Machine learning algorithms are drawing attention for modelling processes in the chemical and biochemical industries. Due to a lack of fundamental understanding of complex processes and a lack of reliable real-time measurement methods in bio-based manufacturing, machine learning approaches have become more important. Hybrid modelling approaches that combine detailed process understanding with machine learning can provide an opportunity to integrate prior process knowledge with various measurement data for efficient modelling of the (bio) chemical processes. In this study, the application of a hybrid modelling framework that combines various first-principles models with machine learning algorithms is demonstrated through a laboratory-scale case of flocculation of silica particles in water. Since flocculation is a process that occurs across length- and time scales, an integrated hybrid multi-scale modelling framework can improve the phenomenological understanding of the process. The first-principles models utilized in this study are molecular scale particle surface interaction models such as combined with a larger-scale population balance model.
|Conference||31st European Symposium on Computer Aided Process Engineering (ESCAPE 31)|
|Period||06/06/2021 → 09/06/2021|
|Series||Computer Aided Chemical Engineering|
- Hybrid modelling
- Surface interactions
- Interfacial tension energy