.png){kind=icon}
.png){kind=icon}
Return
Morphological characterization and process optimization of gas-atomized powders based on synchrotron X-ray computed tomography and machine learning
.png){kind=icon}
PRE
AI
S
Y
Z
M
Y
H
DOI：10.1016/j.addma.2026.105227.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Metal powder is a key raw material for additive manufacturing (AM), with its quality directly determining the performance of AM components. Gas atomization (GA) is a dominant powder production method due to its high efficiency and cost-effectiveness, yet it frequently generates defective powders (e.g., hollow, satellite, and irregular particles) that significantly degrade the mechanical and fatigue properties of AM parts. Existing powder quality quantification methods primarily focus on particulate and bulk properties, but overlook the correlation between powder morphology and the GA process—failing to provide effective feedback for process optimization. This study aims to comprehensively evaluate GAed powders via morphological characteristic analysis. Synchrotron X-ray computed tomography (SXCT) and 3D reconstruction are used to characterize powders, which are classified into five types (irregular, satellite, normal, hollow, multi-defect) based on the formation mechanisms of defective powders in the GA process. By using 2,598 powder samples and 9 morphological parameters, the machine learning models are trained. Through principal component and correlation analyses, 4 key defect-distinguishing parameters are identified, and all models achieved high accuracy. To assess GAed powders, the particle defect distribution (PDD) is built for quality evaluation. Guided by these findings, an improved nozzle is designed, manufactured, and used for powder preparation. PDD results indicated about 300% increase in the mass ratio of normal powders with high sphericity.
Keywords:
Gas atomization
Powder morphology
Synchrotron X-ray computed tomography
Machine learning
Process optimization
Journal
IF:
11.1
Papers:
4.5K
Citations:
4.9W
Organization
Cited Papers