Neural Topology Optimization Via Active Learning for Efficient Channel Design in Turbulent Mass Transfer

The data and code for the paper C. Kou, Y. Yin, M. Zhu, S. Jia, Y. Luo, X. Yuan, L. Lu. Neural Topology Optimization Via Active Learning for Efficient Channel Design in Turbulent Mass Transfer, Advanced Science, 12(37), e08386, 2025.

Data

• The training dataset of neural operators can be downloaded via OneDrive. The dataset is generated by compiling UDF.c and then running fluent.cas.
• The raw experimental data can be found in experiment_data.

Code

All code are in the folder src.

• Run the train_GC.py and train_GP.py to train the neural operators.
• Run the main.py to perform neural topology optimization using the pretrained nueral operators.

Cite this work

If you use this data or code for academic research, you are encouraged to cite the following paper:

@article{kou2025neural,
  author  = {Kou, Chenhui and Yin, Yuhui and Zhu, Min and Jia, Shengkun and Luo, Yiqing and Yuan, Xigang and Lu, Lu},
  title   = {Neural Topology Optimization Via Active Learning for Efficient Channel Design in Turbulent Mass Transfer},
  journal = {Advanced Science},
  volume  = {12},
  issue   = {37},
  pages   = {e08386},
  year    = {2025},
  doi     = {https://doi.org/10.1002/advs.202508386}
}

Questions

To get help on how to use the data or code, simply open an issue in the GitHub "Issues" section.