High-performance simulations of colliding inertia-particles in isotropic turbulence

Ryo Onishi, Yuya Baba, Keiko Takahashi

Research output: Contribution to journalArticlepeer-review

Abstract

This study develops an efficient parallel simulation of colliding particles in stationary isotropic turbulence. Flow is computed with a fourth-order finite-difference method and particles are tracked with the Lagrangian method. Particle collisions are efficiently detected by the cell-index method, which is often used in molecular dynamics simulations. The developed code is written in Fortran 90 in conjunction with MPI library. Auto-parallelization directives are inserted in the code for shared-memory parallelization, making it possible to run the code in the so-called hybrid-parallelization mode. The code is designed to minimize the MPI communication, which results in a high parallel performance. The present simulation has been run on up to 512 3 grids with 10 million particles. The largest simulation has successfully obtained the collision frequencies for the Taylor-scale-based Reynolds number, Re γ, of 209, which is much larger than previously attained maximum Rs γ of 85 in particle collision simulations. The collision frequencies obtained in this study have shown that a collision frequency model underestimates the frequency in high Re γ flows.

Original languageEnglish
Pages (from-to)1879-1891
Number of pages13
JournalNihon Kikai Gakkai Ronbunshu, B Hen/Transactions of the Japan Society of Mechanical Engineers, Part B
Volume77
Issue number782
DOIs
Publication statusPublished - 2011
Externally publishedYes

Keywords

  • Cell-index method
  • High-performance computing
  • Isotropic turbulence
  • Multi-phase flow
  • Parallel computing
  • Particle collisions

ASJC Scopus subject areas

  • Condensed Matter Physics
  • Mechanical Engineering

Fingerprint

Dive into the research topics of 'High-performance simulations of colliding inertia-particles in isotropic turbulence'. Together they form a unique fingerprint.

Cite this