| About this Abstract |
| Meeting |
2021 Annual International Solid Freeform Fabrication Symposium (SFF Symp 2021)
|
| Symposium
|
Special Session
|
| Presentation Title |
Heterogenous Sensing and Scientific Machine Learning for In-process Anomaly Detection in Laser Powder Bed Fusion – A Single-track Study |
| Author(s) |
Aniruddha Gaikwad, Brian Giera, Ibo Matthews, Prahalada K. Rao |
| On-Site Speaker (Planned) |
Aniruddha Gaikwad |
| Abstract Scope |
The objective is to predict the build quality of a track of fused material as a function of process signatures derived from a pyrometer and high-speed optical video camera integrated into a laser powder bed fusion additive manufacturing system. The central hypothesis of this work is that the accuracy of anomaly prediction improves significantly when machine learning models incorporate process signatures that are based on fundamental knowledge of the process regime, as opposed to purely data-driven machine learning algorithms, such as deep learning convolutional neural networks. To test the efficacy of such a scientific machine learning hypothesis over 1000 single tracks (stainless steel 316L) of length 5 mm were fused under 121 different laser power and laser velocity combinations. During the process of fusing the single tracks, meltpool-level signatures were acquired using an in-situ pyrometer and high-speed optical video camera located coaxial to the laser. |
| Proceedings Inclusion? |
Definite: Post-meeting proceedings |