FSI oscillating plate
 

Dear users,

I've succesfully made a FSI of an oscillating plate where I apply a pressure on one side of the plate (using CFX and ANSYS Workbench flexible dynamics), and then release it, then the plate will move like the first modal bending mode.

http://img12.imageshack.us/img12/6376/1604200925507.jpg

However now I want it to let it move on the first bending mode without decaying amplitude, so I applied a pressure function:
100*sin(360*50*time)

However the problem is that there is an interference with the resonance frequency of the plate. When the plate resonance frequency is slightly different to the frequency of the applied pressure you get a result like this:

http://img12.imageshack.us/img12/940...2009211210.jpg

The above picture gives the amplitude of the tip. Is there a way to avoid this and obtain beautiful sinusoidal deflections?

I tried: using other material properties; changing pressure frequency to match exactly to the frequency of the plate (found via modal analysis), however still stays; randomly changing the pressure frequency.
Update: I changed the pressure function in 0.1/time*sin(time*360*50) and now I get a better result:

http://img19.imageshack.us/img19/7725/1604200940644.jpg

But its not the best solution (non uniform amplitude in beginning) and very dependant on the frequency difference.

Hope to hear from you soon. Thanks alot!!!

Hi,

If the points on the chart represent your timestep size then you need to make your timestep much finer. Do a timestep sensitivity analysis to work out how small the timestep needs to be for accurate results. While you are at it, do likewise for convergence and mesh size and the FSI parameters too.

Glenn Horrocks

Thanks for your reply. Could you provide me a link/reference to the timestep size analysis (I'm not familiar with it).

Hi,

The basic idea behind any sensitivity analysis is very simple. For timestep sensitivity, do a simulation with some time step size, then do the same simulation with a significantly different timestep, maybe half the timestep. Compare the important output parameter between the two simulations. If the change in output is acceptable relative to the accuracy you desire then you can use the coarser timestep. If not accurate enough then use a finer timestep again and keep refining the timestep until you achieve the level of accuracy you wish.

This is only the most basic description of it. A much more thorough description can be found (particularly concentrating on mesh sensitivity) in "Computational Fluid Dynamics" by Roache, or a brief version at http://journaltool.asme.org/Template...umAccuracy.pdf

Glenn Horrocks

Hi
I do not know what is your real objective behind constant amplitude. There may be a workaround depending on what you finally want to achieve. If you are just interested in just a quasi-steady (periodic solution or without any variation over one cycle to the other) solution and not in real dynamic response of the structure, you may just set the timint,off in the Ansys solver.
The consequence of this action: you will not see any dynamic response from the system. The present approach may be applied to the problems which are harmonic or periodic in nature. In this case your amplitude of the structure, will always be constant and you still can investigate steady-state influence of fluid or other forces on the structure. However you have to apply a periodic function (as you have done in your case) to excite your structure at a given frequency.

Hope this will help.

Vivek