[Louis.C]

Hi everybody,

I'm currently creating a RANS model of a cube in the air. I'm in high Re and I'm trying to model the flow around this cube in k-omega-SST.
As you can see in the picture below, my residuals are strange... My residuals react like if I change something in fvSchemes or fvSolution. The model seems to converge and then blows up !
I'm sure of my BC, and k-epsilon works without any problems.
Do you have any ideas or suggestions about my problem ?

Thank you in advance.

Louis

OpenFOAM 2.2.0

[RodriguezFatz]

This indeed looks like you changed the fvSchemes... Can you tell us a bit more about your setup? What about b.c. and fvSchemes?
Did you have any "bounding errors" in the log file before the blow ups?

[Louis.C]

Thanks for your reply.

My "bounding error" appear after the blow ups. But only for the "little" blow ups ( approximately times 1600 and 2900), the others don't involve "bounding error"... Despite the fact that they're more important, it's very strange.

Concerning my BC, in inlet I've a logarithmic profile for the velocity (which corresponds to a field data). Same thing for the TKE (Turbulent Kinetic Energy) and omega.

On the wall I keep the wall-functions even if I have my y+ E [0,4].

And for my fvSchemes, this residuals correspond to an upwind resolution. I've tried linearUpwind grad(U), the same blow ups appear... so I keep 1st order for the moment.

[RodriguezFatz]

Did you try to run the case with constant (fixed value) boundary condition? Maybe the profile is the source of your pain...

[Louis.C]

I didn't try, I will tomorrow.

But another case that I studied was in low-Re, and exactly the same thing occurred (with inlet fixed value).
It's only with k-omega-sst, my k-epsilon model don't have this problem I don't know why. At that time, I thought it was openFOAM which didn't like the k-omega-sst in low-Re... But I was clearly wrong.

I will try tomorrow to keep a fixed value in inlet, I will let you know if it works. If not, you will have to find another solution for me

[RodriguezFatz]

I just recalled, that i saw exactly what happend in your case, when i had bad posed periodic b.c. Do you use any periodics at all?

[Louis.C]

When you say periodic b.c, you mean like a velocity which change periodically during the simulation ?
Sorry, but I have never used this kind of b.c... maybe an option is wrong.

I'm going to try with a fixed value at the inlet, if it works, it might be the logarithmic profile which is wrong.

In "0/U" at the inlet I have :

inlet
{
type timeVaryingMappedFixedValue;
setAverage off;
offset (0 0 0);
}

And my velocity profile is in "constant/boundaryData/inlet/0/U". This is well read by OpenFOAM, I find a good profile at the inlet of my system. Omega get the same option.

[Louis.C]

Well... I don't know really why, but I've found my problem.

First I've tried a fixed value at the inlet, it's doing the same thing than before...

To make the convergence more easy for my model, I've initialize my model with low-Re, and then I increased it (otherwise it blows up).

This morning I tried another way to make my model converged, I've resolved it with y+ ~ 100, and then I've used mapFields to paste my solution on my heavy mesh (y+ ~ 1).

It seems that the blow ups on my residuals are linked to my mesh or k-omega in low-Re, I don't really know, maybe twice.

Well... Rodriguez thanks for your help !

Good continuation, and if one day I've another problem, I'll keep you informed

[RodriguezFatz]

Great to hear that!