A Gaussian mixture modelling approach to the data-driven estimation of atomistic support for continuum stress

Manfred Hannes Ulz, Sean James Moran

Research output: Contribution to journalArticlepeer-review

Abstract

Recent developments in multiscale modelling include the treatment of atomistic scale interactions via molecular dynamics simulations. The atomistic stress definition at a given continuum point contains a space-averaging volume over nearby atoms to provide an averaged macroscopic stress measure. Previous work on atomistic stress measures introduce the size of this volume as an a priori given parameter. In this contribution we let the atomistic data speak for itself by hypothesizing that the influence between atoms can be effectively estimated from their relative spatial position and stress. Atoms with highly similar spatial position and stress should therefore be contained within the same space-averaging volume. We motivate the application of Gaussian mixture modelling as a principled probabilistic means of estimating this similarity directly from the atomistic data. This model enables the discovery of homogeneous sub-populations of atoms in an entirely data-driven manner. The size of the space-averaging volume then follows naturally from the average maximum extent of the sub-populations. Furthermore, we demonstrate how the model can be used to compute the stress at arbitrary continuum points. Thorough evaluation is conducted on a numerical example of an edge dislocation in a single crystal. We find that our results are in excellent agreement with the corresponding analytical solution.
Original languageEnglish
Article number 065009
JournalModelling and Simulation in Materials Science and Engineering
Volume20
Issue number6
DOIs
Publication statusPublished - 2012

Fields of Expertise

  • Advanced Materials Science

Treatment code (Nähere Zuordnung)

  • Basic - Fundamental (Grundlagenforschung)
  • Theoretical

Fingerprint

Dive into the research topics of 'A Gaussian mixture modelling approach to the data-driven estimation of atomistic support for continuum stress'. Together they form a unique fingerprint.

Cite this