/PIFNOeikonal

Primary LanguageJupyter Notebook

PIFNOeikonal

This repository releases the codes related to the paper "Seismic traveltime simulation for variable velocity models using physics-informed Fourier neural operator" submitted to IEEE TGRS.

Software requirement

Python Version: 3.8.16

Pytorch Version: 2.0.0+cu117

Code explanation

PIFNO_eikonal_example.ipynb: Solving the eikonal equation using PIFNO for various models from OpenFWI.

FMM_T_T0.ipynb: Generate reference traveltimes and background traveltimes using FMM.

Overview

We have developed an innovative multi-source seismic traveltime simulation method adaptable to various velocity models, employing an advanced deep-learning technique known as the physics-informed Fourier neural operator (PIFNO). curvelet50_train_T Training data from the CurveVel-A family: velocity models (first column), numerical traveltime from FMM (second column), predicted traveltime from PIFNO (third column), and traveltime difference (fourth column) for a source in the middle.

Citation information

If you find our codes and publications helpful, please kindly cite the following publications.

@article{song2024pinnpstomo,

title={Seismic traveltime simulation for variable velocity models using physics-informed Fourier neural operator},

author={Song, Chao and Zhao, Tianshuo and Waheed, Umair bin and Liu, Cai},

journal={arXiv preprint arXiv:2311.03751},

year={2023}

}

contact information

If there are any problems, don't hesitate to get in touch with me through my emails: chao.song@kaust.edu.sa;chaosong@jlu.edu.cn