[SimonHb]

Hi,

I am trying to simulate the drag & lift of a 3D surf foil in water (5° inclination and 5.56 m/s).
I use simpleFoam (steadyState) with komegaSST turbulence model. I generated 3 layers (94% layer coverage) making my average y+ around 40-50 (when I run the post process utility YPlus).

My problem is the following: I simulated different layer thicknesses in order to check whether my results would remain similar with a different y+ average value (but that would stay between 30 and 300) and it is clearly not the case. If the drag does not change too much, the lift strongly depends on the mesh changes (more than 20% difference, and 5-10% for the drag).

I've read lots of threads about that 30<y+<300 requirement for accurate simulations with komega SST turbulence model but it still looks unclear to me.

Is the average y+ provided by the postprocess utility YPlus relevant ?
In reality, y+ values vary from 1e-2 to 200 around the foil (still provided by the YPlus postprocess utility).
What could I check in my model other than the mesh that could affect the accuracy and make the simulation so y+ sensitive ?
My BCs looks ok, my fvSchemes as well so I don't really know where to look at now.

I attached the mesh and geometry so that you could get an idea of my simulation. Don't hesitate to tell me if you need more info about my case.

Thanks,

Simon

[mAlletto]

If you are using wallfunctions all Y+ should lay between 30 and 300. You can use the nutUSpalding which are valid also in the viscose sublayer. If you have a recirculation region wallfunctions don't work well.

Best Michael

[David*]

I agree with Michael that the average y+ is not relevant, but the values for each cell. 1e-2 is too low for a wall function to work.


Do you use a recent OpenFOAM version? In old versions (.org 3.0 and older), the y+ utility did not work consistently: Yplus and ystar

[JNSN]

Hi all,


I don't agree with Michael and David (at least not completely ). You can not avoid very small yPlus values in some cells, as you will have a stagnation point with velocity == 0 and therefore yPlus -> 0.


But as Michael said, definitely worth to check nutUSpalding wall function.



Best,
Jan

[SimonHb]

Hi all,

Thanks for your answers.
For information :
- I use OF8
- I did use nutUSpalding wall function
- Indeed, at some points I get very low y+ values that I can’t avoid near the stagnation point
In attachment you’ll find visuals plotting y+ over the foil surface to get an idea.
And to illustrate my problem more concretely, I ran two simulations at 10° inclination (no stall yet so no significant recirculation area) and 5.56 m/s water flow with the following mesh parameters :
- Mesh 1 : 7 refinements on foil surface / 3 layers / 0.3 as finalLayerThickness (relativeSize true)
- Mesh 2 : 7 refinements on foil surface / 3 layers / 0.6 as finalLayerThickness
- Mesh 3 : 7 refinements on foil surface / 3 layers / 0.15 as finalLayerThickness
Eventually I get the following results :
- Mesh 1 : avg y+ ~36 (min 1.8 & max 369) / Fx = 64 N / Fz = 584 N
- Mesh 2 : avg y+ ~66 (min 3 & max 387) / Fx = 66 N / Fz = 552 N
- Mesh 3 : avg y+ ~22 (min 1 & max 400) / Fx = 65.5 N / Fz = 604 N
You’ll see in attachment the y+ plotted on the surface of the foil : extrado (scale -100 to 300) and intrado (scale 0 to 100).

So the drag is quite stable but the lift is significantly varying from one mesh to another. And I’m surprised to get such a difference between mesh 1 and 2 whereas the Y+ is mainly above 30 and below 300. What do you think ?

Best,
Simon

[SimonHb]

And the last visual