We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:


Current browse context:


Change to browse by:

References & Citations


(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Condensed Matter > Materials Science

Title: Accelerating discrete dislocation dynamics simulations with graph neural networks

Abstract: Discrete dislocation dynamics (DDD) is a widely employed computational method to study plasticity at the mesoscale that connects the motion of dislocation lines to the macroscopic response of crystalline materials. However, the computational cost of DDD simulations remains a bottleneck that limits its range of applicability. Here, we introduce a new DDD-GNN framework in which the expensive time-integration of dislocation motion is entirely substituted by a graph neural network (GNN) model trained on DDD trajectories. As a first application, we demonstrate the feasibility and potential of our method on a simple yet relevant model of a dislocation line gliding through a forest of obstacles. We show that the DDD-GNN model is stable and reproduces very well unseen ground-truth DDD simulation responses for a range of straining rates and obstacle densities, without the need to explicitly compute nodal forces or dislocation mobilities during time-integration. Our approach opens new promising avenues to accelerate DDD simulations and to incorporate more complex dislocation motion behaviors.
Subjects: Materials Science (cond-mat.mtrl-sci); Machine Learning (cs.LG)
Cite as: arXiv:2208.03296 [cond-mat.mtrl-sci]
  (or arXiv:2208.03296v1 [cond-mat.mtrl-sci] for this version)

Submission history

From: Nicolas Bertin [view email]
[v1] Fri, 5 Aug 2022 17:35:07 GMT (675kb,D)

Link back to: arXiv, form interface, contact.