Geometry-agnostic data-driven thermal modeling of additive manufacturing processes using graph neural networks

Mojtaba Mozaffar; Shuheng Liao; Hui Lin; Kornel Ehmann; Jian Cao

doi:10.1016/j.addma.2021.102449

Geometry-agnostic data-driven thermal modeling of additive manufacturing processes using graph neural networks

Mojtaba Mozaffar, Shuheng Liao, Hui Lin, Kornel Ehmann, Jian Cao^*

^*Corresponding author for this work

Mechanical Engineering

Research output: Contribution to journal › Article › peer-review

47 Scopus citations

Abstract

Additive manufacturing (AM) is commonly used to produce builds with complex geometries. Despite recent advancements in data-driven modeling of AM processes, the generalizability of such models across a wide range of geometries has remained a challenge. Here, a graph-based representation using neural networks is proposed to capture spatiotemporal dependencies of thermal responses in AM processes. Our results tested on the Directed Energy Deposition process, indicate that our deep learning architecture accurately predicts long thermal histories for unseen geometries in the training process, offering a viable alternative to expensive computational mechanics or experimental solutions.

Original language	English (US)
Article number	102449
Journal	Additive Manufacturing
Volume	48
DOIs	https://doi.org/10.1016/j.addma.2021.102449
State	Published - Dec 2021

Funding

The authors declare the following financial interests/personal relationships which may be considered as potential competing interests: Jian Cao reports financial support was provided by US Department of Defense. Jian Cao reports financial support was provided by National Institute of Standards and Technology. Kornel F. Ehmann reports financial support was provided by National Science Foundation. Mojtaba Mozaffar reports financial support was provided by Department of Defense. Shuheng Liao reports financial support was provided by National Institute of Standards and Technology. This work was supported by the Vannevar Bush Faculty Fellowship N00014-19-1-2642 , National Institute of Standards and Technology (NIST) – Center for Hierarchical Material Design (CHiMaD) under grant No. 70NANB14H012 , and the National Science Foundation (NSF) – Cyber-Physical Systems (CPS) under grant No. CPS/CMMI-1646592 .

Keywords

Additive manufacturing
Graph neural network
Machine learning

ASJC Scopus subject areas

Biomedical Engineering
General Materials Science
Engineering (miscellaneous)
Industrial and Manufacturing Engineering

UN SDGs

This output contributes to the following UN Sustainable Development Goals (SDGs)

Access to Document

10.1016/j.addma.2021.102449

Cite this

@article{44a70ff44f344197ab0ab33a5fcd3480,

title = "Geometry-agnostic data-driven thermal modeling of additive manufacturing processes using graph neural networks",

abstract = "Additive manufacturing (AM) is commonly used to produce builds with complex geometries. Despite recent advancements in data-driven modeling of AM processes, the generalizability of such models across a wide range of geometries has remained a challenge. Here, a graph-based representation using neural networks is proposed to capture spatiotemporal dependencies of thermal responses in AM processes. Our results tested on the Directed Energy Deposition process, indicate that our deep learning architecture accurately predicts long thermal histories for unseen geometries in the training process, offering a viable alternative to expensive computational mechanics or experimental solutions.",

keywords = "Additive manufacturing, Graph neural network, Machine learning",

author = "Mojtaba Mozaffar and Shuheng Liao and Hui Lin and Kornel Ehmann and Jian Cao",

note = "Publisher Copyright: {\textcopyright} 2021 Elsevier B.V.",

year = "2021",

month = dec,

doi = "10.1016/j.addma.2021.102449",

language = "English (US)",

volume = "48",

journal = "Additive Manufacturing",

issn = "2214-8604",

publisher = "Elsevier B.V.",

}

TY - JOUR

T1 - Geometry-agnostic data-driven thermal modeling of additive manufacturing processes using graph neural networks

AU - Mozaffar, Mojtaba

AU - Liao, Shuheng

AU - Lin, Hui

AU - Ehmann, Kornel

AU - Cao, Jian

PY - 2021/12

Y1 - 2021/12

N2 - Additive manufacturing (AM) is commonly used to produce builds with complex geometries. Despite recent advancements in data-driven modeling of AM processes, the generalizability of such models across a wide range of geometries has remained a challenge. Here, a graph-based representation using neural networks is proposed to capture spatiotemporal dependencies of thermal responses in AM processes. Our results tested on the Directed Energy Deposition process, indicate that our deep learning architecture accurately predicts long thermal histories for unseen geometries in the training process, offering a viable alternative to expensive computational mechanics or experimental solutions.

AB - Additive manufacturing (AM) is commonly used to produce builds with complex geometries. Despite recent advancements in data-driven modeling of AM processes, the generalizability of such models across a wide range of geometries has remained a challenge. Here, a graph-based representation using neural networks is proposed to capture spatiotemporal dependencies of thermal responses in AM processes. Our results tested on the Directed Energy Deposition process, indicate that our deep learning architecture accurately predicts long thermal histories for unseen geometries in the training process, offering a viable alternative to expensive computational mechanics or experimental solutions.

KW - Additive manufacturing

KW - Graph neural network

KW - Machine learning

UR - http://www.scopus.com/inward/record.url?scp=85119606282&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85119606282&partnerID=8YFLogxK

U2 - 10.1016/j.addma.2021.102449

DO - 10.1016/j.addma.2021.102449

M3 - Article

AN - SCOPUS:85119606282

SN - 2214-8604

VL - 48

JO - Additive Manufacturing

JF - Additive Manufacturing

M1 - 102449

ER -

Geometry-agnostic data-driven thermal modeling of additive manufacturing processes using graph neural networks

Abstract

Funding

Keywords

ASJC Scopus subject areas

UN SDGs

Access to Document

Other files and links

Fingerprint

Cite this