Applications of Radial-Basis Function and Generalized Regression Neural Networks for Modeling of Coagulant Dosage in a Drinking Water-Treatment Plant: Comparative Study

Abstract

The coagulation process, which involves many complex physical and chemical phenomena, is one of the most important stages in water-treatment plants. The coagulant dosage rate is nonlinearly correlated to raw water characteristics such as turbidity, conductivity, and pH. The coagulation reaction is hard or even impossible to control satisfactorily by conventional methods. The coagulant dosage has typically been determined through the jar test, which requires a long experiment time in a field-water-treatment plant. Modeling can be used to overcome these limitations. In this study, a model for the approximation of coagulant dosage rates in water-treatment plants in Algeria has been developed using artificial neural network (ANN) techniques. Two different ANN techniques, the generalized regression neural network (GRNN) and the radial-basis function neural network (RBFNN), were tested for this purpose. The trained GRNN model outperforms the corresponding RBFNN model.