Machine Learning-Based Prediction of Complex Combination Phases in High-Entropy Alloys

High-entropy alloys (HEAs) have emerged as a novel class of materials that exhibit a wide range of desirable properties making them a focal point for research and potential applications across various industries. The complexity and variability in the compositions of HEAs pose a significant change in predicting their phases, which is crucial for determining their applicability and performance in specific applications. Accurate phase prediction is essential for determining the ideal combination of elements required to design HEAs with targeted properties. This study proposes a machine learning (ML) based approach to predict the phase structure of HEAs utilizing experimental data containing features derived from the chemical composition and corresponding phases. A Boolean vector technique was employed to represent the presence or absence of multiple phase combinations enhancing the model’s ability to accurately capture complex phase structures. Four robust ML algorithms consisting of support vector machine (SVM), k-nearest neighbors (KNN), random forest (RF) and neural network (NN), were employed to develop models capable of classifying the phases of HEAs. The performance of these models was rigorously evaluated through testing on unseen data samples. The findings revealed that both NN and KNN demonstrate superior performance achieving a remarkable test accuracy of 84.85%. This study underscores the potential of ML as an effective tool for predicting the phases of HEAs, offering a new avenue for more innovative approaches to material design and discovery in the future.