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 |
Real-Time Implementability Evaluation of FPGA-Based Trackwise Model Predictive Control for Laser Powder Bed Fusion |
| Author(s) |
Chen-Wei Yang, Zhuo Yang, Gi Suk Hong, Lu Yan |
| On-Site Speaker (Planned) |
Zhuo Yang |
| Abstract Scope |
In additive manufacturing, Multi-Scale Model Predictive Control has the potential for ensuring part quality. Trackwise closed-loop control aims to detect and repair overfusion or lack of fusion, often observed by image-based melt pool monitoring through the analysis of melting overlap between adjacent tracks. While most trackwise control strategies exist at the design stage, real-time implementation remains challenging due to the need for both complex computation and control performance, which is difficult to achieve using general-purpose PCs. In this work, we leverage the NIST AM Data Integration Testbench to evaluate the implementability of real-time trackwise control using FPGA-based processing on a frame grabber of a 10KHz melt pool image acquisition rate. We implement (1) trackwise overlap ratio calculation and (2) trackwise neural network inference. The results demonstrate that both tasks can be completed on time on the FPGA by applying a hardware-friendly design, achieving a balance between latency and resource utilization. |
| Proceedings Inclusion? |
Planned: Post-meeting proceedings |