Publications

Abstract : The characterization of nano-scale materials in the lab setting requires a huge cost, precision and time. The molecular simulations rule out the cost and precision but still these simulations are computationally expensive and intensive. In this regard, we present a support vector machine (SVM)-based molecular dynamics simulation of monolayer graphene to predict its temperature and strain rate-dependent fracture strength. The design of experiments algorithm for full factorial design with six levels of variation in both input settings (temperature and strain rate) is used to create the sample space for training the machine learning model. The prediction capability of machine learning model is further tested by utilizing separate samples generated with SOBOL sequence sampling technique. The accuracy of prediction is assessed by observing correlation coefficient (R2) and error analysis (probability density function (PDF) plots). To construct the model, temperature and strain rate are used as the input features and the desired response quantity is fracture strength of graphene.