CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   CFX (http://www.cfd-online.com/Forums/cfx/)
-   -   Flow around a cylinder with k-epsilon model (http://www.cfd-online.com/Forums/cfx/117290-flow-around-cylinder-k-epsilon-model.html)

 CFX13 May 6, 2013 07:22

Flow around a cylinder with k-epsilon model

5 Attachment(s)
Hello,
I'm trying to simulate a simple flow around the cylinder using CFX. The Reynolds number is 2000, and I use the k-epsilon model as turbulence model. the temperature difference between the cylinder wall and the fluid is 1K. the simulation is running and it also converges but there is no karman vortex Street in this case. But when I simulate the same problem as laminar, then I can see this swirl. the following fotos shows the most important settings for this problem.
regards

 Lance May 6, 2013 07:46

Re = 2000 and using turbulence model? I would guess that the turbulence model introduce additional dissipation that removes the vortex street. That would explain why you see the vortex street when you are using a laminar approach.

Why are you using gravity? Is buoyancy important in your flow?

 RicochetJ May 6, 2013 08:29

You have low Reynolds number flow with questionable turbulence. Also it's flow over a cylinder so I assume you have some sort of vortex shedding. I would'nt use k-eps if you were to use a turbulence model. Try k-omega SST, with low turbulence intensity at the boundaries, also try it with and without the gamma-theta transitional model switched on. That transitional model is optimized for low external Re flows.

But before doing that I'd question if it's even turbulent.

 CFX13 May 7, 2013 03:59

1 Attachment(s)
hi,
first of all thanks for your tipps. i have increased the raynolds number to 4000, but still no karman street can be observed (see the attachement). i also tried K-omega SST model without considering buoyant effect and obtained the similar results as using K-epsilon model.
regards.

 ghorrocks May 7, 2013 19:41

You need a low-dissipation numerical model to get the vorticies. Are you using a second order space and time discretisation scheme? And time steps small enough (adaptive timestepping is STRONGLY recommended)?

 irinsun May 8, 2013 11:58

we used high resolution scheme for this problem.

To reduce the additional dissipation caused by turbulence model, now I am going to try CDS for advection scheme, second order backward euler for transient scheme.
which scheme will u suggest for turbulence numerics. Should i also use second order since it seems that first order is recommended for turbulence equations.

thanks.

irinsun

 oj.bulmer May 8, 2013 14:14

Are you using scalable wlal funcition? How about using automatic wall function, and using Y+ values of - say close to 1 etc? This will capture the phenomena near wall well and will also work if there are laminar/transition flow structures around your cylinder.

I wouldn't consider k-eps here because of low turbulence and its dissipation. kw-SST is a better choice. Also, refining the mesh might help in reducing dissipation.

OJ

 RicochetJ May 8, 2013 16:40

Quote:
 Originally Posted by oj.bulmer (Post 426167) Are you using scalable wlal funcition? How about using automatic wall function, and using Y+ values of - say close to 1 etc? This will capture the phenomena near wall well and will also work if there are laminar/transition flow structures around your cylinder. I wouldn't consider k-eps here because of low turbulence and its dissipation. kw-SST is a better choice. Also, refining the mesh might help in reducing dissipation. OJ
Hi OJ. Surely having a Y+ less than 11 using k-eps is pointless as it will use scalable wall functions.

 oj.bulmer May 8, 2013 16:57

I just suggested that omega based model with automatic wall function might fare better than epsilon based model with scalable wall function in this particular case, as he was considering. The use of k-eps is to be avoided in this case, primarily because of disspation it will introduce, eating up all the transient/turbulent instabilities that give rise to vortices. In cases of higher Re, k-eps may fare better.

OJ

 ghorrocks May 8, 2013 22:48

Agreed, SST or k-w are the turbulence models to try here.

But second order time stepping is a MUST. You will need this.

 RicochetJ May 9, 2013 06:36

Quote:
 Originally Posted by oj.bulmer (Post 426210) I just suggested that omega based model with automatic wall function might fare better than epsilon based model with scalable wall function in this particular case, as he was considering. The use of k-eps is to be avoided in this case, primarily because of disspation it will introduce, eating up all the transient/turbulent instabilities that give rise to vortices. In cases of higher Re, k-eps may fare better. OJ
Oh absolutely, I do agree. In fact I did recommend the k-omega SST in my earlier post. I just thought you were referring to a Y+ of less than 1 for the k-eps.

 All times are GMT -4. The time now is 21:42.