In this study, trends from an adsorption process simulator are compared against experimental results obtained from a Rapid Pressure Swing Adsorption (RPSA) pilot plant for the separation of air over a packed bed of LiLSX pellets. The primary purpose of this study was to examine the impact of two model formulations for intrapellet mass transfer on predicted process performance, the linear driving force model and a rigorous pellet model (the viscous flow plus dusty gas model intrapellet flux equation). Non-isothermal and non-isobaric behavior is maintained within the modelformulation. Nine PSA data sets at cyclic steady state were measured, with total cycle time varying from 8 to 50 s. At the short cycle time, large discrepancies between the simple lumped model and the rigorous pellet model arose, with the rigorous pellet model tracking qualitative trends in pilot plant data more faithfully than the simple model. The accepted value for the dimensionless cycle time theta(i) below which linear driving force models fail is 0.1. However, our data and models suggest that a more appropriate value would be 1.0. Below theta(i) = 1.0 the simple linear driving force model started to deviate from experimental data and the rigorous pellet model. This suggests that conclusions from studies of single pellets may not apply rigorously to packed beds of pellets in which each pellet is exposed to rapidly varying and coupled boundary conditions in which composition, pressure, and temperature all change with time. (c) 2006 American Institute of Chemical Engineers.