[Nek5000-users] Periodic BC on non-parallel boundaries

[nek5000-users at lists.mcs.anl.gov]

Hi Stefan,

What do you mean by "remapping a periodic-side domain in userdat2"?
>

I build a mesh using genbox with dimensions [0,1]^3.  The BC are periodic in
x.  Then in userdat 2 I remap the (x,y) cross-section from a square to an
annulus:

      call rescale_x(xm1,zero,pi2)
      call rescale_x(ym1,zero,one)

      n = nx1*ny1*nz1*nelv   ! Turn box into annulus
      do i=1,n
         th = xm1(i,1,1,1)
         r  = r0 + (r1-r0)*(1.-ym1(i,1,1,1))
         xm1(i,1,1,1) = r*cos(th)
         ym1(i,1,1,1) = r*sin(th)
      enddo

This gives me an annulus centered on the z-axis with interior radius r0 and
exterior radius r1.  (I tried centering it on the y-axis instead with the
periodic boundary lying in the xy-plane, but the result was the same.)  My
velocity BC in what are initially the y and z directions are all SYM, so I
need to use the stress formulation.

This procedure worked in 2D for no-slip or stress-free BC, but in 3D it is
not working for stress-free BC.  I can visualize the 3D geometry, and the
annular cylinder appears to be properly constructed, but the fluid solution
blows up.

I understand now that my fractional-annulus cases blow up because non-cyclic
periodic BC at the theta-ends are incompatible with stress-free walls, but
this should not be an issue for the full 2π annulus, even in 3D.  Can you
think of a reason that non-cyclic periodic BC would work in 2D but not in
3D?

Sure, even with IFCYCLIC=F you'll get a result. But is probably not
> the one you had in mind, simply because the BCs are different i.e.
> periodic vs cyclic.
>

Ah yes, you`re right that the results are actually different.

Thanks a lot,

David
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/nek5000-users/attachments/20111011/5b733be5/attachment.html>