Predicting grain growth kinetic in steels using machine learning and XAI for mechanical properties

Area of Science:

• Materials Science
• Metallurgy
• Computational Materials Science

Background:

• Optimizing steel mechanical properties requires understanding grain growth kinetics during thermomechanical processing.
• Traditional models struggle with complex interactions between composition and processing parameters.

Purpose of the Study:

• Develop a novel machine learning (ML) framework to predict austenitic grain growth behavior.
• Utilize chemical composition and process conditions for accurate predictions.

Main Methods:

• A comprehensive dataset of 1039 experimentally validated samples was used.
• XGBoost algorithm was employed, achieving high R2 after hyperparameter optimization.
• Feature selection and SHAP analyses identified key parameters influencing grain growth.

Main Results:

• The XGBoost model achieved an R2 of 0.9728, demonstrating exceptional predictive capability.
• Temperature, initial grain size, and holding time were identified as dominant factors.
• Experimental validation on 316L stainless steel showed strong agreement between predicted and measured grain sizes.

Conclusions:

• This study presents the first integrated ML and experimental approach for predicting steel grain growth kinetics.
• The developed framework offers a powerful tool for alloy design and process optimization.
• Future work will expand the framework to include more variables and alloy systems.