Deep learning for melt pool depth contour prediction from surface thermal images via vision transformers

Anomalous melt pools during metal additive manufacturing (AM) can lead to deteriorated mechanical and fatigue performance. In-situ monitoring of the melt pool subsurface morphology requires specialized equipment that may not be readily accessible or scalable. Therefore, we introduce a machine learni...

Full description

Saved in:
Bibliographic Details
Main Authors: Francis Ogoke, Peter Pak, Alexander Myers, Guadalupe Quirarte, Jack Beuth, Jonathan Malen, Amir Barati Farimani
Format: Article
Language:English
Published: Elsevier 2024-12-01
Series:Additive Manufacturing Letters
Subjects:
Online Access:http://www.sciencedirect.com/science/article/pii/S2772369024000513
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1846127100318187520
author Francis Ogoke
Peter Pak
Alexander Myers
Guadalupe Quirarte
Jack Beuth
Jonathan Malen
Amir Barati Farimani
author_facet Francis Ogoke
Peter Pak
Alexander Myers
Guadalupe Quirarte
Jack Beuth
Jonathan Malen
Amir Barati Farimani
author_sort Francis Ogoke
collection DOAJ
description Anomalous melt pools during metal additive manufacturing (AM) can lead to deteriorated mechanical and fatigue performance. In-situ monitoring of the melt pool subsurface morphology requires specialized equipment that may not be readily accessible or scalable. Therefore, we introduce a machine learning framework to correlate in-situ two-color thermal images observed via high-speed color imaging to the two-dimensional profile of the melt pool cross-section. We employ a hybrid CNN-Transformer architecture to establish a correlation between single bead off-axis thermal image sequences and melt pool cross-section contours measured via optical microscopy. Specifically, a ResNet model embeds the spatial information contained within the thermal images to a latent vector, while a Transformer model correlates the sequence of embedded vectors to extract temporal information. The performance of this model is evaluated through dimensional and geometric comparisons to the corresponding experimental no-powder melt pool observations. Our framework is able to model the curvature of the subsurface melt pool structure, with improved performance in high energy density regimes compared to analytical models. Additionally, the use of ratiometric temperature estimates improves the accuracy of the model predictions compared to monochromatic imaging. This work establishes a framework extensible towards powder-based AM builds.
format Article
id doaj-art-8ea513ade1ab43c6bc8a4c267904085f
institution Kabale University
issn 2772-3690
language English
publishDate 2024-12-01
publisher Elsevier
record_format Article
series Additive Manufacturing Letters
spelling doaj-art-8ea513ade1ab43c6bc8a4c267904085f2024-12-12T05:24:14ZengElsevierAdditive Manufacturing Letters2772-36902024-12-0111100243Deep learning for melt pool depth contour prediction from surface thermal images via vision transformersFrancis Ogoke0Peter Pak1Alexander Myers2Guadalupe Quirarte3Jack Beuth4Jonathan Malen5Amir Barati Farimani6Department of Mechanical Engineering, Carnegie Mellon University, Pittsburgh, 15213, PA, USADepartment of Mechanical Engineering, Carnegie Mellon University, Pittsburgh, 15213, PA, USADepartment of Mechanical Engineering, Carnegie Mellon University, Pittsburgh, 15213, PA, USADepartment of Mechanical Engineering, Carnegie Mellon University, Pittsburgh, 15213, PA, USADepartment of Mechanical Engineering, Carnegie Mellon University, Pittsburgh, 15213, PA, USADepartment of Mechanical Engineering, Carnegie Mellon University, Pittsburgh, 15213, PA, USADepartment of Mechanical Engineering, Carnegie Mellon University, Pittsburgh, 15213, PA, USA; Department of Chemical Engineering, Carnegie Mellon University, Pittsburgh, 15213, PA, USA; Machine Learning Department, Carnegie Mellon University, Pittsburgh, 15213, PA, USA; Correspondence to: Carnegie Mellon University, 5000 Forbes Avenue, Pittsburgh, 15213, PA, USA.Anomalous melt pools during metal additive manufacturing (AM) can lead to deteriorated mechanical and fatigue performance. In-situ monitoring of the melt pool subsurface morphology requires specialized equipment that may not be readily accessible or scalable. Therefore, we introduce a machine learning framework to correlate in-situ two-color thermal images observed via high-speed color imaging to the two-dimensional profile of the melt pool cross-section. We employ a hybrid CNN-Transformer architecture to establish a correlation between single bead off-axis thermal image sequences and melt pool cross-section contours measured via optical microscopy. Specifically, a ResNet model embeds the spatial information contained within the thermal images to a latent vector, while a Transformer model correlates the sequence of embedded vectors to extract temporal information. The performance of this model is evaluated through dimensional and geometric comparisons to the corresponding experimental no-powder melt pool observations. Our framework is able to model the curvature of the subsurface melt pool structure, with improved performance in high energy density regimes compared to analytical models. Additionally, the use of ratiometric temperature estimates improves the accuracy of the model predictions compared to monochromatic imaging. This work establishes a framework extensible towards powder-based AM builds.http://www.sciencedirect.com/science/article/pii/S2772369024000513Deep learningin-situ monitoringVision transformersLack-of-fusion
spellingShingle Francis Ogoke
Peter Pak
Alexander Myers
Guadalupe Quirarte
Jack Beuth
Jonathan Malen
Amir Barati Farimani
Deep learning for melt pool depth contour prediction from surface thermal images via vision transformers
Additive Manufacturing Letters
Deep learning
in-situ monitoring
Vision transformers
Lack-of-fusion
title Deep learning for melt pool depth contour prediction from surface thermal images via vision transformers
title_full Deep learning for melt pool depth contour prediction from surface thermal images via vision transformers
title_fullStr Deep learning for melt pool depth contour prediction from surface thermal images via vision transformers
title_full_unstemmed Deep learning for melt pool depth contour prediction from surface thermal images via vision transformers
title_short Deep learning for melt pool depth contour prediction from surface thermal images via vision transformers
title_sort deep learning for melt pool depth contour prediction from surface thermal images via vision transformers
topic Deep learning
in-situ monitoring
Vision transformers
Lack-of-fusion
url http://www.sciencedirect.com/science/article/pii/S2772369024000513
work_keys_str_mv AT francisogoke deeplearningformeltpooldepthcontourpredictionfromsurfacethermalimagesviavisiontransformers
AT peterpak deeplearningformeltpooldepthcontourpredictionfromsurfacethermalimagesviavisiontransformers
AT alexandermyers deeplearningformeltpooldepthcontourpredictionfromsurfacethermalimagesviavisiontransformers
AT guadalupequirarte deeplearningformeltpooldepthcontourpredictionfromsurfacethermalimagesviavisiontransformers
AT jackbeuth deeplearningformeltpooldepthcontourpredictionfromsurfacethermalimagesviavisiontransformers
AT jonathanmalen deeplearningformeltpooldepthcontourpredictionfromsurfacethermalimagesviavisiontransformers
AT amirbaratifarimani deeplearningformeltpooldepthcontourpredictionfromsurfacethermalimagesviavisiontransformers