| About this Abstract |
| Meeting |
2020 TMS Annual Meeting & Exhibition
|
| Symposium
|
Materials Processing Fundamentals
|
| Presentation Title |
M-71: Machine Learning Approaches for Prediction Mechanical Properties of Austenitic Stainless Steels |
| Author(s) |
Liping Yang, Sen Liu |
| On-Site Speaker (Planned) |
Liping Yang |
| Abstract Scope |
Austenite Stainless steels (ASSs) has been increasingly used in architectural and structural applications since its high corrosion resistance, aesthetic appearance and ease of maintenance. The correlations between functional properties of ASSs and compositions and process parameters are complex and time-consuming to explore. This paper presents mechanical properties data of various compositions of ASSs and processing parameters. It applies machine learning (ML) to ultimate tensile strength (UTS) and elongation datasets for establishing the composition-processing-properties relationships. The importance and correlations of chemical compositions and processing features were analyzed based on mutual information (MI) score and Pearson correlation matrix. The overfitting of training data was prevented using bias-variance trade-off techniques. Results of ML model can achieve prediction accuracy with R2 values over 0.99 for UTS and 0.86 for elongation. The developed model can effectively guide practitioner to prepare most promising chemical compositions, preprocessing and heat treatments for developments of high-performance ASSs. |
| Proceedings Inclusion? |
Planned: |