Florentina Anca Caliman

Neural Networks and Genetic Algorithms Optimization of the Photocatalytic Degradation of Alcian Blue 8GX

Abstract An artificial neuronal network (ANN) was developed for predicting the color removal efficiency of the TiO2 assisted photodegradation of the cationic dye Alcian Blue 8 GX. The data used for neural network modeling have been obtained by performing experiments in a batch photocatalytic reactor at laboratory scale. The selected input variables of the neural network were the concentrations of the pollutant, photocatalyst and hydrogen peroxide, as well as the initial pH of the solution, while the output variables examined were the UV-A irradiation time and the color removal efficiency. A Jordan Elman neural network with context units that remember past activity was used. This model showed good prediction ability, demonstrating that the proposed model is reliable for prediction of the process efficiency under varying conditions within the experimental region. The Jordan Elman neural model was then included into an optimization procedure solved with genetic algorithms for determining the optimal working conditions that lead to maximum efficiency and minimum reaction time.