Artificial neural networks for mnodelling the degradation of emerging contaminants Process

Abstract

Diclofenac sodium is an emerging contaminant that can be harmful for ecology and human health. This substance can be degraded by a heterogeneous Photo-Fenton process, CoFe2O4 as catalyst, H2O2 as oxidant and UV radiation. The aims of the work are the comparison of different artificial neural networks to characterize the relationship between diclofenac degradation and H2O2 consumption, with the Total Organic Carbon achieved in the mineralization of the drug and the testing of the selected model capacity to predict the Total Organic Carbon concentration, by employing the reused catalyst. The best performing backpropagation neural network was constituted with a ten neurons hidden layer with sigmoid transfer function and one linear neuron, as output. It was determined that the model can approximate the trend between the input data (Absorbance and H2O2 concentration) and output ones (Total Organic Carbon concentration) when it was validated with data from reactions employing CoFe2O4 for second and third time. The development of these models is of interest due to the consequent reduction of time and costs in experimental work. It represents a study of the evolution of chemical indicators in the treatment of emerging contaminants.