Optimization, state estimation, and model predictive control of simulated moving bed processes

Simulated Moving Bed (SMB) processes are highly efficient continuous chromatographic processes for the separation of components due to a counter-current flow of the liquid phase and the solid bed. Especially for temperature sensitive components or for substances with similar thermodynamic properties, the SMB process poses an important process option. The counter-current flow is established by switching the inlet ports and outlet ports periodically in the direction of the liquid flow. Since SMB processes are characterized by mixed discrete and continuous dynamics, spatially distributed state variables with steep slopes, and slow and strongly nonlinear responses of the concentration profiles to changes of the operating parameters, they are difficult to optimize and to control. In order to exploit the full economic potential of such processes, advanced optimization and control approaches based upon rigorous nonlinear process models employing efficient algorithms for simulation and optimization are needed. This thesis considers the optimization of SMB processes including the variants ModiCon SMB and the reactive Hashimoto SMB process based on direct mathematical optimization. In this work, the superior performance of these variants in terms of reduced eluent consumption and higher feed throughput is demonstrated. In order to control the product purities of SMB processes in the presence of plant disturbances, an advanced control scheme is developed incorporating a moving horizon state and parameter estimation scheme (MHE) and a nonlinear model predictive control (NMPC) scheme. The NMPC scheme directly optimizes the economic process performance which is referred to as optimizing control.