| About this Abstract |
| Meeting |
MS&T23: Materials Science & Technology
|
| Symposium
|
Additive Manufacturing: Equipment, Instrumentation and In-Situ Process Monitoring
|
| Presentation Title |
Melt Pool Scale Process Monitoring for Laser Hot Wire Additive
Directed Energy Deposition |
| Author(s) |
Brandon Abranovic, Elizabeth Chang-Davidson, Jack L Beuth |
| On-Site Speaker (Planned) |
Jack L Beuth |
| Abstract Scope |
In order to reliably print quality parts using additive manufacturing, process monitoring is a crucial step in the build process. During deposition, large quantities of data are collected and must be analyzed in order to detect and eliminate process anomalies. Unsupervised deep learning techniques are valuable in executing this analysis due to their ability to recognize flaws without the need for vast quantities of labeled data. A convolutional long-short term memory autoencoder model was trained on process data from a laser hot wire additive manufacturing process. This model used, as input, data from both a visible-light camera and an infrared camera, to encompass melt pool disturbances as well as near-melt pool part temperatures. This model is shown to be feasible as a real-time monitoring technique capable of detecting known characteristic process flaws, as well as a post-deposition data analysis tool for directing part testing towards suspected flaw areas. |