The PowerFeed process is a modification of the simulated moving bed (SMB) technique, where the flow rates in the unit are allowed to change during a switching period following a predefined pattern. Through multiobjective optimization, using a genetic algorithm, it is shown that PowerFeed achieves better performances than SMB in several cases of interest for applications. Moreover, its experimental feasibility and effectiveness is assessed by running and comparing SMB and PowerFeed separations of the enantiomers of alpha-ionone, on a cyclodextrin based chiral stationary phase. (C) 2004 American Institute of Chemical Engineers.