Surrogate Modeling for Fluid Flows Based on Physics-Constrained Label-Free Deep Learning
Luning Sun, Han Gao, Shaowu Pan, Jian-Xun Wang
TensorFlow and PyTorch implementation of Physics-Constrained Label-Free Deep Learning
| Parametric Pipe Flow |
|---|
| Small Aneurysm | Middle Aneurysm | Large Aneurysm |
|---|---|---|
 |
 |
 |
- python 3
- PyTorch 0.4 and above
- TensorFlow 1.15
- matplotlib
- seaborn
-
Install PyTorch, TensorFlow and other dependencies
-
Clone this repo:
git clone https://github.com/Jianxun-Wang/LabelFree-DNN-Surrogate.git
cd LabelFree-DNN-Surrogate
Perform UQ tasks, compare the distribution of Quantity of Interest (QoI) between DNN model and OpenFOAM benchmar, including:
- Parametric Pipe Flow
- Parametric Geometry Aneurysm (To Be Added)
Example :
cd Tutorial
python pipe_post.py
| Parametric Pipe Flow |
|---|
Train a parametric DNN surrogate for pipe flow
cd Tutorial
python poiseuillePara.py
Train a parametric DNN surrogate for aneurysmal flow
cd ParametricAneurysm
python main.py
If you find this repo useful for your research, please consider to cite:
@article{SUN2020112732,
title = "Surrogate modeling for fluid flows based on physics-constrained deep learning without simulation data",
journal = "Computer Methods in Applied Mechanics and Engineering",
volume = "361",
pages = "112732",
year = "2020",
issn = "0045-7825",
doi = "https://doi.org/10.1016/j.cma.2019.112732",
url = "http://www.sciencedirect.com/science/article/pii/S004578251930622X",
author = "Luning Sun and Han Gao and Shaowu Pan and Jian-Xun Wang"
}Thanks for all the co-authors and Dr. Yinhao Zhu for his valuable discussion.
Code is inspired by cnn-surrogate