Details
| Original language | English |
|---|---|
| Article number | 100327 |
| Journal | Additive Manufacturing Letters |
| Volume | 15 |
| Early online date | 8 Oct 2025 |
| Publication status | Published - Dec 2025 |
Abstract
Wire-arc directed energy deposition (DED) has emerged as a promising additive manufacturing (AM) technology for large-scale applications. However, the complex thermal dynamics inherent to the process present challenges in ensuring structural integrity and mechanical properties of fabricated components. Finite element method (FEM) simulations have been conventionally employed to predict thermal history during deposition. However, their high computational demand increase significantly with scale. Given the necessity of multiple repetitive simulations for heat management and the determination of optimal printing strategy, FEM simulation quickly becomes unfit. Instead, advancements have been made in using trained neural networks as surrogate models for rapid prediction. However, traditional data-driven approaches necessitate large amounts of relevant and verifiable external data, either from simulation, experimental, or analytical solutions, during the training and validation of the neural network. Regarding large-scale wire-arc DED, none of these data sources are readily available in quantities sufficient for an accurate surrogate. The introduction of physics-informed neural networks (PINNs) has opened up an alternative simulation strategy by leveraging the existing physical knowledge of the phenomena with advanced machine learning methods. However, the practical application of PINNs for real-world large-scale wire-arc DED has been rarely explored, particularly within the context of structural engineering. This study investigates one of the necessary steps for up-scaling PINN with a focus on advanced and effective sampling of collocation points — a critical factor controlling both the training time and the performance of the model. The results affirm the potential of PINNs to outperform FEM in terms of wall-clock times, while maintaining the desired accuracy and offering resolution-agnostic evaluation. Further discussion provides an outlook on the future steps for improving the PINNs for wire-arc DED simulations.
Keywords
- Data-free modeling, Large-scale modeling, PDE-based learning, Physics-informed neural networks, Sobol’ sequences, Thermal simulation, Wire-arc additive manufacturing
ASJC Scopus subject areas
- Materials Science(all)
- Materials Science (miscellaneous)
- Engineering(all)
- Engineering (miscellaneous)
- Engineering(all)
- Mechanics of Materials
- Engineering(all)
- Industrial and Manufacturing Engineering
Cite this
- Standard
- Harvard
- Apa
- Vancouver
- BibTeX
- RIS
In: Additive Manufacturing Letters, Vol. 15, 100327, 12.2025.
Research output: Contribution to journal › Article › Research › peer review
}
TY - JOUR
T1 - Physics-informed machine learning surrogate for scalable simulation of thermal histories during wire-arc directed energy deposition
AU - Ryan, Michael
AU - Baqershahi, Mohammad Hassan
AU - Moshayedi, Hessamoddin
AU - Ghafoori, Elyas
N1 - Publisher Copyright: © 2025 The Authors
PY - 2025/12
Y1 - 2025/12
N2 - Wire-arc directed energy deposition (DED) has emerged as a promising additive manufacturing (AM) technology for large-scale applications. However, the complex thermal dynamics inherent to the process present challenges in ensuring structural integrity and mechanical properties of fabricated components. Finite element method (FEM) simulations have been conventionally employed to predict thermal history during deposition. However, their high computational demand increase significantly with scale. Given the necessity of multiple repetitive simulations for heat management and the determination of optimal printing strategy, FEM simulation quickly becomes unfit. Instead, advancements have been made in using trained neural networks as surrogate models for rapid prediction. However, traditional data-driven approaches necessitate large amounts of relevant and verifiable external data, either from simulation, experimental, or analytical solutions, during the training and validation of the neural network. Regarding large-scale wire-arc DED, none of these data sources are readily available in quantities sufficient for an accurate surrogate. The introduction of physics-informed neural networks (PINNs) has opened up an alternative simulation strategy by leveraging the existing physical knowledge of the phenomena with advanced machine learning methods. However, the practical application of PINNs for real-world large-scale wire-arc DED has been rarely explored, particularly within the context of structural engineering. This study investigates one of the necessary steps for up-scaling PINN with a focus on advanced and effective sampling of collocation points — a critical factor controlling both the training time and the performance of the model. The results affirm the potential of PINNs to outperform FEM in terms of wall-clock times, while maintaining the desired accuracy and offering resolution-agnostic evaluation. Further discussion provides an outlook on the future steps for improving the PINNs for wire-arc DED simulations.
AB - Wire-arc directed energy deposition (DED) has emerged as a promising additive manufacturing (AM) technology for large-scale applications. However, the complex thermal dynamics inherent to the process present challenges in ensuring structural integrity and mechanical properties of fabricated components. Finite element method (FEM) simulations have been conventionally employed to predict thermal history during deposition. However, their high computational demand increase significantly with scale. Given the necessity of multiple repetitive simulations for heat management and the determination of optimal printing strategy, FEM simulation quickly becomes unfit. Instead, advancements have been made in using trained neural networks as surrogate models for rapid prediction. However, traditional data-driven approaches necessitate large amounts of relevant and verifiable external data, either from simulation, experimental, or analytical solutions, during the training and validation of the neural network. Regarding large-scale wire-arc DED, none of these data sources are readily available in quantities sufficient for an accurate surrogate. The introduction of physics-informed neural networks (PINNs) has opened up an alternative simulation strategy by leveraging the existing physical knowledge of the phenomena with advanced machine learning methods. However, the practical application of PINNs for real-world large-scale wire-arc DED has been rarely explored, particularly within the context of structural engineering. This study investigates one of the necessary steps for up-scaling PINN with a focus on advanced and effective sampling of collocation points — a critical factor controlling both the training time and the performance of the model. The results affirm the potential of PINNs to outperform FEM in terms of wall-clock times, while maintaining the desired accuracy and offering resolution-agnostic evaluation. Further discussion provides an outlook on the future steps for improving the PINNs for wire-arc DED simulations.
KW - Data-free modeling
KW - Large-scale modeling
KW - PDE-based learning
KW - Physics-informed neural networks
KW - Sobol’ sequences
KW - Thermal simulation
KW - Wire-arc additive manufacturing
UR - http://www.scopus.com/inward/record.url?scp=105018230629&partnerID=8YFLogxK
U2 - 10.1016/j.addlet.2025.100327
DO - 10.1016/j.addlet.2025.100327
M3 - Article
VL - 15
JO - Additive Manufacturing Letters
JF - Additive Manufacturing Letters
M1 - 100327
ER -