| About this Abstract |
| Meeting |
MS&T24: Materials Science & Technology
|
| Symposium
|
Additive Manufacturing Modeling, Simulation, and Machine Learning: Microstructure, Mechanics, and Process
|
| Presentation Title |
Process Design for Metal Additive Manufacturing Through High-Speed Imaging and Vision Transformers |
| Author(s) |
David Guirguis, Conrad Tucker, Jack L Beuth |
| On-Site Speaker (Planned) |
David Guirguis |
| Abstract Scope |
The full potential of metal additive manufacturing remains untapped due to inherent uncertainties and variability in the printing outcome. While process mapping is critical for optimizing printing parameters, conventional approaches rely on ex-situ characterization and design of experiments. Unfortunately, these methods fall short in capturing real-time dynamics during printing. On the other hand, in situ approaches face limitations due to restricted observable features and the need for complex, high-cost setups to measure temperature accurately or incorporate material-specific features. These limitations hinder the generalizability of machine learning models.
To address these challenges, we propose a deep learning-based method. By incorporating temporal features from molten metal dynamics during laser-metal interactions using video vision transformers and high-speed imaging, we create in situ process maps. These maps directly quantify defects and variability within the printing process. Importantly, our approach demonstrates robustness across different alloy compositions and intrinsic thermofluid properties. |