A Global State Feedback Linearization and Decoupling MPC of a MIMO Continuous MSMPR Cooling Crystallization Process

Continuous manufacturing of pharmaceuticals requires effective control strategies to meet the tightly regulated critical product quality attributes, which has triggered a wide interest in model predictive control (MPC) techniques. This study applies a linear form of MPC to a nonlinear seeded continuous MSMPR crystallization process of acetaminophen in water, by using a global linearization technique; input-output state-feedback linearization (SFL). The two novelties are the application of MPC with SFL to continuous crystallization for SISO and MIMO systems; and the handling of the constraints as part of the SFL framework. For a single-input single-output system (SISO), MPC with SFL is applied to achieve a supersaturation reference trajectory by manipulating the coolant temperature subject to bounds and constraints. Additionally a multiple-input multiple-output (MIMO) case is presented using MPC with SFL and decoupling (SFLD). The supersaturation control from SISO was implemented in conjunction with crystal mean size control by manipulating the seed loading rate within bounds. The SISO configuration managed to control the crystallization process to within 1% of a prescribed set-point value within 2 residence times. In contrast, the MIMO controller was only able to stay within 4% of the set-point after 10 residence times, likely due to the highly coupled effects of the inputs on the outputs, which are lost in part due to the SFLD.