Machine-Learning-Based Monitoring of Laser Powder Bed Fusion
- Lawrence Livermore National Lab. (LLNL), Livermore, CA (United States)
A two-step machine learning approach to monitoring Laser Powder Bed Fusion (LPBF) additive manufacturing is demonstrated that enables on-the-fly assessments of laser track welds. First, in situ video melt pool data acquired during LPBF is labeled according to the (1) average and (2) standard deviation of individual track width and also (3) whether or not the track is continuous, measured post-build through an ex situ height map analysis algorithm. This procedure generates three ground truth labeled datasets for supervised machine learning. Using a portion of the labeled 10-millisecond video clips, a single Convolutional Neural Network architecture is trained to generate three distinct networks. With the remaining in situ LPBF data, the trained neural networks are tested and evaluated and found to predict track width, standard deviation, and continuity without the need for ex situ measurements. This two-step approach should benefit any LPBF system – or any additive manufacturing technology – where height-map-derived properties can serve as useful labels for in situ sensor data.
- Research Organization:
- Lawrence Livermore National Laboratory (LLNL), Livermore, CA (United States)
- Sponsoring Organization:
- USDOE National Nuclear Security Administration (NNSA)
- Grant/Contract Number:
- AC52-07NA27344
- OSTI ID:
- 1481059
- Alternate ID(s):
- OSTI ID: 1468826
- Report Number(s):
- LLNL-JRNL-748383; 933429
- Journal Information:
- Advanced Materials Technologies, Vol. 3, Issue 12; ISSN 2365-709X
- Publisher:
- WileyCopyright Statement
- Country of Publication:
- United States
- Language:
- English
Web of Science
Design and validation of a ten nanosecond resolved resistive thermometer for Gaussian laser beam heating
|
journal | December 2019 |
Toward the digital twin of additive manufacturing: Integrating thermal simulations, sensing, and analytics to detect process faults
|
journal | January 2020 |
Similar Records
(Machine) learning to do more with less
Multi phenomena melt pool sensor data fusion for enhanced process monitoring of laser powder bed fusion additive manufacturing