when I read the papers of DNS and LES, such as, 1. "Zonal Embedded Grids for Numerical Simulations of Wall-Bounded Turbulent Flows". A.G. Kravchenko. 2. Kim John, " Turbulence statistics in fully developed channel flow at low Reynolds number". 3. Parviz Moin " Numerical investigation of turbulent channel flow ".

I think the channel half width (delta) and boundary layer thickness used there should be a constant. But I really cannot figure out what value it is.

Please help me if you have any points about it.

Many thanks.

(1). In a fully-developed channel flow, you don't have a boundary layer at all. (2). The half width or the channel size is no longer a parameter. Instead, you are talking about the Reynolds number. (3). I am no expert in DNS or LES, but in either case, the difference is in the governing equations. DNS is transient Navier-Stokes equations, LES is transient Navier-Stokes equations + sub-scale modeling. And RANS is Reynolds average of transient Navier-Stokes equations + global Reynolds stress modeling. (4). I think, LES is a troublesome concept.

Most LES of channel flows are temporal simulations, i.e., they use periodic boundary conditions for velocities in both axial and spanwise directions. The pressure BC is step periodic and the pressure difference drives the flow.

After you attain a statistically stationary state, you have fully developed turbulent flow and there is no boundary layer thickness (the top and the bottom boundary layers merge to produce the fully developed flow profile). The only parameter in the simulation is Re. You can set any value for the half-width as long as the Re based on half-width, centerline mean velocity and kinematic viscosity gives you the Re you need. Also, the axial length of the domain should atleast be 4 to 5 times the channel width. Also not that sometimes the Re quoted is the friction Reynolds number based on friction velocity rather than bulk velocity.

I set x direction Lx=12; boundary layer thickness delta=2; To follow exactly the Reynolds number, I think, it should be okay for me gain simularity.

Thanks for you two valuable suggestions.

peter