[JR22]

Hi,

I have been fighting for a while trying to bring the residuals for a model I am running. I am using simpleFoam and SST k-omega turbulence model. I decided to plot the residuals for the simpleFoam motorBike (OF2.2.0) and their residuals don't look too good either. I've followed some of the strategies in the forum to bring them down, and the strategies have worked to some degree (the best I've gotten is about one order of magnitude better than the image for my residuals). If I make a very rough model of a sphere suspended in the air with the same settings as my model and a fairly coarse mesh, I get amazingly low residuals.

Any comments?
Is there a general criteria for residuals for SST k-omega?

Thanks in advance


Here are the residuals for the OF2.2-simpleFoam motorBike tutorial. You could run this sucker forever, and the residuals will stay moreless the same:


Here are my residuals:



Thanks very much in advance.

[fredo490]

Check the yPlus values of your models. To do so you can use the yplusRAS command to calculate it. To use the k omega you need to have a maximum yPlus of about 4.

[gillimaniac]

Hey JR,

I am currently facing the same problems regarding k-omega-sst with simpleFoam.
Other than fredo490 I don't think that you need an y+ = 4.
It is nowhere said definitely that the implemented k-omega-sst model works for LowRe only. Most threads say that it is capable as well for HighRe and LowRe simulations.

But I have to say that your residuals dont seem "amazingly low" to me like you said.
This is a residual plot from my benchmarking regarding wall treatment, different y+ values and boundary layer mesh which i am currently struggling with: https://www.dropbox.com/s/9q8ubx9fg3...st%20order.png

Please let me know if you get any further with this because i'm really interested in getting k-omega-sst to work reliable

[fredo490]

K-Omega SST works with y+ > 4 because it switchs to a high Reynolds model. So don't expect to take advantages of K-omega SST if you have a y+ > 4 ! Just take a standard k-epsilon and it's done. Moreover, try to avoid to have 4 < y+ < 20 because the model might give you some strange behavior (it is also true on Fluent). You have to choose: y+ < 4 or y+ > 20.

The only way to predict the transition laminar/turbulent is to use a y+ < 1 with K-omega SST

Plus, there are some effects that are impossible to model if your y+ > 4.

[gillimaniac]

So you're saying that k-omega-sst with wall functions on a mesh with Y+> lets say 30 gives no advantage to standard k-epsilon?
Why then did henry state out that the high-re-version of k-omega-sst is currently implemented in openfoam?

What would be the use of a high-re k-omega-sst which is - as you say - just some instable version of standard k-epsilon with no additional use?

I'm rather confused right now.

[fredo490]

I've never said that it is an "instable version of standard k-epsilon"...

1) I said that no matter what, it is dangerous to work with a 4 < y+ < 20 (which is the area near y+ = 11 that is the limit of the log law) ! If you work with a y+ between 4 and 20, the program will locally switch from a wall function to a "classic" boundary condition and this might give you a messy results. Moreover, many cases report some mesh dependency when the y+ is between 1 and 4 for k-Omega SST .

2) K-Omega SST theoretically works for all y+ < 300... So yes, it works for y+ > 4 but in this case the results are fairly (very ?) similar to a k-epsilon model since it will use a similar wall functions and that the free stream will use a k-epsilon behavior.

3) The main purpose of using K-Omega SST is to predict the adverse pressure gradients and separating flow areas (but for this, you need a y+ < 4 and ideally y+ < 1).

[gillimaniac]

Thanks a lot for this clarification and sorry that i misunderstood your points