[bastiencucuel]

Hello,
I'm Bastien and it's my first post on this forum.
I'm currently working to simulate the drag coefficient for differents 3D objects, but before doing that, I wanted to test my mesh with a very known case; the cylinder.
So I've done a 5mm diameter cylinder for a length of 15cm.
My fluid is air at 25 degres celcius and my inlet velocitie is 10m/s.

I want to make a 3D study because my other objects won't be symetrical like the cylinder.

I've done for my mesh a structured mesh far from the object and a finer and unstructured mesh near the object.
My turbulence model is SST model.

I've try a lot of simulations with differents meshes (more or less accurate) and I dont achieve to obtain a correcdt drag coefficient. I obtain one about 0.45 whereas it should be about one.
I've also tried steady and transient simulations with the quasi same result.

Thank you very much for your help
Bastien
PS: I'm surry if there are some mistakes in my post, I'm not quite fluent in english.

[singer1812]

It is critical that you get the separation correct on that type of geometry in order to get the drag coefficient correct.

Mesh independance in the boundary layer and the wake region is of vital importance.

Are you sure you are getting those correct?

[bastiencucuel]

thanks for the answer.
I don't know what you mean by "Mesh independance in the boundary layer and the wake region"

I've just done an unstrucutred mesh in a box near the object and a structured mesh outside this little box.
I join you a picture of my mesh in the little box
little box.jpg.

[singer1812]

You need to resolve the frictional and the pressure drag accurately to get a good answer on this.

Frictional drag will be a function of how well you resolve or model the boundary layer and how well you accurately predict the separation.

Pressure drag will be a funciton of how well you predict the sepatation and the how well you resolve the wake region physics. You might have issues dealing with the vortical region behind the cylindar.

Both of those are highly dependent on your mesh, inlet conditions, and turbulence model choice.

From the look of your pic, it doenst even look like you have any inflation on the cylindar surface. Also, it looks like you just used an octree mesh on the interior, as generated, from ICEM. I suggest, use blocking on the entire domain and produce hex mesh with elements packed in near surface (look up near wall suggestions in CFX help).

I also suggest you use a turbulence model that gets the separation correct and near wall physics suitable to your situation. You might want to start with SST, and depending on what you need this for and how accurate you need, you might have to switch to a transient simulation or even a DES/LES simulation.

[bastiencucuel]

thanks,
If I well understood you, you suggest to use a strctured mesh throughout the all domain, ins't it?
but I think it's important for my case to keep an unstructured mesh near the object because this cylinder test is just a test before other cases with much more complicated shapes, for whiches a structured mesh won't work.

[singer1812]

Ok, keep the unstructured, but use good inflation. And switch to a delauney mesh. You will get smoother mesh transition

[bastiencucuel]

OK thanks.
To do the inflation I have to go to compute mesh--> prism mesh or there are another option

[bastiencucuel]

I'm using ICEM

[singer1812]

Yes that will work, but only give you the default inflation parameters, which probably will not be sufficient.

I suggest you read up on inflation (boundary layers) in ICEM and CFX help.

If you still have mesh questions regarding inflation (prism), look into the ICEM section of the forum.

[bastiencucuel]

OK thanks
I 'll let you know if I have better result

[bastiencucuel]

when O go to prism mesh to make an inflation it gives me a structured mesh, and I don't think it will work for my other objects:
inflation.jpg

[singer1812]

Your prism layer is way to large. It should only be the thickness to capture the boundary layer size you expect. And I think you dont understand the neccessity of using inflation near boundries.

I think you need to read up some introduction to cfd materials.

Might I suggest "An Introduction to Computational Fluid Dynamics- The Finite Volume Method"; HK Versteeg & W Malalasekera

[bastiencucuel]

ok
thank you very much