Model-based quality control in continuous manufacturing of pharmaceutical granules (QC4CM)
Professor Dr.-Ing. Dirk Abel
Rheinisch-Westfälische Technische Hochschule Aachen
Fakultät für Maschinenwesen Institut für Regelungstechnik (IRT)
Professor Dr. Jörg Breitkreutz
Institut für Pharmazeutische Technologie und Biopharmazie
The continuous wet granulation and drying process chain is an important step in pharmaceutical manufacturing. Whereas classical manufacturing relies on batch-wise processes, continuous processing promises to increase productivity and efficiency. In this proposal we consider a process in which powder containing active ingredients and excipients is dosed with water into a twin screw granulator. Within the screw granulator, two important unit operations wetting and kneading are performed yielding pharmaceutical granules. These granules are then dried using a vibrated fluidized bed dryer. Overall, particle processing as well as product formulation take place during the considered process. The added complexity of continuous manufacturing requires more advanced control concepts for a robust process behaviour. This research project aims to develop an overarching control concept to control the critical quality attributes (CQAs) of the described process. Process analytical technologies are developed that allow access to the relevant CQAs like moisture content or drug concentration in real-time. For this purpose, near-infrared spectroscopy and multi-frequency microwave resonance measurements will be used. The resulting system is then reduced in cost and improved in robustness for the application in a realistic production environment. Measured signals are used for the generation of physics and data-based real-time capable grey box models. These models are defined for each unit operation. The models consist of a white box model structure based on first principles combined with a data-driven representation of any intractable relations. Methods are developed that allow the continuous adaption of these models towards changes in the process or materials. The resulting adaptive models are embedded into a robust nonlinear model predictive control framework. This framework allows to observe and control the intermediate CQAs of each unit operation facing model and measurement uncertainties. Novel models are then derived that describe the dynamic behaviour of the sensitive active ingredient Enalapril inside the process. The ingredient decays in contact with water and at high temperatures. The resulting models are used to close a high level control loop including multiple unit operations. In this high level optimization, the Enalapril concentration at the end of the production line is optimized. To allow the continuous validation of the applied methods and models, stochastical digital twins of the processed granules are developed. These digital twins are used to create predictions of CQA values which can be compared with ex-situ samples taken during production. Agreement of the prediction and samples validate the process control. All methods and models will be tested on a laboratory wet granulation path of the type QbCon1 present at HHU Düsseldorf.