PSI - Issue 76

Luca Esposito et al. / Procedia Structural Integrity 76 (2026) 50–58

55

Fig. 1. Fatigue behavior of AlSi10Mg specimens in the Z-direction under R = 0 . 1: (a) Comparison between the ASTM E739 method and the proposed single-mode log-Weibull model for Z-as-built samples; (b) Comparison between the ASTM E739 method and the proposed single-mode log-Weibull model for Z-machined samples. Arrows indicate runout specimens.

points marked with arrows indicate runouts. The single-mode log-Weibull formulation shows improved agreement with the experimental trend, particularly in the transition region between high and very high cycle fatigue. Unlike the ASTM E739 linear regression, which tends to underestimate the endurance limit, the proposed model exhibits an asymptotic behavior consistent with the experimental runouts beyond 10 6 cycles. For Z-as-built specimens, the fatigue data exhibited relatively low scatter, and the unimodal Weibull approach provided a satisfactory fit to the observed trend. Supporting this result, fractographic analyses revealed that all failures in the as-built condition were initiated at the rough surface rather than from internal defects.

Table 1. Single-mode log-Weibull model parameters Parameters

Z-as-built

Z-machined

S e (MPa) β (Cycles)

82.2 4328 22.33

164

3188 4.22

γ

In contrast to the as-built condition, the Z-machined specimens exhibited a markedly higher scatter in fatigue life, as illustrated in Figure 1b). Although the median curve of the log-Weibull model closely follows the experimental data, the expected fatigue life at a 95% survival probability is, in many stress regimes, shorter for the Z-machined specimens than for the as-built ones. The parameters estimated for the single-mode log-Weibull model are summarized in Table 1. Notably, the Z-machined specimens exhibit a higher fatigue limit ( S e ) but a significantly lower Weibull modulus ( γ ), which reflects the greater scatter in fatigue life observed for this condition. The wide dispersion of the experimental data, particularly in the high-cycle regime, could not be adequately captured by either the ASTM E739 regression or the single-mode log-Weibull model. The poor agreement between both models and the observed data, especially in terms of survival bands, suggests the presence of multiple failure mechanisms. This observation motivated the adoption of a bimodal log-Weibull formulation, aimed at accounting for the coexistence of two distinct defect-driven fatigue behaviors—one associated with porosity-type features and the other with lack-of-fusion defects. Figure 2 compares the fatigue life predictions obtained for the Z-machined dataset using both the single-mode and the bimodal log-Weibull models. As shown in the left panel, the bimodal formulation automatically identifies two distinct median S – N curves, which separately capture the fatigue behavior of specimens failing at shorter and longer lifetimes. This separation aligns with the underlying assumption that distinct defect populations (e.g., lack-of-fusion versus porosity)