[Nek5000-users] issue w/ insulated wall?

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

Hello all,

Following the issue that I have had with temperature field in my  
problem (see https://lists.mcs.anl.gov/mailman/htdig/nek5000-users/ 
2014-June/002883.html), I have been performing several test cases and  
trying to solve the issue. I just wanted to share two of these tests,  
and wondered if anyone have any comments/thoughts on them.

I have a 3D cylindrical domain with several embedded horizontal holes  
in it and an inflow (at the top) and an outflow (at the bottom). I  
solved both momentum and heat equations for this problem, but no heat  
source was added to the domain.

In the first test, the problem was initialized with a temperature  
(equal to the inflow temperature) and all walls were set as no-slip  
insulated boundary condition. For this setup, one expects temperature  
to remain unchanged forever. However, after around t=200 (of  
convective time scales), the temperature starts to grow gradually  
leading to NAN value. This usually occurs for elements attached to  
the wall and inflow opening.

In the second test, everything is identical to the first test, except  
that I changed the walls temperature boundary condition to a constant  
value equal to initial temperature. Temperature remains constant so  
far (after t = 1000).

Although, these two examples are physically different (in terms of  
boundary conditions), but it might raise the question whether there  
is any issue with insulated (i.e., 'I  ' flag in .rea) boundary  
condition. It, however, certainly shows that the issue begins from  
walls. I should note that the walls are curved surfaces (i.e., not  
parallel to coordinate planes). I searched the users archive and  
apparently, similar issue with insulated wall was reported:

https://lists.mcs.anl.gov/mailman/htdig/nek5000-users/2012-February/ 
001688.html

Regards,
Mohsen