About this Abstract |
| Meeting |
2025 Annual International Solid Freeform Fabrication Symposium (SFF Symp 2025)
|
| Symposium
|
2025 Annual International Solid Freeform Fabrication Symposium (SFF Symp 2025)
|
| Presentation Title |
Accurate Detection of Local Porosity in Laser Powder Bed Fusion Through Deep Learning of Physics-Based In-Situ Infrared Camera Signatures |
| Author(s) |
Berkay Bostan, Shawn Hinnebusch, David Anderson, Albert To |
| On-Site Speaker (Planned) |
Berkay Bostan |
| Abstract Scope |
Porosity critically affects the reliability of metal laser powder bed fusion (LPBF) parts. This study presents a deep learning framework to predict local porosity in IN718 parts manufactured under standard conditions, with overall porosity below 0.03%. The model uses in-situ infrared (IR) camera signatures and achieves over 90% balanced accuracy for detecting local porosity over 34 μm. Input features include six physics-based IR signatures and local scan vector length, all related to porosity formation. Spatial interactions across each pixel and its 26 neighbors are incorporated. Custom convolutional filters compensate for the IR camera’s limited resolution and frame rate, especially near stripe boundaries and part edges. Ground truth pore data are obtained through serial sectioning and microscopy. The model outperforms traditional machine learning approaches in speed, generalization, and model size. Shapley Additive Explanations provide physical insight into pore formation, enabling accurate in-situ porosity detection and improving understanding of defect mechanisms in LPBF. |
| Proceedings Inclusion? |
Planned: Post-meeting proceedings |