About this Abstract |
| Meeting |
2027 TMS Annual Meeting & Exhibition
|
| Symposium
|
Modeling, AI Applications, and Method Development in Nuclear Materials
|
| Presentation Title |
A Strength-Augmented Phase-Field Fracture Framework Using Arbitrary Data-Driven Strength Surfaces: Development and Application to Nuclear Fuels |
| Author(s) |
K M Zaheen Nasir, Tianchen Hu, Wen Jiang |
| On-Site Speaker (Planned) |
K M Zaheen Nasir |
| Abstract Scope |
Predicting crack nucleation in nuclear fuels remains a central challenge for fuel performance and safety, where, among other environmental factors, material strength plays a significant role. Classical Griffith-based phase-field fracture (PFF) captures crack propagation but ties the nucleation stress to toughness, elastic properties, and the regularization length instead of the actual material strength. In this work, we present a strength-augmented PFF framework that incorporates an arbitrary—including data-driven—strength surface directly into the phase-field evolution while preserving Griffith-type propagation. Strength surfaces are inferred from limited failure data using physics-constrained input-convex neural networks enforcing convexity, symmetry, regularity, and a consistent elastic domain. We apply our approaches to several nuclear fuel cracking problems, such as UO2 fuel cracking and TRISO buffer-layer fracture, demonstrating a path to microstructure and irradiation informed strength descriptions for engineering-scale fuel performance analysis. |
| Proceedings Inclusion? |
Planned: |
| Keywords |
Machine Learning, |