Prof. Peter SCHMID
King Abdullah University of Science and Technology (KAUST)
9 allée de la Découverte
4000 Sart-Tilman, Liège
Data-based reductions and closure models
for dynamical systems
The reduced-order description of physical systems from observed data is an important component in the analysis of many dynamical processes occurring in nature. In the first part of this talk, we will introduce a data-based reduction using transfer operators and graph theory, which will result in a Markov-model description of the underlying process. It will be exemplified on a magnetic Taylor-Green model for the Earth's magnetic field reversal. The second part will address the stability of closure approaches in highly truncated model reductions. A residualization, rather than truncation, technique will be discussed and applied to the closure of the harmonically balanced Navier-Stokes equations using a machine-learning approach.