| About this Abstract |
| Meeting |
MS&T23: Materials Science & Technology
|
| Symposium
|
Materials Informatics for Images and Multi-dimensional Datasets
|
| Presentation Title |
Predicting the Occurrence and Mechanism of Liquid Metal Embrittlement Using Machine Learning |
| Author(s) |
Benjamin A. Begley, Justin E Norkett, Cameron Frampton, Victoria M Miller |
| On-Site Speaker (Planned) |
Benjamin A. Begley |
| Abstract Scope |
Predicting whether liquid metal embrittlement (LME) will occur between a given pair of liquid and solid metals is key to enabling the use of liquid metals in a wide variety of applications. Most previous attempts have considered LME binary: LME or no LME. Recent developments have separated liquid metal embrittlement into multiple mechanisms; in this work, data was extracted from over 1000 publications spanning over 100 years of LME research, separated by mechanism, then several machine learning classification techniques, including k-nearest neighbors (KNN) classifiers and decision tree classifiers, were used to develop a model that predicts the occurrence and mechanism of LME. Specifically, a KNN classifier using a 1-vs-1 approach for predicting each LME mechanism is more than 80% accurate, surpassing predictive models in the literature. Additionally, a 1-vs-1 decision tree model is used to investigate the features of the liquid-solid pair that are most important for predicting LME.
|