Systems Engineering for Biopharmaceutical Manufacturing Processes with Phase Transitions

Abstract

The biopharmaceutical industry has been growing significantly over the past few years. Various biotherapeutics have been developed and played a crucial role in advancing global health, e.g., monoclonal antibodies, cell and gene therapy, and messenger ribonucleic acid (mRNA) vaccines. Nevertheless, the current biopharmaceutical manufacturing is facing several challenges associated with production scalability, operational flexibility, and quality control. For example, many biopharmaceutical processes are still operated in batch mode, while continuous manufacturing has already been adopted in other major industrial sectors. As such, significant efforts have been recently dedicated to the development and integration of advanced manufacturing technology for biopharmaceutical processes.

This thesis presents an array of novel computational tools and methods across process systems engineering (PSE) domains developed for biopharmaceutical manufacturing. Several PSE applications are discussed and demonstrated, including process modeling, monitoring and state estimation, model-based control and optimization, and uncertainty quantification. Two important biopharmaceutical processes that involve phase transitions are studied, namely (1) lyophilization and (2) cell thawing, serving as the two main parts of this thesis.

The first part of this thesis centers on lyophilization. Lyophilization (also known as freeze drying) is a low-temperature, low-pressure dehydration process used for improving the stability of various biotherapeutics, including its recent application to mRNA vaccines. In this thesis, we first present a set of mechanistic models for various lyophilization designs, including conventional/batch lyophilization, microwave lyophilization, hybrid lyophilization, and the state-of-the-art lyophilization technology, namely continuous lyophilization of suspended vial. The validated models can accurately predict the evolution of the critical process parameters for the entire lyophilization cycle, including the product temperature, ice/water fraction, sublimation front position, and concentration of bound water (residual moisture). Second, we showcase a state observer that estimates the concentration of bound water by using temperature measurement as an input. This state observer allows for real-time tracking of the residual moisture, which is typically present in trace amounts and difficult to measure online. Third, we describe a highly efficient framework for incorporating the probabilistic uncertainty into our model via polynomial chaos theory (PCT). Our PCT-based lyophilization model is demonstrated for fast uncertainty quantification and stochastic control. Finally, we propose a new, efficient way of solving optimal control problems via reformulation and simulation of a hybrid discrete/continuous system of mixed-index differential-algebraic equations. The proposed algorithm is several orders of magnitude faster than the traditional optimization-based approaches while maintaining similar/better accuracy. The algorithm is also demonstrated for finding the optimal control policies for lyophilization under different conditions.

The second part of this thesis focuses on cell thawing. Cell thawing a critical step in cell therapy before cells are introduced to the patients. Traditionally, the thawing process requires human intervention to ensure proper progression and completion before the cells can be administered to the patients. In this thesis, we build a thermal imaging-based system that facilitates the monitoring, control, and automation of cell thawing. First, a mechanistic model is proposed to guide the design of cell thawing systems. Then, a thermal imaging-based state observer is developed for real-time process monitoring, which keeps track of the phase transition (melting/thawing) and accurately identifies the endpoint of the process. Lastly, we present a novel optimal control architecture for accelerating the process and controlling the final product quality corresponding to the thawing protocols.

Finally, with all the tools and methods established, we discuss the development of a digital twin for biopharmaceutical manufacturing. The digital twin receives and processes data from its physical counterpart, performs several complex computations in real time, and provides instructions/feedback to the physical system, serving as a virtual representation of the physical system of the manufacturing system. Ultimately, this thesis provides a complete PSE framework and case studies for advanced biopharmaceutical manufacturing that involves phase transitions.