[Nek5000-users] Uservp, eddy diffusivity and explvis

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

Howdy,

I was trying to test an eddy viscosity model for an internal flow 
problem with heat transfer - set up very similar to turbChannel. The 
domain contains a stagnating region, from which the attached picture 
shows a slice (flow in x from left, wall at bottom and right side is 
also wall, the flow goes around this cylinder in/out of the 
in-plane/y-direction).
The viscosity computation kicks in after the 2nd time step, before that 
diffusivity is just equal param(2) (rho is unity). What happens though 
is that as soon as I start the simulation with uservp (param 30 > 0), 
the velocity goes nuts in the stagnation region, starting from step 0, 
even though viscosity does not (yet) change.
After some debugging, I found that this behavior is caused by the 
following line which I took from turbChannel:
"if(ifsplit .and. param(30).gt.0) ifexplvis = .true."

ifsplit is true in my case, so this condition is met and yields the 
weird behavior as seen in the plot.

When I comment it out, however, everything seems normal and even the 
eddy viscosity computation seems to work well, although I haven't done 
long term testing yet.

What would you advise me to do? I am using rev548.

Thanks,
Markus
-------------- next part --------------
A non-text attachment was scrubbed...
Name: explvis.png
Type: image/png
Size: 94794 bytes
Desc: not available
URL: <http://lists.mcs.anl.gov/pipermail/nek5000-users/attachments/20100718/a8363693/attachment.png>