CFD Online Discussion Forums

CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   OpenFOAM Running, Solving & CFD (http://www.cfd-online.com/Forums/openfoam-solving/)
-   -   Temperatures too low near inlets (http://www.cfd-online.com/Forums/openfoam-solving/59826-temperatures-too-low-near-inlets.html)

helmut February 13, 2007 14:29

Hi, I have two ducts separa
 
Hi,

I have two ducts separated by a solid slab. I solve fluid flow in the ducts (on sub meshes) and use the resulting velocities to solve temperature on the whole mesh. The inlets and exterior walls all have fixed value boundary conditions, while the outlets have zero gradient conditions. There is a jump discontinuity where the duct inlets(300K) meet the slab(400K). In the ducts, near the inlets, I get temperatures (291<T<300) below the lowest boundary value. Any ideas?

Thanks,
Helmut

hjasak February 13, 2007 14:56

Add some non-orthogonal correc
 
Add some non-orthogonal correctors - if you remember, I've sent you an E-mail on this a while back.

Regards,

Hrv

hjasak February 13, 2007 15:16

Hello Helmut, Forgot to men
 
Hello Helmut,

Forgot to mention: the unboundedness you mention is due to the non-orthogonal correction in the Laplacian. If you feel your mesh should be orthogonal, you can remove the unboundedness by using the following setting for the equation in system/fvSchemes - it specifies that the non-orthogonal correction will not be added:

laplacianSchemes
{
laplacian(gamma,T) Gauss linear uncorrected;
}

(please adjust the name of the laplacian, I cannot guess it).

Also, just using a uniform initial guess for T will help very much - if I remember correctly, you are starting with a step-profile.

Sorry for my forgetfulness,

Hrv

helmut February 13, 2007 15:49

Hi Hrv, Thanks for the sugg
 
Hi Hrv,

Thanks for the suggestions. I'm using a uniform initial temperature, having seen the folly of the step-profile earlier (thanks!). The mesh should be orthogonal (rectangular blocks and simple uniform grading). Anyway, I tried with some nonorthogonal correction, as in your first reply above, and then using Gauss linear uncorrected laplacian schemes, as suggested next. Same results.

H.


All times are GMT -4. The time now is 22:22.