| Author(s) |
Steven Storck, Mike Brown, Brandan Croom, Mary Daffron, Ari Lax, Li Ma, Robert Mueller, Victor Leon, Samuel Gonzalez, Brad Bazow, Jackson Pittman, Vince Pagan, Jade Traiger, Ranjit LoboPrabhu, Morgan Trexler, Mark Foster, Colin Goodman |
| Abstract Scope |
Defect formation in the AM process has caused a great deal of uncertainty in qualifying parts and put a burden on post manufacturing inspection. In-situ monitoring can relieve this challenge enabling inspection in process, yet this vision has had limited success due to chaotic synthesis and rapid solidification of AM. A high-speed spectral sensor was developed to measure changes in processing dynamics and combined with a FPGA to augment the process in about one microsecond. Combining high-speed data collection with ML results in defect identification better than 99% for lack of fusion and 94% for keyholes. An active high-speed control system aided by efficient ML techniques has been developed with a response time faster than the Nyquist sampling threshold. The fusion of high-speed sensing, ML based model, and sub microsecond process augmentation has the potential to quantify defects in process, and mitigate their formation, enabling direct qualification of critical parts. |