LabelFree-DNN-Surrogate

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

Dependencies

• python 3
• PyTorch 0.4 and above
• TensorFlow 1.15
• matplotlib
• seaborn

Installation

• Install PyTorch, TensorFlow and other dependencies

• Clone this repo:

git clone https://github.com/Jianxun-Wang/LabelFree-DNN-Surrogate.git
cd LabelFree-DNN-Surrogate

Uncertainty Propagation

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

Training

Parametric Pipe Flow

Train a parametric DNN surrogate for pipe flow

cd Tutorial
python poiseuillePara.py

Parametric Aneurysmal Flow

Train a parametric DNN surrogate for aneurysmal flow

cd ParametricAneurysm
python main.py

Citation

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"
}

Acknowledgments

Thanks for all the co-authors and Dr. Yinhao Zhu for his valuable discussion.

Code is inspired by cnn-surrogate