Abstract
The increasing complexity of industrial biotechnology demands advanced modeling techniques capable of capturing the intricate dynamics of bioreactors. Traditional regression-based and empirical methods often fall short when confronted with the highly nonlinear behavior and limited experimental data characteristic of bioprocesses. Addressing these challenges requires a more intelligent approach—one that leverages domain knowledge to model complex bioprocess dynamics effectively, even with sparse data, while maintaining interpretability and robustness. In this study, we introduce a process-informed, data-driven methodology for modeling the dynamics of industrial bioreactors, leveraging the capabilities of the rising field of Scientific Machine Learning (SciML). Our approach leverages Physics-Informed Neural Networks (PINNs) to seamlessly integrate domain knowledge encoded in physical laws with sparse experimental data and deep learning techniques, enabling precise simulation and modeling of bioreactor operations. The framework is validated using real-world experimental data, demonstrating its capability for state and space estimation, essential for optimizing biomanufacturing processes. Our findings underscore the potential of Physics Informed Neural Networks to address the challenges of dynamic modeling in bioprocesses, paving the way for the development of intelligent, autonomous, and data-driven solutions in industrial biotechnology. This fusion of data-driven and mechanistic modeling represents a transformative step toward enhancing the efficiency, scalability, and sustainability of modern biomanufacturing practices.