PSI - Issue 53

S. Leonardi et al. / Procedia Structural Integrity 53 (2024) 327–337

335

S. Leonardi et al. / Structural Integrity Procedia 00 (2023) 000–000

9

Fig. 5. (a) Internal porosity computed for all the numerical and the experimental test samples of this work. For the 3D-printed test sample, internal porosity is computed after image post-processing. (b) Results of the FE simulations for the homogenized Young’s modulus E normalized by the corresponding matrix value E m . Computed values of this parameter are reported for all the numerical and the fabricated test samples of this work (see Tables 1 and 4).

is noted that here we choose to compare the internal porosity as the pores on the sample edges are not accurately 3D printed (see Figures 3,4). As seen from Figure 5a, the AlSi10Mg test samples can better reproduce the porosity of their numerical model than the Inconel 625 cellular architectures. Nevertheless, the AlSi10Mg cellular materials exhibit a higher polydispersity in pore size and a lower pore circularity. This observation can be rationalized on account of the base material composition. Notably, the AlSi10Mg powder has higher thermal conductivity than that of the Inconel 625 powder, thereby resulting into larger solidified melt pools. To assess and quantify the impact of the observed topological features on the resulting mechanical properties, we used computational homogenization to estimate the e ff ective mechanical properties (see Section 2.4). Specifically, the homogenized elastic moduli were computed by means of FE simulations carried out onto image-based numerical models. The latter were constructed by means of the open-source Gmsh software, whereby the FE model of every experimental micrographs was built using the topological descriptors (i.e. pore major and minor axis and center coordinates) extracted from the image-analysis program. Here too, the pores at the cell edges are disregarded due to the poor accuracy of their manufacturing. The results of the FE simulations for the elastic Young’s modulus E , normalized by the corresponding matrix modulus E m , are reported in Figure 5b. Specifically, Figure 5b gives the relative value of this parameter computed for all the numerical and experimental test samples investigated in this study (see Tables 1 and 4). As seen, the FE results are consistent with the trend of the internal porosity reported in Figure 5a. Notably, a sti ff er response is computed for the experimental test samples containing a lower degree of geometrical porosity, compare e.g. the values of E / E m for the Al-30 and In-30 fabricated using the same numerical model N-30. It is noted in passing that - in the interest of comparison - the FE models of the defect-free numerical samples N-20 to N-60 contain only the internal pores. Interestingly, data in Figure 5b also indicate that polydispersity in pore size does not a ff ect the resulting mechanical properties. This is in agreement with the findings of prior studies (Hooshmand-Ahoor et al. (2022); Tarantino et al. (2019); Zerhouni et al. (2019, 2018)) and can be seen, e.g., by comparing pairwise sample Al-20 with its corresponding numerical model N-20. The comparison between these two samples is here highlighted as they contain a similar amount of internal porosity (see Figure 5a). Likewise, the same comparison shows that - at similar internal porosity - the small degree of ellipticity introduced by the LPBF manufacturing is found to have no influence on the homogenized elastic modulus. On the other hand, ellipticity of the 3D-printed pore is likely to result into anisotropy of the mechanical properties. The latter was not quantified in this study. Collectively, the results of this work therefore reveal that cellular materials with random pore features are less sensitive to geometrical defects induced by the LPBF process than period cellular lattices (Liu et al. (2017); Dallago et al. (2018)).