Machine Learning Enabled Wall Modeling for LES of Turbulent Boundary Layers including Laminar Precursors

Abstract

The proposed effort aims to significantly enhance the accuracy, efficiency, and robustness of wall-~~modeled Large Eddy Simulations"" (LES) at high Reynolds number. In two closely coordinated projects at JHU and MIT, we propose to achieve these improvements by deve""loping an integrated approach to wall modeling, bridging first-~~principles based wall models for the initial laminar and post-~~tra""nsition fully turbulent portions, as well as the challenging transitional regime. The various complexities of the flow, i.e. in lami""nar boundary layers with cross stream pressure gradients, the structural complexity in the transition and turbulent regions and the" lack of analytical models motivates the use of machine learning tools. These will be used to derive efficient means to predict the instability mechanisms and local skin friction given complex conditions and parameters that characterize these conditions. The integrated wall modeling approach will be tested on available data including application to boundary layers on NACA 65airfoil.

Document Details

Document Type
DoD Grant Award
Publication Date
Sep 29, 2017
Source ID
N000141712959

Entities

People

  • Qiqi Wang

Organizations

  • Massachusetts Institute of Technology
  • Office of Naval Research
  • United States Navy

Tags

Fields of Study

  • Physics

Readers

  • Computational Fluid Dynamics (CFD)
  • Fluid Mechanics and Fluid Dynamics.

Technology Areas

  • AI & ML
  • AI & ML - Neural Networks