Machine learning assisted prediction of creep life in Ni-based alloys: An ensemble-based stacking regressor approach

Creep deformation beyond the design tolerance is a major cause of material fracture in the hot end components of aviation engines, which limits their service life. So, creep-oriented alloy design has always been crucial to prevent the material fracture caused by creep failure. But creep testing is time-consuming and costly, limiting the use of trial-and-error approaches for alloy design. To address this issue, this study establishes a Machine Learning (ML) based model known as Hybrid Ensemble Regressor (HER) which employs a stacking ensemble architecture combining Decision Tree (DT), Light Gradient Boosting Machine (LightGBM) and Gaussian Process Regressor (GPR) as base models. The predictions of these three base models are then fed into a linear regression meta-model. This approach enhances the creep rupture life prediction accuracy in high-performance alloys by leveraging the strength of each individual model, thus reducing overfitting and improving generalization. This yielded coefficient of determination ( ), Mean Absolute Percentage Error (MAPE) and Symmetric Mean Absolute Percentage Error (SMAPE) values of 0.980, 0.319 and 0.473 respectively which were best when compared among other machine learning models such as GPR, Gradient Boosting Machine (GBM), LightGBM, Support Vector Regressor (SVR), Ridge, Lasso, Extreme Gradient Boosting (XGBoost), Support Vector Machine (SVM), K-Nearest Neighbor (KNN), DT, Random Forest (RF) and Artificial Neural Network (ANN), thus demonstrating the significance of HER model. The HER model’s R² value is 0.980, close to 1 represents the accuracy of the model outperforming previous researches highlighting its relevance.