Modelling, Simulation, and Control of Polymorphic Crystallization

Abstract

Polymorphism is a phenomenon where multiple crystal forms exist for the same chemical compound and each form has distinct physical properties. Therefore, controlling polymorphism to ensure consistent production of the desired polymorph is important especially in drug manufacturing where safety is paramount. This study investigates the modelling, simulation, and control of polymorphic crystallization of L-glutamic acid.To facilitate the determination of optimal operating conditions and speed process development, a kinetic model for polymorphic crystallization of L-glutamic acid is developed using Bayesian inference. The developed kinetic model appears to be the first to include all of the transformation kinetic parameters including dependence on the temperature. In addition, the distributions of the model parameters are obtained.Selecting an efficient and sufficiently accurate computational method for simulating the model is important to ensure the behaviour of the numerical solution is determined by the assumed physical principles and not by the chosen numerical method. In this study, the weighted essentially non-oscillatory (WENO) methods are investigated and shown to give better computational efficiency than the standard numerical methods to simulate the polymorphic crystallization model developed in this thesis.Finally, control strategies are developed for the polymorphic transformation of L-glutamic acid from the metastable alpha-form to the stable beta-form crystals. These include the T-control and C-control strategies which are the most widely used control strategies in non-polymorphic crystallization, a computational efficient nonlinear predictive control (NMPC) strategy which does not rely on nonlinear programming, and the integrated NMPC and batch-to-batch (NMPC-B2B) control strategy based on a hybrid model in which the NMPC performs online control which handle constraints effectively while the batch-to-batch control refines the model by learning from the previous batches. Simulation results are presented to demonstrate and compare the performance and robustness of each control strategy.