[anubasu]

Hi everyone,

I am a bit confused. For ke turbulence model y+ should be around 30, but do we need to give inflation layers? I am modeling fully turbulent partially-filled pipe flow using interFOAM. If without inflation layers I can get y+ around 30 is it not sufficient? I am a bit confused. ALso for komega, y+ should be around 1, so for that is inflation layer required?

Any help is greatly appreciated. Thanks

[NotOverUnderated]

Originally Posted by anubasu

Hi everyone,

I am a bit confused. For ke turbulence model y+ should be around 30, but do we need to give inflation layers? I am modeling fully turbulent partially-filled pipe flow using interFOAM. If without inflation layers I can get y+ around 30 is it not sufficient? I am a bit confused. ALso for komega, y+ should be around 1, so for that is inflation layer required?

Any help is greatly appreciated. Thanks

Using a reasonable y+ is NOT sufficient. You must put sufficient number of cells to resolve the boundary layer itself.

[anubasu]

Originally Posted by NotOverUnderated

Using a reasonable y+ is NOT sufficient. You must put sufficient number of cells to resolve the boundary layer itself.

Ok, thanks. How to decide the first layer thickness? I must keep reducing it until it reaches the desired y+ value right? And if I am interested near free surface of the flow, will it cause any changes if I use inflation layers or not?

Thanks.

[JBeilke]

Originally Posted by NotOverUnderated

Using a reasonable y+ is NOT sufficient. You must put sufficient number of cells to resolve the boundary layer itself.

You either use wall functions and an Y+ of 30 .. 200 OR you use a low-reynolds approach, where you resolve the boundary layer with a suitable fine mesh.

[NotOverUnderated]

The boundary layer is composed of:

- Inner layer: viscous sublayer + buffer layer + logarithmic layer
- Outer layer: from the edge of the logarithmic layer to the edge of the boundary layer itself.


It does not matter if you're using wall functions or resolving the viscous sublayer; you must have a sufficient number of cells that cover the entire boundary layer (inner layer + outer layer).

Imagine this scenario: You use a mesh with a first cell center from the wall that satisfies y+ < 1 but with a growth rate of 5.0. Yes, the y+ is satisfied, but certainly, there will be no sufficient number of cells that will cover the entire BL.

The bottom line: You must use a proper y+ with a growth rate that guarantees that you sufficiently resolve the entire boundary layer.

[anubasu]

Originally Posted by NotOverUnderated

The boundary layer is composed of:

- Inner layer: viscous sublayer + buffer layer + logarithmic layer
- Outer layer: from the edge of the logarithmic layer to the edge of the boundary layer itself.


It does not matter if you're using wall functions or resolving the viscous sublayer; you must have a sufficient number of cells that cover the entire boundary layer (inner layer + outer layer).

Imagine this scenario: You use a mesh with a first cell center from the wall that satisfies y+ < 1 but with a growth rate of 5.0. Yes, the y+ is satisfied, but certainly, there will be no sufficient number of cells that will cover the entire BL.

The bottom line: You must use a proper y+ with a growth rate that guarantees that you sufficiently resolve the entire boundary layer.

Thanks NotOverUnderated. How to know the first layer thickness and growth rate? I am doing a free surface pipe flow using interFOAM. So I have guessed the first layer thickness as 1e-5 and calculated y+, and keep changing it until it becomes around 30 for ke model right? And for the growth rate? Should I study any book?

Many Thanks

[arjun]

I would want first point at Yplus of 30. Just where the buffer layer ends.
Then i would use law of wall from here.
If i have to have grid points starting from yplus of 1 (or 5) to some points in 5 to 30 range then then some points after that then i do not really need any wall law. Everything is pretty much resolved for shear rate.

[anubasu]

What I can tell is, for internal flows wall function can do the job to resolve the BL but for external flows where drag is very important we cannot rely on wall functions-we have to resolve the BL by making a very fine mesh physically.
Yes growth rate is very important to resolve the BL accurately. I hope I am correct.

[anubasu]

I am a bit confused about the y+value that I will take while using ke and komega models? is it around 30 and 1 respectively for pipe flow?

[NotOverUnderated]

Originally Posted by anubasu

I am a bit confused about the y+value that I will take while using ke and komega models? is it around 30 and 1 respectively for pipe flow?

Let me try to alleviate the confusion:

k-epsilon is not a single turbulence model, there are hundreds of these models and you should know that what is referred to as k-epsilon is the standard k-epsilon model which is a high Reynolds number model. Be careful though that High Reynolds number name here is misleading and has nothing to do with the Reynolds number of your object. A high-Re model is a model that can't be integrated all the way to the wall and it requires using Wall Functions. However, there are versions of the k-epsilon models that can be used all the way down to the wall (achieved by using damping functions). You guessed it, these models are termed "Low-Re models". (advice: never use low-Re kepsilon models, they suffer from a number of numerical issues and should be avoided all together).

On the other hand the k-omega family are low-Re models by construction (they are valid all the way down to the wall) which means that they can be used without wall functions. Hence they require y+ to be less than 5. But it also important to keep in mind that they can be used with wall functions as well.

Conclusion.

- The standard k-epsilon model is a High-Re turbulence model and requires the usage of wall functions (y+ must be greater than 30).

- The k-Omega model is a low-Re model that can be used with and without wall functions.

I hope this helps.

[arjun]

Originally Posted by anubasu

What I can tell is, for internal flows wall function can do the job to resolve the BL but for external flows where drag is very important we cannot rely on wall functions-we have to resolve the BL by making a very fine mesh physically.
Yes growth rate is very important to resolve the BL accurately. I hope I am correct.

You are correct. In separating flows there can be issues so its better to resolve the boundary layer if possible. (then we do not really need wall functions).

[anubasu]

Hi arjun and NotOverUnderated,

Thanks, What I understood by reading some lectures is- if I can resolve the BL by a very fine mesh, then no need of wall functions.
But to save CFD time, we use a wall function for certain flow types near the wall without resolving the mesh. In this case, there is no need of inflation layers and growth rate.