Thermal wall function.

deandrade87 · Post by **deandrade87** » Tue Jun 14, 2016 6:05 pm

Hello, I was checking clptur.f90 when I saw this bit of code:

    ! Dirichlet on the scalar, with wall function
    if (iturb.ne.0.and.icodcl(ifac,ivar).eq.5) then

      call hturbp(iwalfs,prdtl,sigmas(iscal),yplus,dplus,hflui,ypth)

      ! Compute (y+-d+)/T+ *PrT
      yptp = hflui/prdtl
      ! Compute lambda/y * (y+-d+)/T+
      hflui = rkl/distbf *hflui

      ! Neumann on the scalar, with wall function (for post-processing)
    elseif (iturb.ne.0.and.icodcl(ifac,ivar).eq.3) then
      call hturbp(iwalfs,prdtl,sigmas(iscal),yplus,dplus,hflui,ypth)
      ! y+/T+ *PrT
      yptp = hflui/prdtl
      hflui = hint

    else
      ! y+/T+ *PrT
      yptp = 1.d0/prdtl
      hflui = hint
    endif

I don't understand two things (among many others

):

- why would the Neumann b.c. use hint (which is defined a bit earlier directly from the turbulent viscosity) and not hflui coming from the evaluation of the wall function?

- is there a particular advantage in using the Arpaci Larsen wall function when the turbulent diffusivity is known? I mean, following the theory guide, the Arpaci Larsen wall function is used to evaluate the term:

$K = \frac{y^+}{Y^+} = \frac{1}{\sigma} + \frac{1}{\sigma_T}\frac{\nu_t}{\nu}$

but if nu_t and sigma_t are known, K is known as well. Does this imply that using the wall function is an improvement of the linear eddy viscosity model for scalar diffusion? Do you know in which cases using icodcl(ifac,ivar)=5 gives better results than icodcl(ifac,ivar)=1?

Thanks

Jacopo