Exact non-overlapping domain decomposition for near-wall turbulence modeling

Sergey Utyuzhnikov, Chen Wang

Research output: Contribution to journalArticlepeer-review

4 Citations (Scopus)

Abstract

The near-wall non-overlapping domain decomposition (NDD) method proved to be very efficient for turbulence modeling. It represents a trade-off between accuracy and computational time. Meanwhile, the current NDD presumes the use of a simplified model near the wall. Although the approximate NDD method has successfully been applied to the Reynolds-averaged Navier–Stokes (RANS) models, it has some limitations regarding to the accuracy. In this paper, for the first time an exact NDD method is proposed, which can be applied to RANS equations without any simplification. Similar to the approximate NDD, the approach is based on a Robin-to-Dirichlet algorithm. Its convergence is proven for a multidimensional linear elliptic equation. In the new algorithm, the approximate NDD is effectively used to construct a preconditioner which immediately takes into account the near-wall flow physics. The algorithm is implemented in a low-Reynolds-number k−ε model to simulate a one-dimensional channel flow. A comparison against the approximate NDD method is provided. The results demonstrate that the approximate NDD approach performs not very well in predicting the turbulent kinetic energy and turbulent dissipation. As a contrast, the exact NDD approach can provide practically accurate prediction of the velocity field regardless of the location of the interface boundary while reducing the original computational time by one order of magnitude.

Original languageEnglish
Pages (from-to)283-291
Number of pages9
JournalComputers and Fluids
Volume181
DOIs
Publication statusPublished - 15 Mar 2019
Externally publishedYes

Keywords

  • Interface boundary condition
  • Low-Reynolds-number model
  • Near-wall turbulence modeling
  • Non-overlapping domain decomposition

Fingerprint

Dive into the research topics of 'Exact non-overlapping domain decomposition for near-wall turbulence modeling'. Together they form a unique fingerprint.

Cite this