Note: This content is accessible to all versions of every browser. However, this browser does not seem to support current Web standards, preventing the display of our site's design details.


Controlling Simulated Moving Beds


G. Erdem

Seminars on Advanced Separation Processes, Process Engineering Department, ETH-Zürich, Switzerland

Nowadays, the Simulated Moving Bed (SMB) technology is adopted in the food, agrochemical, pharmaceutical and biotechnology industries for difficult applications, such as the resolution of racemates, and it is considered attractive for complex separation tasks, such as bioseparations or separation of natural compounds involving a number of difficult-to-characterize compounds. Therefore, there are now a large number of potential small-scale applications of the SMB technology that call for a new SMB paradigm, which exploits the flexibility and versatility of the technology. Proper implementation of SMBs in production will require the application of robust control techniques. The issue of process control under uncertainties, e.g. the competitive adsorption behavior of multi-component mixtures that can never be measured precisely, will have to be addressed. Recently more and more new SMB schemes like the VARICOL process, multicomponent separations and solvent gradient applications, which require more information for design purposes, are in focus. Especially those processes will largely benefit from appropriate process control. SMBs are constituted of several chromatographic columns with inlets and outlets, whose position within the column carousel switches periodically. Therefore SMBs reach only a cyclic steady state, where compositions change periodically and exhibit nonlinear dynamics with dead-times and lengthy analytic techniques for product quality assessment. These features pose fundamental questions and challenges on both SMB technology and control theory. In this work a new Model Predictive Control (MPC) concept for the SMB is proposed: An SMB process should follow a periodic reference trajectory, due to the periodic behavior of the system. The manipulated variables are the flowrates, whereas the output concentrations at the extract and raffinate are measured. This measurement output together with the information provided by a linearized reduced order model for the SMB plant goes to a Kalman filter, which is able to handle model errors and disturbances. Then the Model Predictive Controller calculates the new flowrates according to a previously defined objective function. The realization of this concept is discussed and the implementation on a virtual SMB platform including the developed modeling tools is assessed.


Type of Publication:


M. Morari

No Files for download available.
% No recipe for automatically generating a BibTex entry for (06)Talk
Permanent link