Fourier-DeepONet for full waveform inversion (FWI)

The data and code for the paper M. Zhu, S. Feng, Y. Lin, & L. Lu. Fourier-DeepONet: Fourier-enhanced deep operator networks for full waveform inversion with improved accuracy, generalizability, and robustness. Computer Methods in Applied Mechanics and Engineering, 416, 116300, 2023.

Datasets

• FlatVel-B (FVB)
• CurveVel-A (CVA)
• CurveFault-A (CFA)
• Style-A (STA)

Run data_gen_f.py, data_gen_loc.py, and data_gen_loc_f.py to generate seismic data of FWI-F, FWI-L, and FWI-FL, respectively.

Code

In train.py and test.py, change the arguements 'dataset' and 'task' in main function as needed.

Run train.py for training, and then run test.py for testing.

Cite this work

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

@article{zhu2023fourier,
  title   = {{Fourier-DeepONet}: Fourier-enhanced deep operator networks for full waveform inversion with improved accuracy, generalizability, and robustness},
  author  = {Zhu, Min and Feng, Shihang and Lin, Youzuo and Lu, Lu},
  journal = {Computer Methods in Applied Mechanics and Engineering},
  volume  = {416},
  pages   = {116300},
  year    = {2023},
  doi     = {https://doi.org/10.1016/j.cma.2023.116300}
}

Questions

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