[wc34071209]

Hi Foamers,

I use the same grid with k-epsilon models (standard, realizable and RNG) and k-omega SST model respectively. The calculated y+ by the yPlus function object is so different for different turbulence models.

For one patch, the calculated y+ for k-epsilon models is around average 0.7. But for k-omega SST model, it is around 0.07.

I understand that there should be some differences. However, 10 times is, I think, too big.

Anyone has clue why this happens?

Thank you.

[arvindpj]

Correct computation of y+ involves, correct computation of wall shear stress. Each turbulence model is apt for a certain type of flow and their prediction of flow patterns like flow separation zones may not be correct. Check your wall shear stress values.

Cheers,
-J

[wc34071209]

Originally Posted by arvindpj

Correct computation of y+ involves, correct computation of wall shear stress. Each turbulence model is apt for a certain type of flow and their prediction of flow patterns like flow separation zones may not be correct. Check your wall shear stress values.

Cheers,
-J

Thanks. But I think both the calculation of wall shear stress and y+ are dependent on wall functions.

It seems that if I change the wall functions for mut from 'mutkWallFunctions' to 'mutUWallFunctions', the calculated y+ of k-omega model increases from 0.07 to 0.3.

I also asked another question, regarding the wall functions, y+ and the grid. I appreciate you could also make some comment.
What is the consequence of using a high-Re turbulence model on a low-Re grid

[Vignesh2508]

Interesting behaviour. What was the value when you tried calculating the y+ value yourself with the velocity. To which turbulence model y+ value does it match?

Vicky

[piu58]

The y+ value differs at the surface. Which one do you mention? If you choose some extrem value such large differences may be possible, because the numerical treatment of extrem points is difficult.

[wc34071209]

Originally Posted by Vignesh2508

Interesting behaviour. What was the value when you tried calculating the y+ value yourself with the velocity. To which turbulence model y+ value does it match?

Vicky

Hi Vicky,

Thank you for your reply. I am using buoyantSimpleFoam to solve the heat transfer problem. The geometry is an ordinary room with a constant temperature cylinder-like heat source inside. The free stream velocity is 0.3 m/s. The heat source and the wall have the convective heat transfer with the air. The first layer thickness of the heat source is 0.5 mm.

However, as different models calculate different y+, I don't know whether I should increase or decrease the first layer thickness.

Thank you.

/Yuehan

[wc34071209]

Originally Posted by piu58

The y+ value differs at the surface. Which one do you mention? If you choose some extrem value such large differences may be possible, because the numerical treatment of extrem points is difficult.

I mean the average yPlus of different surfaces. The average y+ is given by the yPlusRAS function object of OpenFOAM.

[wc34071209]

I think it is the wall function for mut that matters. When using k-omega SST model on a low-Re grid, the mutkWallFunction tends to give a very low y+ value, while the mutUSpaldingWallFunction or mutLowReWallFunction give the correct value of y+.

[Vignesh2508]

I think, that depending on the wall function, the local fluid velocity will change and hence the y+ would also vary along with that. So first, you have to decide on the turbulence model for your problem and then choose that particular y+ value.

Thanks

Vicky

[wc34071209]

Originally Posted by Vignesh2508

I think, that depending on the wall function, the local fluid velocity will change and hence the y+ would also vary along with that. So first, you have to decide on the turbulence model for your problem and then choose that particular y+ value.

Thanks

Vicky

Hi Vicky,

Thank you for your reply. I am using k-omega SST model. I believe that y+ is a physical property which in principle should not depend on the wall function or turbulence model chosen.

The way that "mutkWallFunction" calculates y+ relies on the assumptions that hold only in the log-law (or overlapped, correct me if I am wrong) region.

So if I use a low-Re gird, then there is a risk that "mutkWallFunction" may not be able to give correct y+ values.

[lpz456]

Originally Posted by wc34071209

I think it is the wall function for mut that matters. When using k-omega SST model on a low-Re grid, the mutkWallFunction tends to give a very low y+ value, while the mutUSpaldingWallFunction or mutLowReWallFunction give the correct value of y+.

haha, after five years, I meet the same question as you, the reason for so low yPlus calculated by SST model is exactly by nutwallFunction(low-Re model should choose nutLowRewallfunction)