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/)
-   -   rhoSimpleFoam - Residuals do konverge, but oscillate all the way down (http://www.cfd-online.com/Forums/openfoam-solving/128500-rhosimplefoam-residuals-do-konverge-but-oscillate-all-way-down.html)

okroud January 14, 2014 05:30

rhoSimpleFoam - Residuals do konverge, but oscillate all the way down
 
3 Attachment(s)
Hello guys,

I am calculating a simple laminar channel flow, with one inlet, one outlet, two walls, and periodic bc's in z-direction. Since i had stability issues, I'm starting with first order Schemes and a relaxation of 0.1 for p and 0.7 for U. Fortunetaly, the residuals do konverge (finally!). But they show oscillations as shown in the attached picture.

http://postimg.org/image/g2unbkoj1/

Could someone give me a hint about the way I could change my fvSchemes or fvSolution to dampen these oscillations. My fvSchemes and fvSolutions are attached in this Post.

Thanks in advance!

okroud

Tobi January 14, 2014 07:55

Hi,

can you take a picture of your mesh and Initial conditions / velocity?
Additionally a slice of U at 1000, 1010 and 1020 iterations?

okroud January 14, 2014 09:57

5 Attachment(s)
Quote:

Originally Posted by Tobi (Post 469964)
can you take a picture of your mesh and Initial conditions / velocity?
Additionally a slice of U at 1000, 1010 and 1020 iterations?

Hi Tobi,
thank you for your interest. I will be able to post these things on Thursday.


Edit: Unfortunately i only have solutions of Timestep 1800,1900 an 2000. There is no visible difference between them. Indeed these timesteps could be interesting regarding the p-residual, I will reproduce them over the weekend. Pictures of the mesh are attached in this post.

Boundary conditions for U:

dimensions [0 1 -1 0 0 0 0];

internalField uniform (0 0 0);

boundaryField
{
in
{
type flowRateInletVelocity;
flowRate 0.17359;
value uniform (0 0 0); // placeholder
}
out
{
type zeroGradient;
}

lid
{
type fixedValue;
value uniform (0 0 0);
}
bottom
{
type fixedValue;
value uniform (0 0 0);
}

PER_MERGED //periodic bc's
{
type cyclic;
}
}

Boundary conditions for p:

dimensions [1 -1 -2 0 0 0 0];

internalField uniform 101325;

boundaryField
{
in
{
type zeroGradient;
}
out
{
type fixedValue;
value uniform 101325;
}

lid
{
type zeroGradient;
}
bottom
{
type zeroGradient;
}

PER_MERGED //periodic bc's
{
type cyclic;
}
}

okroud January 28, 2014 05:05

I found the Solution to my Problem: The relaxation factor of the oscillating Solution for rho was 0.2. I deleted it completely (which i guess means that it is now 1). Now the residuals konverge without any oscillation:). But this rises further questions:

1. Does anyone have a clue why this relaxation of rho is producing such great oscillations?

My guess: Since i use the thermoPhysical Model:

hPsiThermo<pureMixture<sutherlandTransport<specieT hermo<hConstThermo<perfectGas>>>>>;

rho is just calculated with the equation of state for perfect gas. So after rho is underrelaxated, the solution of rho is not anymore in harmony with the equations of momentum and energy, which are solved in the following iteration.

2. But why is the underrelaxation of rho implemented in the solver? What is it's use?

Would be great to hear some opinions :)


All times are GMT -4. The time now is 11:37.