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 |
Data-Efficient Bayesian Optimization for Layer-wise Dimensional Accuracy Control in Directed Energy Deposition |
Author(s) |
Zijue Chen, Dayalan Gunasegaram |
On-Site Speaker (Planned) |
Zijue Chen |
Abstract Scope |
In Directed Energy Deposition (DED), controlling bead height (H) and height-to-width (HW) ratio is essential for meeting dimensional and surface quality requirements. We present a data-efficient Bayesian Optimization framework that efficiently identifies optimal layer-wise combinations of laser power and scanning speed under fixed powder feed rate. A 2D scanner captures each bead’s cross-section, which is compared to an ideal profile using a weighted error metric. This error guides the optimization process to suggest the following parameter set that seeks to correct the difference as the subsequent layer is deposited. Experimental results show that our tuning method can quickly converge to parameter regions and producing increasingly accurate bead profiles, with the desired H and HW ratio. Heatmaps are provided to visually confirm the evolving accuracy landscape. |
Proceedings Inclusion? |
Planned: Post-meeting proceedings |