G'day,
I've tried several different settings to find a optimal scheme for 3D VIV simulations. Basicly, I'll keep using PISO since it's highly recommended for transient flow. (and with Green-Gauss Node Based gradient). But what about pressure interpolation(PRESTO!, body force weighted) and under-relaxation factors for pressure and momentum? I tried both pressure interpolation schemes, but didn't benefit from them as the pic shows:
"equations converged but mesh motion hasn't reached convergence criterion (1e-05)."
Although the user guide says that it's recommended to increase under-relaxation factors for all equations near 1 when using PISO, it seems to be too aggressive?
Does anyone know how to make the calculation stable? Please share with me. Thank you!
Note that my mesh was built well with minimum orthogonal quality of 0.3, and dynamic mesh was set correctly.
Hi,
Orthogonal quality of 0.3 is sufficient, try increasing it. The more it is closer to 1 the better results you get. I see you are using a time transient simulation, so, try using k-epsilon realizable and reduce turbulence intensity in the boundary condition. I guess that should work.
Arvind Pallela Hi,
I’ve tried many time to enhance mesh quality but failed. Face sizing, inflation and Multizone are used, and the inflation part consist of prisms; as for other meshes are tetrahedral. Do you have any better scheme to make quality higher?
To reduce complexity, the flow is laminar, namely, turbulence is not considered.
You could use body sizing or decrease the minimum element size and use adaptive mesh. It is more about keeping the mesh more uniform and fine. If you want better results on the surface you could use face sizing and inflation but the flow domain mesh size should be good enough. I usually use 1-2mm mesh size for flow domain. For large models I don't exceed by more than 5 mm minimum mesh size.
Laminar is used only in cases where Reynolds number is less than 2300. It means after a particular velocity your simulation will diverge. If you don't want much complex use Spalart-Allmaras. It's a one equation model so it's simple and would give you good results.
Arvind Pallela hi,
but my CV is 3D, if mesh size is too small, it'll take plenty of time to build meshes and computationally expensive as well. I just tried a test reducing element size from 8mm to 4.5mm, and then quality only increased by 0.03.
The Reynolds number in my case is around 500. It's a simple model.
Well, I wonder how far the distance between the cylinder and outlet is appropriate? (at least) In other words, if the diameter is D, then the distance of 6D is far away enough? Or is there any empirical formula for calculating that distance? (Considering computational resource, I can't build the domain too large.)
Milne Ciou hi,
I understand that you are concerned about the computational expense but you need to maintain minimum quality to ensure proper results. You could check it during the meshing process by using report mesh quality.
Also, in your case it is recommended you use at least Spalart-Allmaras or k-epsilon because the flow isn't laminar.
Arvind Pallela hi,
here is the video showing the cylinder gradually oscillating. Could you tell me how to distinguish if the flow is laminar or turbulence? I’m just so confused. Because usually we would say that Re=500 can be laminar.
Arvind Pallela hi,
thank you for reminding me that; I've found a book discussing 3D vortex shedding after cylinder. yes, once the Reynolds number exceeds 200, the flow will transit from laminar to turbulence.
but I noticed that even I reduce Re below 200, there was still downwash wake near the top of the cylinder. Is this laminar?
Milne Ciou hi,
It is difficult to say exactly if the flow is laminar or turbulent as critical Reynolds number differs from geometry to geometry and also the flow domain hydraulic diameter and cross section. Though low Reynolds number vortex street is possible it may not completely eliminate the wake.
- Badges
- Science topic
- Similar topics
- Computer Communications (Networks)
- Simulators