PREDICTION OF PROPERTIES OF HIGH-ENTROPY ALLOYS USING NEURAL NETWORKS

10.25712/ASTU.1811-1416.2026.01.007

Abstract

The article discusses the application of neural networks for predicting the properties of high-entropy alloys (HEAs) characterized by complex chemical composition and multiphase structure. Traditional prediction methods (phenomenological rules, CALPHAD, and molecular dynamics) often require significant computational resources and have limited accuracy. It is proposed to use neural networks capable of identifying complex dependencies in data to solve this problem. The methodology for developing the model is described, including data preprocessing, network architecture with fully connected layers and Dropout, as well as the training and evaluation process. The results show high accuracy of predictions, confirmed by the mean absolute error (MAE) for the following properties: microhardness, Young's modulus, yield strength, and tensile strength. The study demonstrates the potential of neural networks in accelerating the development of new HEAs with desired properties and opens up prospects for further research in materials science.