Transferability of anomaly data to fatigue properties of PBF-LB AlSi10Mg parts with different volumes

Fatigue of metallic materials produced by Laser Powder Bed Fusion has been extensively studied for both standard testing specimens and application relevant components. However, fatigue properties of specimens and components suffer from large scatter, mainly due to the presence of volumetric anomalies, residual stresses and surface roughness. Therefore, the transferability from specimens to components is still an open point that must be addressed. In this work, fatigue properties of notched components are investigated and compared with the standard specimens. Then, the properties of the components are inferred from the standard specimens’ data using simple models. The maxima anomalies’ distributions are estimated with a competing risk approach using Machine learning-assisted Extreme Value Statistics. Finally, the S-N curves of the investigated components, predicted employing the Shiozawa model for finite life and the El-Haddad model for the fatigue limit, closely matched the experimental results.