Speaker
Description
The imaging and quantitative analysis of defects such as porosity and cracks are essential steps in optimizing the laser-powder bed fusion (L-PBF) process parameters for a given material. Most studies to date rely on optical microscopy, which provides only two-dimensional surface cross-sections. X-ray computed tomography (XCT) offers added advantages, including three-dimensional information and access to multiple cross-sectional views. However, XCT imaging of pores and cracks remains underexplored, particularly during parameter optimization for the L-PBF processing of NiTi alloys. XCT images require segmentation in order to quantify features in 3D, but this task is often complicated by limited resolution and image artifacts. Here, we develop and apply a four-class segmentation model based on a 2.5D U-Net neural network architecture to classify and extract cracks and pore defects. This approach reveals the true 3D morphology of the defects and accurately distinguishes cracks from pores, enabling their quantification. We perform a quantitative analysis to investigate how remelting of in situ alloyed NiTi and variations in laser energy density affect defect formation across a series of 52 NiTi samples fabricated by L-PBF. Furthermore, the obtained XCT cross-sections and corresponding segmentations illustrate the influence of scan vector rotation between layers compared to a no-rotation scanning strategy, offering an internal view of cracks, pores, lack of fusion, and balling effects. Our findings indicate that samples produced with a 45° rotation of scan vectors are prone to balling and delamination, whereas samples without scan rotation exhibit high levels of cracking. Finally, compositional analysis using scanning electron microscope and energy dispersive spectroscopy suggests that a substantial amount (99.42%) of the detected cracks are located at the contours, with mean penetration lengths of 0.18 mm for no rotation and 0.11 mm for scan vector rotation.
| Apply for student award at which level: | PhD |
|---|---|
| Consent on use of personal information: Abstract Submission | Yes, I ACCEPT |