Nonlinear Modeling for the Degradation of Aqueous Azo Dyes by Combined Advanced Oxidation Processes Using Artificial Neural Networks

Abstract

One-hidden-layer artificial neural networks (ANNs) using a back-propagation structure have been trained on different sets of experimental data to identify and evaluate the degradation of different azo dyes (Reactive Yellow 84, Reactive Blue 19, Direct Red 23, Direct Red 28, and Acid Blue 193) by photo-Fenton process and combined ozonation and ultrasonolysis processes. Different input variables such as pH, initial concentrations of dyes and ozone, reaction time, ultrasonic power density, and initial concentrations of hydrogen peroxide and ferrous in aqueous solution were employed to model the degradation rates of azo dyes based on the decolorization efficiency and the removal rate using chemical oxygen demand (COD) and total organic carbon (TOC). A new model expression is developed to find the effect of individual parameters and their interactions on the efficiency of organic degradation by advanced oxidation processes.