Modeling the treatment of industrial wastewater using anaerobic baffled bioreactors with artificial neural network

In this study, the Artificial Neural Network model was used to model the behavior of anaerobic baffled reactors. In this model influent COD, hydraulic retention time and time to reach steady state condition were considered as input, while the output were the effluent COD and the concentration of volatile fatty acids (VFA). One hidden layer with 6 nodes related the input to the output data. Thus the structure of the proposed model is 3-6-2. The transform functions tried for the input neurons were the Sigmoid and Tangent functions and a Tangent function was used for the output. The Batch Back Propagation method (BBP) was used to train and test the neural network. Out of about 200 experimental data points used in this study, 80% were used for training and 20% for testing. The tests showed that the sigmoid function for the input data and the Tangent function for the output data performed well and are suitable. The network’s Average Relative Deviation (ARD) error for the predicted effluent COD and VFA was 2.97% and 6.09% respectively, demonstrating good accuracy and suggesting that the use of Neural Network modeling for the treatment of industrial wastewater using anaerobic bioreactors has great promise.

کلیدواژه ها:

Anaerobic Baffled Bioreactors ، Modeling ، Neural Network ، Wastewater Treatment