Predictions of Lattice Parameters in NiTi High-Entropy Shape-Memory Alloys Using Different Machine Learning Models.

Area of Science:

• Materials Science
• Computational Materials Science
• Machine Learning Applications

Background:

• Shape-memory materials (SMMs) are crucial for high-temperature applications.
• Predicting the lattice parameters of the monoclinic B19' phase in SMMs is vital for their performance.
• Existing methods for lattice parameter prediction can be computationally intensive or less accurate.

Purpose of the Study:

• To evaluate the efficacy of three machine learning models—linear regression (LR), random forest (RF), and support vector regression (SVR)—in predicting lattice parameters.
• To compare the predictive accuracy of these models across two distinct datasets: ZrO2-based shape-memory ceramics (SMCs) and NiTi-based high-entropy shape-memory alloys (HESMAs).
• To explore a combined approach of ML models for enhanced prediction accuracy.

Main Methods:

• Application of linear regression (LR), random forest (RF), and support vector regression (SVR) models.
• Training and validation using two distinct datasets of ZrO2-based SMCs and NiTi-based HESMAs.
• Comparative analysis of model performance based on prediction accuracy for lattice parameters (a_c, a_m, b_m, c_m, and β_m).

Main Results:

• Linear regression (LR) demonstrated the highest accuracy for predicting a_c, a_m, b_m, and c_m in NiTi-based HESMAs.
• Random forest (RF) excelled in predicting β_m for both ZrO2-based SMCs and NiTi-based HESMAs.
• Support vector regression (SVR) exhibited the largest deviations between predicted and actual lattice parameter values across both datasets.

Conclusions:

• Machine learning models, particularly LR and RF, offer a viable and accurate approach for predicting lattice parameters in shape-memory materials.
• A combined RF and LR approach can further enhance the accuracy of lattice parameter predictions for martensitic phases.
• These findings support the development of advanced shape-memory materials for stable high-temperature applications through accurate computational predictions.