Enhancing phase characterization of AlCuCrFeNi high entropy alloys using hybrid machine learning models: A comprehensive XRD analysis

Abstract

High-Entropy Alloys (HEAs) offer exceptional mechanical and thermal properties, driving advancements in aerospace, energy, and biomedical applications. However, their complex phase diagrams present challenges for accurate phase prediction. This study introduces the Tree-Neural Ensemble Classifier (TNEC), a hybrid model integrating tree-based models and neural networks within a boosting framework to enhance phase classification accuracy. Experimental data from X-ray Diffraction (XRD) analysis of AlCuCrFeNi HEAs, subjected to heat treatments at 800 °C, 950 °C, and 1100 °C, alongside untreated samples, were used for model training. Preprocessing techniques, including noise reduction and feature extraction, ensured high-quality datasets. TNEC achieved an accuracy of 92 %, significantly outperforming Random Forest (RF) at 85 %, Support Vector Machine (SVM) at 80 %, and state-of-the-art Gradient Boosting and XGBoost models 89 %. Principal Component Analysis (PCA) confirmed structural transformations induced by temperature variations. These results highlight TNEC's capability to accurately predict phase compositions and structural transitions, providing a significant step toward real-time phase optimization in HEAs, accelerating materials discovery, and enhancing manufacturing efficiency.