Discussion Forum

I'm investigating an FSI problem.
I'm interested in the frequency response of a membrane of a reservoir with water but I have found that a frequency analysis with FSI can be a bit tricky.

Instead, I have done a pulse response of the membrane. By applying a force pulse using a smoothed step function on the top of the membrane. I then decrease the time steps and smoothness of the pulse to find the point at which the membrane will 'ring'. From this ringing, I can deduce the first resonant frequency.

The ringing can be seen by plotting the centre deflection (fsi.disp) as a function of time. I've included this plot, as a reference, the applied step function is plotted as well, (scaled so you can compare it).

Currently, I'm using a 2D model with 'coarser' mesh.

For the solver, I clear the update scaled absolute tolerance in the time stepping field.
I find I need to switch to 'fully coupled' solver at shorter time intervals for the simulation to work.

I would like to expand this simulation to a 3D problem. Currently I'm solving for: 29730 dof and the solution time is 344 s. (5 minutes, 44 seconds). (core i3, winXP 64-bit, 8GB of ram)

When I switch to 3D the computing time is going to increase a lot.

The first question is: Is this a valid method of simulating the device, to get an idea of the resonance frequency?

Secondly: what can I do to decrease the computation time, keeping in mind that in 3D the mesh will be much larger.