Physics-Informed Recurrent GAN for Flow Field Reconstruction

This code implements a physics-informed recurrent super-resolution generative reconstruction model for use in rotating detonation combustors.

Dependencies

To install the necessary dependencies, run:

pip install -r requirements.txt

Dataset

The dataset is generated using the in-house code TurfSIM of SCP and is approximately 50GB in size.

Downloading Data

1. The source files are listed in the source directory. You can download the .dat source files from the following link: Download Source Files.

2. After downloading the .tif files, run generate_data.py located in the source directory. This script will process the source files and construct an index for the dataset.

Example Usage

To initiate distributed training with a batch size of 64 for 200 epochs, use the following command:

python train.py --batchSize 64 --nEpochs 200

For a complete evaluation of the test set, use:

python test.py --dic #path_to_model

Citation

If you find this work useful, please cite it as follows:

@article{Wang2024,
    title = {Physics-informed recurrent super-resolution generative reconstruction in rotating detonation combustor},
    journal = {Proceedings of the Combustion Institute},
    volume = {40},
    number = {1},
    pages = {105649},
    year = {2024},
    issn = {1540-7489},
    doi = {https://doi.org/10.1016/j.proci.2024.105649},
    author = {Xutun Wang and Haocheng Wen and Quan Wen and Bing Wang},
    keywords = {Flow-field reconstruction, Physics-informed machine learning, Generative adversarial network, Recurrent neural network, Rotating detonation combustor}
}