APPLICATION OF ARTIFICIAL NEURAL NETWORKS FOR PREDICTING COD REMOVAL EFFICIENCIES OF ROTATING DISKS AND PACKED-CAGE RBCS IN TREATING HYDROQUINONE

In this study, an artificial neural network (ANN) was applied to predict the performance of two rotating biological contactor (RBC) systems in removal of hydroquinone (a toxic aromatic compound). The first system was a two-staged conventional RBC and the second one was a one-staged packed-cage RBC with bee-cell 2000 biofilm carriers. Both systems had a total area of about 2 m² for biofilm attachment. The main aim is to predict COD removal efficiencies in both systems using ANN. Efficiency evaluation of the reactors was obtained at different influent COD from 200 to 5000 mg/L. Exploratory data analysis was used to detect relationships between the data and the evaluated dependents. The appropriate architecture of the neural network models was determined using several steps of training and testing the models. The modeling results showed that there is a good agreement between the experimental data and the predicted values with a correlation coefficient (R²) of 0.998 and 0.997 for RBC with rotating disks and packed-cage RBC, respectively.