Composition optimization of AlFeCuSiMg alloys based on elastic modules: A combination method of machine learning and molecular dynamics simulation

High entropy alloys (HEAs) has attracted much attention owning to its excellent mechanical properties. However, the application of these HEAs is limited by the uncertain element ratio, high cost and low efficiency preparation methods. In this work, we optimize the composition of AlFeCuSiMg HEAs based on elastic modulus as a prediction index through machine learning (ML) and molecular dynamics (MDs) simulation. The training sets and test sets are prepared by MDs. By comparing the average R2 and RMSE values of different ML models, we selected support vector regression (SVR) model and random forest (RF) regression model to predict the elastic modulus of AlFeCuSiMg HEAs. The prediction results are consistent with MDs and confirm the effectiveness of this method for component design, which provided a guidance for the design of HEAs and accelerate the development of new alloy materials.