"Discovering Structure-Property Relationships in Metallic Materials through Latent Space Mapping of Plasticity and Microstructure"
M. Calvat, C. Bean, D. Anjaria, J.C. Stinville
Understanding and predicting the mechanical properties of metallic materials relies on insights into how microstructural features at multiple scales influence plastic flow. These properties are often governed by complex, heterogeneous interactions within the microstructure, spanning from the nanoscale to the macroscale. Capturing these interactions is critical to accelerating materials discovery and design.
Experimental advances have equipped researchers with high-resolution tools capable of mapping these multiscale heterogeneities. Techniques like Electron Backscatter Diffraction (EBSD) provide rapid, detailed characterization of microstructural variations, while High-Resolution Digital Image Correlation (HR-DIC) allows nanometer-scale tracking of plastic deformation over large fields of view during mechanical loading. However, linking these vast datasets to macroscopic properties requires innovative analytical frameworks that can efficiently extract underlying structure-property relationships.
In this work, we develop a transformative methodology to encode and map microstructural and plasticity information into a latent feature space, compressing the data to its most impactful features while maintaining spatial information. This approach identifies specific microstructural and plasticity heterogeneities that drive mechanical behavior, ultimately enabling the rapid prediction of mechanical properties. By unveiling these critical structure-property relationships, our study paves the way for accelerated discovery and optimization of metallic materials with tailored mechanical properties.