[Nek5000-users] velocity projection (p94) for linear simulations

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

Hi Nima,

At present, projection is not turned on for the
perturbation velocity.   When we originally set up
the perturbation scheme we had envisioned (and occasionally
used) as many as 50 perturbation fields at one time -- the
thinking was that it would take too much memory to save projection
fields for all of those.

Right now there is a general sense that we need to restructure
the projection code so that it can readily handle the variety of
use cases that we've been encountering of late.  I'll talk with
the developers here and hopefully we can come up with a fix soon.

Best, Paul


On Wed, 5 Feb 2014, nek5000-users at lists.mcs.anl.gov wrote:

> Hi Neks,
>
> I am running a linear simulation with fixed dt, p93, 94, 95 > 0,
> filtering and dealiasing.
> However, it seems velocity projection is not affecting number of
> Helmholtz iterations which is fixed after a while.
>
> I tested for p94=0 and p94>0 and practically got the same number of
> iterations (and even a slight increase in time for time step when p95>0).
> However in case of p95=0 and p95>0 I can see the improvement clearly.
> So, pressure projection is working fine.
> I need tight tolerances, div=1.e-10, hmh=1.e-11, and reducing Hmholtz
> iterations can help a lot.
>
> Would you please clarify the situation on velocity projection in linear
> cases?
>
> Here is a sample time step :
> Step1122816, t= 2.8264667E+02, DT= 2.5173018E-04, C=  0.390 6.0349E+04
> 5.1942E-02
>  1122816 2.8264667E+02 Perturbation Solve:    1
>    1122816    Hmholtz VELX:     18   8.3081E-12   1.0343E-01   1.0000E-11
>    1122816    Hmholtz VELY:     17   8.0312E-12   2.3905E-01   1.0000E-11
> *****   11 alpha:  1.5898E-06  2.6447E-09  1.6940E-09  1.5196E-09
> 3.6127E-10  8.4097E-10  6.6903E-10 -2.2649E-10 -1.5484E-09 -3.4627E-10
> ******  11  4.0897E-09  1.8192E-10  2.2480E+01 alph12
>   1122816 U-PRES gmres:      7  9.4727E-11  1.0000E-10  1.8192E-10
> 6.8312E-03  1.5541E-02
>
> Best,
> Nima
>
> -- 
> Nima Shahriari
> Linné Flow Center, Mechanics KTH
> SE-100 44, Stockholm, Sweden
> Phone: +46 8 7906876
> E-mail: nima at mech.kth.se
>
>