Enter Note Done
Go to previous page in this tab
Session
  • Presentation | H52C: AI Advances in Subsurface Hydrology and Energy II Oral
  • Oral
  • Bookmark Icon
  • H52C-04: Discrete Spatial Diffusion Framework and Assessment Metrics for Rock Image Generation
  • Schedule
    Notes
  • Location Icon243-244
    NOLA CC
    Set Timezone
  •  
    View Map

Generic 'disconnected' Message
Author(s):
Qianqian Zhou, University of Texas at Austin (First Author, Presenting Author)
Javier E. Santos, Los Alamos National Laboratory
Agnese Marcato, Los Alamos National Laboratory
Masa Prodanovic, University of Texas at Austin
Michael Pyrcz, University of Texas at Austin


Understanding how fluids move through underground rocks is important for fields like energy production, water storage, and environmental cleanup. However, getting detailed images of these rocks is difficult, expensive, and sometimes impossible. To solve this, we created a new computer-based method to generate realistic rock images that look and behave like the real thing. Our approach uses a step-by-step process that carefully mimics how natural rock structures form, while also keeping important physical features like how much empty space (porosity) they contain. We tested the method by comparing the fake images to real ones and found that they match well in both shape and internal structure. This means we can now create large sets of realistic rock images to help scientists study underground systems more effectively and at a lower cost.



Scientific Discipline
Suggested Itineraries
Neighborhood
Type
Main Session
Discussion