About this Abstract |
Meeting |
2020 TMS Annual Meeting & Exhibition
|
Symposium
|
Additive Manufacturing: Advanced Characterization with Synchrotron, Neutron, and In Situ Laboratory-scale Techniques
|
Presentation Title |
Unsupervised Learning Applied to Powder Metals for Additive Manufacturing |
Author(s) |
Ryan Cohn, Andrew Kitahara, Srujana Rao Yarasi, Elizabeth Holm |
On-Site Speaker (Planned) |
Ryan Cohn |
Abstract Scope |
Current methods of characterizing powders used for additive manufacturing do not provide a complete understanding of powder flow (particle size distribution, hall flow) or are expensive and time consuming (rheology measurements.) In an attempt to provide more reliable high throughput measurements of powder characteristics a computer vision pipeline is proposed. Powder images are segmented so each particle may be analyzed individually. A pre-trained convolutional neural network is used to extract quantitative visual information for each particle. Unsupervised learning is used to cluster the powder particles into distinct groups. The distribution of particles in each group then provides a fingerprint which can be compared between different powders, including powder that is recycled during the additive manufacturing process. |
Proceedings Inclusion? |
Planned: Supplemental Proceedings volume |