Dresden 2020 – wissenschaftliches Programm
Die DPG-Frühjahrstagung in Dresden musste abgesagt werden! Lesen Sie mehr ...
Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe
DY: Fachverband Dynamik und Statistische Physik
DY 3: Fluid Physics of Turbulence
DY 3.3: Vortrag
Montag, 16. März 2020, 10:15–10:30, ZEU 118
Machine learning in subcritical plane Couette flow — •Stefan Zammert — Philipps-Universität Marburg
Plane Couette flow shows transient turbulence for Reynolds numbers where the laminar flow is linearly stable. In this so-called subcritical range the time evolution of the flow is deterministic but a turbulent trajectory eventually returns to the laminar state without any obvious precursor.
We study small periodic domains of plane Couette flow and use neural networks to predict if a turbulent trajectory returns to the laminar state within a fixed time T. The performances of the network for variations of the input variables are compared with the goal to minimize the amount of input variables necessary for a good prediction.
Having a reliable and fast method to predict the the decay of turbulence by using a limited set of input quantities which is also easily accessible in experiments might for example be helpful for active turbulence control.