A machine-learning-enhanced hierarchical multiscale method for bridging from molecular dynamics to continua

In the community of computational materials science, one of the challenges in hierarchical multiscale modeling is information-passing from one scale to another, especially from the molecular model to the continuum model. A machine-learning-enhanced approach, proposed in this paper, provides an alternative solution. In the developed hierarchical multiscale method, molecular dynamics simulations in the molecular model are conducted first to generate a dataset, which represents physical phenomena at the nanoscale. The dataset is then used to train a material failure/defect classification model and stress regression models. Finally, the well-trained models are implemented in the continuum model to study the mechanical behaviors of materials at the macroscale. Multiscale modeling and simulation of a molecule chain and an aluminum crystalline solid are presented as the applications of the proposed method. In addition to support vector machines, extreme learning machines with single-layer neural networks are employed due to their computational efficiency.