This study presents a control strategy of carbon dioxide capture from flue gas by VacuumPressure Swing Adsorption (VPSA).The key objective of this work is to control the purityand recovery of the adsorption product (carbon dioxide) by manipulating the step duration ofthe cyclic process.gPROMS modeling platform is used for a single fixed-bed 6-step VPSAprocess simulation using 5A zeolite as adsorbent for CO2 capture from gas mixture with 15％CO2 and 85％ N2 (resembling post-combustion flue gases of power stations).To evaluate theperformance of VPSA under consideration, and to testify the influences of cycle stepsduration on product purity and recovery, dynamic simulations are carried out.The feed timeand purge have inverse impact on purity and recovery, both feed and purge duration time aretaken as system inputs (manipulated variables) while product purity and recovery are systemoutputs (controlled variables).To obtain the input-output mathematical model, systemidentification toolbox in Matlab software is used to identify a mathematical model that best fitthe step response data which is obtained by conducting various input step tests to each input.　　For control purpose, Model Predictive Control (MPC) is applied to handle the inherentnon-linear nature and discontinuous operation of the VPSA process.The MIMO controlscheme istested and showed good results in terms of system stability and fast tracking of theset-points.Finally, PSOalgorithm is applied to optimize the feed and purge durations tomaximize the purity and recovery, the resultsshows that with such VPSA process, over 98％of CO2 is recovered with purity of 59％, at tfeed =202s and tpurge =114s.　　Keywords: VPSA control; MIMO-MPC; CO2 capture