About this Abstract |
Meeting |
2025 TMS Annual Meeting & Exhibition
|
Symposium
|
Advanced Characterization Techniques for Quantifying and Modeling Deformation
|
Presentation Title |
How to Make Material Textures Amenable to Analysis by Neural Networks |
Author(s) |
Marc J. De Graef |
On-Site Speaker (Planned) |
Marc J. De Graef |
Abstract Scope |
Orientation Distribution Functions (ODFs) represent the volume fraction of material that has a certain orientation with respect to an external reference frame. They are typically defined using Bunge Euler angles but can be expressed using any other orientation representation, in particular quaternions and related neo-Eulerian representations. In this contribution we will show that the ODF can be reduced to a single RGB image using a projection from the unit quaternion sphere to the Clifford Torus and then to the 2D Square Torus (ST) map without loss of information. After describing the projection process, we will show how the ST representation can be used to generate training images for neural network applications, thus making material textures amenable to analysis by neural networks. |
Proceedings Inclusion? |
Planned: |
Keywords |
Modeling and Simulation, Mechanical Properties, Characterization |