SVM-based fast 3D pore-scale rock-typing and permeability upscaling for complex rocks using minkowski functionals

Document Type

Journal Article

Publication Title

Advances in Water Resources

Volume

183

Publisher

Elsevier

School

School of Engineering

RAS ID

64713

Funders

National Natural Science Foundatoin of China / China Postdoctoral Science Foundation / Australian Research Council / State Key Laboratory of Coal Mine Disaster Dynamics and Control in Chongqing University

Grant Number

ARC Numbers : DP160104995, DP200103548

Grant Link

http://purl.org/au-research/grants/arc/DP160104995 http://purl.org/au-research/grants/arc/DP200103548

Comments

Jiang, H., Arns, C., Yuan, Y., & Qin, C. Z. (2024). SVM-based fast 3D pore-scale rock-typing and permeability upscaling for complex rocks using minkowski functionals. Advances in Water Resources, 183, article 104605. https://doi.org/10.1016/j.advwatres.2023.104605

Abstract

The rapid advancement of digital core analysis has greatly promoted the research progress of flow and transport in porous media. However, complex analytical process with exceeding computational load impedes the application on large data volume. Considering the strong heterogeneity of the underground porous media, the integration of pore-scale information into continuum scale is widely concerned for the future development of digital physical analysis. For hierarchical porous structures, pore-scale rock-typing and upscaling of petrophysical properties is a promising solution towards the issue, and morphological and topological descriptors associating data clustering methods are popularly utilized. However, the size of the regional support through which the parameter fields are generated heavily affects the descriptive capacities of the parameters and the following partitioning process. We propose in this work a robust integrated pore-scale rock-typing and upscaling technology for 3D porous structures which uses Minkowski functionals as the descriptive parameters. A fast-computational method utilising fast Fourier transform has been applied for efficient generation of the parameter fields. A comparative study between the two different classification methods of Support Vector Machine (SVM) and Gaussian Mixture Model (GMM) has been conducted on two complex artificial porous systems and a laminated sandstone through various regional support sizes. Throughout the test, SVM has illustrated obvious advantage of overcoming regional support size effect even with limited labelling information. The Upscaling of permeability on the natural sandstone sample based on the rock type distribution has demonstrated excellent accuracy comparing with full scale direct computation.

DOI

10.1016/j.advwatres.2023.104605

Access Rights

subscription content

Link to Full Text

Share

 
COinS