Mixed dimensionsal modeling

[nore57]

Dear all,

I am currently working on coupling between beam and shell elements to build a FEA model for framed structures. I do need to model details such as joints with shell elements and the remaining structure with beam elments. The transition between the two structral elements is quite not straightforward for me. However I managed in Abaqus and Nastran to use the native tools (rigid bodies or mbc beams) to have some nice transition though some spurious stresses exist (which is expectable according to the threads I have read on forums).

Now I have to do the same procedure in a software that lacks all that kind of mixed dimensional modeling tools. So I kinf of use so beams of infinite rigidity (very high young modulus) at the transition forming a pider-shape element. However i get wrong results in terms of bending stresses. I even get some torsional stresses that are not supposed to exist in my full 1D model

Is anyone familiar with this kind of problems please? How can one guarante proprer tranfer of loads when lucking MPC or RBE elements?


Many thanks in advance.

I have been on this for weeks trying releasing some DOF I don't want for example, changing rigidities and so on in vain.

 

[GregLocock]

It's a nasty problem. One trick is to terminate the beam well short of the shell, and then build a shell model of the beam so you can build a realistic joint between the beam and the rest of the structure.

You can then ignore any odd stresses in the beam element end of the shell-beam.

Cheers

Greg Locock


New here? Try reading these, they might help FAQ731-376

http://eng-tips.com/market.cfm?

 

[nore57]

Thanks for your answer. I don't understand your proposal very well.

You mean I should extend the shell mesh of the joint so that I can reproduce the beam cross section then connect it directly to the beam

I did that actually. however I am interested in comparing the results of my full 1d model to those of my coupled model and coupled 1d/2d model.

Let me explain on an example. Say I take a long beam under uniform loading. I model it fully in 2d (using shell elements) and fully in 1D using beam elements. Then I slice it at 1/3 and represent this slice with beam elements, then from the 1/3 to 2/3 I use shell elements, then from 2/m to full length I model it using beam elements.

Then I compare stress distribution on the member O-1/3L (where L is beam length) of the coupled model, full 2D and full 1D. Unfortunately the coupled 1D and full 1D do not match.

In what you suggested you said I should ignore stress distribution where please?

Actually i make those comparisosn to validate my approach. If stress distribution is underestimated or overestimated, this is due to the fact I do not transmot my loads correctly in my model.


What would be the reason according to you?Is there any way to have an accurate model?

Thanks again.