IFPAC Digital 2021 – Topics and Synergies with Bioprocess Technology at TU Wien

A key aspect in our Bioprocess Simulation & Control working group is to condense as much process knowledge as possible from experimental data into mathematical models and digital twins. We participated on this year’s IFPAC meeting which took place digitally from February 28^th until March 5^th in order to increase our understanding of the current challenges and industrial demands in the digital bioprocess technology. We collected some key topics that are highly related to our field of research.

General benefits of incorporating digital twins into biotechnological processing

The capability to simulate process behaviors and expand control strategies by the use of digital twins poses a number of benefits, some of which are:

Greatly reduced experimental workload in Research & Development
Multi-objective optimization of Critical Quality Attributes (CQAs) and economic factors like costs, space-time-yields etc. possible
Prediction of process failures and other risks possible
Platform for the exchange of process knowledge between different stakeholders

Bioprocess digital twins based on mechanistic and hybrid models

The advantages of incorporating available scientific knowledge and engineering principles into process models rather than using just data driven statistical models is a common ground in the modern understanding of the Quality by Design paradigm. With statistical MVDA models the combined effects of the critical process parameters (CPPs) on the critical quality attributes (CQAs) used to be determined by a physical DoE with respect to a few factors. However, there are limitations when it comes to knowledge transfer to other equipment and scales. Mechanistic models allow to put meaningful constraints on the model and thereby increasing robustness but it requires a much bigger effort to map complex relationships and maintain flexibility at the same time. Also, hybrid models which capture statistical and mechanistic relationships have been addressed several times to overcome limitations of both modeling paradigms for improved interpretation, predictive accuracy and simulation capabilities.

Advanced Monitoring and Process Analytical Technology

Improvements in process monitoring are essential for effective QbD and PAT realization with both hardware and software sensors. On the hardware sensor side spectroscopic methods like Raman spectroscopy were frequently mentioned as a rich data source to gain process knowledge from. Soft sensors can be based on those measurements to increase the overall observability or to detect deviating behavior of a batch early on in the process. Hybrid models again have been often proposed here to combine existing process knowledge with statistical machine learning models to predict cell growth, cell viability, substrate uptake and the formation of product.

Data architecture

Several recommendations have been presented regarding the architecture and management of data, which highly influences the quality of system interfaces and the depth of information which can be gathered from this data. Points that were discussed include:

What leads to good design and functional data architecture?
How to modernize an existing data environment?
Considerations and trade-offs when building data architectures for multiple applications and use-cases
(industrial) Internet of Things (IoT), Lab digitalization, Intelligent connections between laboratory devices and software

For further information on the International Forum on Process Analytical Chemistry visit: www.ifpacglobal.org