| About this Abstract |
| Meeting |
TMS Specialty Congress 2024
|
| Symposium
|
Accelerating Discovery for Mechanical Behavior of Materials 2024
|
| Presentation Title |
Crystal Plasticity Finite Element Method Accelerated by Efficient GPU-computing |
| Author(s) |
Fanglei Hu, Fan Chen, Stephen Niezgoda, Tianju Xue, Jian Cao |
| On-Site Speaker (Planned) |
Fanglei Hu |
| Abstract Scope |
Crystal Plasticity Finite Element Method (CPFEM) has emerged as a powerful tool for establishing structure–property relationships. However, because of the need to handle the Newton iteration over the stress residual and calculate the Jacobian based on constitutive laws, CPFEM is computationally expensive. In this study, we propose the application of GPU to accelerate the three-dimensional CPFEM simulation, focusing on the mechanical response of metal polycrystals under different loading conditions while employing phenomenological constitutive models. The simulation code is developed based on our open-source code JAX-FEM, an affordable platform implemented with pure-Python while scalable to efficiently solve problems with moderate to large sizes (>100,000 degrees of freedom). Three simulation examples are performed for the cases of face-centered cubic (fcc) copper, body-centered cubic (bcc) tantalum, and bcc 316L steel. A thorough evaluation of the GPU implementation's acceleration performance in comparison to the fundamental CPU computations and Abaqus UMAT will be presented. |
| Proceedings Inclusion? |
Definite: Other |