Using Machine Learning for Improving a Non-Linear k-eps Model: A First Attempt
Report, 2023
the friction velocity and the viscosity or the turbulence kinetic energy and its dissipation.
The input variables are computed using DNS of developing boundary layer at Re_theta=8180. The NN model is created using Python's pytorch. It turns out that the NN model acts as a low-Re model, replacing all low-Re modifications which are present in non-ML non-linear eddy-viscosity turbulence models.
After having trained the NN model in the boundary-layer flow, it is validated in two channel flows at Re_tau = 550
and Re_tau = 5200$. Good agreement is obtained.
turbulence modelleing
non-linear models
Machine Learning
CFD
Author
Lars Davidson
Chalmers, Mechanics and Maritime Sciences (M2), Fluid Dynamics
Subject Categories (SSIF 2011)
Energy Engineering
Fluid Mechanics and Acoustics