Development of kinetic models for photocatalytic ozonation of phenazopyridine on TiO₂ nanoparticles thin film in a mixed semi-batch photoreactor

The degradation of phenazopyridine hydrochloride (PhP), an analgesic pharmaceutical, through photocatalytic ozonation process was modeled using three types of kinetic approaches. The experiments were performed using a semi-batch photoreactor where TiO₂ nanoparticles thin film was coated on ceramic plates irradiated by UV-A light in the proximity of ozone. The surface morphology, topography and roughness of the bare and TiO₂ coated ceramic plates were analyzed using scanning electron microscopy (SEM) and atomic force microscopy (AFM). The photocatalytic ozonation kinetic characteristics were experimentally investigated under the different operational parameters including pH, PhP initial concentration, ozone inlet flow rate, light intensity, ozone and oxygen dissolved concentrations. It was found that the pseudo-first-order rate constant (k_app) decreased with increase in the initial PhP concentration and increased with increasing ozone inlet flow rate solution pH and light intensity. Based on the generally accepted intrinsic elementary reactions for photocatalytic ozonation process, a novel kinetic model was proposed and validated for predicting PhP removal efficiency. The developed kinetic model explicitly explains the dependency of the apparent kinetic constant on PhP initial concentration, light intensity, ozone and oxygen dissolved concentrations. A good agreement among the predicted values of PhP removal efficiency using the developed model with experimental results was observed (R²=0.987, MSE=0.0011 and MAE=0.0290). In order to profoundly evaluate and compare the accuracy of the proposed intrinsic kinetic model, an empirical kinetic model as function of main functional parameters and an artificial neural network model (ANN) by 3-layer feed-forward back propagation network with topology 5:14:1 were developed. The performance of the three models was compared based on the error functions and analysis of variance (ANOVA). Comparison based on the errors function demonstrated that the experimental data were fitted reasonably well by all the three proposed models with an adequate accuracy. Also, ANOVA results showed that there is no significant discrepancy among the predicted values of the three proposed models, implying that the kinetic model have been developed based on the proper specification of the intrinsic reactions occurred in the process.