[cecilia_xiao559]

Hey all,

I am modeling a complicated flow domain with rotating walls, inlet velocity in components. the Reynolds number in my model is in a big range (as low as hundreds to 10^6) with some part of the geometry being a stagnate zone. I started with a relatively fine mesh and the initial run resulted in max y plus number in hundreds. I used the y plus adaption method to modify the mesh for the wall treatment. but I am not sure what would be a good range for this particular case because the flow is not uniform (varies a lot in the domain).

anyone have any suggestions for the y plus value?

Thank you!

[spl]

The Y+ you need depends on the turbulence model you're using and what your trying to get out of the simulation. Is it important to accurately capture the boundary layer? What is it your modelling?

[cecilia_xiao559]

Originally Posted by spl

The Y+ you need depends on the turbulence model you're using and what your trying to get out of the simulation. Is it important to accurately capture the boundary layer? What is it your modelling?

Hello,

Thank you for your reply.

I am trying to simulate the flow in a rotating rotor. we are more focused on the flow pattern but at the same time, we want to see how much the flow cools the rotor. So we want to get a relatively accurate result/estimate from the simulation. I started with k-epsilon to save computing time. Is this model okay? what would you recommend? what would be the corresponding Y plus value for the model?

Thank you,

[spl]

This might be of some help -

http://www.cfd-online.com/Wiki/Best_...omachinery_CFD

K-e uses wall functions so a larger Y+ is acceptable. K-w SST is commonly used in turbomachinery, this is low Reynolds turbulence model and requires a Y+~1. The fluent theory guide covers the approaches of different turbulence models in the near wall region so take a look at that too.

[FMDenaro]

if you need an evaluation of the heat flux at walls, you need to refine much the grid near walls... Unfortunately, 2-3 computational nodes could be necessary within y+ =1 which drive to a very expensive computational cost.

On the other hand, if your geometry has some flow symmetry, you can assume a 2D geometry considering that the flow is statistically two-dimensional. That will reduce your computational cost

[cecilia_xiao559]

Originally Posted by spl

This might be of some help -

http://www.cfd-online.com/Wiki/Best_...omachinery_CFD

K-e uses wall functions so a larger Y+ is acceptable. K-w SST is commonly used in turbomachinery, this is low Reynolds turbulence model and requires a Y+~1. The fluent theory guide covers the approaches of different turbulence models in the near wall region so take a look at that too.

Thank your for the explanation. That is a very helpful link.

So in my model, the highest Re can go up to 3E6. Should this be considered to be a low Re turbulence model? the Re varies in a very large range so i wasn't sure which model works the best. Would you suggest me to change to K-w model for the Re or keep using K-e for the wall function?

[cecilia_xiao559]

Originally Posted by FMDenaro

if you need an evaluation of the heat flux at walls, you need to refine much the grid near walls... Unfortunately, 2-3 computational nodes could be necessary within y+ =1 which drive to a very expensive computational cost.

On the other hand, if your geometry has some flow symmetry, you can assume a 2D geometry considering that the flow is statistically two-dimensional. That will reduce your computational cost

Hello,

Thank you for your reply.

2d is not a option for my project so I will have to keep it a 3d model. I am also thinking about adding inflation layer to the walls for better mesh. is it gonna to solve the 2-3 computational node requirement? what would you recommend on finer mesh within y+=1?

Thank you~

[FMDenaro]

Originally Posted by cecilia_xiao559

Hello,

Thank you for your reply.

2d is not a option for my project so I will have to keep it a 3d model. I am also thinking about adding inflation layer to the walls for better mesh. is it gonna to solve the 2-3 computational node requirement? what would you recommend on finer mesh within y+=1?

Thank you~

try first to estimate how many computational nodes you would have fulfilling the requirement....

[cecilia_xiao559]

Originally Posted by FMDenaro

try first to estimate how many computational nodes you would have fulfilling the requirement....

Hello,

I am a little confused about why I am aiming for y+=1. (I should also mention in my model the Re can go up to 3E6.) Is this a estimate for proper mesh on the walls? is it for a surface mesh or inflation layer? why 1? and also, how do i estimate it?

thank you~

[FMDenaro]

Originally Posted by cecilia_xiao559

Hello,

I am a little confused about why I am aiming for y+=1. (I should also mention in my model the Re can go up to 3E6.) Is this a estimate for proper mesh on the walls? is it for a surface mesh or inflation layer? why 1? and also, how do i estimate it?

thank you~

in the standard theory of turbulent boundary layer, you have the viscous sublayer at y+=O(1). Therefore, your grid should be fine enough to describe the behaviour of the sublayer if you want an accurate estimation of the stress (and heat) flux at wall.

[spl]

A low Reynolds turbulence model is one that fully solves the boundary layer through the viscous sub layer to give a laminar boundary condition. It therefore doesn't need a wall function. Its not necessarily for low Reynolds number flows. And as FMDenaro says for thermal losses through the walls a small y+ is essential.

[cecilia_xiao559]

Originally Posted by FMDenaro

in the standard theory of turbulent boundary layer, you have the viscous sublayer at y+=O(1). Therefore, your grid should be fine enough to describe the behaviour of the sublayer if you want an accurate estimation of the stress (and heat) flux at wall.

thank you for explaining this! it makes a lot more sense to me now.

I calculated the wall spacing aiming for y+=1 and it result in 0.0000035 m. how should i use this for meshing? Can you be more specific on how to add enough computational nodes on the wall? should i do a surface mesh or inflation layer or sth else?

Thank you

[spl]

This should be the height of your first cell if your aiming for a y+=1. And you'll need to do a mesh study to ensure you have enough nodes whatever method you use.

[FMDenaro]

usually, some law as cosine or hyperbolic tangent is used for the grid stretching at wall. The first computational node should be lower that y+=1 in such a way to have 2-3 nodes before the end of the viscous sublayer