Authors:
Kamil Dedecius
1
and
Pavel Ettler
2
Affiliations:
1
Academy of Sciences of the Czech Republic, Czech Republic
;
2
COMPUREG Plzeň s.r.o., Czech Republic
Keyword(s):
Bayesian modelling, Hierarchical model, Parameter estimation, Cold strip rolling, Rolling mills.
Related
Ontology
Subjects/Areas/Topics:
Informatics in Control, Automation and Robotics
;
Sensors Fusion
;
Signal Processing, Sensors, Systems Modeling and Control
;
Time Series and System Modeling
Abstract:
Industrial model-based control often relies on parametric models. However, for certain operational conditions either the precise underlying physical model is not available or the lack of relevant or reliable data prevents its use. A popular approach is to employ the black box or grey box models, releasing the theoretical rigor. This leads to several candidate models being at disposal, from which the (often subjectively) prominent one is selected. However, in the presence of model uncertainty, we propose to benefit from a subset of credible models. The idea behind the multimodelling approach is closely related to hierarchical modelling methodology. By using several modelling levels, it is possible to achieve relatively high quality and robust solution, providing a way around typical constraints in industrial applications.