Physics informed data-driven near-wall modelling for lattice Boltzmann simulation

PINN provides a data driven model to solve wall models in turbulence modelling with LB method. The trainig process is realized by pytorch.

Raw data

The traing and test dataset are obbtained from a high precision simulation. Data of two samplings ('sparse' and 'dense') are provided in the path './data' .

The file 'y2d' is the position of the grid in normal-to-wall direction. The files 'u' and 'yp-instant' are the velocities and position of the samples.

Training

Two training methods are provided in the folder './training'. The script 'NNWC' is the training method process without PDF correction, and the script 'NNPC'

is the training process with PDF correction.

Model implementation and visualization

The model is visualized by prediction. The prediction is realized by script 'plots' in './visualization'

Cite this work

@article{xue2024physics,
  title={Physics informed data-driven near-wall modelling for lattice Boltzmann simulation of high Reynolds number turbulent flows},
  author={Xue, Xiao and Wang, Shuo and Yao, Hua-Dong and Davidson, Lars and Coveney, Peter V},
  journal={arXiv preprint arXiv:2402.08037},
  year={2024}
}