About this Abstract |
| Meeting |
2026 Annual International Solid Freeform Fabrication Symposium (SFF Symp 2026)
|
| Symposium
|
2026 Annual International Solid Freeform Fabrication Symposium (SFF Symp 2026)
|
| Presentation Title |
A Neural ODE Approach for Thermomechanical Field Prediction in Directed Energy Deposition |
| Author(s) |
Dhruba Aryal, Praveen S. Vulimiri, Todd Sparks, Albert C. To |
| On-Site Speaker (Planned) |
Dhruba Aryal |
| Abstract Scope |
In metal directed energy deposition, accurate prediction of thermal and mechanical fields such as temperature, residual stress and displacement enables improved design and optimization of parts and process parameters without relying on expensive experimental trials. High fidelity thermo-mechanical simulations using the Finite Element Method can provide such predictions. However, these simulations remain computationally expensive. This work proposes a deep learning method by training a neural ordinary differential equation (ODE) model that leverages features derived from the governing coupled thermo-mechanical equations to predict both thermal and mechanical responses. The model demonstrates excellent agreement with unseen datasets for both fields, accurately capturing residual stresses and deformations while achieving significantly faster inference compared to full finite element simulations. Additionally, the model enables localized prediction of thermo-mechanical fields at regions of interest without requiring full-domain simulation, offering further computational savings for targeted analysis. |
| Proceedings Inclusion? |
Undecided |