A new adaptive modeling strategy based on optimal control for atomic-to-continuum coupling simulations

H. Ben Dhia, Ludovic Chamoin, J. Tinsley Oden, Serge Prudhomme*

*Corresponding author for this work

Research output: Contribution to journalArticlepeer-review

15 Scopus citations


A new adaptive method based on an optimal control approach is proposed for adaptive modeling of an atomic-to-continuum coupling method constructed from the Arlequin framework. The coupling method provides for an approximation of the solution to a fully atomic model in which errors may arise due to the misplacement of the overlap region. The objective is thus to determine, a posteriori, the best position of this overlap region. More precisely, the problem is to find the optimal size of the atomic region that one needs to consider in the coupled formulation in order to accurately estimate predefined quantities of interest. In this new adaptive process, the position of the overlapping domain between the two models is conveniently parameterized and iteratively determined by searching for the optimal parameters. The performance of the method is demonstrated on one-dimensional and two-dimensional test problems. In particular, it is observed that the approach yields better convergence with respect to quantities of interest when compared to a classical adaptive approach based on a posteriori estimates of the modeling error.

Original languageEnglish (US)
Pages (from-to)2675-2696
Number of pages22
JournalComputer Methods in Applied Mechanics and Engineering
Issue number37-40
StatePublished - Sep 1 2011
Externally publishedYes


  • Adaptive modeling
  • Arlequin method
  • Goal-oriented error estimation
  • Modeling error
  • Optimal control

ASJC Scopus subject areas

  • Computational Mechanics
  • Mechanics of Materials
  • Mechanical Engineering
  • General Physics and Astronomy
  • Computer Science Applications


Dive into the research topics of 'A new adaptive modeling strategy based on optimal control for atomic-to-continuum coupling simulations'. Together they form a unique fingerprint.

Cite this