AMR vs High Order Schemes

  • Authors:

  • Leland Jameson

  • Affiliations:

  • Lawrence Livermore National Laboratory, University of California, P.O. Box 808, MS L-039, Livermore, California 94551. jameson3@llnl.gov

  • Venue:
  • Journal of Scientific Computing
  • Year:
  • 2003

Quantified Score

Hi-index 0.02

Visualization

Abstract

Adaptive Mesh Refinement (AMR) schemes are generally considered promising because of the ability of the scheme to place grid points or computational degrees of freedom at the location in the flow where truncation errors are unacceptably large. For a given order, AMR schemes can reduce work. However, for the computation of turbulent or non-turbulent mixing when compared to high order non-adaptive methods, traditional 2nd order AMR schemes are computationally more expensive. We give precise estimates of work and restrictions on the size of the small scale grid and show that the requirements on the AMR scheme to be cheaper than a high order scheme are unrealistic for most computational scenarios.