About this Abstract |
Meeting |
MS&T23: Materials Science & Technology
|
Symposium
|
Materials Processing and Fundamental Understanding Based on Machine Learning and Data Informatics
|
Presentation Title |
Optimizing Heat Treatment Routes for Ni-based Alloys Using Monte Carlo Tree Search |
Author(s) |
Vickey Nandal, Sae Dieb, Dmitry S. Bulgarevich, Toshio Osada, Toshiyuki Koyama, Satoshi Minamoto, Masahiko Demura |
On-Site Speaker (Planned) |
Vickey Nandal |
Abstract Scope |
This work aims to design more flexible non-isothermal heat treatment (non-IHT) routes to achieve better 0.2% proof stress than the conventional isothermal heat treatment (IHT) in two-phase Ni-based alloys using a machine learning technique. In our previous study, we found 110 non-IHT routes out of 1620 that outperformed IHT benchmark, and we identified the common trend in the top 5 non-IHT routes from which we extracted the new idea of two-step heat treatment. In this study, we examined more non-IHTs other than the top 5 patterns, which still outperform the IHT benchmark value. Interestingly, we found entirely different characteristics from the top 5 non-IHT routes, although this exhibits a little lower performance than the top 5 non-IHTs. Finally, we summarize the classifications of the pattern that we found so far into the three categories. The insights gained from these findings may assist in the more flexible designing of non-IHT approaches. |