Speaker
Description
In this presentation, we discuss an ongoing effort to use machine learning for the design of porous architectures that regulate liquid-metal behavior in fusion reactor first walls. Liquid metals are attractive plasma-facing materials because they can continuously renew the surface and mitigate irradiation damage. However, controlling liquid exposure to the plasma while limiting evaporation and maintaining thermal stability remains a major challenge. We explore the use of architected porous media as a geometric control layer that governs liquid-metal retention, exposure, and transport under high-temperature conditions. Our focus is on the development of machine learning-based surrogate and inverse-design models that capture the relationship between pore-scale geometry, connectivity, and surface topology and coupled heat and mass transfer processes relevant to liquid-metal systems. These models are designed to operate in regimes where strong thermal gradients, phase change tendencies, and radiative effects are expected to play an important role. The presentation will outline the proposed design framework, including geometry parameterization, training strategies for data-driven models, and performance metrics used to evaluate candidate porous architectures. We will also discuss how this approach enables systematic exploration of complex design spaces that are difficult to access using purely physics-based optimization. This work aims to establish a foundation for data-driven porous media design tailored to extreme energy environments and to highlight the role of machine learning in guiding the development of future liquid-based plasma-facing components for fusion systems.
| Country | United States |
|---|---|
| Green Housing & Porous Media Focused Abstracts | This abstract is related to Green Housing |
| Acceptance of the Terms & Conditions | Click here to agree |
