[pi120]

Hello,

Anybody knows how the field function 'Wall Y+' is defined in STAR-CCM+? From literature it should be:

y+ = u_tao * y / nu

with (each value taken at the first cell of the prism layer, i.e. the one adjacent to the wall):

y = distance from the wall of the cell's centroid [m]
nu = kinematic viscosity [m^2/s]
u_tao = sqrt(tao_w/rho) [m/s]

with
tao_w = wall shear stress [Pa]
rho = density [kg/m^3]

If I create a field function with this definition the value is much different from that of the STAR-CCM+ default field function 'Wall Y+'. At least this happens in the 2D simulation I am running.

Anybody who knows how 'Wall Y+' is defined?


Thanks
Pietro

[pi120]

it seems like instead of using the exact distance between wall and centroid of first cell, STAR-CCM+ uses the 'Wall Distance', which is however limited to 1e-6 m...

[LuckyTran]

You have cells smaller than 1e-06? You can freely change that limit. It's there because most CAD software writes in a ascii UTF8 format which doesn't have any better precision.

[pi120]

thanks for the tip. yes I have up to 2e-8 m and I used SolidWorks for the CAD.
Does this mean that it does not make sense to have the first centroid at a distance lower than 1e-6m?

[LuckyTran]

You're saying you have cells smaller than the precision of your CAD... The mesh already makes no sense, we can forget about the results entirely. At least, that's what it sounds like.

If you truly have and need mesh resolution of that scale, then you can change these options. But please make sure you give the CAD and mesh proper treatment. It's atypical to have that kind of precision. Possible, but atypical.

[pi120]

Yes you're right... I just checked the wall shear stresses and they reach ~3000. Now that I think about it, wall shear stresses of 3000 do not make much sense, as they are the product of viscosity (1e-5) times velocity gradient (which has to be 1e8 to give such value...).

This high wall shear leads my y+ to steeply increase and therefore I need a lower cell size normal to the wall. That's why I was using 2e-8.

Any idea of why this happens? The only reason of this unphysical value I can think of is the very low aspect ratio of the first prism layer cell (AR = 0.05)...

Couple info about the sim:

Normalized residuals converge all below 1e-5. Flow features and aerodynamic coefficients are physical.
It's a 2D supersonic flow at M = 2.5 over a capsule of diameter 0.03m at sea level (Re = 2e6). As I am using k-w SST, I would like to reach y+ = 1 at the wall. This is easy for the wake region and close to the stagnation point. Before the flow expansion it increases due to wall shear increase.

[pi120]

also, the simulation is steady and I am using the coupled implicit solver

[pi120]

as I said before, wall shear stress is mu*du/dy

mu = 1e-5 Pa-s
du/dy = 1e8 s^-1

specifically
du = 100 m/s
dy = 1e-6 m

so the problem lies in a du which is very high between the first two layers of the prism layer....

[LuckyTran]

It's okay to not have y+ <1 everywhere. It's okay if a few cells violate this. There will be singularity like situations where you have super high gradients due to there not being any boundary layer. Take the boundary layer over a flat plate for example, you basically have infinite shear stress on the leading edge always.

In the image showing your body, I can't see the prisms layers and that is a issue. Yet I count 16+ prism layers in your zoomed in pic. That means you have ridiculously thin layers. The prism layers should span most of the boundary layer and blend smoothly with the core mesh. Packing of a ton of prisms into a very small region next to the wall just to get a low y+ value is not always better.

[pi120]

Yes my prism layer looks too thin… even though the boundary layer is all inside of it, it is true that it does not blend smoothly with the core mesh. What I think is even worse is the low aspect ratio of the wall cell, AR=0.05… is this still ok as long as the fluid is parallel to the wall?

Regarding the high wall shear: in my opinion it does not look like a singularity. The contour shows shear stresses higher than 1000 for most of the non wake region.

I think there is a chance that the results could make sense. Reynolds is relatively high (2.1M) and therefore the BL thickness is low. Add to this a supersonic flow: you will need to pass from a zero velocity at the wall to a supersonic velocity outside the BL in a short distance. Also the flow is assumed turbulent, with higher gradient at the wall than if it was laminar. All these effects might lead to such high gradients…

The problem is that, if the results are actually correct, the only way to go might be with a high y+ treatment, as I have y+>1 for most of the non-wake geometry and not for only a few cells…