Perpustakaan Digital ITB

High-entropy alloys (HEAs), particularly FeNiCrCoAl, have attracted significant attention due to their exceptional mechanical performance across a wide temperature range. However, optimising the composition to achieve desirable mechanical characteristics remains challenging due to the complex interactions between elements. This study aims to investigate the effect of compositional and temperature variations on the tensile and elastic properties—specifically, ultimate tensile strength (UTS), Young’s modulus (E), and hardness—using computational approaches. The scope includes simulations across composition ranges of 5–35 at% for each element and temperature variations from 100 K to 1200 K. Molecular dynamics (MD) simulations were employed to conduct uniaxial tensile tests and calculate elastic constants (C11, C12, and C44), from which UTS, E, and hardness values were derived. The simulations were carried out using the LAMMPS software, covering a total of 1812 datasets with different compositions across 12 temperature points each (100 K increment). The resulting dataset was used to train machine learning (ML) models—namely, random forest (RF), artificial neural networks (ANNs), and support vector machine (SVM)—to predict mechanical properties and evaluate feature importance. The models were assessed based on evaluation metrics and supported by correlation analysis. The results demonstrated that the mechanical properties of FeNiCrCoAl deteriorate with increasing temperature, with each elemental composition showing distinct influences. Fe, Ni, and Co primarily affect tensile strength, Al and Fe dominate Young's modulus, and Cr and Al impact hardness most significantly. Among the evaluated ML models, random forest achieved optimal performance for UTS prediction with an R² score of 0.973 and RMSE of 0.23. At the same time, feature importance and Spearman's rank correlation analysis confirmed the dominant elemental contributions to UTS, Young's modulus, and hardness. By integrating molecular dynamics and machine learning approaches, this work established a comprehensive framework for mechanical property design in FeNiCrCoAl high-entropy alloys, identifying optimal compositions from both mechanical and statistical perspectives: for improved UTS (Cr 5–9 at.% and Al 5–8 at.%), enhanced Young's modulus (Cr 5–8 at.% and Co 5–8 at.%), with alternative formulations via (FeNiAl)100-x-yCoxCry, and increased hardness (Cr and Al adjusted to 15–35 at.%).