The separation of xylene isomers is an important application in separation processes that is based on their adsorption properties on different adsorbents. In this work, the Price and Danner method was employed with a neural network to investigate the adsorption behavior of binary systems of p-xylene/m-xylene, p-diethyl benzene/m-xylene, and p-diethyl benzene/p-xylene and the ternary system of p-diethyl benzene/rn-xylene/P-xylene at 130 and 175 degrees C. The Redlich-Kister, Wilson, and NRTL models were used to determine the activity coefficients in the adsorbed phase. Comparison with experimental data from the literature indicated that the proposed thermodynamic model would best determine surface excess when it is used along with the Redlich-Kister activity coefficient model. (C) 2010 Elsevier B.V. All rights reserved.