Optimizing model predictive control of the chromatographic multi-column solvent gradient purification (MCSGP) process

The multi-column solvent gradient purification (MCSGP) process is a new continuous chromatographic process specifically designed for the purification of proteins and peptides. Due to its countercurrent purification principle, the performance increase with respect to the batch process can be up to 10× in productivity and even more at higher yields. Various successful applications of the MCSGP-technology have been reported, e.g. the purification of monoclonal antibodies from high-titer supernatants with ion-exchange resins and the purification of polypeptides.However, optimal MCSGP operation is a challenge and the current practice is to operate the MCSGP units at sub-optimal conditions to guarantee robustness and improve the separation performance by tuning the operating conditions off-line. As a result, systematic optimization tools like control and automation of MCSGP is of great interest in order to exploit the full economic potential of this process. An automatic control algorithm for MCSGP units that guarantees an optimal, robust operation with product purities in specification is a challenging problem because of the uncertainty related to the adsorption behavior of the mixture to be separated and the complex dynamics involved in this process, i.e. its cyclic and hybrid nature due to the inlet/outlet port switching with strong nonlinearities and delays in the feedback information.In this work, a control algorithm based on previous work for control of continuous chromatographic process is developed for the MCSGP-technology. The flow rates as well as the modifier gradients have been chosen as manipulated variables. The suitability of the controller is proven by means of two simulation case studies, i.e. the separation of monoclonal antibody variants and of a mixture of peptides.