| About this Abstract |
| Meeting |
2023 TMS Annual Meeting & Exhibition
|
| Symposium
|
Algorithm Development in Materials Science and Engineering
|
| Presentation Title |
Characterizing Microstructure Evolution in Latent Space for Machine Learning Applications |
| Author(s) |
Saaketh Desai, Ankit Shrivastava, Marta D'Elia, Habib Najm, Remi Dingreville |
| On-Site Speaker (Planned) |
Ankit Shrivastava |
| Abstract Scope |
Characterizing and quantifying microstructure evolution is critical to forming quantitative relationships between process conditions, resulting microstructure, and observed properties. Machine-learning methods such as recurrent networks can accelerate the development of these relationships by accelerating materials simulations, while techniques such as reinforcement/active learning can help improve representations and target specific microstructures/properties. However, these methods rely on the non-trivial task of identifying low-dimensional microstructural fingerprints that effectively relate process conditions to properties. In this work, we survey and discuss the ability of various linear/non-linear dimensionality reduction methods such as Principal Component Analysis, Karhunen Loeve Expansion, autoencoders/variational autoencoders, and diffusion maps to quantify and characterize the learned latent space microstructural representations and their time evolution. We target microstructure evolution problems such as spinodal decomposition, thin film deposition, and grain growth. This work paves the way to identify representation schemes that handle a variety of microstructural features across length scales for various machine-learning applications. |
| Proceedings Inclusion? |
Planned: |
| Keywords |
Machine Learning, Computational Materials Science & Engineering, Modeling and Simulation |