Large eddy simulation of transitional channel flow using a machine learning classifier to distinguish laminar and turbulent regions

Abstract

Breakdown to turbulence in wall-bounded flows takes place through sporadic bursts of turbulent spots. Wall-modelled large-eddy simulations (LES) of transition to turbulence must dynamically identify the nascent turbulent regions, track their evolution, and apply the appropriate wall stress within and outside the turbulent/non-turbulent (T-NT) interface. Self-organized maps (SOM), a machine learning classifier, objectively and efficiently captures the T-NT interface. Wall-modeled LES with SOM interface identification predicts both orderly and bypass transition.