Using Machine Learning Approaches to Predict Atomic-Scale Glass Failure in Environmental Conditions

Silicate glasses are widely applied in fields including medicine, optics, electronics, telecommunication and energy. These materials often fail due to their inherently brittle characteristics, while the relationships between atomic structure and the fracture nucleation process remain incompletely understood. Our team used classical ReaxFF molecular dynamics simulations to model fracture nucleation of silica-based glasses in aqueous environments. Developing structure-property relationships using this simulation method is computationally expensive, limiting the number of samples and making interpretation challenging. This project aims to use a recurrent neural network with the long-short term memory mechanism to learn the dynamic processes behind fracture nucleation and predict likely locations of fracture nuclei. We interpret the learned model to glean new physical insight into how local structure influences failure in this critical material.
Sandia National Laboratories is a multi-mission laboratory managed and operated by National Technology and Engineering Solutions of Sandia, LLC., a wholly owned subsidiary of Honeywell International, Inc., for the U.S. Department of Energy’s National Nuclear Security Administration under contract DE-NA0003525.
SAND2020-11508 A