About this Abstract |
| Meeting |
2026 TMS Annual Meeting & Exhibition
|
| Symposium
|
Algorithms Development in Materials Science and Engineering
|
| Presentation Title |
Performance and Accuracy Analysis of Peridynamics and Phase‑Field Fracture Models |
| Author(s) |
Iran Shepherd, Sam Reeve |
| On-Site Speaker (Planned) |
Sam Reeve |
| Abstract Scope |
We consider the computational cost and accuracy tradeoffs for two fracture modeling frameworks: particle-based peridynamics and finite element based phase-field. CabanaPD is a peridynamics application built on the Cabana particle library and Kokkos performance portability library. RACCOON is a phase-field for fracture application with a complete fracture model based on a material’s strength built on the MOOSE finite element framework. We compare the performance of CabanaPD and RACCOON across various architectures, focusing on the Frontier supercomputer. The benchmark problem selected was the Kalthoff-Winkler experiment where a pre-notched plate is impacted by a projectile. The result is propagating cracks at each notch at angles verifiable for accuracy. The computational metrics considered included runtime and prediction error for both methods across number of computational points (grid or particles), local computational interactions (grid stencil or particle neighbors), and model forms. This work concludes with recommendations for efficient fracture predictions at the exascale. |
| Proceedings Inclusion? |
Planned: |
| Keywords |
Computational Materials Science & Engineering, Modeling and Simulation, Other |