Use of artificial neural networks: predicting trickling filter performance in a municipal wastewater treatment plant

Artificial neural networks (ANN) are used as an alternative function approximation tool for predicting the performance of trickling filter treatment process in a municipal wastewater treatment plant,Solon, Ohio, USA, which uses a trickling filter followed by an activated sludge process. The treatment plant had an average monthly inflow flow rate of 2.92 mgd (million gallons per day). The average raw, settled,and final BOD (biochemical oxygen demand) was 449, 235 and 4.8 mg/l,respectively, while the corresponding value for TSS (total suspended solids) was 296, 131, and 6.1 mg/l. The overall removal efficiency for BOD and TSS was 98.93 per cent and 97.95 per cent respectively. The best ANN model for predicting the trickling filter effluent BOD and TSS has a prediction error of 31.45 per cent and 32.54 per cent respectively. The number of input variables, as well as number of nodes in hidden layer seemed not to have a definite effect on the prediction error for the ANN model. The prediction errors obtained with ANN models were lower than those obtained by multiple regression analysis.