The goal of this paper is to remove radioactive Kr from Xe based on the development of a viable PSA (pressure swing adsorption) process, because it normally requires a few years for Kr-85 to achieve a significant decay level of this isotope. A general purpose model is built to remove Kr-85 in the cyclic PSA process with partial differential equations. Because there are no steady state and constant operating variables in the operation of the PSA process, the design of PSA at each step is critical for satisfactory performance. Thus, an optimization technique based on a genetic algorithm is used. The optimization leads to significant improvement for PSA. Moreover, for the physical model based design, changes in the operation of the process mean the original design condition may no longer be a true reflection of the desired quality. An iterative learning design is also proposed. This work aims to find out the optimal design condition for PSA. (C) 2013 The Institution of Chemical Engineers. Published by Elsevier B.V. All rights reserved.