An experimental study has been conducted to evaluate the bilevel optimizing control technique for simulated moving beds (SMBs) that was proposed in a previous article (Kim et al. Ind. Eng. Chem. Res. 2010, 49, 3689-3699). Off-line optimization provides optimum time-varying flow rates for feed and desorbent in the upper level, and repetitive model-based predictive control (RMPC) is performed for regulation of the extract and raffinate purities in the lower level. Experiments have been carried out in a four-zone SMB system packed with Dow 50WX4 400-mesh resin to separate L-ribose and L-arabinose at 99.7 vol% from their mixture dissolved in water. The optimizing control technique was implemented just as in the numerical study in the previous work, except for some minor customization. In the experiments, RMPC revealed quite satisfactory tracking and disturbance rejection performance. Optimization resulted in a 45% increase in productivity and a 12% decrease in desorbent consumption from an arbitrary initial point, and the cascaded RMPC law tightly maintained the product purities at their set points by effectively compensating the changes in the feed and desorbent flow rates.