[Gerhard]

Good day,


In a situation where you have walls where the near-wall flow is of importance and other walls where the near-wall flow is not important, how should you treat the unimportant walls?


Here is an example: Flow enters a large plenum chamber, then flows through a fan, and finally discharges to the atmosphere through a diffuser.


Not wanting to waste cells at the plenum walls (which are not that important), the y+ values there might be in excess of 1000.


Since the near-wall flow in the vicinity of the fan and especially in the diffuser is important, more cells will be used to achieve 30 < y+ < 100 for wall-functions, or even y+~1 if integration through the sub-layer is desired.


The question: what should the boundary conditions of the unimportant walls be where the y+-values are, let's say y+ > 500?


These are my thoughts for the k-omega turbulence model using a wall-function approach:


Option 1

U
Important -> noSlip

Unimportant -> slip



k
Important -> kqRWallFunction

Unimportant -> zeroGradient



omega
Important -> omegaWallFunction

Unimportant -> zeroGradient

Option 2

U
Important -> noSlip

Unimportant -> noSlip



k
Important -> kqRWallFunction

Unimportant -> fixedValue of 1e-10



omega
Important -> omegaWallFunction

Unimportant -> fixedValue of 1e10

Any suggestions?
Thanks.

[Gerhard]

If I may reply to my own question...


This is my thinking. Please tell me if you do not agree.


Whether you treat these "unimportant" walls using either Option 1 or 2 (provided in the previous post), it should not influence the key results.
However, if Option 1 and 2 do not yield similar key results, then these walls are in fact not "unimportant".
In other words, they are important, and if that is the case, one should aim to achieve 30 < y+ < 100 if you which to model the near wall flow using wall-functions.

[Tobi]

Hi Gerhard,

a tricky topic. In my personal optinion, I would stick with option 3.

• A wall is treated as a wall so -> noSlip
• As we have a wall there, I would use wall functions even if y+ >> 100
• The important thing is that the first cell center is not in the viscouse sub-layer ( y+ > 17 -> Reference to Ferziger & Peric if I do have the number correclty in my head)

However, I believe that every guy who works in the turbulence field will tell us that all approaches are wrong

[Gerhard]

Hi Tobi,

Haha, I like your remark about the turbulence guys

I agree with what you say about using noSlip for walls.
Using wall functions for like y+ ~ 1000 just feels strange: You are then in the defect layer and wall functions are not designed to describe the flow that far away from the wall... But what else, right?

And as I mentioned earlier that you should probably just aim to have 30 < y+ < 100, it can really be a huge challenge when you are working with big systems of relatively complex geometry. Getting quality layers near the wall can then really be a pain and time consuming.

Thanks for your inputs. I appreciate it.

Regards,
Gerhard