The performance of the simulated moving bed (SMB) technology and its modification, the Varicol process, was optimized using an experimentally verified model for the enantioseparation of SB-553261 racemate. Single and multiobjective optimizations have been carried out for both existing as well as design stage and their efficiencies were compared. The optimization problem involves a relatively large number of decision variables, both continuous variables such as flow rates, switching time and length of the columns, as well as discrete variables like number and distribution of columns. A state-of-the-art new optimization technique based on a genetic algorithm (nondominated sorting genetic algorithm with jumping genes) was utilized which allows handling of these complex optimization problems. The optimization results showed that significant improvement could be made to the chiral drug separation process using both the SMB and the Varicol process. It was found that the performance of a Varicol process is superior to that of a SMB process in terms of treating more feed using less desorbent or increasing productivity while at the same time achieving better product quality. Optimum results were explained using equilibrium theory by locating them in the pure separation region.
Copyright 2004 Wiley Periodicals, Inc.