Hi Dear Friends,

In recent days I am trying to simulate the vortex sheeding behind square cilinder, in case of turbulent regime. My questions are:

a) What is the domain in order to capturate the wake; b) What are the correct boundary contions in this case.

Thanks. Valdemir.

(1). If you put a square cylinder in a very confined space, then it will restrict the wake oscillations. (2). As a matter of fact, people have being using a splitter plate attached to the back end of the body to reduce the wake oscillation. (3). So, my suggestion is: increase the computational domain as large as possible. This can be accomplished by using non-uniform mesh. (4). With a large computational domain, the effect of boundary condition will be minimized. That is, you can use simple boundary conditions in any way you like. If you reduce the computational domain, you will have a lot of problems with the boundary conditions. Don't try that, unless it is absolutely necessary.

What is a square cylinder? Is it rectangular or is it round or is it a rounded rectangular?

Astrid

Streamwise direction :

Total : 21.0 D

Upstream : 5.0 D Downstream : 16.0 D

Acrosss Stream direction

Total : 16.0 D

Spanwise direction : periodic b.c.

Total : 2.0 D

Here, D is a dimension of square cylinder section.

I am wondering how or where you get this dimensions. I have a feeling that none of these is sufficient to minimize boundary effects. This is based on my experience with circular cylinders.

Yes, You are right. The bigger computational domain, the better results you can get. But with this computational domain, I got reasonable results. The discrepancy between experimental results and LES output of mine is not only the problem of the size of computational domain, but also turbulence model, control volume size, ... etc. Please refer other papers (maybe the domain size is similar to my suggestion).