PSI - Issue 76

Daniele Rigon et al. / Procedia Structural Integrity 76 (2026) 35–42

38

where 0.728 is the maximum value of α for a semi-circular crack. It is worth citing that all the present evaluations of the initial crack size can be made using the profile parameter R v by applying the correction proposed in Rigon et al. (2024). Once a e ff is known, one can use Eq. (1) to estimate the fatigue threshold ∆ σ g , th of the i-th specimen in the case that the subsurface and internal defects are not relevant in terms of e ff ective size with the as-built surface. The Murakami parameter, √ area , coupled with a shape factor α = 0 . 5 for internal defects and α = 0 . 65 in the case of a surface or subsurface defect, can be used to evaluate the mode I SIF (Murakami (2019)). Thus, a e ff takes the following form: a e ff = α a =   0 . 5 2 √ area for interal defects 0 . 65 2 √ area for surface or subsurface defects (6) 3. Materials and Methods Cylindrical fatigue specimens with a 6-mm-long gauge section and nominal diameter equal to 6 mm were manufac tured by means of PBF-LB / M using a Sisma MySint100 system (Sisma S.p.A., Italy). The specimens were oriented in the built volume with their longitudinal axis parallel to the build direction (i.e., normal to the build platform). After the specimens were removed from the platform, the ends of the specimens were turned to avoid secondary bending e ff ects in the axial fatigue tests. The dimensions of the samples are reported in Figure 1a. The samples were then annealed by ramping up to 950 ◦ C for 120 minutes, holding for 30 minutes, and then cooling in the furnace to room temperature, all under an inert atmosphere. The resulting microstructure is shown in Fig. 1(b). The Vickers microhardness of the materials is 348.5 kg f / mm 2 and the mean value of the α lamellae is 3 µ m . These two values can be used to determine the parameters ∆ σ 0 and a 0 of Eq. (1) using the empirical equation reported in Rigon and Meneghetti (2020, 2022). 3.1. Areal measurements The areal measurements were performed using the experimental setup developed in previous investigations ( Mioli et al. (2024, 2025)), where a comparison was made between Optical Profilometry (OP) and X-ray Computed To mography (CT) measurements (see Fig. 2a). This setup allows to entirely measure the central gauge section of the specimens by performing 16 partially overlapped scans with a size of 1.75x7.92 mm 2 by using an optical profiler Sensofar X Neox. For the k-th scan, a surface filtering procedure consistent with the workflow of ISO 25178-2 Annex G was applied to isolate the roughness component using MountainsMap software. First, the noise was removed using an S -filter (high-pass filter) by retaining only the wavelength components larger than λ s = 2 . 5 µ m. Then, the form of the surface was eliminated by subtracting a third-degree polynomial through an F -operation. Finally, waviness was removed using an L -filter (low-pass filter) by retaining only the wavelength components smaller than λ c = 800 µ m. In order to mitigate edge e ff ects introduced by the use of Gaussian filters during surface processing, a thickness border region λ c / 2 = 0 . 4 mm was removed from the periphery of each scan prior to filtering (see the relevant dimensions in Figure 2b). The entire surface was measured as schematically illustrated in Figure 2c which also reports the size of the net area of each scan (equal to 6.76 mm 2 ) and the size of A ref within which the S mea v , max , i parameter was extracted for each specimen. Subsequently, a control area A 0 = 0.16 mm 2 was established to extract a random selection of S v , j from the 16 scans with a sample size ranging from 3 to 240. For each extraction, the Gumbel parameter λ and δ were determined using the moment method and then S est v , max , i for each random selection was determined from a return period A ref / A 0 = 675, as schematically depicted in Fig. 2d.

3.2. X-Ray Computed Tomography

CT scans were performed using a metrological micro-CT system (MCT225, Nikon Metrology, UK), with an X-ray tube voltage of 170 kV and a filament current of 41 µ A. A total of 1800 projections were acquired, with an exposure time of 1.4 s per projection. A 0.1 mm thick copper filter was used to reduce beam hardening. A voxel resolution of 4 µ m was achieved, allowing to gather the entire gauge length surface at a suitable resolution for other scopes