skip to main content
OSTI.GOV title logo U.S. Department of Energy
Office of Scientific and Technical Information

Title: Machine-Learning-Based Monitoring of Laser Powder Bed Fusion

Abstract

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.

Authors:
 [1];  [1];  [1];  [1];  [1];  [1];  [1]; ORCiD logo [1]
  1. Lawrence Livermore National Lab. (LLNL), Livermore, CA (United States)
Publication Date:
Research Org.:
Lawrence Livermore National Lab. (LLNL), Livermore, CA (United States)
Sponsoring Org.:
USDOE National Nuclear Security Administration (NNSA)
OSTI Identifier:
1481059
Alternate Identifier(s):
OSTI ID: 1468826
Report Number(s):
LLNL-JRNL-748383
Journal ID: ISSN 2365-709X; 933429
Grant/Contract Number:  
AC52-07NA27344
Resource Type:
Journal Article: Accepted Manuscript
Journal Name:
Advanced Materials Technologies
Additional Journal Information:
Journal Volume: 3; Journal Issue: 12; Journal ID: ISSN 2365-709X
Publisher:
Wiley
Country of Publication:
United States
Language:
English
Subject:
42 ENGINEERING

Citation Formats

Yuan, Bodi, Guss, Gabriel M., Wilson, Aaron C., Hau-Riege, Stefan P., DePond, Phillip J., McMains, Sara, Matthews, Manyalibo J., and Giera, Brian. Machine-Learning-Based Monitoring of Laser Powder Bed Fusion. United States: N. p., 2018. Web. doi:10.1002/admt.201800136.
Yuan, Bodi, Guss, Gabriel M., Wilson, Aaron C., Hau-Riege, Stefan P., DePond, Phillip J., McMains, Sara, Matthews, Manyalibo J., & Giera, Brian. Machine-Learning-Based Monitoring of Laser Powder Bed Fusion. United States. doi:10.1002/admt.201800136.
Yuan, Bodi, Guss, Gabriel M., Wilson, Aaron C., Hau-Riege, Stefan P., DePond, Phillip J., McMains, Sara, Matthews, Manyalibo J., and Giera, Brian. Wed . "Machine-Learning-Based Monitoring of Laser Powder Bed Fusion". United States. doi:10.1002/admt.201800136.
@article{osti_1481059,
title = {Machine-Learning-Based Monitoring of Laser Powder Bed Fusion},
author = {Yuan, Bodi and Guss, Gabriel M. and Wilson, Aaron C. and Hau-Riege, Stefan P. and DePond, Phillip J. and McMains, Sara and Matthews, Manyalibo J. and Giera, Brian},
abstractNote = {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.},
doi = {10.1002/admt.201800136},
journal = {Advanced Materials Technologies},
issn = {2365-709X},
number = 12,
volume = 3,
place = {United States},
year = {2018},
month = {9}
}

Journal Article:
Free Publicly Available Full Text
This content will become publicly available on September 5, 2019
Publisher's Version of Record

Save / Share: