About this Abstract |
| Meeting |
MS&T25: Materials Science & Technology
|
| Symposium
|
Materials Processing and Fundamental Understanding Based on Machine Learning and Data Informatics
|
| Presentation Title |
B-3: Machine Learning of the Creep Life of Heat-Resistant Steel and Thermodynamical Analysis Using Generative AI |
| Author(s) |
Taehyeon Cho, Yoon Suk Choi, Wangrok Seok |
| On-Site Speaker (Planned) |
Taehyeon Cho |
| Abstract Scope |
Heat-resistant steels are mainly used for power plants, including pipes, turbines, and heat exchangers, where they face high temperatures and pressures. Consequently, creep resistance is a key property. In this study, we developed a forward machine learning protocol to predict the creep life of heat-resistant steels based on their composition and heat treatment conditions. The training process involved extracting datasets under specific test conditions, applying dimensionality reduction for alloy group clustering, and constructing a composition–creep life prediction model using an artificial neural network (ANN). A synthetic composition–creep life dataset was then generated with a diffusion model and validated using the trained ANN. To investigate the relationship between thermodynamic properties and creep life, we calculated thermodynamic parameters with Thermo-Calc for the synthetic compositions and analyzed their correlation with the predicted creep life. This approach provides insights into alloy design strategies for enhancing the long-term performance of heat-resistant steels. |