The prediction of material failure, encompassing fatigue, fracture, and progressive damage, remains a cornerstone of engineering design and structural integrity assessment. While traditional physics-based models have laid a solid foundation, they often struggle with complexity, multi-scale phenomena, and material uncertainties. The emergence of machine learning (ML) offers a powerful paradigm shift.

However, purely data-driven approaches can lack physical consistency and generalizability.
This session aims to spotlight the latest advancements in integrating mechanistic understanding with ML techniques to create robust, predictive models for failure analysis. We encourage submissions that leverage ML not as a black box, but as a tool to discover, enhance, or accelerate physics-based models. 

Topics of interest include, but are not limited to:

• Physics-informed neural networks (PINNs) for predicting crack propagation and fatigue life.
• ML-enhanced constitutive models for damage and plasticity.
• Multi-scale modeling bridged by machine learning.
• Discovery of failure-related governing equations from experimental or simulation data.
• Fusion of heterogeneous data (e.g., from digital image correlation, acoustic emission, microscopy) for damage state diagnosis and prognosis.
• Uncertainty quantification in ML-predicted failure modes.
   

We seek contributions that demonstrate a synergistic coupling between data-driven methods and the underlying physics of material deformation and failure, ultimately leading to more reliable and trustworthy engineering solutions against failure.