| About this Abstract |
| Meeting |
2022 TMS Annual Meeting & Exhibition
|
| Symposium
|
Environmental Degradation of Multiple Principal Component Materials
|
| Presentation Title |
NOW ON-DEMAND ONLY – Interpretable Machine Learning to Understand Corrosion in Complex Compositional Alloys |
| Author(s) |
Timothy Q. Hartnett, Angela Gerard, Prasanna Balachandran, John Scully |
| On-Site Speaker (Planned) |
Timothy Q. Hartnett |
| Abstract Scope |
Machine learning (ML) is rapidly becoming an important computational tool for exploring the structure-property relationships in complex compositional alloys (CCAs), including the high entropy alloys (HEAs). When studying the corrosive behavior of these materials, variability in processing and characterization leads to heterogenous datasets that adds an additional layer of complexity to training ML models. In addition to demonstrating the generalizability of machine learning models, it is critical to probe the learned models to glean insights into the source of model predictions. Here we use novel local interpretability techniques to explore the behavior of ML models trained to predict the passivation current density of CCAs. These techniques offer a detailed look into how a model thinks each training feature impacts the passivation current density. The results offer a new approach for understanding the complex behavior of heterogeneous systems using ML. |
| Proceedings Inclusion? |
Planned: |
| Keywords |
Machine Learning, Environmental Effects, High-Entropy Alloys |