Predicting Dynamic Response of Adsorption Columns with Neural Nets

Abstract

Artificial neural networks have been applied to a limited number of environmental engineering problems. This paper investigates the feasibility of utilizing the concept of neural nets in developing networks for predicting the breakthrough curves of fixed-bed adsorbers. Close agreement is observed between the breakthrough curves predicted by the developed neural network and those obtained from the mathematically based adsorption model (HSDM). The advantages of using neural networks in modeling environmental processes over the commonly used traditional methods are addressed. Also, further improvements and generalization of the developed predictive fixed-bed adsorption neural network are discussed.