Neural network modelling of a depollution process

Abstract

In this paper an artificial neural network is developed to model a new depollution process that uses sequential cultures of anaerobic bacteria and yeasts to efficiently remove both carbon and nitrogen from wastewaters. A set of batch experimental runs are used to train and test various neural network topologies. It is shown that the neural network accurately tracks the dynamics of the biological species of the yeast reactor in the process and account for the influence of butyric acid, ammonia and pH on the overall efficiency of purification.