Neural network modeling of the photocatalytic degradation of 2,4-dihydroxybenzoic acid in aqueous solutionE. Oliveros1, F. Benoit-Marquié2, E. Puech-Costes2, M.T. Maurette2 and C.A.O. Nascimento3
1 Lehrstuhl für Umweltmesstechnik, Engler-Bunte-Institut, Universität Karlsruhe, 76128 Karlsruhe, Germany
2 Laboratoire des IMRCP, Université Paul Sabatier, 31062 Toulouse Cedex, France
3 Escola Politecnica, Universidade de São Paulo, 01000 São Paulo, SP, Brazil
Artificial neural networks have been used for modeling the TiO2 photocatalytic degradation of 2,4-dihydroxybenzoic acid, chosen as a model water contaminant, as a function of the concentrations of substrate and catalyst. The experimental design methodology was applied to the choice of an appropriate set of experiments well distributed in the experimental region (Doehlert uniform array). Contrary to a classical treatment of the data, based on apparent rate constants modeled by a quadratic polynomial function, neural network analysis of the same experimental data does not require the use of any kinetic or phenomenological equations and allows the simulation and the prediction of the pollutant degradation as a function of irradiation time, as well as prediction of reaction rates, under varying conditions within the experimental region.
Key words: Artificial neural networks / 2,4-dihydroxybenzoic acid (DHBA) / experimental design / modeling / photocatalytic degradation / TiO2.
© EDP Sciences, Wiley-VCH 1998