Decolorization of the Reactive Blue 19 from Aqueous Solutions with the Fenton Oxidation Process and Modeling with Deep Neural Networks

Abstract

The decolorization of Reactive Blue 19 (RB19) from aqueous solutions using the Fenton oxidation process was researched. The effects of different operating parameters, e.g., H2O2, Fe(II), initial dye concentration, pH, and solution temperature, on the decolorization of RB19 were investigated. Increasing, the H2O2 concentration and temperature increased the rate of the decolorization; however, increasing initial RB19 concentration reduced the decolorization. Additionally, modeling of the decolorization obtained by the Fenton oxidation process was researched based on deep neural networks (DNN) architecture providing the best performance in terms of optimum hidden layers and neuron numbers in addition to ideal activation and optimization function pairs. The performances of the models were analyzed on the training, validation, and test data. According to the experimental results, the seven hidden layers DNN model with “relu” activation function and “RMSProp” optimization function provided the best performance with root mean square error (RMSE) of 3.39 and correlation coefficient (R2) of 0.99.