The photocatalytic degradation kinetics of the anti-inflammatory drug ibuprofen in aqueous solution under UV/TiO2 system and neural networks modeling

Abstract

Abstract In this study, the kinetic degradation of the anti-inflammatory drug Ibuprofen in aqueous solution by heterogeneous TiO2 photocatalytic was investigated. The data obtained were used for training an artificial neural network. Preliminary experiments of photolysis and adsorption were carried out to assess their contribution to the photocatalytic degradation. Both, direct photolysis and adsorption of Ibuprofen are very low-efficient processes (15,83% and 23,88%, respectively). The degradation efficiency was significantly elevated with the addition of TiO2 Catalyst (>94%). The photocatalytic degradation followed a pseudo-first-order reaction according to the L-H model. The hydroxyl radicals and photo-hole (h+‏) were found to contribute to the Ibuprofen removal. The higher the initial concentration of Ibuprofen resulted in the lower percentage of degradation. This can be credited to the fact that the created photon and radicals were constant. The higher the initial concentration of Ibuprofen the fewer radicals were shared for each Ibuprofen molecular and so the lower percentage of degradation. The maximum photoactivity from the available light is accomplished when the concentration of catalyst reaches to 1 g/L (0.8 g), which was adopted as the optimal amounts. Compared to the removal of ibuprofen, the mineralization was relatively lower. This decrease is due to the organic content of the treated solution, which is mainly composed of recalcitrant intermediate products. The network was planned as a Levenberg-Marquardt algorithm with three layer, four neurons in the input layer, fourteen neurons in the hidden layer and one neuron in the output layer (4:14:1). The artificial neural network was trained until the MSE value between the simulated data and the experimental results was 10−5. The best results (R 2 = 0.999 and MSE = 1.5 × 10−4) were obtained with a log sigmoid transfer function at hidden layer and a linear transfer function at output layer.