PSI - Issue 77

Bastian Roidl et al. / Procedia Structural Integrity 77 (2026) 119–126 Author name / Structural Integrity Procedia 00 (2025) 000 – 000

125

7

designs: at R =0.1, their fatigue performance ranks among the worst, while at R =-1, they rank among the best. This contrasting response is reflected in the mean-stress sensitivity plot in Fig. 3. Even when A- and G-type specimens swap positions at different lifetimes, the H-type remains the highest, resulting in the highest sensitivity to mean stress. This behavior might be seen as surprising, because the specimens of types G and H were intentionally designed to have the same perimeter at the critical cross-section (the outer perimeter of H-type specimens was turned to decrease its importance compared to the inner as-built surface). What differs is their total cross-sectional area, which is 3 times larger in H-type specimens than in G-type specimens. By blindly applying the size effect as commonly understood in high-cycle fatigue, one would expect the H-type specimens to exhibit fatigue strengths below those of G-type specimens. However, this is not observed in the present case. Freudenthal already noted this paradox in 1946 [22]. For the crack initiation phase, larger volumes are associated with a higher probability of failure, leading to a decrease in fatigue strength with increasing volume. However, when analyzing crack growth, the specimens with larger diameters exhibit slower crack growth, because the induced bending stress due to asymmetry caused by the growing crack in its root is smaller. This likely explains the S-N curves for the R =-1 case in Fig. 2. G-type specimens have small cross-sections, and any crack initiated likely has an immediate effect on the stress equilibrium at the affected cross-section. The induced bending stress is high due to the short arm over which the bending moment acts. In contrast, a crack, initiated in an H-type specimen, propagates into regions of larger radius, and even if it penetrates the wall, the bending modulus of the remaining cross section remains much higher than in G-type specimens. Consequently, crack growth in H specimens is slower than in G-specimens. A question remains why this effect is not observed at R =0.1. This probably relates to the maximum stress being 2.222 x the amplitude. Even at high-cycle fatigue, the induced nominal stress is high, and the local stress at the induced crack root is significantly elevated. Consequently, once a crack is initiated, the crack growth phase is very short for both G- and H-type cases, as their diameter or wall thickness, respectively, is too small compared to the A-type specimens. All these proposals share a common expectation: although the specimens are printed as unnotched, they behave as if they were notched or even pre-cracked. Previous papers by authors (e.g., [14]) support that assumption by reporting steep slopes in the S-N curves, typically around k =5, a value characteristic of notched components. As a consequence, the mean stress effect of AM-parts evaluated here is not the only factor in estimating the fatigue strength; it also depends strongly on specimen geometry. The mean stress correction methods benchmarked in this study aim to interpolate fatigue life across different load ratios. While some models (e.g., Dietmann and Half-Slope) perform better on average, their accuracy varies by geometry. This underscores the challenge of capturing geometry dependent mean stress effects at AM-parts with a single formula and highlights the need for careful evaluation of the influence of geometry in fatigue prediction models. 4. Conclusion The present study evaluates the mean stress effect on the fatigue behavior of Laser Powder Bed Fusion manufactured AlSi10Mg specimens of various designs. Key conclusions include: • The response of AM-parts resembles that of notched specimens. • Likely due to frequent pores or lack of fusion, acting as stress raisers, the crack growth phase is of prime importance. In its evaluation, geometry (distance from the crack initiation spot to the opposite wall) plays a decisive role. • The size effect commonly applied in fatigue analysis – where larger specimens are expected to exhibit lower fatigue strength – may not necessarily be the appropriate approach in these cases. • For the same reason, no common mean stress effect model covers all experimental variants described, as these models were defined rather for cases of unnotched components and a significantly smaller proportion of the crack growth phase. This work improves the understanding of the influence of geometry and mean stress on the fatigue properties of additively manufactured components and helps manufacturers to develop safer and more durable additively manufactured components.