Multi-objective optimization and machine learning-based prediction of tensile properties of an armchair graphene sheet

Abstract

Graphene is a widely used nanoparticle in different industries, especially in nanocomposite applications. Prediction of its properties is of great importance for engineers. Therefore, a comprehensive study is first performed in this study to investigate the effect of temperature, strain rate, number of layers, and dimensions on the tensile properties of graphene nanosheets including elastic modulus (E), yield strength (YS), and ultimate tensile strength (UTS) using molecular dynamics simulation. Then, E, YS, and UTS are simultaneously optimized via response surface methodology. To determine the simplest and the most accurate machine learning algorithm for the prediction of tensile properties, three algorithms of Decision Tree (DT), Random Forest (RF), and Gradient Boosted Tree (GBT) are compared and the GBT algorithm is introduced as the best one. Furthermore, the architecture of each algorithm was optimized via the Taguchi design of experiment method to enhance the prediction accuracy. The DT algorithm with a maximum depth of 8 was obtained as the most accurate one.