Adnan KhanAdnanProduct & AI
San Francisco Bay Area
Get in touch

Source note3 min read

Physics-Guided Generative AI for Property-Targeted 3D Porous Media Design

Source

arXiv

arxiv.org/abs/2607.24274v1

Read the source

Source

Plain-Language Brief

The paper proposes a new framework using physics-guided generative AI to design three-dimensional porous media. The goal is to improve applications in various fields such as filtration, energy storage, and biomedical scaffolds by creating structures that meet specific property requirements. The framework combines several models to learn from a compact, physically informative latent design space, generate structures based on target porosity and directional permeability, and refine the generated structures using property-level feedback.

What the Document Says

  • The paper discusses the inverse design of three-dimensional porous media for applications in various fields.
  • The challenge lies in finding distinct pore geometries with similar porosity or permeability, and how small structural changes can significantly impact transport behavior.
  • The paper proposes a physics-guided generative AI framework for property-targeted porous media design.
  • The framework includes a property-aware variational autoencoder, a conditional latent diffusion model, and an independently trained differentiable structure-to-property surrogate.
  • The goal is to learn a compact, physically informative latent design space.
  • The framework generates porous structures conditioned on target porosity and directional permeability.
  • The generated samples are refined using property-level feedback during denoising and decoding.
  • Experiments show improved target-property matching, directional permeability control, and property correlation compared to baselines.

Key Concepts Mentioned

  • Inverse design
  • Three-dimensional porous media
  • Filtration
  • Catalysis
  • Energy storage
  • Fuel cells
  • Thermal management
  • Biomedical scaffolds
  • Physics-guided generative AI
  • Property-aware variational autoencoder
  • Conditional latent diffusion model
  • Differentiable structure-to-property surrogate
  • Target porosity
  • Directional permeability

Claims Made by the Document

  • The proposed framework improves target-property matching, directional permeability control, and property correlation compared to baselines.

Evidence Used

  • Experiments on procedurally generated structures and real micro-CT porous-media datasets show improved results.

Methods / Approach

  • Physics-guided generative AI framework
  • Property-aware variational autoencoder
  • Conditional latent diffusion model
  • Differentiable structure-to-property surrogate

Limitations Stated by the Authors

  • UNKNOWN (Not stated in source)

Limitations Noted by the Agent

Agent interpretation — not from the source.

  • The study does not provide real-world application results, only experiments on procedurally generated structures and micro-CT datasets.
  • It is unknown how the proposed method will perform with more complex and diverse porous media structures.

Open Questions

  • How well will this approach work in real-world applications?
  • Can the framework handle extremely complex and diverse porous media structures?

Notes for Future Q&A

May include interpretation. Labeled where it does.

  • The paper focuses on improving the design process, but does not provide specific examples of real-world applications.

Raw Abstract / Excerpt

Inverse design of three-dimensional porous media is central to applications in filtration, catalysis, energy storage, fuel cells, thermal management, and biomedical scaffolds, but remains challenging because many distinct pore geometries can share similar porosity or permeability while small structural changes can strongly affect transport behaviour. This paper proposes a physics-guided generative AI framework for property-targeted porous media design, combining a property-aware variational autoencoder, a conditional latent diffusion model, and an independently trained differentiable structure-to-property surrogate. The framework learns a compact, physically informative latent design space, generates porous structures conditioned on target porosity and directional permeability, and refines generated samples using property-level feedback during denoising and decoding. Experiments on procedurally generated structures and real micro-CT porous-media datasets show improved target-property matching, directional permeability control, and property correlation compared with representative property-aware variational-autoencoder and latent-diffusion baselines. The results demonstrate a scalable route towards controllable inverse design of complex porous geometries and establish a foundation for simulation-informed generative AI tools in engineering and advanced materials discovery.