Effects of data normalization and inherent-factor on decision of optimal coagulant dosage in water treatment by artificial neural network

The artificial neural network (ANN) has been applied to the nonlinear relationship between accumulated input and output numerical data for the coagulation processes in water treatment. However, the high turbidity of the raw water may affect the predicting ability. Therefore, it is necessary to enhance the precision of ANN. In this study, inherent-factor was devised, and the prediction capabilities of the ANN were compared in terms of a root mean square error. Weight decay regularization was adapted to avoid over-fitting in this research, along with the use of the Levenberg-Marquardt method in the development of ANN models. The predicting ability of ANN was improved by the inherent-factor and without data normalization. The Pearson correlation coefficient was sufficient to select the optimal input variables for predicting the optimal coagulant dosage for ANN. The final input variables included the raw water turbidity and coagulant dosage on Day t − 1, which was built on the transfer function from input layer to hidden layer with a tan-sigmoid function and the transfer function from hidden layer to output layer with a linear function. The input data was not normalized.