Wall Modeling in LES of Turbulent Flows Using Reinforcement Learning

  • Vadrot, Aurélien (Aarhus University)
  • Yang, Xiang (Pennsylvania State University)
  • Abkar, Mahdi (Aarhus University)

Please login to view abstract download link

This work seeks to design a wall-model (WM) using unsupervised machine learning (ML) tools that can first recover the law of the wall in equilibrium flows before capturing the dynamics of non-equilibrium flows. In this study, a WM built on reinforcement learning and trained to take non-equilibrium effects into account is shown. This model will initially be compared on equilibrium half-channel flow with existing ML-based WM found in the literature. Then, non-equilibrium half-channel flows with small, medium, and strong pressure gradients applied in both the spanwise and streamwise directions will be used to evaluate the model.