Using Machine Learning for formulating new wall functions for Large Eddy Simulation: A First Attempt

Preprint, 2022

Machine learning is used for developing wall functions for Large Eddy Simulations (LES). I use Direct Numerical Simulation (DNS) of fully-developed channel flow at frictional Reynolds number of 800 to create a database. This database is using as a training set for the machine learning method (support vector regression). The input data (i.e. the influence parameters) are the local Reynolds number, the non-dimensional velocity gradient and the timeaveraged y + value. The machine learning method is trained to predict the wall shear stress. 
The support vector regression methods in Python are used. The trained machine learning model is saved to disk and it is subsequently uploaded into the Python CFD code pyCALC-LES (Davidson, 2021). LES is carried out on coarse – and semi-course – near-wall meshes and the wall-shear stress is predicted using the developed machine learning models.