Circular cylinder vortex shedding/strouhal number too high
 

I am trying to simulate the vortex shedding off a circular cylinder.

The mesh is a 3D mesh of 2 elements thick. I am using all tetra elements. Symmetry is specificed for the wall boundary of the domain to simulate a 3D case.

I attempted to simulate with SST model and LES.

With the SST model, I can observe the changes in the lift and drag. The lift osscilates with a constant amplitude. However, the strouhal number calculated suggests the shedding to be occuring too quickly. I am getting strouhal number of 0.28 when literatures suggest it to be closer to 0.20.

With LES model, strouhal is about 0.24, but amplitude of lift is constantly changing with each lift variation cycle.

My questions are:

1. What could be causing the higher than actual strouhal number?

2. Can I use tetra elements for such simulation? Or is hexa required?

3. Since SST is a rans model, does it mean it is not recommended for simulation of vortex shedding as compared to LES?

Les result

You cannot do LES with a mesh only 2 elements thick. You need a full 3D mesh.

This question is an FAQ: http://www.cfd-online.com/Wiki/Ansys..._inaccurate.3F

Are you in a Re regime where the shedding is 2D or 3D?

Can I use SST for vortex shedding with a 2 element thick mesh in CFX?

I read that LES does not work with tetra. Can SST work with tetra in transient cases?

Lastly, will a tetra mixed with hex mesh, as shown in ICEM tutorial, be suitable for my application?

If the flow is 2D then a 2 element thick mesh can be used for SST models. But not if it is a 3D turbulent flow.

Both LES and SST work fine with tet grids. You will have problems getting the dissipation low enough to work with LES but in principle it can be done. The more fundamental issue is you cannot do LES with a mesh which is smaller than the turbulence length scale - so 2 elements thick is not going to work for LES.

For flow over a circular cylinder you should be able to produce a proper 1 element thick mesh if you want to model this 2D or a 3D mesh with all elements hexas. With a simple geometry like this you should be able to achieve that.

As far as the Strouhal number is concerned:
If the computational domain is not wide enough (the distance of the side walls from the cylinder), the flow around the cylinder is faster than the inlet velocity. Calculating the Strouhal number from the frequency of the vortices and the inlet velocity will give too high values.

Hello,
can you support this statement by the literature or experimental results????? I really liked to find more information.


Many thanks.

The first part of the statement (that the flow is faster than it should be due to blockage factor) is pretty elementary. The second part of the statement (that it will result in too high Strouhal number) is true most of the time as Strouhal number generally increases with increasing Reynolds number (ref: https://en.wikipedia.org/wiki/Strouhal_number).