THERMAL PROBE SCANNING OF LASER-POWDER BED FUSION ADDITIVE MANUFACTURING 316 STAINLESS STEEL
Date Issued
May 2026
Author(s)
Joslin, Chase
Oak Ridge Associated Universities, Oak Ridge National Laboratory, University of Tennessee Knoxville
Advisor(s)
Bradley H. Jared
Additional Advisor(s)
Brett G. Compton
Sudarsanam S. Babu
Abstract
Laser-powder bed fusion (L-[hyphen]PBF) additive manufacturing (AM) continues to be a promising tool for fabricating complex geometries. However, both in-situ monitoring and ex-situ evaluation techniques cannot yet fully certify and qualify components due in part to the complex nature of melting dynamics at the build plane and laser interaction zone. This study began using a reflection configuration Flash Thermography (FT) to investigate the viability of detecting subsurface porosity ex-situ. From the FT results, active thermography was utilized to observe changes in thermal signatures on 316SS parts printed by L-PBF AM. By performing a low-power laser scan as a post-melt heat source, the surfaces of printed parts were thermally probed without remelting. A design of experiments (DOE) was performed to investigate parameters for the thermal probe scan (TPS). Imaging the TPS with a low frame rate near-infrared camera and an on-axis photodiode was found to identify near-surface pores in-situ. A dynamic multiscale convolutional neural network (DMSCNN) was trained to classify 2-dimensional (2D) visible-light, near-infrared, and photodiode images. A regression model was fit to predict percent porosity using anomaly class predictions from the DMSCNN as inputs. Data were segmented into 1mm x 1mm x 0.1mm zones. Parts printed with engineered pores had a regression correlation coefficient, R²[squared], value of 0.795. While the regression model did not accurately predict percent porosity for control parts using anomaly class predictions. However, the TPS does show viability to detect large scale lack-of-fusion porosity in L-PBF in-situ process monitoring.
Disciplines
Degree
Master of Science
Major
Mechanical Engineering
File(s)![Thumbnail Image]()
Name
JoslinThesis_2026.pdf
Size
3.74 MB
Format
Adobe PDF
Checksum (MD5)
6b5ade917d3c4fc50952ec17bd89097c