| About this Abstract |
| Meeting |
2024 TMS Annual Meeting & Exhibition
|
| Symposium
|
Novel Strategies for Rapid Acquisition and Processing of Large Datasets from Advanced Characterization Techniques
|
| Presentation Title |
Utilizing Advanced Computer Vision Techniques Based on Machine Learning and Artificial Neural Networks to Process Micrographs of Ni-base Superalloys |
| Author(s) |
Pascal Thome, Luis Arciniaga, Alexander Richter, Sammy Tin |
| On-Site Speaker (Planned) |
Pascal Thome |
| Abstract Scope |
Understanding the microstructure is key to the further development of high-performance Ni-base superalloys. Microscopy-based characterization techniques have greatly contributed to advances in materials science over the years. With the proliferation of more powerful microscopes, and resulting increased volumes of microstructure data, automatic image processing approaches become increasingly relevant. We demonstrate a procedure for evaluating a dataset of serial sectioned optical micrographs of dendritic microstructures of single crystal Ni-base Superalloys based on object detection and semantic segmentation via artificial neural networks. In order to create a digital twin, the three-dimensional data is transformed into a microstructure database using relational geometric ontology. Furthermore, we present another method that allows for efficient batch processing of combined EBSD and EDS scans of polycrystalline Ni-base Superalloys. By applying machine learning clustering algorithms on the EDS data, chemical heterogeneities can be detected automatically, as well as their influence on grain size evolution during manufacturing. |
| Proceedings Inclusion? |
Planned: |
| Keywords |
Characterization, Machine Learning, Computational Materials Science & Engineering |