[Nek5000-users] Selective frequency damping
nek5000-users at lists.mcs.anl.gov
nek5000-users at lists.mcs.anl.gov
Thu Sep 15 03:27:50 CDT 2011
It indeed solved my problems.
Thanks a lot.
On 13 September 2011 16:56, <nek5000-users at lists.mcs.anl.gov> wrote:
> Hi Jean-Christophe,
>
> A couple of comments.
>
> 1). I would use declaration of arrays vxold,...,qxold,...,f_x,...
> similar to declaration of vx, vy,... in
>
> nek5_svn/trunk/nek/SOLN :
>
>
> COMMON /VPTSOL/
> ...
> $ , VX (LX1,LY1,LZ1,LELV)
> ...
>
>
> 2). To set an array on velocity mesh to zero, please use
>
> ntot = nx1*ny1*nz1*nelv
> call rzero(vxold,ntot)
>
> instead of
>
> vxold = 0.0D+00
>
>
> 3). You do need local element number for setting the forcing in userf,
> namely
>
> e = gllel(eg)
> ffx = f_x(ix,iy,iz,e)
>
> Otherwise it will work correctly only on one core.
>
>
> Let me know if it helps.
> Best,
> Aleks
>
>
>
>
>
> On Tue, 13 Sep 2011, nek5000-users at lists.mcs.anl.**gov<nek5000-users at lists.mcs.anl.gov>wrote:
>
> Hi Nek's,
>>
>> I've been trying to implement the selective frequency damping (Akervik E.,
>> Brandt L., Henningson D.S., Hoepffner J., Marxen O., Schlatter P. 2006.
>> *Steady
>> solutions of the Navier-Stokes equations by selective frequency
>> damping.*Physics of fluids
>> *18*) which enables to compute stationnary solution. Here is what I have
>> written in my .usr file:
>>
>> In userchk:
>>
>> common /vxold/ vxold(size(vx,1),size(vx,2),**size(vx,3),size(vx,4))
>> common /vyold/ vyold(size(vy,1),size(vy,2),**size(vy,3),size(vy,4))
>> common /vzold/ vzold(size(vz,1),size(vz,2),**size(vz,3),size(vz,4))
>>
>> common /qxold/ qxold(size(vx,1),size(vx,2),**size(vx,3),size(vx,4))
>> common /qyold/ qyold(size(vy,1),size(vy,2),**size(vy,3),size(vy,4))
>> common /qzold/ qzold(size(vz,1),size(vz,2),**size(vz,3),size(vz,4))
>>
>> common /for_x/ f_x(size(vx,1),size(vx,2),**size(vx,3),size(vx,4))
>> common /for_y/ f_y(size(vy,1),size(vy,2),**size(vy,3),size(vy,4))
>> common /for_z/ f_z(size(vz,1),size(vz,2),**size(vz,3),size(vz,4))
>>
>> real chi, omega_c
>> integer step_filtr
>>
>>
>> chi = .25
>> omega_c = .1
>> step_filtr = 020000
>>
>> if(istep.EQ.step_filtr.AND.**nid.EQ.0) write(*,*) 'Filtrage'
>>
>> if(istep.EQ.0) then
>> vxold = 0.0D+00
>> vyold = 0.0D+00
>> vzold = 0.0D+00
>> endif
>>
>> if(istep.LT.step_filtr) then
>> qxold = 0.0D+00
>> qyold = 0.0D+00
>> qzold = 0.0D+00
>>
>> f_x = 0.0D+00
>> f_y = 0.0D+00
>> f_z = 0.0D+00
>>
>> call opcopy(vxold,vyold,vzold,vx,**vy,vz)
>> endif
>>
>> if(istep.GE.step_filtr) then
>> qxold = qxold + dt*omega_c * (vxold - qxold)
>> qyold = qyold + dt*omega_c * (vyold - qyold)
>> qzold = qzold + dt*omega_c * (vzold - qzold)
>>
>> f_x = -chi * (vxold - qxold)
>> f_y = -chi * (vyold - qyold)
>> f_z = -chi * (vzold - qzold)
>>
>> call opcopy(vxxold,vyyold,vzzold,**vxold,vyold,vzold)
>> call opcopy(vxold,vyold,vzold,vx,**vy,vz)
>> endif
>>
>> and in userf:
>>
>> common /for_x/ f_x(size(vx,1),size(vx,2),**size(vx,3),size(vx,4))
>> common /for_y/ f_y(size(vx,1),size(vx,2),**size(vx,3),size(vx,4))
>> common /for_z/ f_z(size(vx,1),size(vx,2),**size(vx,3),size(vx,4))
>>
>>
>> integer e,f,eg
>> c e = gllel(eg)
>>
>>
>> c Note: this is an acceleration term, NOT a force!
>> c Thus, ffx will subsequently be multiplied by rho(x,t).
>>
>>
>> ffx = f_x(ix,iy,iz,eg)
>> ffy = f_y(ix,iy,iz,eg)
>> ffz = f_z(ix,iy,iz,eg)
>>
>>
>> I have tried to compute with such routines the base flow for a 2D
>> lid-driven
>> cavity flow at Re = 8500, however depending on the number of spectral
>> elements I use, I sometimes get a stationnary solution that looks slightly
>> as the one I'm looking for (but some differences near the lid) and
>> sometimes
>> I get one which is actually not stationnary. I applied the same algorithm
>> for a finite-difference code and encountered no problem, so I was
>> wondering
>> then if I missed something specific to spectral elements or Nek 5000 (such
>> as a mass matrix multplication for instance) or to F77 (which I'm not at
>> ease with, see my definition of the common for instance)?
>>
>> Sincerely yours,
>>
>> --
>> Jean-Christophe
>>
>> ______________________________**_________________
> Nek5000-users mailing list
> Nek5000-users at lists.mcs.anl.**gov <Nek5000-users at lists.mcs.anl.gov>
> https://lists.mcs.anl.gov/**mailman/listinfo/nek5000-users<https://lists.mcs.anl.gov/mailman/listinfo/nek5000-users>
>
--
Jean-Christophe
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/nek5000-users/attachments/20110915/14de9422/attachment.html>
More information about the Nek5000-users
mailing list