BeamSolver3D: wave propagation

[palyvar]

Hi,
I am a newbie in FEM and I am attempting to simulate elastic wave propagation using transient simulation and BeamSolver3D. However, the waves do not travel. I was able to make it work using StressSolve and with a full 3d model of the beam but I am wondering if it is possible using 1D beams. I am attaching a GIF of the simulation result (U), the SIF file and an archive of the whole setup (sif + mesh).

[kevinarden]

I can only conclude that the beam solver does not include the dynamic forces of acceleration and velocity. The motion on the far side of the circle is moving but only due the the rings stiffness as opposed to the solid model which is excited by dynamic forces. I plot the displacement of the far side node.
beam model

beam.png

(76.54 KiB) Not downloaded yet

solid model

elementp2.png

(70.32 KiB) Not downloaded yet

I test this by leaving the load on only for the first 5 steps, then turn it off, the problem should still continue to oscillate, the beam model does not
but the solid model does.

solidring.png

(72.09 KiB) Not downloaded yet

[palyvar]

Thank you very much for your response! Then, I'll simply continue only with the solid model.

I was confused because the model's documentation says something about effects of inertial forces in transient simulation, but the effects are probably different from my expectations (the Density parameter has observably some impact).

[kevinarden]

I think by inertial in the beam solver they may mean gravity turned into a static body force, not acceleration.

[raback]

Hi

I looked at the code and at least the M(d^2u/dt^2) terms are there for transient simulation.

I looked at the sif file and one thing that caught my eye was the sloppy linear system tolerances. If your timestep is very short with sloppy tolerances it could be that the previous solution fulfills the equation and nothing is done. I would recommend using "direct" instead of "iterative" for 1D meshes as they result to very narrow bandwidth.

-Peter

[kevinarden]

I used direct, mumps, on my studies. And P:2 elements on the solid models.

[kevinarden]

I have tested an tried everything I can think of, but the solid model and the bean model yield very different results. The solid model appears to be a more expected result.

solid.zip

(13.24 KiB) Downloaded 34 times

beam.zip

(6.21 KiB) Downloaded 35 times

[kevinarden]

I take test case
https://github.com/ElmerCSC/elmerfem/tr ... Cantilever

and change from steady state to transient and the solution fails immediately

ERROR:: ComputeChange: Numerical Error: Norm of solution appears to be NaN

[raback]

Hi

How about if you set for the BeamSolver section:

Code: Select all

Use Global Mass Matrix = True

-Peter

[kevinarden]

Yes that fixes the problem, solutions look good now.
Thanks.