Artificial neural network and response surface methodology modeling for diclofenac removal by quaternized mesoporous silica SBA-15 in aqueous solutions

Abstract

Diclofenac (DCF) is a non-steroidal, anti-inflammatory drug commonly prescribed for humans and animals. DCF has been widely detected in aquatic environments and wastewater, and so various adsorbents have been applied for DCF removal from aqueous solutions. Here, we explored DCF removal by quaternary ammonium-functionalized mesoporous silica SBA-15 (q-SBA-15) in aqueous solutions using artificial neural network (ANN) and response surface methodology (RSM) modeling. For this purpose, SBA-15 was synthesized and quaternized with dimethyloctyl[3-(trimethoxysilyl)propyl] ammonium chloride to obtain q-SBA-15. The physicochemical characteristics of q-SBA-15 were examined using various instruments including field emission scanning electron microscopy, transmission electron microscopy, elemental analysis, nitrogen gas adsorption-desorption analysis, nuclear magnetic resonance spectroscopy, X-ray diffractometry, Fourier-transform infrared spectrometry (FTIR), and X-ray photoelectron spectrometry (XPS). N2 adsorption-desorption analysis indicated that q-SBA-15 had a BET surface area of 125.1 m2/g, an average pore diameter of 3.86 nm, and a mesopore volume of 0.121 cm3/g. In the FTIR spectra, several new peaks appeared after DCF removal, confirming DCF adsorption onto the surface of q-SBA-15. The XPS spectra illustrated that DCF was adsorbed onto N+ surface sites on the quaternary ammonium moiety of q-SBA-15 through anion exchange between anionic DCF− and chloride ion. Single-parameter experiments were conducted in terms of initial pH, adsorbent dosage, reaction time, and initial DCF concentration. DCF removal rate decreased gradually in the initial pH of 6.0–10.0. DCF removal reached equilibrium at 12 h with the maximum removal capacity of 593 mg/g. Multi-parameter experiments were designed using the central composite design method with four input parameters and two output parameters (DCF removal rate and final pH). Fifty-six experiments (28 experimental conditions in duplicate) were conducted to examine the simultaneous effects of the input parameters on the output parameters. Based on the experimental data, RSM and ANN models were developed to predict the DCF removal rate and final pH. The additional experimental data (4 experimental conditions in duplicate) demonstrated that the developed ANN model had better predictability than the RSM model for the output parameters.