| About this Abstract |
| Meeting |
2025 TMS Annual Meeting & Exhibition
|
| Symposium
|
Novel Strategies for Rapid Acquisition and Processing of Large Datasets from Advanced Characterization Techniques
|
| Presentation Title |
Microstructure Informatics: Automated Microstructure Characterization and Neural Network Based Modeling of Processing-Structure-Property Relations |
| Author(s) |
Pascal Thome, Luis Arciniaga, Michael Edward Madigan, Sammy Tin |
| On-Site Speaker (Planned) |
Pascal Thome |
| Abstract Scope |
Materials science research projects often aim to understand the relationship between process parameters, microstructure, and macroscale properties of materials. While it is relatively straightforward to determine process parameters and properties through standardized measurement methods, it has always been challenging to fully represent the microstructure quantitatively. Through modern approaches in the field of computer vision technology, it is now possible to detect microstructural entities on micrographs in an efficient and accurate manner by employing advanced image processing filters and clustering algorithms. By utilizing these detected entities, translation-invariant quantitative metrics can be obtained, serving as a "digital fingerprint" of the microstructure. These fingerprints can then be used as target or feature vectors to establish correlative relationships between process parameters and the resulting microstructures. Taking advantage of such a digital infrastructure, we describe the study of polycrystalline Ni-based superalloys synthesized by both powder metallurgy as well as additive manufacturing. |
| Proceedings Inclusion? |
Planned: |
| Keywords |
Characterization, Additive Manufacturing, High-Temperature Materials |