Element releases

Hi,

First of all, thanks for a great plugin. I have a quick question regarding karamba3d: is there a way of applying releases between beam elements in a model? I.e. disabling moments to be transferred between beams in some local axis. A solution where extra elements don't have to be added would be preferable.

In my case I have four elements meeting in one node - two of type A and two of type B. I want moments to be transferred between the As and moments to be transferred between Bs. However, I do not want moments to be transfered between As and Bs - they should be pinned relative to each other (like a scissor). Any ideas?

Thanks,

Emil

Replies to This Discussion

Permalink Reply by Karamba3D on October 29, 2015 at 1:01am

Hi Emil,

you will have to add a spring between groups A and B.

Have a look at 'ZeroLengthElement_Scissor.gh' in the examples that come with Karamba (see 'C:\Program Files\Rhinoceros 5.0 (64-bit)\Plug-ins\Karamba\Examples\TestExamples\Zero_Length_Element').

Best,

Clemens

Permalink Reply by emlplsn on October 29, 2015 at 1:37am

OK! Thanks alot for fast reply Clemens. It's kind of hard to imagine how a logical release object would work in the GH environment.

Permalink Reply by emlplsn on October 29, 2015 at 1:05pm

Just out of curiosity - would it be possible to mix with the DOF connectivity in Karamba using the API to somehow create this kind of release without adding elements?

Permalink Reply by Karamba3D on October 29, 2015 at 1:24pm

For the rotation joint is is necessary to add a constraint in one form or the other. The spring is equivalent to a penalty approach. Another approach would be to work with Lagrangian multipliers but this would have to be implemented in the calculation core.

The spring elements are not very costly from a computational point of view.

Best,

Clemens

Permalink Reply by emlplsn on October 30, 2015 at 1:33am

Interesting, never heard about Langragian multipliers in this context before. Maybe I am oversimplifying this the way I think about it, but isn't one of the fundaments in FEM theory how global DOFs are (dis/) connected to element local DOFs? I manage to dig out this example from one of my old text books:

In the node where the pin joint is applied two rotational DOFS are present - one for the beam to the left and one for the frame to the right. This enables them to rotate freely relative each other (without adding elements). The global stiffness matrix will later be of size (n+1) x (n+1) where n is the number of DOFs without the extra rotational DOF. In this way one could potentially experiment with mixed pinned / clamped joints (in a 3d context of course). For instance releasing Mxx and Myy, but locking Mzz (torsion) between elements.

I do realize that this discussion might depart from Karamba as the focus, and more approach a general FEM discussion. Anyway, if there is anyone who knows about this stuff it's probably you Clemens! (Hence why I'm asking.)

Cheers,

Emil

Permalink Reply by Karamba3D on November 4, 2015 at 12:00pm

Emil,

to introduce additional DOFs to simulate joints would be a possibility. It is also possible to eliminate these DOFs on the element level so that they do not increase the number of solution variables on the system level. The technique is discribed in 'Non-linear Finite Element Analysis of Solids and Structures' by M.A. Crisfield, Volume 2, page 252 ff.

Best,

Clemens

Permalink Reply by emlplsn on November 9, 2015 at 2:35am

Hi Clemens,

Many thanks again for your informative response. Much appreciated! Very interesting approach with eliminating DOFs on element level to avoid increasing complexity. I will definitely check out the reference you mentioned.

All the best,

Emil

RSS