About this Abstract |
| Meeting |
MS&T25: Materials Science & Technology
|
| Symposium
|
Integrated Computational Materials Engineering for Physics-Based Machine Learning Models
|
| Presentation Title |
A GNN based Finite Element Simulations Emulator: Application to Parameter Identification for Aluminum Alloy 6DR1 |
| Author(s) |
Ossama Abou Ali Modad, Georges Ayoub |
| On-Site Speaker (Planned) |
Ossama Abou Ali Modad |
| Abstract Scope |
In this study, a novel graph neural network based finite element surrogate model is developed to emulate 3D finite element (FE) tensile testing simulations and accelerate parameter identification for aluminum 6DR1 alloy. The GNN model was able to accurately replicate the local and global mechanical responses as well as rapidly and efficiently identify material parameters compared to the traditional parameter identification method. Sensitivity analysis highlighted key material parameters that impact model performance, while uncertainty quantification verified prediction robustness and determined the parameter space where the GNN predictions are robust. |