Hybrid neural model for decoloration by UV/H 2O 2 involving process variables and structural parameters characteristics to azo dyes

Abstract

The purpose of this work is to obtain a neural model of the Acid Brown 75, Acid Orange 52, Acid Orange 10 and Direct Red 28 dyes decoloration process. Dyes aqueous solutions were individually treated in plug-flow reactor, with ultraviolet radiation and hydrogen peroxide. The decoloration process was evaluated in function of the absorbance reading in each dye maximum absorbance wavelength. The input variables corresponding to the input neurons of the neural feedforward backpropagation model used in the work were comprised by structural parameters characteristic of each dye (number of azo bonds and sulphonate groups) and also by process operational variables (temperature, initial pH, hydrogen peroxide volume, reactor operation time and dyes concentration). The combination of structural and operational parameters provided a hybrid character in relation to the input variable nature. The correlation coefficients (approximately 0.96 for the data total, validation and test sets) showed the good model prediction capacity. The neural model obtained also provided, via Garson Partition Method, the determination of the influence of the decoloration process input variables.