Comparative analysis of the response surface methodology (RSM) and artificial neural network (ANN) modelling for the removal of diclofenac potassium from synthesized pharmaceutical wastewater using a palm sheath fiber nano-filtration membrane and optimization

Abstract

Palm sheath fiber obtained from the petiole of palm tree leaf was defatted and characterized by X-ray diffraction (XRD), revealing a crystalline composition of 75 % calcite, 10.5 % quartz, 4 % periclase, and 10.2 % lime. Further analysis of pore area, volume, and diameter confirmed the membrane as an adsorptive nanofiltration material. A stock solution of Diclofenac potassium was prepared and was filtered varying four process factors: temperature (30–50 °C), pH (6–10), flow-rate (1–5 ml/min), and initial concentration (40–120 mg/L). The removal efficiency of Diclofenac Potassium was analyzed and compared using two optimization models; Methodology (RSM) and Artificial Neural Networks (ANN), focusing on the influence of these factors. The performance and sensitivity of these models were assessed using statistical metrics such as correlation coefficients (R²), Absolute Average Relative Deviation (AARD), and Mean Absolute Error (MAE). Both models demonstrated strong correlation with the experimental data, with the ANN model providing the best predictive accuracy. Optimization of the process via genetic algorithms yielded the optimal membrane removal efficiency of 84.78 %, achieved at initial concentration (102 mg/L), pH (8.8), temperature (40.6 °C), and flow rate (3.6 ml/min). The validation of these optimized parameters was carried out through triplicate experiments, resulting in an average confirmatory removal efficiency of 84.67 %, which validated the ANN prediction. Additionally, adsorption isotherm analysis revealed that the Dubinin-Radushkevich (D-R) model was the best fit for the experimental data, with an R² value of 0.9839. The adsorption kinetics suggested that the process followed pseudo-first-order kinetics as the rate-limiting mechanism.