| About this Abstract |
| Meeting |
TMS Specialty Congress 2025
|
| Symposium
|
The 7th International Congress on 3D Materials Science (3DMS 2025)
|
| Presentation Title |
Unsupervised Learning for Structure Detection in Plastically Deformed Crystals |
| Author(s) |
Armand Barbot, Riccardo Gatti |
| On-Site Speaker (Planned) |
Armand Barbot |
| Abstract Scope |
Molecular Dynamics is a powerful method allowing to simulate different materials at the particle scale such as glassy materials or metallic nanocrystals. To determine the local structure at the particle-scale, several approaches were developed, mainly relying on local order parameters to describe the surrounding environment of each particle. However, they are mostly relying on hand chosen criteria and thus only works for already known structures.
In this study, we present an unsupervised learning method to automatically study and detect the different substructures appearing at the atomistic scale within a crystal under plastic deformation. This approach combines autoencoder, clustering and classification methods.
By applying our method on a Nickel FCC nanocrystal plastically deformed under uniaxial compression, we were able to detect more sub-structures associated with plasticity and with a higher degree of precision than traditional hand-made criteria. This study was published on Computational Materials Science. |
| Proceedings Inclusion? |
Undecided |