A Physics-based and Geometry-aware Reduced-order Framework for Accelerated Part-Scale Thermomechanical Simulation of Additive Friction Stir Deposition Process
Abstract
Additive friction stir deposition (AFSD) is attractive for naval and aerospace sustainment because it enables high rate, solid state near net shape fabrication and repair of large metallic components. However, full part scale thermomechanical simulations remain computationally demanding because of the moving heat source, progressive material deposition, and evolving geometry. This work presents a physics-based and geometry-aware reduced order framework for rapid prediction of temperature and residual stress fields in part scale AFSD builds. The transient temperature field is represented using a tensor product quadratic B spline basis on a fixed Cartesian background grid, while arbitrary part geometries are embedded using a signed distance function (SDF) and a smooth material indicator. Residual stresses are recovered through an inherent strain formulation in which an offline thermo elasto plastic representative volume element (RVE) simulation supplies a calibrated eigenstrain field that is spatially scaled using the part scale peak temperature history. The thermal formulation is evaluated against published numerical and experimental AFSD data and two sample multilayer builds are simulated as demonstrations. For a four layer thin-wall build, the thermal solution requires approximately 22 s compared to 806 s for the finite element model, while the complete thermo-mechanical framework requires approximately 86.4 s compared to 2,406 s. For a larger 20 layer hollow cylindrical build, the reduced order thermal simulations requires 0.148 to 0.152 h compared with 1.8 h for the finite element model, corresponding to thermal speed-ups of approximately 11.8 to 12.2 times. The complete thermomechanical framework requires 0.346 h compared to 5.4 h for the full finite element analysis, providing an overall speed-up of approximately 15.6 times.
Related
- A GPU-Accelerated and Geometry-Aware Quantum-Classical Framework for Optimized PES Scans with CUDA-Q
- Phase-Field Fracture Simulation of Highly Deformable Thin Structures via a Discrete Differential Geometry Framework
- A thermomechanical framework for strongly nonlocal continua
- Passage of particles through matter and the effective straggling-function: High-fidelity accelerated simulation via Physics-Informed Machine Learning
- Surface chemistry as data for level-set simulation of etching and deposition
BibTeX
@misc{sarker2026physicsbased,
title = {{A Physics-based and Geometry-aware Reduced-order Framework for Accelerated Part-Scale Thermomechanical Simulation of Additive Friction Stir Deposition Process}},
author = {Apurba Sarker and Sourav Saha},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2610.22375},
eprint = {2610.22375},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2610.22375},
note = {CC BY-NC-SA 4.0},
}
RIS
TY - UNPB
TI - A Physics-based and Geometry-aware Reduced-order Framework for Accelerated Part-Scale Thermomechanical Simulation of Additive Friction Stir Deposition Process
AU - Apurba Sarker
AU - Sourav Saha
PY - 2026
DA - 2026/10/01
PB - arXiv
DO - 10.48550/arxiv.2610.22375
UR - https://arxiv.org/abs/2610.22375
AB - Additive friction stir deposition (AFSD) is attractive for naval and aerospace sustainment because it enables high rate, solid state near net shape fabrication and repair of large metallic components. However, full part scale thermomechanical simulations remain computationally demanding because of the moving heat source, progressive material deposition, and evolving geometry. This work presents a physics-based and geometry-aware reduced order framework for rapid prediction of temperature and residual stress fields in part scale AFSD builds. The transient temperature field is represented using a tensor product quadratic B spline basis on a fixed Cartesian background grid, while arbitrary part geometries are embedded using a signed distance function (SDF) and a smooth material indicator. Residual stresses are recovered through an inherent strain formulation in which an offline thermo elasto plastic representative volume element (RVE) simulation supplies a calibrated eigenstrain field that is spatially scaled using the part scale peak temperature history. The thermal formulation is evaluated against published numerical and experimental AFSD data and two sample multilayer builds are simulated as demonstrations. For a four layer thin-wall build, the thermal solution requires approximately 22 s compared to 806 s for the finite element model, while the complete thermo-mechanical framework requires approximately 86.4 s compared to 2,406 s. For a larger 20 layer hollow cylindrical build, the reduced order thermal simulations requires 0.148 to 0.152 h compared with 1.8 h for the finite element model, corresponding to thermal speed-ups of approximately 11.8 to 12.2 times. The complete thermomechanical framework requires 0.346 h compared to 5.4 h for the full finite element analysis, providing an overall speed-up of approximately 15.6 times.
ER -