| About this Abstract |
| Meeting |
2025 TMS Annual Meeting & Exhibition
|
| Symposium
|
2025 Technical Division Student Poster Contest
|
| Presentation Title |
SPG-64: Foundational Computer Vision Models for Automated Powder Morphology Characterization |
| Author(s) |
Stephen Price, Kiran Judd, Kyle Tsaknopoulos, Elke Rundensteiner, Danielle L Cote |
| On-Site Speaker (Planned) |
Stephen Price |
| Abstract Scope |
Powder properties, such as their morphology (shape, size, surface texture, etc.), are influential in the overall quality of additively manufactured materials. Changing a powder’s morphology can impact the flowability, deposition quality, deposition efficiency, and porosity of manufactured parts. Image analysis can be used to quantify a powder’s morphology, but doing so by hand is quite time-consuming. Computer vision can mitigate this challenge, offering an automated approach to quantifying morphology, but requires extensive data to train a model, and frequently has lower accuracy than manual labeling. As a result, this work presents DualSight, a novel multi-stage framework to refine traditional computer vision models, reducing the gap in accuracy between manual and automated labels without requiring any additional data or model training. This enables more accurate morphological characterizations and a more informed manufacturing process. |
| Proceedings Inclusion? |
Undecided |
| Keywords |
Machine Learning, Characterization, Additive Manufacturing |