Abstract:Macroscopic transport properties of porous media essentially rely on the geometry and topology of their pore space. The premise of predicting these transport properties is to construct an accurate 3D pore space. To date the methods of modeling porous media are divided into two main groups, direct measurements by some equipment and stochastic statistical methods. Direct measurements of pore structure can be acquired with current equipment such as X-ray computed tomography and laser scanning confocal microscopy, but the unavailability of the equipment and the high cost of the measurement make their widespread application impossible. Many stochastic statistical methods, such as truncated Gaussian random field and simulated annealing methods, reconstruct 3D porous media based on some 2D thin sections by means of lower-order statistical functions. However these functions cannot reproduce the long-range connectivity of pore structure. Therefore, this paper will present a stochastic technique of reconstructing 3D pore space using multiple-point statistics with the purpose of solving the proposed problems. The single normal equation simulation algorithm (SNESIM), one of the most common methods for discrete variable simulation in multiple-point statistics, is the main tool to reproduce the long-range feature of pore space. To test the method, Berea sandstone was used as a sample. In the simulation process, a 2D thin section was taken as the training image for providing patterns of pore structure and some pixels were extracted from it as the conditioning data. The models were reconstructed using the SNESIM algorithm that serves as the simulation engine. In order to test the accuracy of these reconstructed models, pore geometry and topology and transport properties of the reconstructed models were compared with those of the real model obtained by X-ray computed tomography scanning. The comparison result shows that the reconstructed models are good agreement with the real model obtained by X-ray computed tomography scanning in the two-point correlation function, the pore space features and single- and two-phase flow permeabilities, which verifies that the long-range connectivity of pore space can be reproduced by this method. Comparing with other stochastic methods, a more accurate stochastic technique of reconstructing 3D porous media is put forward when only some 2D thin sections are available.

Slice-to-voxel stochastic reconstructions on porous media with hybrid deep generative model

3D-Aware Image Synthesis Via Learning Structural and Textural Representations

Super-Resolution Reconstruction of Porous Media Using Concurrent Generative Adversarial Networks and Residual Blocks

Stochastic reconstruction of porous media based on attention mechanisms and multi-stage generative adversarial network

Reconstruction of three-dimensional porous media using generative adversarial neural networks

Stochastic reconstruction of 3D porous media from 2D images using generative adversarial networks

3D stochastic reconstruction of porous media based on attention mechanisms and residual networks

Three-dimensional reconstruction of granular porous media based on deep generative models.

Sliced Wasserstein Distance-Guided Three-Dimensional Porous Media Reconstruction Based on Cycle-Consistent Adversarial Network and Few-Shot Learning

Improved recurrent generative model for reconstructing large-size porous media from two-dimensional images

Fast and accurate reconstruction of large-scale 3D porous media using deep learning

Using Physics Informed Generative Adversarial Networks to Model 3D porous media

Porous Media Reconstruction Using a Cross-Section Image and Multiple-Point Geostatistics

3D Microstructure Reconstruction and Characterization of Porous Materials Using a Cross-Sectional SEM Image and Deep Learning

3D-porous-GAN: a high-performance 3D GAN for digital core reconstruction from a single 3D image

New 2D to 3D Reconstruction of Heterogeneous Porous Media via Deep Generative Adversarial Networks (GANs)

Stochastic reconstruction of geological reservoir models based on a concurrent multi-stage U-Net generative adversarial network

Deep learning method of stochastic reconstruction of three-dimensional digital cores from a two-dimensional image

Reconstruction of 3D Porous Media Using Multiple-Point Statistics Based on a 2D Training Image

Constrained Transformer-Based Porous Media Generation to Spatial Distribution of Rock Properties

Pore-scale modeling of multiphase flow in porous media using a conditional generative adversarial network (cGAN)