Creep Life Prediction for Superalloys Using Gradient Boosting Decision Trees

Ensuring the reliability of superalloys in high-temperature environments hinges on accurate creep life prediction. In this paper, we introduce a Gradient Boosting Decision Trees (GBDT) method aimed at improving both the predictive accuracy and computational efficiency of existing models. We benchmark the GBDT model against the Random Forest Regression (RFR) approach, utilizing a dataset that includes features such as alloy composition, temperature, and stress. Our findings reveal that GBDT outperforms RFR in terms of both accuracy and processing speed. Furthermore, SHapley Additive exPlanations (SHAP) analysis is used to enhance the interpretability of the model, highlighting the most influential factors in predicting creep life. Overall, the GBDT model emerges as a more effective solution for creep life prediction in superalloys,