A machine learning model was developed to predict the oxidation resistance of Ti-V-Cr burn-resistant titanium alloy, and the natural logarithm of the parabolic oxidation rate constant (lnkp) was utilized as the model output. The results show that the two algorithms based on multiple learners, gradient boosting decision tree (GBDT) and eXtreme Gradient Boosting (XGBoost), show better performance. The coefficient of determination R2 of the models are 0.98 and the maximum error is 6.57 and 6.40%, respectively. The importance and interpretability of the input features were analyzed. The trend of the model analysis results was the same as that of the experimental conclusions, which further revealed the mechanism of the influence of element content and temperature changes on the oxidation resistance of Ti-V-Cr alloys and verified the effectiveness of the model. This study is of great significance for the discovery, prediction, and quantification of new high-temperature oxidation-resistant Ti-V-Cr alloys.
Read full abstract