A Stacking-Enhanced Support Vector Regression Model for Predicting the Hot Deformation Flow Stress of TC18 Alloy

Area of Science:

• Materials Science
• Mechanical Engineering
• Computational Modeling

Background:

• Understanding the hot deformation behavior of TC18 alloy is crucial for optimizing manufacturing processes.
• Accurate constitutive models are needed to predict material response under varying conditions.

Purpose of the Study:

• To develop and validate a highly accurate predictive model for the hot deformation behavior of TC18 alloy.
• To systematically optimize a Stacking-Support Vector Regression (SVR) model using experimental data.

Main Methods:

• Experimental investigation of TC18 alloy hot deformation at temperatures of 720-840 °C and strain rates of 0.001-1 s^-1.
• Development and optimization of a Stacking-SVR model, including kernel selection and hyperparameter tuning via grid search and cross-validation.
• Comparative analysis against other models like Poly-SVR, Arrhenius, XGBoost, and MLP.

Main Results:

• The optimized Stacking-SVR model achieved high accuracy, with R² values of 0.9973 (training) and 0.9982 (test).
• The model demonstrated strong generalization ability, with 98.7% (training) and 94.83% (test) of samples having relative errors within 5%.
• The polynomial kernel function and specific hyperparameters (C=1000, coef0=1, d=5, ε=1, γ=1) were identified as optimal.

Conclusions:

• The proposed Stacking-SVR framework provides a robust and accurate method for modeling TC18 alloy's hot deformation behavior.
• The model exhibits excellent predictive accuracy and generalization capabilities, suitable for practical applications.
• This systematic optimization approach enhances the reliability of material constitutive modeling.