| About this Abstract |
| Meeting |
2025 TMS Annual Meeting & Exhibition
|
| Symposium
|
Characterization of Minerals, Metals and Materials 2025: In-Situ Characterization Techniques
|
| Presentation Title |
In-Situ Sensor Monitoring and Multiclass Porosity Defects Prediction for Laser Powder Bed Fusion With Machine Learning |
| Author(s) |
Sandesh Giri, Sen Liu, Nick Calta, Christopher Tassone |
| On-Site Speaker (Planned) |
Sandesh Giri |
| Abstract Scope |
Laser Powder Bed Fusion (LPBF) has emerged as a promising technology for advanced manufacturing and environmental sustainability. However, LPBF of Aluminum alloy is prone to porosity defects which affect the structural integrity and mechanical properties of manufactured components. Ensuring real-time, accurate defect detection and classification is critical for advancing the scalability of LPBF. We have developed a convolutional neural network (CNN) for real-time porosity detection in LPBF process with the signal registering of high-speed X-ray imaging. In contrast to traditional binary prediction of keyhole-porosity formation, our method defines porosity generation as multi-class (no pore, gas pore, and keyhole pore). Statistical analyses confirmed the efficacy of the multi-class approach, while time series regression further helped differentiate the relationships between the different pore types. CNN model successfully classified multi-class porosity with an accuracy of 85.71%. This research shows the potential for cost-effective, scalable flaw detection and process monitoring in additive manufacturing. |
| Proceedings Inclusion? |
Planned: |
| Keywords |
Additive Manufacturing, Characterization, Machine Learning |