Abstract The development of high entropy alloys (HEA) challenge the conventional alloy paradigm with equiatomic, multi-component alloy. However, HEA design is hindered by the vast compositional space possibility. Here we present a molecular dynamics (MD) simulation to study mechanical behaviour of Fe-Ni-Cr-Co-Al HEAs and machine learning (ML) models to study the mechanical behavior and the pattern from input parameter towards output parameters from the simulation data. The simulation is carried out with tensile tests of the HEAs under certain parameters, such as strain-rate, timestep, and temperature, while the Fe, Ni, and Al compositions are varied from 5% to 35%. The configuration allows one to study the effect of each elemental composition to the mechanical response upon tensile test. The stress-strain data generated within the simulation is utilized as the dataset for training the ML model, such as linear regression, neural networks, support vector machines, and k-nearest neighbors. The results shows that Al content is the primary element affect the tensile properties through phase partitioning (FCC dominant FCC+B2/BCC with HCP-like structure), with Ni and grain size providing secondary contributions and Fe showing minor incremental effect within the sampled window. The increase of Al content induces phase transition and decrease the tensile strength and vice versa, while decreasing grain size increase the strength, although Al content have more dominant effect. For the ML model evaluation, the neural network model has accurately predicted yield and ultimate tensile strengths with R² = 0. 99931 and RMSE = 0. 03440. This study has successfully demonstrate the practical application of molecular dynamics simulation coupled with machine learning to study the mechanical behavior of complex HEAs.
Zaenudin et al. (Mon,) studied this question.