About this Abstract |
Meeting |
6th International Congress on 3D Materials Science (3DMS 2022)
|
Symposium
|
6th International Congress on 3D Materials Science (3DMS 2022)
|
Presentation Title |
Deep Learning-Based 3D Damage Quantification for Natural Cellular Solids |
Author(s) |
Ziling Wu, Ting Yang, Ling Li, Yunhui Zhu |
On-Site Speaker (Planned) |
Ziling Wu |
Abstract Scope |
Quantitative descriptions of deformation processes in 3D are important for the understanding of the mechanical properties of structural materials. In particular, cellular solids represent a challenging group of lightweight complex structural materials. In this work, we present a deep learning-based computational framework for quantitative detection, registration, and analysis of different forms of damages during the entire deformation process collected from high-resolution synchrotron-based X-ray computed microtomography of a natural cellular solid, sea urchin spines. We developed two neural networks, CrackNet and DensifyNet in order to learn and detect the features of the destroyed struts, respectively. Our automatic detection method proves to be accurate and dramatically more efficient compared to manual labeling. The detected damage regions can be further registered to the cellular network to study the formation of damage and propagation patterns. This method could also be applied to other cellular solids’ damage characterization to investigate their deformation mechanisms. |
Proceedings Inclusion? |
Definite: Other |