Molecular dynamics simulation and machine learning-based analysis for predicting tensile properties of high-entropy FeNiCrCoCu alloys

High entropy alloys (HEAs) attract many researchers due to their unique and desirable properties in comparison to conventional alloys, and their potential for advanced applications. Because of the complexity of designing HEAs, several attempts have been conducted to integrate experimental and computational studies with machine learning (ML) algorithms to predict their mechanical properties. Yet, few studies have considered a set of input parameters including atomic concentrations, grain size, operating temperature, and strain rate. Therefore, this study considers these combined predictors to forecast the tensile properties of FeNiCrCoCu HEAs, including Young's modulus, yield strength, and ultimate tensile strength based on molecular dynamics (MD) and ML algorithms. 918 datasets of polycrystalline HEAs were generated by MD simulations. Some of the MD datasets were selected as representative samples and assessed by checking the isotropy of mechanical properties. Also, the MD simulations provided data that reasonably agreed with previously published results. All the generated datasets were used afterward to train Artificial neural networks (ANN), support vector machine, and Gaussian process regression models. The proposed ANN models revealed the most accurate predictions among the other ML models, and their performances were evaluated on new datasets containing different predictor variables' values that were not used to build the models. It was found that the ANN models were most sensitive to the strain rate predictor variable. The proposed ANN models can assist in guiding the experimental work to optimize the search for HEAs with desired tensile properties.