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 |
MODAL: A Scalable Surrogate Modeling Framework for Scanwise Thermal Simulations in Laser Powder Bed Fusion |
| Author(s) |
Berkay Bostan, Praveen Vulimiri, Dhruba Aryal, Albert To |
| On-Site Speaker (Planned) |
Berkay Bostan |
| Abstract Scope |
Laser Powder Bed Fusion (LPBF) can produce complex metal parts but is prone to defects due to rapid and localized thermal behavior. In-situ monitoring provides useful insight but requires specialized hardware and generates large data volumes, limiting its practicality. High-fidelity thermal simulations are accurate but computationally expensive. This work introduces MODAL (MODular Assembly Learning), a scalable surrogate model for scanwise thermal simulations in LPBF. MODAL is trained on a small, diverse block set and uses randomized padding of nodes and heat sources to enable prediction over much larger, more complex domains. On average, test parts are 14 times larger than training blocks in spatiotemporal scale and include thermal predictions for up to three melting events per node, covering the main and two upper layers. MODAL achieves 10% mean absolute percentage error, 46.7 °C mean absolute error, and an R² score of 0.92, while delivering up to 1421× speedup using four GPUs. |
| Proceedings Inclusion? |
Planned: Post-meeting proceedings |