| About this Abstract |
| Meeting |
2024 Annual International Solid Freeform Fabrication Symposium (SFF Symp 2024)
|
| Symposium
|
2024 Annual International Solid Freeform Fabrication Symposium (SFF Symp 2024)
|
| Presentation Title |
Inverse Generation of Metamaterial using Graph Neural Network |
| Author(s) |
Jier Wang, Ajit Panesar |
| On-Site Speaker (Planned) |
Ajit Panesar |
| Abstract Scope |
In this work, a framework that applies Graph Neural Networks (GNN) in generating the metamaterials with desired mechanical and thermal properties is presented. A GNN-based inverse generator is developed to facilitate the creation of truss lattices considering the input target properties. The graph-based representation of the truss lattice enables the model to handle diverse lattice topologies simultaneously, thereby significantly enhancing design flexibility for a more generally applicable inverse generator. GNN is also superior to pixel-based convolutional neural networks as they require smaller data sizes and fewer training parameters in the model. The effectiveness of this inverse generator is validated through comparison with the conventional de-homogenisation method, highlighting its advantages in reducing computational costs and expanding design variety. |
| Proceedings Inclusion? |
Definite: Post-meeting proceedings |