| About this Abstract |
| Meeting |
2021 TMS Annual Meeting & Exhibition
|
| Symposium
|
Phase Stability, Phase Transformations, and Reactive Phase Formation in Electronic Materials XX
|
| Presentation Title |
Using Machine Learning to Predict Hardness of Sn-based Alloys |
| Author(s) |
Yu-Chen Liu, Chih-han Yang, Hannah Carillo, Chuan-cheng Lin, Shih-kang Lin |
| On-Site Speaker (Planned) |
Yu-Chen Liu |
| Abstract Scope |
In this study, we employed the machine learning method with gaussian kernel ridge regression (GKRR) to develop a model for predicting the as-casted Sn-based alloy hardness. The 5-fold (leave-out (LO)-alloy-group) cross-validation (CV) test showed RMSE/σ of 0.43 ± 0.02 (0.38 ± 0.28) and R2 of 0.81 (0.79). The LO-element CV test suggested that our model was able to fully or partially extrapolate to some of the unknown elements space which were not shown in the data set, but was still difficult to do so without the information of some elements. The cross-plot analysis showed that our model was able to predict well within the composition range of the training data set. By using this model, we were able to design solders with hardness as high as 40.7 and as low as 2.7 Hv, and to explore the hardness change vs. types of doping elements in technology-relevant systems for future applications. |
| Proceedings Inclusion? |
Planned: |