• Corpus ID: 238857286

Unsupervised Learning of Full-Waveform Inversion: Connecting CNN and Partial Differential Equation in a Loop

@article{Jin2021UnsupervisedLO,
  title={Unsupervised Learning of Full-Waveform Inversion: Connecting CNN and Partial Differential Equation in a Loop},
  author={Peng Jin and Xitong Zhang and Yinpeng Chen and Sharon Huang and Zicheng Liu and Youzuo Lin},
  journal={ArXiv},
  year={2021},
  volume={abs/2110.07584}
}
  • Peng Jin, Xitong Zhang, +3 authors Youzuo Lin
  • Published 14 October 2021
  • Computer Science, Engineering, Physics
  • ArXiv
This paper investigates unsupervised learning of Full-Waveform Inversion (FWI), which has been widely used in geophysics to estimate subsurface velocity maps from seismic data. This problem is mathematically formulated by a second order partial differential equation (PDE), but is hard to solve. Moreover, acquiring velocity map is extremely expensive, making it impractical to scale up a supervised approach to train the mapping from seismic data to velocity maps with convolutional neural networks… 

Figures and Tables from this paper

On the Robustness and Generalization of Deep Learning Driven Full Waveform Inversion
  • Chengyuan Deng, Youzuo Lin
  • Computer Science, Engineering
    ArXiv
  • 2021
TLDR
For robustness, the upper bounds of the deviation between the predictions from clean and noisy data are proved and a norm-based generalization error upper bound is proved via a stability-generalization framework.
OpenFWI: Benchmark Seismic Datasets for Machine Learning-Based Full Waveform Inversion
  • Chengyuan Deng, Yinan Feng, +4 authors Youzuo Lin
  • Computer Science, Engineering
    ArXiv
  • 2021
TLDR
OpenFWI is the first-of-its-kind in the geoscience and machine learning community to facilitate diversified, rigorous, and reproducible research on machine learning-based FWI.

References

SHOWING 1-10 OF 40 REFERENCES
Data-Driven Seismic Waveform Inversion: A Study on the Robustness and Generalization
TLDR
This article developed a real-time data-driven technique and it is called VelocityGAN, to reconstruct accurately the subsurface velocities and develops a transfer-learning strategy based on VelocityGAN to alleviate the generalization issue.
Deep-Learning Inversion of Seismic Data
  • Shucai Li, Bin Liu, +4 authors Peng Jiang
  • Computer Science, Mathematics
    IEEE Transactions on Geoscience and Remote Sensing
  • 2020
TLDR
End-to-end seismic inversion networks (SeisInvNets) with novel components to make the best use of all seismic data are proposed with promising performance on synthesized and proposed SeisInv data set according to various evaluation metrics.
Well-Logging Constrained Seismic Inversion Based on Closed-Loop Convolutional Neural Network
TLDR
Close-loop CNN can not only learn from labeled data but also extract information contained in unlabeled data and has a better performance than both traditional methods and other deep learning-based methods on the synthetic data set and also can be efficiently applied on the real seismic data set.
Semi-supervised Sequence Modeling for Elastic Impedance Inversion
TLDR
A semi-supervised sequence modeling framework based on recurrent neural networks for elastic impedance inversion from multi-angle seismic data, which achieves an average correlation of 98% between the estimated and target EI using 10 well logs for training on a synthetic data set.
Physics-guided deep learning for seismic inversion with hybrid training and uncertainty analysis
The determination of subsurface elastic property models is crucial in quantitative seismic data processing and interpretation. This problem is commonly solved by deterministic physical methods, such
Deep-learning inversion: a next generation seismic velocity-model building method
TLDR
A novel method based on the supervised deep fully convolutional neural network (FCN) for velocity-model building (VMB) directly from raw seismograms is investigated, showing promising performances in comparison with conventional FWI even when the input data are in more realistic scenarios.
Prestack and poststack inversion using a physics-guided convolutional neural network
An inversion algorithm is commonly used to estimate the elastic properties, such as P-wave velocity ([Formula: see text]), S-wave velocity ([Formula: see text]), and density ([Formula: see text]) of
InversionNet: An Efficient and Accurate Data-Driven Full Waveform Inversion
  • Yue Wu, Youzuo Lin
  • Computer Science, Mathematics
    IEEE Transactions on Computational Imaging
  • 2020
TLDR
This work builds a convolutional neural network with an encoder-decoder structure and employs the conditional random field (CRF) on top of the CNN to generate structural predictions by modeling the interactions between different locations on the velocity model.
Deep-learning tomography
TLDR
This work proposes and implements a novel concept that bypasses these demanding steps, directly producing an accurate gridding or layered velocity model from shot gathers, and relies on training deep neural networks.
An overview of full-waveform inversion in exploration geophysics
TLDR
This review attempts to illuminate the state of the art of FWI by building accurate starting models with automatic procedures and/or recording low frequencies, and improving computational efficiency by data-compression techniquestomake3DelasticFWIfeasible.
...
1
2
3
4
...