[cons013]

Hello,

I have been doing some DDES simulations in fluent over a wing. I have a close refinement region but would like to ensure the length is enough for reasonable isotropy in the wake to get captured by the subgrid model. I am struggling to find resources online on this topic, and have very limited computing power available to run an independence study on this.

Is there a way to know how long the refinement region should extend?

Cheers

[LuckyTran]

The short answer is make sure you have 10 pts (or x pts) across the wake.


The long answer is you need to estimate the taylor microscale at every location and determine where that scale grows larger than your refinement. The faster way of doing this is to eyeball the size of the wake at each region and get the Reynolds number.


The most likely outcome though is you'll find out you always need refinement downstream of the wake (within the wake) but laterally you can coarsen the grid closer to the freestream.

[cons013]

Originally Posted by LuckyTran

The short answer is make sure you have 10 pts (or x pts) across the wake.


The long answer is you need to estimate the taylor microscale at every location and determine where that scale grows larger than your refinement. The faster way of doing this is to eyeball the size of the wake at each region and get the Reynolds number.


The most likely outcome though is you'll find out you always need refinement downstream of the wake (within the wake) but laterally you can coarsen the grid closer to the freestream.

Thanks LuckyTran,

When you say 10 points, could you please elaborate? 10 grid points seems very small... This is a 3d simulation by the way.

[LuckyTran]

See Von Neumann's elephant.

Typically in CFD we say 10 points across a feature, although the rule of 10 gets broken by a lot of people, very often. It sounds like you are ready to break this rule already without even attempting the problem.

Yes 10 points can be very small. That is my point (pun intended). You can come up with all the complicated theories you want, but at the end of the day, you quickly find out that you are constrained by much simpler rules. I've only mentioned 10 points and already you complain that it is too much...


I'm willing to compromise to 8 points.

[cons013]

Originally Posted by LuckyTran

See Von Neumann's elephant.

Typically in CFD we say 10 points across a feature, although the rule of 10 gets broken by a lot of people, very often. It sounds like you are ready to break this rule already without even attempting the problem.

Yes 10 points can be very small. That is my point (pun intended). You can come up with all the complicated theories you want, but at the end of the day, you quickly find out that you are constrained by much simpler rules. I've only mentioned 10 points and already you complain that it is too much...


I'm willing to compromise to 8 points.

Apologies I should have worded that differently. What I meant was that 10 points seems like a very small number. To clarify is that 10 cells? Meaning the wake only needs 10 cells worth of spacing to capture it? I'm not fully understanding.

[LuckyTran]

10 pts is good enough for DNS. Across the wake means across the (thin) shear layer thickness. You repeat this at every streamwise location. 10 points across a feature is a statement of resolution.

It definitely does not mean 10 cells.