Machine-learning-augmented domain decomposition method for near-wall turbulence modeling

Physical Review Fluids, vol. 9

Abstract

In this work, we developed a novel framework for incorporating the near-wall non-overlapping domain decomposition (NDD) method with the machine learning technique. It allows the solution to be calculated with a Robin-type (slip) wall boundary condition on a relatively coarse mesh and then be corrected in the near-wall region by solving the thin boundary-layer equations on a fine subgrid. Through an estimated turbulent viscosity profile provided by a neural network, the proposed method can be easily extended to different turbulence models and achieve commendable accuracy for the test cases of turbulent wall-bounded flows at various Reynolds numbers.

Authors 3

1. Shiyu Lyu corresponding

   Technical University of Munich

   Affiliation as printed

   Chair of Aerodynamics and Fluid Mechanics, School of Engineering and Design, Technical University of Munich, Boltzmannstrasse 15, 85748 Garching, Germany

2. Jiaqing Kou Aachen Institute of Aerodynamics

   RWTH Aachen University

   Affiliation as printed

   Institute of Aerodynamics, RWTH Aachen University, Wüllnerstrasse 5a, 52062 Aachen, Germany

3. Nikolaus A. Adams

   Technical University of Munich

   Affiliation as printed

   Chair of Aerodynamics and Fluid Mechanics, School of Engineering and Design, Technical University of Munich, Boltzmannstrasse 15, 85748 Garching, Germany

   Munich Institute of Integrated Materials, Energy and Process Engineering, Technical University of Munich, Lichtenbergstrasse 4a, 85748 Garching, Germany