About this Abstract |
Meeting |
2023 TMS Annual Meeting & Exhibition
|
Symposium
|
Additive Manufacturing of Metals: Applications of Solidification Fundamentals
|
Presentation Title |
Automatic Melt Pool Segmentation and Tracking in the X-ray Image Sequence |
Author(s) |
Maede Maftouni, Bo Shen, Andrew Law, Rongxuan Wang, Zhenyu Kong |
On-Site Speaker (Planned) |
Bo Shen |
Abstract Scope |
The melt pool is the region of the superheated melted metal generated by laser irradiation on the powder bed surface, whose shape and evolution monitoring plays a critical role in unveiling the manufactured part's microstructural characteristics. Yet, melt pool boundary detection from in-situ X-ray images is challenged by the high noise level, brightness fluctuations, and lack of a sharp image boundary between the melted and unmelted hot metal regions. Here, we present a novel and robust-to-noise computer vision model that automates the melt pool segmentation and tracking in the X-ray image sequences to facilitate the melting process inspection and laser parameter calibration. This work is the first to use the video object segmentation deep learning methodology for melt pool segmentation, with an improved melt pool detection performance over the baseline edge detection techniques. |
Proceedings Inclusion? |
Planned: |
Keywords |
Additive Manufacturing, Machine Learning, |