The imaging and quantitative analysis of defects such as porosity and cracks are crucial steps in optimizing the fabrication parameters of powder bed fusion using laser beam (PBF‐LB) for a particular material type. For this purpose, most studies resort to optical microscopes, which allow analysis of 2‐dimensional surface cross sections. X‐ray computed tomography (XCT) is an alternative method with added benefits such as 3‐dimensional information and multi–cross‐sectional data. However, XCT imaging of pores and cracks has not yet been explored in detail particularly during parameter optimization of the PBF‐LB process of NiTi alloy. XCT images require segmentation in order to quantify different features in 3D, but this is often complex due to limited resolution and image artifacts. Here, we develop and employ a four‐class segmentation model based on a 2.5D U‐Net to classify and extract cracking and pore defects. This reveals the true 3D nature of the defects and distinguishes cracks from pores accurately, allowing their quantification. Quantitative analysis is carried out to investigate the effects of rescanning of in situ alloyed NiTi and varying energy densities on the extent of the produced defects for a series of 52 NiTi samples fabricated in the PBF‐LB process. Moreover, the obtained XCT cross sections and corresponding segmentations highlight the effect of rotation of scan vectors between layers and no‐rotation scanning strategies, providing an inside view of the cracks, pores, lack of fusions and balling effect. The findings revealed that samples fabricated using 45° rotations of scan vectors are susceptible to balling and delaminations, whereas samples with no rotations of scan vectors show high levels of cracking. Additionally, compositional analysis with SEM and EDS is used to surmise how a considerable amount (99.42%) of discovered cracks are located at the contours with mean penetration length of 0.18 mm for no rotation and 0.11 mm for rotation of scan vectors.
Ledwaba et al. (Thu,) studied this question.