3D Carbonate Digital Rock Reconstruction Using Progressive Growing GAN

You, Nan; Li, Yunyue Elita; Cheng, Arthur

doi:10.1029/2021jb021687

Cited by 31 publications

(15 citation statements)

References 59 publications

(104 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…If the data are continuous, such as the heterogeneous field of geological parameters, it is certain that the first two moments can be conditioned on simultaneously since they are uncorrelated. Furthermore, we only consider the reconstruction of binary structures in this work, which are certain to be less realistic than gray-scale reconstructions, such as those in You et al (2021). However, the digital rock reconstruction in this study is mainly developed for the subsequent multiphase flow modeling and uncertainty analysis, which only requires binary structures.…”

Section: Discussionmentioning

confidence: 99%

“…On top of GANs, Shams et al (2020) integrated it with auto-encoder networks to produce sandstone samples with multiscale pores, enabling GANs to predict inter-grain pores while auto-encoder networks provide GANs with intragrain pores. Some other representative applications also exist, such as adopting GANs to reconstruct shale digital cores (Zha et al 2020), utilizing GANs to augment resolution and recover the texture of micro-CT images of rocks (Wang et al 2019(Wang et al , 2020, and reconstructing three-dimension structures from two-dimension slices with GANs (Feng et al 2020;Kench and Cooper 2021;You et al 2021).…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Digital Rock Reconstruction with User-Defined Properties Using Conditional Generative Adversarial Networks

Zheng

Zhang

2022

Transp Porous Med

View full text Add to dashboard Cite

Uncertainty is ubiquitous with multiphase flow in subsurface rocks due to their inherent heterogeneity and lack of in-situ measurements. To complete uncertainty analysis in a multi-scale manner, it is a prerequisite to provide sufficient rock samples. Even though the advent of digital rock technology offers opportunities to reproduce rocks, it still cannot be utilized to provide massive samples due to its high cost, thus leading to the development of diversified mathematical methods. Among them, two-point statistics (TPS) and multi-point statistics (MPS) are commonly utilized, which feature incorporating low-order and high-order statistical information, respectively. Recently, generative adversarial networks (GANs) are becoming increasingly popular since they can reproduce training images with excellent visual and consequent geologic realism. However, standard GANs can only incorporate information from data, while leaving no interface for user-defined properties, and thus may limit the representativeness of reconstructed samples. In this study, we propose conditional GANs for digital rock reconstruction, aiming to reproduce samples not only similar to the real training data, but also satisfying user-specified properties. In fact, the proposed framework can realize the targets of MPS and TPS simultaneously by incorporating high-order information directly from rock images with the GANs scheme, while preserving low-order counterparts through conditioning. We conduct three reconstruction experiments, and the results demonstrate that rock type, rock porosity, and correlation length can be successfully conditioned to affect the reconstructed rock images. The randomly reconstructed samples with specified rock type, porosity and correlation length will contribute to the subsequent research on pore-scale multiphase flow and uncertainty quantification.

show abstract

Section: Discussionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Digital Rock Reconstruction with User-Defined Properties Using Conditional Generative Adversarial Networks

Zheng

Zhang

2022

Transp Porous Med

View full text Add to dashboard Cite

show abstract

“…You et al. (2021) developed a progressive growing generative adversarial network (GAN) to increase the vertical resolution by combination with the technique of GAN inversion. The deep learning methods are powerful for resolution enhancement of micro‐CT images and have high perceptive accuracy compared to traditional interpolation algorithms (e.g., nearest neighborhood and bicubic interpolation).…”

Section: Introductionmentioning

confidence: 99%

“…To circumvent this limitation, Niu et al (2020) proposed a cycle-in-cycle generative adversarial network (GAN) to deal with the unpaired training data for boosting lateral resolution of micro-CT images. You et al (2021) developed a progressive growing generative adversarial network (GAN) to increase the vertical resolution by combination with the technique of GAN inversion. The deep learning methods are powerful for resolution enhancement of micro-CT images and have high perceptive accuracy compared to traditional interpolation algorithms (e.g., nearest neighborhood and bicubic interpolation).…”

mentioning

confidence: 99%

Multiscale Fusion of Digital Rock Images Based on Deep Generative Adversarial Networks

Liu

Mukerji

2022

Geophysical Research Letters

View full text Add to dashboard Cite

Computation of petrophysical properties on digital rock images is becoming important in geoscience. However, it is usually complicated for natural heterogeneous porous media due to the presence of multiscale pore structures. To capture the heterogeneity of rocks, we develop a method based on deep generative adversarial networks to assimilate multiscale imaging data for the generation of synthetic high‐resolution digital rocks having a large field of view. The reconstructed images not only honor the geometric structures of 3‐D micro‐CT images but also recover fine details existing at the scale of 2‐D scanning electron microscopy images. Furthermore, the consistency between the real and synthetically generated images in terms of porosity, specific perimeter, two‐point correlation and effective permeability reveals the validity of our proposed method. It provides an effective way to fuse multiscale digital rock images for better characterization of heterogeneous porous media and better prediction of pore‐scale flow and petrophysical properties.

show abstract

“…Machine learning (ML) techniques, such as deep neural networks (DNN), can be applied to solve geophysical problems. For instance, the generative adversarial network has been used for digital rock reconstruction (You et al., 2021) and seismic waveform synthesis (N. Wang et al., 2021; T. Wang et al., 2021). Qadrouh et al.…”

Section: Introductionmentioning

confidence: 99%

Data‐Driven Design of Wave‐Propagation Models for Shale‐Oil Reservoirs Based on Machine Learning

Xiong

Gei

et al. 2021

JGR Solid Earth

View full text Add to dashboard Cite

The exploration and exploitation of shale oil is an important aspect in the oil industry. Seismic properties and well‐log data are essential to establish wave‐propagation models. Specifically, the description of wave dispersion and attenuation under complex geological conditions needs proper lithological and petrophysical information. This complex physical mechanism has to be considered if a traditional modeling approach is adopted. In this sense, machine learning (ML) techniques provide new possibilities for this purpose. We compare two deep‐neural‐network (DNN)‐based wave propagation models. In the first (pure data‐driven), a DNN is trained to connect seismic attributes, such as wave velocities, to multivariate functions of rock‐physics properties. By training DNNs with different initial parameters, the uncertainty of the proposed method can be quantified. The second method assumes the form of the wave equations. Then, the elastic constants of the constitutive relations are predicted by DNNs. The resulting dynamical equations describe the dispersion and attenuation and wavefield simulations can be performed to obtain more information. On the basis of a test, the two kinds of wave‐propagation models yield acceptable estimations of the seismic properties, with the second approach showing a broader application because the DNN is trained without S wave data. The methodologies illustrate that the new wave‐propagation model based on ML has high precision and can be general in terms of rheological description.

show abstract

3D Carbonate Digital Rock Reconstruction Using Progressive Growing GAN

Cited by 31 publications

References 59 publications

Digital Rock Reconstruction with User-Defined Properties Using Conditional Generative Adversarial Networks

Digital Rock Reconstruction with User-Defined Properties Using Conditional Generative Adversarial Networks

Multiscale Fusion of Digital Rock Images Based on Deep Generative Adversarial Networks

Data‐Driven Design of Wave‐Propagation Models for Shale‐Oil Reservoirs Based on Machine Learning

Contact Info

Product

Resources

About