Hybrid bioprocess model towards the development of a digital twin for an industrial fermentation process

Marc Lemperle*, Pedram Ramin, Julian Kager, Benny Cassells, Stuart Stocks, Krist V. Gernaey

*Corresponding author for this work

Research output: Chapter in Book/Report/Conference proceedingArticle in proceedingsResearchpeer-review

160 Downloads (Orbit)

Abstract

This study presents a bioprocess hybrid model towards the usage in a digital twin framework for an industrial fermentation process involving filamentous fungi. The hybrid model consists of a Light Gradient Boost Machine Model (LGBM) that predicts a viscosity value based on seven online features enabling an accurate prediction of a dynamic profile of for the volumetric oxygen mass transfer coefficient (𝐾𝐿𝑎). The features with the greatest impact on the viscosity predictions are accumulated ammonia flow, accumulated carbon evolution rate (CER) and accumulated air flow. The results highlight the importance of investing in online data, e.g. online viscosity, and digital technologies such as AI and data-driven modelling, to improve prediction accuracy.
Original languageEnglish
Title of host publicationESCAPE 35: 35th European Symposium on Computer Aided Process Engineering 2025 : Book of Short Papers
Editors Jan F. M. Van Impe, Grégoire Léonard, Satyajeet S. Bhonsale, Monika E. Polanska, Filip Logist
PublisherEUROSIS
Publication date2025
Pages199-200
Article number1589
ISBN (Electronic)978-9-492859-36-5
Publication statusPublished - 2025
Event35th European Symposium on Computer Aided Process Engineering (ESCAPE 35) - KU Leuven Campus Ghent, Ghent, Belgium
Duration: 6 Jul 20259 Jul 2025

Conference

Conference35th European Symposium on Computer Aided Process Engineering (ESCAPE 35)
LocationKU Leuven Campus Ghent
Country/TerritoryBelgium
CityGhent
Period06/07/202509/07/2025
SeriesSystems & Control Transactions
Volume4
ISSN2818-4734

Keywords

  • Hybrid model
  • Light gradient boost machine model
  • Online viscosity model
  • Oxygen mass transfer coefficient

Fingerprint

Dive into the research topics of 'Hybrid bioprocess model towards the development of a digital twin for an industrial fermentation process'. Together they form a unique fingerprint.

Cite this