An integrated approach for c-control of antisolvent crystallization processes

Abstract

Concentration control (C-control) strategy for (semi-)batch antisolvent crystallization processes has been recently developed with the aid of new sensors that measure in situ process variables. This control strategy gives better robustness over traditional owrate/ temperature control in the presence of process disturbances. However, the setpoint value for the existing C-control is determined through trial-and-error procedure and hence gives sub-optimal product quality in most of the cases. This motivates the current study to develop a two-staged integrated data-based approach to facilitate the determination of setpoint value for the implementation of C-control strategy in anti-solvent crystallization processes. In the first stage of the proposed design, a k-nearest neighborhood LSSVM (k-NN LSSVM) is used to select a subset of the training database that resembles the current batch dynamics as the relevant training database, which is subsequently used in the second stage for determination of the setpoint value of C-control. Simulation results show that the performance of proposed design is near to the best achievable performance of C-control strategy obtained with a known first-principles model for controlling the semi-batch antisolvent crystallization process. © 2012 IFAC.