Stochastic Modeling of Anisotropic Strength Surfaces from Atomistic Simulations
Please login to view abstract download link
In this talk, we present a modeling framework to infer and statistically characterize anisotropic strength surfaces directly from molecular dynamics simulations. Failure data from large-scale tensile loading simulations across stress ratios and orientations are first encoded in a low-dimensional, angle-dependent parametric strength surface. The approach then uses dimensionality reduction and probabilistic modeling to quantify defect-induced strength variability despite limited atomistic sampling. Principal Component Analysis and a Gaussian mixture model are used to sample physically admissible strength surfaces and estimate confidence intervals in stress space. The framework is finally applied to pristine and defective graphene. Our results reveal strong fracture anisotropy, vacancy-induced weakening, and variability governed primarily by defect density.
