PSI - Issue 75

Ninian Sing Kok Ho et al. / Procedia Structural Integrity 75 (2025) 35–42 Ninian Sing Kok Ho/ Structural Integrity Procedia (2025)

38 4

3. Results and Discussion Precision grinding improved the S-N curve ( Erreur ! Source du renvoi introuvable. ) compared to the production grinded samples, most likely due to the removal of surface defects originally caused by the support structure removal (Figure 2a, b, and d). This suggests that the support structure surface is a dominant factor in the S N curve performance. The X-ray CT results of the runout samples from the refined grinding batch, while not detecting new cracks formed after 10m cycles, shed insight into this reason. Figure 4a shows one of the flat faces of a precision grinded sample in the gauge region, which shows the resolution of the X-ray CT scan. The top curved surface comprises a smooth region right at the middle of the sample and much rougher regions on both sides of this central region (Figure 4c). This can be attributed to the melting of an almost flat layer in the middle since the last layer there is tangential to the hourglass curve at the narrowest point. Furthermore, in both samples that were precision grinded, large subsurface defects were still detected beneath the curved face previously in contact with the support structures (Figure 4b), with small defects at most on the other side. In the sample that was production grinded, large craters were left behind on the grinded surface, and again there were no detected subsurface pores on the other side of the sample. These explain the poor quality of parts at the faces previously connected to support structures. Milling led to a slightly lower S-N curve than the precision grinded samples. It is recommended that a minimum amount of material be determined for removal based on X-ray CT for batch production to fully clear the surface and any subsurface defects (Figure 4b). For the best performing horizontal samples (210 mm, H, HT, milled), 6 out of 8 of the failed samples did so at the milled surface due to a surface defect (Figure 5a). While it might seem disproportionate, the other 2 that failed away from the milled surface (Figure 5b) fall close to the S-N curve obtained from these 6 samples, suggesting a comparable resistance on both curved sides. The two points (green triangles) were, however, excluded from the curve fitting ( Erreur ! Source du renvoi introuvable. ) due to a low R 2 caused by a different mechanism. Stress relieving the samples before separating them from the substrate plate to minimize distortion and warpage might have improved the S-N performance. However, fractography revealed that surface defects were still the cause of crack initiation in all the failed samples, in which unmelted powder particles could be seen occasionally (Figures 5 a and b). When compared to vertical samples scaled down by about half (105 mm) with fully as-built surfaces also fabricated using OEM parameters, the S-N curves of the horizontal sample batches survive longer between 200 and 400 MPa. The fatigue limit of the vertical samples was within the range obtained from the various batches of horizontal samples. The S-N curves are in agreement with literature data at R=0.1 plotted in greyscale in Figure 3 [8, 9].