Introducing springs to a point/face as a boundary condition

[mika]

I updated the utility SpringAssembly so that a localized (generalized) mass can be added in a similar way as a spring. The keyword is "Mass k", k=1,2,... . By physical reasons the values of "Mass 1", "Mass 2" and "Mass 3" should be the same for 3-D displacements as they represent the same scalar quantity. With k > 3 a moment of inertia may be given for models for which it's meaningful.

The shell solver was also updated so that it understands the keyword "Mass k" in connection with elements listed as boundary elements (the file mesh.boundary). If the place is nevertheless specified with the keyword "Target Nodes", the only option is still to apply the utility SpringAssembly.

-- Mika

[kevinarden]

EigenSolve: Computed 3 Eigen Values
EigenSolve: --------------------------------
EigenSolve: 1: -1.941983E-02 0.000000E+00
EigenSolve: 2: 3.100790E-02 0.000000E+00
EigenSolve: 3: 1.682511E+07 0.000000E+00
EigenSolve: --------------------------------

I tried to put the mass in this way but the Eigenvalues are the same

Boundary Condition 3
Target Nodes(2) = 1 4
Mass 1 = Real 1
Mass 2 = Real 1
Mass 3 = Real 1
End

[mika]

I modified your sif somewhat and see a mass to change results. I also mention that to get eigenvector output it's better to forget the (default) component names of the shell solver variable and use a single name for the entire solution. This has also been done in the attached file.

-- Mika

[MoTN]

This does everything except for the lumped masses. There are single node elements type 101, but I do not know how to assign them a mass value.

eigenvalue.zip

Kevin

Thank you very much for creating and sharing the file, Kevin.

[MoTN]

Thanks Mika for the wonderful job!
Where can I download the latest version (which also deals with point masses) for a windows OS?

[kevinarden]

The builds are here but the last one for windows was 3/26. The builds for Linux is nightly.

http://www.nic.funet.fi/pub/sci/physics/elmer/bin/

[MoTN]

The builds are here but the last one for windows was 3/26. The builds for Linux is nightly.

http://www.nic.funet.fi/pub/sci/physics/elmer/bin/

The link has been updated on 2020-04-04. I presume it contains the latest update which takes into account node masses. There is a repository named "ElmerFEM-gui-vtk-nompi-2020-04-04". I installed the package on 3 PCs with windows 7/10 64-bit OS, but it does not work on either of them. It gives a bunch of errors for missing .dll files (e.g., libjpeg-8, libtiff-5, liblz4, libpython3.8.dll and so on). I tried to fix them, but it did not work eventually.
I know you are not a big fan of windows, but your help is very much appreciated

[raback]

Hi,

Yes, the vtk versions there turned out to be broken. Try the older ones without vtk. For distros you need all the required dynamic libs and bundling them currently requires some manual steps.

-Peter

[MoTN]

This does everything except for the lumped masses. There are single node elements type 101, but I do not know how to assign them a mass value.

eigenvalue.zip

Kevin

Hi Kevin,

I installed the latest update 2020-04-08 successfully, and it works perfectly!
I just managed to run your example. Something seems odd to me. Given that Elmer reports omega^2, the first natural frequency is negative (-1.941983E-02 0.000000E+00). I am not sure how to interpret that.

I created two tests to understand how the shell solver works:

(a) A shell mounted on a torsional spring at the mid-chord node;
(b) An equivalent solid model (linear elasticity) using two translational springs at the leading and trailing edges and hinged at the mid-chord node.

Here are the results for the two cases:
ELMER eigenvalue analysis (a) sqrt(-5.470754E-04)/(2pi) (Hz) !?
ELMER eigenvalue analysis (b) 5.2010 (Hz)
Analytical calculation 5.20 (Hz)

It seems that the solver still ignores the spring
Spring 5 = Real 4.019e3
Could you please give me some insight into it?

[mika]

Your shell model is not sufficiently constrained to prevent rigid-body motions as there seems to be a pointwise constraint in one node only. The negative eigenvalue is an approximation of zero. Zero eigenvalues are associated with rigid-body motions.

-- Mika