Regression analysis and artificial intelligence for removal of methylene blue from aqueous solutions using nanoscale zero-valent iron

Abstract

This study attempted to investigate the adsorption of methylene blue (MB) onto nanoscale zero-valent iron (nZVI) from aqueous solutions and to determine the correlation between experimental factors and dye removal efficiency. The adsorption mechanism was discussed in combination with results obtained from transmission electron microscopy, X-ray diffraction, and scanning electron microscopy. Results indicated that at pH 6, nZVI dosage 10 g L−1, initial MB concentration 10 mg L−1, temperature 30 °C, and stirring rate 150 rpm, the equilibrium time was 30 min achieving a removal efficiency of approximately 100%. The adsorption data of MB fitted well to Freundlich isotherm (r2: 0.9358) and pseudo-second-order kinetic model (r2: 0.9976). Response surface methodology (RSM) was developed to visualize the effects of independent factors on the adsorption efficiency. The curves of pH, stirring rate, and reaction time were quadratic linear concave up, whereas nZVI dosage attained a linear up plot. Additionally, artificial neural network (ANN) with a structure of 6–10–1 was used to predict the MB removal efficiency. It was revealed that the ANN model (r2: 0.9313) was more accurate than the RSM model (r2: 0.6316) in describing the adsorption of MB onto nZVI. Sensitivity analysis using the connection weights method showed that the reaction time was the most influential parameter with a relative importance of 22.77%. These advanced modeling techniques could be applied to maximize the performance of nZVI for treating dye-contaminated water under different environmental conditions.