Towards unified AI-driven fracture mechanics: the extended deep energy method (XDEM).

XDEM achieves superior accuracy and efficiency in fracture mechanics by unifying discrete and phase-field models with sparse collocation points.

• Why it matters: Current DEM approaches face stability issues and require dense sampling near cracks, limiting their practical application in complex fracture scenarios.
• What they did: The authors developed XDEM, a deep learning framework that integrates displacement discontinuities, crack-tip asymptotics, and coupling of displacement and phase fields, validated on multiple benchmark problems.
• The result: XDEM outperforms standard DEM in accuracy and efficiency, providing a robust foundation for AI-driven fracture prediction and advancing modeling capabilities in engineering and materials science.