(158d) The Role of Machine Learning and Model Based Control Techniques in Continuous “Plug and Produce” Pharmaceutical Micro-Factories.

Micro-factories are small to medium processing units that offer significant benefits including reconfigurability, reduced footprint and efficient processing. They typically employ continuous manufacturing for efficient scale-up, improved safety, lower energy consumption and reduced waste to support sustainability goals. Modular designs combined with plug and play allow rapid reconfiguration to allow a wide range of products to be manufactured. Finally, easier incorporation of ML/AI techniques and Process Analytical Tools (PAT) improve process robustness through model-based control and real-time diversion strategies.

The inherent flexibility of micro-factories also introduces complexity. For example, plug and play modules increase the need for seamless integration and a reconfigurable process automation layer. This flexibility must also be supported in the ML/AI systems, with model-based monitoring and diversion strategies adapting to the equipment configuration and throughput.

In this presentation, we discuss case studies in continuous processing for drug substance and drug product where these challenges are addressed. In the continuous drug substance case study, we adopt the Pharma 4.0 “Plug and Produce” architectural principles for a continuous drug substance process micro-factory at Strathclyde University’s CMAC research center. Applied Materials’ SmartFactory Rx PharmaMV software is integrated with the micro-factory to provide plug and play integration of process and PAT data, delivering ML driven approaches to optimize experimental design. The microservices based integration between various components is used to demonstrate how the challenges associated with complexity seen in continuous manufacturing are addressed. Further, digital maturity towards predictive operations using model based control is enabled.

In a continuous drug product case study, we present an end-to-end continuous direct compaction platform-based solution. Workflows are employed to combine PAT and soft sensor predictions to provide robust real-time diversion for quality control. This case study is the result of working in partnership with The Medicines Manufacturing Innovation Centre (MMIC) which is a collaboration between the Centre for Process Innovation, University of Strathclyde, UK Research & Innovation, Scottish Enterprise and founding industry partners, AstraZeneca, and GSK.

Through these examples we demonstrate how the benefits of micro-factories can be enhanced through the application of flexible, plug and play, model-driven automation.