About this Abstract |
| Meeting |
2025 Annual International Solid Freeform Fabrication Symposium (SFF Symp 2025)
|
| Symposium
|
2025 Annual International Solid Freeform Fabrication Symposium (SFF Symp 2025)
|
| Presentation Title |
Conditional Variational Autoencoder Framework for Geometric-Constrained, Nonlinear Property-Driven Composite Microstructure Design |
| Author(s) |
Lin Cheng, Jay Yen, Zhangxian Yuan, Denis Molchanov |
| On-Site Speaker (Planned) |
Lin Cheng |
| Abstract Scope |
Additive manufacturing (AM) offers unprecedented capabilities in producing composite materials with highly customized fiber deposition, enabling precise tuning of microstructures for desirable nonlinear material properties. However, this design flexibility introduces significant challenges for conventional design methodologies due to the vast and complex design space. To fully harness the potential of AM for customized composite development, new property-driven design frameworks are essential. In this work, we propose a Conditional Variational Autoencoder (CVAE) with a learnable embedding space to generate composite microstructures that satisfy both geometric constraints and target nonlinear material properties. Geometric constraints are incorporated via the conditioning input, while the embedding space allows the model to adapt designs toward desired mechanical responses. Once trained, the model can efficiently generate candidate microstructures without relying on computationally expensive simulations. We demonstrate the model’s effectiveness by applying fiber configuration parameters as geometric constraints and using strain energy density to represent nonlinear material behavior. |
| Proceedings Inclusion? |
Planned: Post-meeting proceedings |