5th Hybrid Modeling Summer School

8th – 10th of September 2021

NOVA  Science and Engineering School, Universidade Nova de Lisboa, Portugal

Why participating?

  • 5th edition of a well-established summer school on the topic of hybrid modelling for process development, characterization and optimization promoted by ESBES

  • Contact with some of the most renowned specialists in academia and industry in the field of hybrid modelling. 

  • Last edition participants included Bilfinger, Glaxo-Smith-Kline, Bayer, Eppendorf, Boehringer Ingelheim, GE Healthcare, Novasign, ROK Bioconsulting, MCL Gmbh, Hovione, IBET, University of Manchester, Imperial Coledge London, TU Wien, TU Berlin, BOKU University, Nova School of Science, Instituto Superior Técnico, University of Coimbra

  • Strong practical component with a two sessions of hands-on training with computers.

  • Free access to a powerful MATLAB hybrid modelling toolbox; free access granted only for academic use;

Who should attend?

  • PhD students with a strong modelling background and MATLAB language.

  • Senior scientists in academia or industry with a strong modelling background
  • The course is suitable for beginners but also for knowledge consolidation
  • Basic knowledge in machine learning topics is advised but not mandatory 

Why hybrid modelling?

It is foreseen that digitalization based on Machine Learning (ML) will mark profoundly process manufacturing efficiency in the future. Digitalization based on ML naturally challenges our traditional mathematical modeling vision in the process industries. First Principles, mechanistic modeling, phenomenological modeling, or even semi-empirical modeling, embody a form of human knowledge abstraction, which is hardly compatible with current ML approaches. Hybrid ML/Physical modeling follows the principle that ML should enhance traditional engineering rather than replace it. It has the potential to reduce the number of experiments required for model development and in addition the extrapolation properties of the hybrid model are typically much better as e.g. those of strictly data driven models. These properties make hybrid modeling approaches particularly interesting for process development, optimization and scale-up and their improved interpretability has been recognized as an advantage in the QbD and PAT concepts. This course is directed towards PhD-students, Postdoctoral researchers and industry experts that have interest in process modeling and seek for methods to improve process modeling, such ultimately enhancing process operation and design in an efficient way.


Course objectives

  • Provide detailed knowledge about the benefits and challenges of hybrid modeling.
  • Enable the participant to independently develop and apply hybrid models.
  • Present and discuss use cases and applications for Process Development, Process Characterization and Process Optimization/closed Loop applications
 

Topics to be covered

The topics, covering both hybrid modeling fundamentals and their application, will be presented by leading experts. While the fundamental part covers questions on “How to develop a hybrid model?”, “How to enhance the model quality by design of experiments?” and “How to compare different hybrid models?”, it will also let the participants in on tips and tricks for the efficient development of high quality hybrid models. Hands on training will support the learning process and result in a profound understanding of the fundamentals. These courses will be held by leading experts. Hands on training will substantiate the presented topics. This will leave the participants with the necessary knowledge and skills to try hybrid modeling on their own. The industrial oriented part covers the application of hybrid modeling for typical bioprocess engineering and its utilization to support the Quality by Design concept, as promoted by the Process Analytic Technology initiative (PAT). Concepts and industrial applications for process development (e.g.: Shake Flasks to 15l Bioreactor), process characterization (e.g.: E.coli fermentation) and ways towards closed optimization (e.g.: model predictive control of fermentation processes) will be presented and discussed with the experts.