Reconstructing porous media using generative flow networks

Journal Article · 07 August 2021 · Computers and Geosciences

DOI: https://doi.org/10.1016/j.cageo.2021.104905 · OSTI ID:1843406

Guan, Kelly M.  ^[1]; Anderson, Timothy I.  ^[2]; Creux, Patrice  ^[3]; Kovscek, Anthony R.  ^[2]

---

1. Stanford Univ., CA (United States); Stanford University
2. Stanford Univ., CA (United States)
3. Universite de Pau et des Pays de l’Adour, Pau (France)

One area of intense scientific interest for the study of sandstones, carbonates, and shale at the pore scale is the use of limited image and petrophysical data to generate multiple realizations of a rock’s pore structure. Such images aid efforts to quantify uncertainty in petrophysical properties, including porosity–permeability transforms. We develop and evaluate a deep learning-based method to synthesize porous media volumes using a so-called generative flow model trained on x-ray microscope images of rock texture and pore structure. These models are optimized on a log-likelihood objective and they synthesize large and realistic three-dimensional images. Further, we demonstrate the rapid generation of sandstone image volumes that display realism as gauged by quantitative comparison of topological features using Minkowski functionals of porosity, specific surface area, and the Euler–Poincaré characteristic (i.e., pore connectivity). We also evaluate the single-phase permeability using Navier–Stokes and lattice Boltzmann methods and show that transport properties of the generated samples match measured trends.

• Research Organization: Stanford Univ., CA (United States)
• Sponsoring Organization: National Science Foundation (NSF); USDOE Office of Science (SC), Basic Energy Sciences (BES)
• Grant/Contract Number: SC0019165
• OSTI ID: 1843406
• Journal Information: Computers and Geosciences, Journal Name: Computers and Geosciences Vol. 156; ISSN 0098-3004
• Publisher: ElsevierCopyright Statement
• Country of Publication: United States
• Language: English