PCP-GAN: Property-Constrained Pore-scale image reconstruction via conditional Generative Adversarial Networks
arXiv:2510.19465v2 Announce Type: replace-cross Abstract: Obtaining truly representative pore-scale images that match bulk formation properties remains a fundamental challenge in subsurface characterization, as natural spatial heterogeneity causes extracted sub-images to deviate significantly from core-measured values. This challenge is compounded by data scarcity, where physical samples are only available at sparse well locations. This study presents a multi-conditional Generative Adversarial Network (cGAN) framework that generates representative pore-scale images with precisely controlled pr
The increasing maturity of generative AI models, specifically cGANs, is now enabling their application to complex scientific challenges where data scarcity and representativeness are major hurdles.
This development allows for the generation of synthetic, yet highly representative, scientific data in critical resource sectors, potentially accelerating discovery and optimization in a data-constrained environment.
The ability to generate accurate pore-scale images constrained by specific physical properties could reduce reliance on exhaustive physical sampling and improve computational modeling outputs for subsurface characterization.
- · Oil and Gas Exploration
- · Geological Sciences
- · Material Science
- · AI/ML Research Institutions
- · Traditional Manual Data Acquisition Methods
Improved accuracy and efficiency in subsurface resource modeling due to better and more abundant pore-scale data.
Faster and cheaper discovery processes for new oil and gas reserves, or other underground resources like geothermal energy.
Enhanced understanding of subsurface fluid dynamics, leading to breakthroughs in fields like carbon capture and storage or groundwater management.
This signal links to a primary source. Continuum Brief monitors and indexes it as part of the live intelligence stream — we do not republish source content.
Read at arXiv cs.LG