Optimization and nonlinear model predictive control of some simulated moving bed process variants

Abstract

The interest in the simulated moving bed (SMB) technology has turned towards its variants due to the higher performance in separating and purifying different compounds such as enantiomers. Those variants are related to the relaxation of some parameters or the non-conventional arrangements of the train of chromatography columns that make up the SMB technology. Some of the most relevant and popular are the ModiCon and the VariCol processes, obtained by the periodic modulations of feed concentration and the length of arrangements of columns. Optimization and control studies for each of these variants are essential as they help discover new features and create guidelines for selecting and operating them. In this work, ModiCon, VariCol, ModiCon+VariCol (hybrid operation) were studied in detail, and their optimal performances were compared. The application of nonlinear model predictive control (NMPC) was also evaluated for the ModiCon and the VariCol processes. Different examples of enantioseparations were considered in the optimization and control studies. As a result, the complexity of the design optimizing problem of the VariCol process was reduced from MINLP (mixed-integer nonlinear programming) to an NLP (nonlinear programming). The ModiCon and the ModiCon+VariCol processes’ optimal throughput was almost doubled that of the conventional SMB process. The controller responded satisfactorily to different disturbances. In all cases, it fulfilled the purity of raffinate and extract and optimized the process’s economics.