Neuronet modeling of VOC adsorption by GAC

Volatile organic compounds (VOCs) are the most frequently occurring pollutants in groundwater which can impart significant health hazards to the end user even at levels of micrograms per liter. Removal of the VOCs by adsorption onto granular activated carbon (GAG) using adsorption columns is the most efficient practice among the various water treatment unit processes. In this paper, a neural network (NN) was developed to determine the breakthrough time of adsorption columns used to treat water polluted with VOCs from a set of parameters pertinent to adsorption isotherm and other related column-operating conditions. a simple linear regression model that enables determination of the adsorption isotherm from a single physical property of the VOC was developed. The advantages of the NN approach in both the modeling effort involved and prediction accuracy were compared with those of the conventional regression methods. Moreover, the various input parameters were investigated, using parametric analysis employed on the developed NN, to address their relative contribution to the breakthrough time.