Reynolds Number Dependence of the Statistics in High Resolution DNS of Turbulence with up to Re~1130

Yukio Kandea

Department of Computational Science and Engineering
Graduate School of Engineering, Nagoya University, Japan

Abstract-
High resolution direct numerical simulations (DNS) of incompressible homogeneous turbulence in a periodic box based on a spectral method with the number of the grid points up to 40963 and Taylor microscale Reynolds number about 1130 were performed on the Earth Simulator, with sustained performance up to 16.4 TFlops. The talk will present some of the results of the analysis of the DNS data, with special emphasis on the Reynolds number dependence of the statistics including (i) the normalized mean energy dissipation rate, (ii) the energy spectrum, and (iii) the statistics of velocity gradients. Visualization of the flow field obtained by the DNS provides us with some intuitive ideas on the structure of the turbulence. It suggests that the velocity field consists of clusters of small eddies, and that the structure of these clusters at the scale of inertial subrange is distinctively different from those of individual small eddies. Analysis has also been made on (i) the statistics of the energy transfer, (ii) eddy viscosity, and (iii) quantitative examination of Kolmogorov's log-normal hypothesis.


GALCIT Home Page
2004-2005 Fluids Seminar Page


Maintained by: James Faddy
EMail: James Faddy
Last modified: September 21, 2004