Cetyltrimethylammonium bromide Modified Magnetic Apricot Shells for Removing Congo Red Dye and its Artificial Neural Network Model

Abstract

This study focused on the synthesis of a cetyltrimethylammonium bromide-modified magnetic (Fe3O4@Ap/CTAB) nanocomposite from apricot shells and utilised an artificial neural network (ANN) to model the process parameters for removing Congo red (CR) dye from the aqueous environment. The as-prepared magnetic nanocomposite was characterised using various instrumental techniques such as Fourier Transform infrared spectroscopy (FT-IR), Scanning electron microscope (SEM), Energy dispersive x-ray analysis (EDX), Transmission electron microscope (TEM), X-ray diffraction spectroscopy (XRD) and Brunauer–Emmett–Teller (BET) analysis. The nanocomposites showed sheet-like clusters of AS with dark particles of Fe3O4 on the surface. The point of zero charge (pHpzc) was found to be 8.04, while the surface area and the pore volume were 18.7842 m²/g and 0.089736 cm³/g, respectively. The Fe3O4@Ap/CTAB nanocomposite was used as an adsorbent for the removal of Congo red (CR) dye under different conditions. The optimal sorption conditions achieved the highest capacity of 37.175 mg g-1 at 25 °C and 44.053 mg g-1 at 45 °C, at pH 6.5, with a dosage of 50 mg and a concentration of 22.3 mgL-1 in 4 hours. The nanocomposite achieved a 95.77% removal efficiency from real wastewater after shaking for 180 minutes at 25 °C under natural wastewater conditions. The data conformed mostly to the pseudo-second-order kinetic model and Langmuir isotherm, with the thermodynamic studies indicating a favourable, endothermic and spontaneous process with ΔH° value of +27.193 KJ/mol. The adsorbent efficiently sequestered the dye from the wastewater via physical adsorption and electrostatic attractions. A multi-layer ‘feed-forward neural network’, comprising 5 input parameters, a hidden layer, and an output parameter, was employed to predict CR removal efficiency. The performance of the network, evaluated by the coefficient of regression (R2), mean squared error, average percentage error, and mean absolute error, demonstrated a high R2 value of 0.9587. The minimal relative percentage error between the predicted and the test data confirms the ANN's effectiveness in capturing the non-linear behaviour of CR removal. This model of ANN can be employed for the prediction and optimisation of the adsorption process parameters for maximising CR removal using Fe3O4@Ap/CTAB, thereby aiding in the provision of pure water.