Application of artificial neural network for modeling wastewater treatment process

Development of reliable model in chemical engineering facilitates all subsequent steps in process optimization and monitoring operation. Many chemical process intended to wastewater treatment can exhibit complex nonlinear behaviour. In this paper, three layered feed forward neural network is used to predict the photo-catalytic degradation yield of solophenyl red, an azo dye widely used in textile industry. The approach adopted to find the optimal topology of the network is based on finding the architectural parameters (the hidden nodes number, the activation function and the training algorithm) that minimize the prediction error. Experimental data required for the development of the network are extracted from the study performed by [1]. In this study a new photo-catalyst have been used to eliminate under solar light the solophenyl red. The result show that the predicted data from the designed optimal neural network architecture was in good agreement with the experimental data. The excellent value of the correlation coefficient attested the accuracy of the model and proved its ability to fit this complex system.