1D beam modelling

[Rich_B]

Hello,

I'm also using Windows, and I usually just install over the old version, no need to uninstall the old version. Right, there is not a 'check for updates' operation for Elmer. So when one wants to update Elmer, download the latest nightly Windows installer and run. I try to update Elmer every month or two, unless there is an interesting change in the meantime.

Rich.

[kevinarden]

The Linux version with the PPA has an update feature.

[amarfayyad]

Hi Peter and Rich,
I did the verification of the eigenvalue analysis for the 1D beam and it came in very good agreement apart from the torsional mode. The sif and verification data are attached. Couldn't attach the mesh as more than 3 attachments are not working. Thanks again!
-Amar

[amarfayyad]

The Linux version with the PPA has an update feature.

Good to know and thanks, Kevin.

[raback]

Nice work in getting the agreement! Have you checked the mesh convergence? I.e. if you increase the mesh density do the results improve? For 1D problems you should be able to increase the density quite a bit...

Note that for 1D problems the filling of the sparse matrix is rather favorable for direct methods. Hence I would always use them. Try:

Code: Select all

linear system solver = direct
linear system direct method = umfpack

or even

Code: Select all

linear system direct method = banded

Using direct methods eliminates the risk of divergence of the linear system.

-Peter

[amarfayyad]

Hi Peter,
Thank you very much. I haven't done mesh convergence as I am doing proof of concept studies. However, I will definitely do one and post the result by using the code you have suggested soon. In the meantime, I have been trying to work on the transient case of the 1D beam. I did the 2D beam tutorial example and I noticed that between the static and transient cases, the difference is these four lines
Timestep intervals(1) = 500
Timestep Sizes(1) = 1.0e-4
Timestepping Method = BDF
BDF Order = 2
. Even though I got the transient case as well as Rayleigh damping right in 2D, I couldn't get even the transient case working for 1D adapting the same lines of code. On checking the solver manual, Timoshenko solver seemed like, it didn't have any time-dependent terms. The StressSolve didn't work for 1D too. So, how can I do a transient case for a 1D beam? Will I also be able to perform an eigenvalue analysis for a free vibration problem? Any help in this regard will be much appreciated. Thanks.
-Amar

[raback]

Hi Amar,

I think the inertia terms are modeled, however you should (ni addition to density) give also there keywords being asked in the code:

Code: Select all

    Area(1:n) = GetReal(Material, 'Cross Section Area', Found)
    Torsional_Constant(1:n) = GetReal(Material, 'Torsional Constant', Found)

    ! If we don't give the moment of area component-wise it is assumed to be the same
    Area_Moment_2(1:n) = GetReal(Material, 'Second Moment of Area', Found)
    IF( Found ) THEN
      Area_Moment_3(1:n) = Area_Moment_2(1:n)
    ELSE
      Area_Moment_2(1:n) = GetReal(Material, 'Second Moment of Area 2', Found)
      Area_Moment_3(1:n) = GetReal(Material, 'Second Moment of Area 3', Found)
    END IF

The code seems to be liberal in tolerating that these are not given. That should be modified.

-Peter

[kevinarden]

viewtopic.php?t=7887&hilit=beam

In this post is example of Timoshenko beam solver transient, one of the posts has beam.zip with the case files.

[amarfayyad]

Hi Kevin and Peter, Thank you very much for the responses. I was trying to run the case from a sif file and followed the example Kevin suggested. However, it didn't work for me. In the case of the beam, I get a response that doesn't make sense and in the case of the Circle, the solver didn't even run. I am attaching the suggested eg in Circle.rar and my case in 1Dbeam.rar. Do you think you can help me in this regard? Many thanks for considering my request.
- Amar

[kevinarden]

Downloaded your files and ran the circle case as is it ran fine. Post the log text or state the error if you are getting one.

In the case of the beam you do not have any loads or constraints as a function of time so it is going to be a constant defelction.