found to be ۸۶.۴ % for ۳.۵ ml min-۱flow rate, ۱۰۰ mg/L concentration, and ۲ cm bed height. ANN and ANFIS predictive models were applied to the study and both the models were found to give good results. ANFIS model showed the highest R۲ value for validation data compared to ANN and hence ANFIS could be applied to the prediction of adsorption compared to ANN for the column studies.

کلیدواژه ها:

CGLAC ، fixed-bed column model ، BDST ، BTC ، Artificial neural network ، Adaptive - Fuzzy Network