Fatigue Crack Growth in Additively Manufactured and Advanced Alloys: From Synchrotron Imaging to Physics-Informed Digital Twins

Predicting fatigue crack growth in structural alloys requires linking processes that span enormous differences in scale, from crystallographic slip at the nanoscale to failure of an entire component. Engineering assessments still rely heavily on phenomenological models such as Paris, Walker and NASGRO. These models are practically useful, but their limitations become apparent under multiaxial loading, in microstructures with pronounced local variation, and in harsh environments where several damage mechanisms act at the same time. This paper reviews recent research in three interrelated areas: high-resolution in situ characterization, multiphysics damage mechanics, and physics-informed data-driven modeling. It first discusses recent advances in real-time three-dimensional and four-dimensional synchrotron X-ray microtomography, high-resolution digital image correlation, and electron backscatter diffraction. These techniques allow researchers to examine crack-tip shielding, local plastic deformation, and the interaction between cracks and grain boundaries. The paper then examines the factors that affect crack growth in emerging material systems, including anisotropy, residual stress, and defect-induced cracking in additively manufactured alloys, and hydrogen-assisted, corrosion, and thermomechanical fatigue. Comparison tables and schematic diagrams place competing mechanisms, measurement techniques, and modeling approaches side by side to make them easier to compare. The paper takes the form of a narrative review rather than following a systematic review protocol. Because purely empirical “black-box” models can extrapolate poorly and are difficult to certify, particular attention is given to physics-informed machine learning, including physics-informed neural networks and neural operators such as deep operator networks and Fourier neural operators. These methods build fracture mechanics constraints, such as the stress intensity factor range threshold ΔKth, the path-independent J-integral, and energy balance into the loss function or the model architecture. Finally, the paper discusses how these models can be integrated into real-time structural health monitoring systems. It proposes a research roadmap for fatigue crack growth digital twins and points out the gaps in existing benchmark datasets and the areas where probabilistic mechanics could contribute to the future of alloy manufacturing.