Structured hybrid modelling: mathematics as key technology for process analysis and optimization
- Andreas Schuppert (Bayer AG Leverkusen, Mathematische Verfahrenstechnik)
Abstract
Model-based control and optimization, hitherto suffering from a severe lack of available models, dramatically progresses using Structured Hybrid Models (SHM). Combining rigorous and black-box models with a knowledge-based flowsheet, SHM's are extrapolable, if the flowsheet depicts the real process structure, if the dimension of the data-base can be reduced, and if the model can be identified using overall process data.
Mathematical considerations show that, even for complex flowsheets, SHM extrapolability only depends on the complexity of the constituting black-boxes. Improving flowsheet structure enhances extrapolability and reduces data demand.
SHM's exhibit significant practical advantages over traditional techniques and are open for new fields of application.