.png){kind=icon}
.png){kind=icon}
Return
Physics-informed process-structure-property prediction model for additively manufactured metallic parts
.png){kind=icon}
PRE
AI
T
M
H
Y
J
Z
J
DOI：10.1016/j.actamat.2026.122630.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Establishing process-structure-property (PSP) relationships is essential for elucidating how process parameters affect microstructural evolution and their influence on mechanical properties in laser powder bed fusion (LPBF). However, prediction of mechanical properties coupled with microstructural evolution remains challenging due to complex thermal histories inherent to LPBF, in addition to time- and cost-intensive efforts for analyzing subsequent heat treatment effects. While conventional data-driven methods have attempted to predict mechanical properties directly from process or structure parameters, they often fail to characterize PSP relationships. Recent rapid advances in artificial intelligence techniques enable new approaches that combine machine learning with physics-informed modeling to overcome these limitations. Here, we propose a physics-informed process-structure-property prediction (PSPP) model for LPBF-fabricated metallic parts by integrating conditional denoising diffusion probabilistic models (cDDPM) with physics-informed neural networks (PINN). The cDDPM reconstructs high-fidelity microstructures under heat treatment conditions, while the PINN accurately predicts tensile behavior from microstructural features. Furthermore, the PSPP model successfully identifies optimum heat treatment conditions for maximizing mechanical performance with a low prediction error (≤ 6.1%). Therefore, this PSPP framework is expected to be widely applicable to other additive manufacturing techniques for dramatically reducing design cycles and experimental costs.
Keywords:
Physics-informed neural networks
Additive manufacturing
Laser powder bed fusion
Process-structure-property relationship
Heat treatment
Journal
IF:
9.3
Papers:
2.0W
Citations:
12.9W
Organization
Cited Papers