doiSerbia

Prediction of doxycycline removal by photo-fenton process using an artificial neural network - multilayer perceptron model

Boucherit Nabila (Biomaterials and Transport Phenomena Laboratory (LBMPT), Yahia Fares University, Medea, Algeria), na_boucherit@yahoo.fr
Hanini Salah (Biomaterials and Transport Phenomena Laboratory (LBMPT), Yahia Fares University, Medea, Algeria)
Ibrir Abdellah (Materials and Environment Laboratory (LME), Faculty of Technology, Yahia Fares University, Medea, Algeria)
Laidi Maamar (Biomaterials and Transport Phenomena Laboratory (LBMPT), Yahia Fares University, Medea, Algeria)
Fissa Mohamed Roubehie (Biomaterials and Transport Phenomena Laboratory (LBMPT), Yahia Fares University, Medea, Algeria)

This paper presents a study on the effectiveness of the Photo-Fenton Process (PF) for removing the doxycycline hyclate (DXC) antibiotic. The experiment showed that the best removal efficiency was achieved (79%) at pH 3 for 2.5 mg/L of DXC, 76.53 mg/L of H2O2, and 86.8 mg/L of Fe2+. The degradation mechanism of DXC by hydroxyl radicals was confirmed by FTIR and HPLC. To model the oxidation reaction of DXC by PF, an multilayer perceptron (MLP) based optimized artificial neural network (OANN) was used, taking into account experimental data such as pH and initial concentrations of DXC, H2O2, and Fe2+. The OANNN predicted removal efficiency results were in close agreement with experimental results, with an RMSE of 0.0661 and an R2 value of 0.99998. The sensitivity analysis revealed that all studied inputs significantly impacted the transformation of DXC.

Keywords: Doxycycline hydrate, Modelling, Photo-Fenton, Optimized Artificial Neural Network, Removal