Coupled Fluid-Structure Eigen Analysis

[raback]

Hi Kevin,

The HelmholtzStructure cases are hierarchical: structure -> acoustics. The FsiShoebox* cases are strongle coupled: structure <-> acoustics.

-Peter

[kevinarden]

using the ShoboxFsiEigen2d test case with
Density = 2710.0
Youngs Modulus = 70e9
Poisson Ratio = 0.3
The beam frequency should be 87 Hz in air and 55 Hz in water

with

$ AirDensity = 1.0
$ SoundSpeed = 300

ElmerSolver reports
EigenSolveComplex: EIGEN SYSTEM SOLUTION COMPLETE:
EigenSolveComplex:
EigenSolveComplex: Number of converged Ritz values is: 10
EigenSolveComplex: Computed Eigen Values:
EigenSolveComplex: 1 (1.04082768848625157E-004,-1.08456037487612890E-008)
EigenSolveComplex: 2 (-7.9012503238339047,-4.0733589099508940)
EigenSolveComplex: 3 (-7.8696137561585102,4.1349833785406878)

Changing the fluid density results in the imaginary part of the complex eigenvalue 1

results.png (15.77 KiB) Viewed 139 times

Any fluid density over 18 results in a solver error, water would be 1000.0

[eschenk]

I finally got some time to build out a simplified 2D version of the guitar model.

I think the fluid structure modal analysis is working correctly, or at least I get results that are consistent with my expecations, and driving changes either the fluid or the structure causes coupled changes in the joint system reaction at each frequency in a scanning simulation.

For the transient simulation of a coupled structure/fluid with the wave solver for the fluid, I get really strange results. I tweak the structure with a transient load, expecting it to vibrate. It goes through 1/2 cycle of vibration, and settles into a position slightly bent in the direction of the original force with exponential decay in the ongoing motion. The fluid oscilates in response, but also goes to zero quite quickly. Changing the fluid density or sound speed appears to have no effect on the structural response, which is incorrect in any case.

Here is a screenshot of a graph comparing displacement for structural oscilaiton without the fluid coupling vs the joint system, where I also show average pressure. The structure only system oscilates quite a bit as expected with a resonable decay of the oscilation. The coupled system (lower graph) stops moving very quickly.

Screenshot 2024-05-06 193951.png

(74.11 KiB) Not downloaded yet

I've also attempted to do the joint Eigen system analysis in 2D here, but the solver crashes on me rather than coming to a solution.

I've attached a zip file of the geometry and the sif files I've been using.

In the zip there are 9 case studies as follows:

case_air_eigen.sif - Find the eigenmodes of the air volume only. Seems to work as expected
case_air_scan.sif - Scan over frequencies with harmonic analysis to find the eigenmodes of the of the air. Agrees on the first mode. Some variance in later modes. See notes in the files for further thoughts.
case_box_eigen.sif - Find the eigenmodes of the guitar box structure only. Seems to work as expected.
case_box_scan.sif - Scan over frequencies with harmonic analysis to find the eigenmodes of the guitar box structure. Matches the structural Eigenmode solution.
case_box_transient_scan.sif - Transient solve of the structure only in response to an impulse. Top of the graph shown above.
case_joint_eigen.sif - Attempt to do an eigenmode solve for the joint system. Crashes.
case_joint_eigen_test.sif - Frequency space simulation of a single frequency of the joint system. Seems to work.
case_joint_scan.sif - Frequency scan of the modes of the coupled system. This appears to generate correct results as mentioned above.
case_joint_transient_scan.sif - Transient solve the joint system in repsonse to an impulse. Bottom of the graph shown above. The results here are not correct as discussed above.

Taken with Kevin's earlier notes, I think that the conclusion is that while the frequency space harmonic analysis coupling works between Wave Solver and the elastic structural solver, the time coupling does not work and the Eigenmode frequency analsysis does not work either.

[eschenk]

One test I would like to do that I have gotten to is to try frequency scanning on the beam in fluid case Kevin gives here to see if the fluid coupling gives the expected results. I probably won't get to that for a week or so given my free time availability for the next bit.

The idea would be to scan and look for peaks in the curve of the average pressure and/or average displacement (or better yet the integral of the pressure and/or displacement). These should correspond to the frequencies of the eigen modes of the joint system.

[eschenk]

I got around to testing the 2d beam case. I'm not sure what I am seeing is what is expected or not.

I ran a scanning simulation of the immersed beam in both air and water.
In Air the first resonance of the beam is at around 84 Hz.
In Water the first resonance of the beam is around 51 Hz.
This is with quadratic structural elements, but only first order fluid elements. When I try quadratic fluid elements as well, the solver crashes.


I don't have an analytic solution at hand for these case, but looking at the results documented here: https://www.comsol.com/blogs/natural-fr ... sed-beams/ this seems roughly in line with expecations.

Kevin, you had mentioned the results should be 87Hz in air and 55 Hz in water, however the vaccum first resonance of the model shows as 84.9 Hz when I solve for the Eigen Frequencies of the beam alone. How did you arrive at the target frequencies you mentioned?

Graph of the pressure peaks in the two simulations, air on the top, water on the bottom.

Screenshot 2024-05-12 133844.png

(47.92 KiB) Not downloaded yet

If I substiute the Wave solver for the Hemlholtz solver, I get about the same results in air, but wildly different results in water, with the resonance being around 2 Hz! I can't explain this, as I would expect the same or slightly different results, as both are solving the Helmholtz equations, the first in frequency space, the second in the time domain. I'm not sure if this indicates a bug in the strong coupling with the wave solver, or a problem in the solver itself.

Zip file for the cases attached. You can switch to the Wave solver by changing around the commented blocks at the start of each solver to see the effect.

I need to go back and revist the simplified guitar case now, as I used the wave solver there, but when I switch to the Helmholtz solver I get a crash in the joint case. I suspect a problem with the fluid/structure boundary in the geometry, but I need to track that down now.

At this point, my conclusion is that for the strong FSI coupling only the Stress Solver and Helmholtz Solver can be succesfully coupled in the time frequency domain right now.

Is this a known bug, or have I got a problem in either my geometry or boundary conditions that is causing trouble with the Wave solver?

Cheers,

Eric