|
发表于 2013-11-4 09:57:36
|
显示全部楼层
回复 52# lwd1981 的帖子
Given the complexity of "turbulence" and for the sake of a concreteness, the question of how to simulate turbulence could be framed more rigorously and restrictively. Suppose we restrict ourselves to the incompressible Navier-Stokes (NS) equations, then this is an easier question.
The incompressible NS equation is a nonlinear PDE. The strength of the nonlinearity determines the number of the degrees of freedom (DoF) to sufficiently describe the flow. Kolmogorov's theory leads to a heuristic estimate of the number of DoF to be Re^{9/4}, which turns out to be an upper bound on the sufficient (but not necessary) number of DoF for 3D (incompressible) turbulent flows. But, let's just use that as an estimate.
The number of DoF is spectral DoF, of course. The above estimate poses the limitation of direct numerical simulation (DNS). Any simulation which does not satisfy the number of DoF requirement cannot be a DNS, thus a modeling of some sort is explicitly or implicitly included in the simulation.
Let us now return to the original question of "mean-free-path" (MFP). It seems that the implied question is that which of the (viscous) boundary layer and the Knudsen layer is more important. THEORETICALLY, the Knudsen number is zero for the incompressible NS equation for which grad u = 0, thus the Mach number Ma is zero, and Kn = Ma / Re = 0. In reality, most likely this is applicable, thus the Knudsen layer should not be a factor in the context of the incompressible NS equations.
As for the COMPRESSIBLE flows with high Mach number, the situation changes, thus the above discussion is no longer valid.
[ 本帖最后由 luo@odu.edu 于 2013-11-4 10:05 编辑 ] |
|