PSI - Issue 83

Ahmad Issmail et al. / Procedia Structural Integrity 83 (2026) 229–238

234

(Table 3). While PBF-LB thermal history-dependent micro- and mesostructural differences have been eliminated through the T6 HT, the observed difference in fatigue performance is mainly attributed to the decrease in surface roughness with specimen size (Fig. 2 and Table 1). It is to be noted that the high surface roughness of m|H does not contradict this conclusion, as its measurements are strongly affected by the support remnants and adhered unfused powder, which are not load-bearing and therefore do not contribute to stress concentrations. In addition, although deeper valleys are observed in the profile height distribution of m|H, these are counterbalanced by their blunter nature, as discussed in [16]. Fig. 4(b) shows that the dominant role of surface roughness in the observed fatigue performance also extends to the effect of build orientation for standard specimens. In fact, consistently with its lower R a and R v values, S|V yielded overall longer fatigue lives than S|45° at all stress levels. This is not the case, though, for miniature specimens where m|H exhibited a comparable fatigue behavior to m|45° despite its significantly higher surface roughness. As already mentioned, this can be attributed to the lower sharpness of the profile height distribution of m|H, as well as to the benign nature of the support remnants and unfused particles that contributed to the high values of the surface roughness parameters registered by m|H. It is to be acknowledged that, in this work, it is not easy to draw clear conclusions on the fatigue superiority between different miniature build orientations, considering the limited amount of data per size orientation category, but especially due to the high surface-to-volume ratio, which results in a higher deviation of the measured effective stress (corrected with the actual load-bearing area measured from the fracture surface) from the nominal experimental stress, leading to a higher difference between experimental stress levels among different build orientations, and consequently limiting the comparability of lives at similar stress values. In a previous work on the same specimen geometries subjected to SR HT, m|H exhibited the joint best fatigue performance among six specimen size-orientation combinations [16]. The fact that a similar observation does not clearly apply to the specimens subjected to the T6 HT highlights that one of the main reasons for the previously observed superior fatigue performance of m|H in [16] was related to microstructural and mesostructural aspects affecting fatigue crack initiation and propagation. In PBF–LB AlSi10Mg, it is reported that fatigue cracks tend to initiate from SiC particles located at the boundaries of coarse grains elongated along the build direction [24], while fatigue crack growth occurs along the MPBs and layer interfaces [8, 25–27]. As such, the orientation of the microstructure and that of the MPBs (and the layer interfaces) potentially influence the overall fatigue life. Specifically, perpendicular microstructural orientation to the load may delay crack initiation, while an unfavorable orientation of the MPBs to the preferential crack growth direction (i.e., normal MPBs to the plane of the maximum principal stress) forces the crack propagation onto a more tortuous path, thereby increasing fatigue life. The effect of the performed T6 HT is to neutralize the MPB orientation variable and isolate the effect of roughness in the observed fatigue differences. As in the studied as-built specimens (both SR and T6 HT), microstructural influence on fatigue crack initiation can be neglected as cracks consistently initiated from notches of the surface roughness, the obtained results highlight that the fatigue crack growth stage played an important role in determining fatigue life when solidification imprints were present in the specimens. The same conclusion can be drawn from a comparison between S|V and S|H at the high stress levels (from approximately 150MPa), at which the fatigue crack growth is more likely to play a more influential role than at lower stresses. In the SR case [16], S|45° displayed an overall worse fatigue performance, but its fatigue life matched or slightly exceeded that of S|V in the high stress region, likely owing to the misalignment between the MPB and the plane of the maximum principal stress (i.e., the plane orthogonal to the load). In the current case, T6 HT erases the MPBs from the microstructure, leading S|V to consistently outperform S|45° even under high loads. Further, even though the number of analyzed T6 HT specimens in this study is lower than that of the SR counterparts, the elimination of the MPB-dependent fatigue crack growth behavior resulted in a smaller scatter in fatigue life across different miniature categories after T6 HT (T σ = 1.16), compared to the SR case (T σ = 1.27) investigated in [16]. Considering this microstructural homogenization after T6 HT, the fatigue results of PBF-LB AlSi10Mg with as-built surface and T6 HT condition were statistically analyzed together, regardless of size or build orientation, as an attempt to obtain a collective S–N fatigue curve (Fig. 4(d)). It is to be noted that, as the adopted MLE approach requires the specimen gauge length as an input parameter, the unified MLE S–N curves in Figs. 4(a) and 4(d) that collectively represent both miniature and standard specimens are determined assuming that all specimens have an average gauge length of 9mm. This is a simplification that leads to an error in the fatigue limit, which,