This workshop introduces participants to the development and application of digital twin technology for bioprocess systems, using Physics-Informed Neural Networks (PINNs) as the core modeling framework. Attendees will survey the digital twin landscape — from foundational concepts to the specific challenges of twinning biological systems — and critically compare conventional data-driven approaches against first-principle hybrid methods. Working through a real-life E. coli bioreactor case study, participants will confront the limitations of purely data-driven strategies under scarce experimental conditions and explore how embedding physical knowledge into the neural network architecture enables robust, data-efficient modeling. The case study then guides attendees through the twinning methodology, demonstrating how a physics-based scaffold is progressively updated with incoming process data to evolve from a digital shadow into a fully predictive digital twin. Participants are expected to bring their own laptop to fully engage in the hands-on sessions.
Participants should also ensure that they can access the workshop material through the Google Colab notebook pinn_cstr_workshop.ipynb – Colab, prior to the workshop, allowing sufficient time for preparation and active participation.
Workshop facilitators:
Κonstantinos Μexis (NTUA), Nikos Trokanas (NTUA), Antonis Kokossis (NTUA)
Organised by:
National Technical University of Athens (NTUA) as part of Horizon Europe project Bioindustry 4.0
Visit Website: https://ipsen.ntua.gr/
Follow us on LinkedIn:
http://linkedin.com/company/industrial-process-systems-engineering-unit