https://github.com/slimgroup/fno4co2
Learned coupled inversion with Fourier neural operators
https://github.com/slimgroup/fno4co2
carbon ccs deep-learning fno inversion julia machine-learning time-lapse
Last synced: 5 months ago
JSON representation
Learned coupled inversion with Fourier neural operators
- Host: GitHub
- URL: https://github.com/slimgroup/fno4co2
- Owner: slimgroup
- License: mit
- Created: 2021-06-01T14:57:02.000Z (almost 4 years ago)
- Default Branch: master
- Last Pushed: 2024-03-14T18:39:24.000Z (about 1 year ago)
- Last Synced: 2024-03-15T06:29:45.822Z (about 1 year ago)
- Topics: carbon, ccs, deep-learning, fno, inversion, julia, machine-learning, time-lapse
- Language: Julia
- Homepage: https://library.seg.org/doi/10.1190/image2022-3722848.1
- Size: 116 MB
- Stars: 19
- Watchers: 4
- Forks: 5
- Open Issues: 1
-
Metadata Files:
- Readme: README.md
- License: LICENSE
- Citation: CITATION.bib
Awesome Lists containing this project
README
# FNO4CO2
[![][license-img]][license-status] [![][zenodo-img]][zenodo-status]
This repository contains the implementation of learned coupled inversion framework and the numerical experiments in [Learned coupled inversion for carbon sequestration monitoring and forecasting with Fourier neural operators](https://arxiv.org/abs/2203.14396), accepted by the International Meeting for Applied Geoscience & Energy 2022.
The aforementioned framework entails a re-implementation of Fourier neural operators from [Fourier Neural Operator for Parameter Partial Differential Equations](https://arxiv.org/abs/2010.08895) authored by Zongyi Li et al. The [original repository](https://github.com/zongyi-li/fourier_neural_operator) is in python.
This code is based on the Julia Language and the package [DrWatson](https://juliadynamics.github.io/DrWatson.jl/stable/) to make a reproducible scientific project named
> FNO4CO2
To (locally) reproduce this project, do the following:
1. Download this code base. Notice that raw data are typically not included in the
git-history and may need to be downloaded independently.
2. Download [python](https://www.python.org/) and [Julia](https://julialang.org/). The numerical experiments are reproducible by python 3.7 and Julia 1.7.
3. Install [Devito](https://www.devitoproject.org/), a python package used for wave simulation.
4. Open a Julia console and do:
```
julia> using Pkg
julia> Pkg.activate("path/to/this/project")
julia> Pkg.instantiate()
```
This will install all necessary Julia packages for you to be able to run the scripts and
everything should work out of the box.
## Examples
The repository currently includes several scripts.
`gen_perm.jl` generates random permeability samples and `gen_conc.jl` generates time-varying CO2 concentration for each of them using the numerical simulator from [FwiFlow.jl](https://github.com/lidongzh/FwiFlow.jl). To save you some time to reproduce our examples, we've also provided the dataset through a dropbox link -- when you run the network training script, the dataset will be downloaded automatically.
`fourier_3d.jl` trains a 3D FNO which maps the permeability to time-varying CO2 concentration governed by two-phase flow equations, with the dataset generated by `gen_perm.jl` and `gen_conc.jl`. You can train the FNO on GPU if available by setting the env variable ``export FNO4CO2GPU=1`` in your environment (e.g. in `~/.zshrc` on my mac, or just do `FNO4CO2GPU=1 julia`). The trained 3D network for the two-phase flow example is provided in the repository under `data/3D-FNO`.
`fourier_3d_grad.jl` script shows how to conduct a learned inversion to estimate the permeability (input of FNO) from the CO2 concentration snapshots. It uses gradient descent with back-tracking line search to iteratively invert the FNO.
`learned_coupled_inversion.jl` script shows how to conduct a learned coupled inversion, i.e. we invert for the permeability from time-lapse seismic datasets. The process involves inverting multiple physics as shown in [Coupled Time-Lapse Full-Waveform Inversion for Subsurface Flow Problems Using Intrusive Automatic Differentiation](https://agupubs.onlinelibrary.wiley.com/doi/abs/10.1029/2019WR027032), while it uses a pre-trained FNO as a surrogate for the fluid-flow solver.
## Citation
If you use our software for your research, we appreciate it if you cite us following the bibtex in [CITATION.bib](CITATION.bib).
## Acknowledgements
We thank the developers from several software packages in the open-source software community, which we based our implementation on. The FNO is re-implemented following Zongyi Li et al's work in [https://github.com/zongyi-li/fourier_neural_operator](https://github.com/zongyi-li/fourier_neural_operator). The two-phase flow dataset is generated by [FwiFlow.jl](https://github.com/lidongzh/FwiFlow.jl). We use [Devito](https://www.devitoproject.org/) and [JUDI.jl](https://github.com/slimgroup/JUDI.jl) for wave simulations. We use [SetIntersectionProjection](https://github.com/slimgroup/SetIntersectionProjection.jl) for constrained optimization.
This research was carried out with the support of Georgia Research Alliance and partners of the ML4Seismic Center.
## Author
Ziyi (Francis) Yin, [[email protected]](mailto:[email protected])
[license-status]:LICENSE
[zenodo-status]:https://doi.org/10.5281/zenodo.6799258
[license-img]:http://img.shields.io/badge/license-MIT-brightgreen.svg?style=flat?style=plastic
[zenodo-img]:https://zenodo.org/badge/DOI/10.5281/zenodo.3878711.svg?style=plastic