CFD Online Discussion Forums

CFD Online Discussion Forums (
-   CFX (
-   -   vortex shedding-perturb flow at beginning (

cfxuser1 April 8, 2009 16:53

vortex shedding-perturb flow at beginning

I am trying to obtain vortex shedding from a body and I cannot get it done. The flow is like symmetrical on both sides of the body, a clean undisturbed flow. I have tried so many things such as decreasing time step to 0.001 (I don't want to decrease any further otherwise it is going to take ages to finish), changing the blend factor from 0.75 to 1, changing flow speed. I am using SST turbulence model. Can anybody help me with this problem? I have read in other posts where people are saying to perturb the flow a bit at the beginning. I am new to cfd and most of the time I don't really know what am doing. Could someone please tell me how to perturb the flow at the beginning?

I am working on a project and I really need to be able to do this as soon as possible. Any help would greatly be appreciated.


ghorrocks April 8, 2009 18:53


In my experience these oscillating flows don't need to be kicked to start up, there is enough numerical noise around to do that by itself.

Are you using second order timestepping? That will make a big difference. Also, what is the Re number? Doing transient vortex shedding combined with turbulence modelling can be tricky for low Re flows.

Glenn Horrocks

cfxuser1 April 8, 2009 19:31

Hi Glenn,

Here is the model below.

This is a 2D simulation. I want the circle to generate vortices. It has a diameter of 4cm and the flow speed is 5m/s. Once the circle generates vortices, the thin rectangular film at the back of the circle will flap. Basically you can think of it as a flag with the pole modelled as the cirle and the flag as the thin strip. The reynolds number is around 14000 taking the circle into consideration. The specified blend factor is 1 and the transient scheme is 2nd order backward euler.

Also in Default domain>Fluid models, I chose SST turbulence model which i think is good and then there is an option for transitional turbulence. I do not understand what this option does but I've ticked it and let it on fully turbulent. I don't know if this is right.

What do you think is the problem? Please help.

cfxuser1 April 8, 2009 19:35

I tried posting the image of the model. It didn't work. Well it is just as i described above. There is a very tiny gap between the circle and the thin rectangle. It is just as a flag with its pole as sketched below.


My only concern is to get the circle to shed vortices.

cfxuser1 April 10, 2009 23:06

Please, can anybody give me an idea about what i can do. I ran a simulation with the circle only and i go vortex shedding. But when i put the flag in front of it and apply the same cfx-pre settings, there is no vortex shedding. Why is that? I don't understand. Please anybody has any idea why is it so?


ghorrocks April 13, 2009 21:29


If you are running with a turbulence model you may need to kick it to start it off. Maybe give the initial condition a small asymmetry? Also you can start running a laminar flow simulation which should go asymmetric quickly then change over to a turbulent simulation.

Also with a turbulence model you will need to carefully select the timestep size to resolve the fluctuating vortex field but not resolve the small scale turbulence.

Glenn Horrocks

cfxuser1 April 16, 2009 06:42

Hi Glenn,
when you say give the initial condition a small asymmetry, can i do that by putting a y-component of velocity at the inlet or does it need to be defined somewhere else? Also when you say initial condition, do you mean initial global conditions? I am sorry if am asking too much, i don't really know much. What if I run a simulation with a x- and y- velocity components long enough and use this as initial guess/values for a 2nd simulation with the only the required normal velocity at the inlet, would this work?

And my timestep is 0.001s which is small enough i think.


ghorrocks April 16, 2009 08:21


Yes, a small y-component somewhere initially should do it. In the global IC is fine. Using a run as an IC can also work.

Why do you say 1ms is small enough? Unless you have done a sensitivity check you cannot know. For the record I start my current series of simulations with timesteps of 1e-11s. If I run them any bigger I get errors creeping in - a sensitivity analysis told me that.

Glenn Horrocks

All times are GMT -4. The time now is 02:00.