[prince_pahariaa]

Dear friends

I am modeling turbulent flow heat transfer model. Re no is in the range of 30,000. Geometry is little complicated. Inlet is simple pipe flow and then it meet an hemispherical geometry in mid way and flow split. From top of hemispherical surface a solid vacuum tube is attached (there is no flow in the tube). Therefore, an annuls geometry formed from that point on.

Hemispherical geometry is little peculiar. It has the thickness of 1.5mm at mid point and 3mm at the top surface. It is again solid and modeled as a wall.

I have used k-epsilon model along with standard wall function in fluent.

My y+ values for the flow rate of 3kg/s is in the range 0f 30<y+<75 which is acceptable. but for flow rate of 35kg/s its range increase from 30<y+<745.

I have tried adapting the grid. And after that i get maximum y+ is 381 but i did not notice any change in my temperature profile. I have tried with k-omega model (which does not require wall function) and there also not much change.

My question is whether my modeling is valid inspite of y+>300 as my reynolds no is very large ?? If not what should be right range of y+. As 30<y+<300 is valid for flat plate... and my geometry is not different.

Also I did not understand why there is an upper limit for y+

[Far]

http://www.linkedin.com/groups/y-in-...032.S.85359857

Few clarifications.

If you are predicting heat transfer then you need to resolve the thermal boundary layer as well. So try to keep the low Y+ as possible (~1) and sufficient no of nodes in boundary layer.

My question is whether my modeling is valid inspite of y+>300 as my reynolds no is very large ?? If not what should be right range of y+. As 30<y+<300 is valid for flat plate... and my geometry is not different.

Do not confuse the low Reynolds number turbulence models with Flow Reynolds number.

[prince_pahariaa]

Hey far,
Thanks for taking interest.

I get your point Far. As heat transfer near walls are important my grid should be very fine to capture thermal boundary layer. And as my Y+ values are very much larger than 1, that means my results are not at all good.

That imply i should not use wall function at all and for this i need to opt for low Re number models. .

That mean i need to re-mesh my geometry and start from the scratch ??
Way to go..

I need to clear one point if u can help me out. In fluent i do not find any wall function for K-omega model. Through some literature i also get to know that k-omega is better in predicting near wall profile. but if its the case why my temperature profile is not changing with K-omega model ?? I also get literature which suggest wall function for k-omega model. Little contradiction here.. and i am confused. if u can tell me about it or point out some reference.

[Far]

I need to clear one point if u can help me out. In fluent i do not find any wall function for K-omega model. Through some literature i also get to know that k-omega is better in predicting near wall profile. but if its the case why my temperature profile is not changing with K-omega model ?? I also get literature which suggest wall function for k-omega model. Little contradiction here.. and i am confused. if u can tell me about it or point out some reference.

With K-omega model works best with the Y+ < 1-10 and there is no point to predict the near wall profile with the wall function as wall function uses the empirical correlations to give boundary condition to wall conditions to solver. On the other hand integration to wall approach solves the flow upto wall.

However with y+ greater than 30, fluent switches to wall function approach automatically.

[LuckyTran]

Originally Posted by prince_pahariaa

Also I did not understand why there is an upper limit for y+

the law of the wall is valid only for the logarithmic layer of the boundary layer. Hence, to correctly apply the law of the wall, your first grid point (or y+) needs to be located in the logarithmic region. For very large y+, it becomes very questionable whether you are in the logarithmic region, hence the validity of using the law of the wall is lost.

Originally Posted by Far

Few clarifications.

If you are predicting heat transfer then you need to resolve the thermal boundary layer as well. So try to keep the low Y+ as possible (~1) and sufficient no of nodes in boundary layer.

A finer mesh always helps, but resolving the boundary layer is not the reason to target a y+ ~ 1. For substances such as air, the themal boundary layer is even larger than the hydrodynamic boundary layer. The thermal boundary layer is even more easily resolved than the hydrodynamic layer. For other materials like water the thermal boundary layer is only slightly smaller. Not unless you are working with very high Prandtl number fluids does the thermal boundary layer become very thin. The reason you want the smaller y+ when solving heat transfer is to more accurately compute the wall temperature gradient, and hence the heat flux.

That said, the temperature boundary conditions at the wall are handled very similar to the the velocity boundary conditions so the use of large y+ is not totally wrong. The wall modelling approach for both high and low Re schemes for the energy equation is nearly identical to the velocity.

[hectorgabriel85]

Originally Posted by Far

With K-omega model works best with the Y+ < 1-10 and there is no point to predict the near wall profile with the wall function as wall function uses the empirical correlations to give boundary condition to wall conditions to solver. On the other hand integration to wall approach solves the flow upto wall.

However with y+ greater than 30, fluent switches to wall function approach automatically.

Hi Far, where I can get a reference for your statement of y+<1-10?

[LuckyTran]

Originally Posted by hectorgabriel85

Hi Far, where I can get a reference for your statement of y+<1-10?

The omega model is designed to be integrated to the wall and the boundary layer completely resolved. There are wall functions for omega at high y+, but that is a bandaid.

If you need a reference, you can consult the book Turbulence Modeling for CFD by David Wilcox who created the k-omega model.

[hectorgabriel85]

Originally Posted by LuckyTran

The omega model is designed to be integrated to the wall and the boundary layer completely resolved. There are wall functions for omega at high y+, but that is a bandaid.

If you need a reference, you can consult the book Turbulence Modeling for CFD by David Wilcox who created the k-omega model.

Hi Tran. I found a reference to y+ < 2.5 in the Wilcox book you recommended. Page 385 3rd Edition 2006.