Shell vs Beam

Hi all,

I have a question regarding the calculation of shell elements.
Comparing a line element with a rectangular section and four shell elements describing the same cross-section, I get quite different results. See example below:

Now in this case the shell elements are not behaving as expected.
I did this test because I found some inconsistencies is a more complex bridge model. Here a schematic line model combined with one plate for the deck gives expected results (verified by hand calculations and another FEM-package called Sofistik), where the more detailed model made of plates gives quite a different outcome.

Do you have any idea what could be going wrong?

It seems that it might have something to do with the shear forces in the shell calculation.



Hi Merijn,
are you certain that they way you have structured your model is correct,
so that the results are comparable? To me it seems that the way you have supported the shell is quite different than what you have with the beam-element.

What I’m trying to say is that on the shell model you have eccentricity, which you are missing in the beam element. Try supporting the shell also from the neutral axis.


1 Like

Hi @merijndeleur, can you upload your definition so that we can take a look at it.

Hi Eero,
I figured as much already, however, changing the support to the neutral axis doesn’t change much (already tried).

Here’s my definition. (82.3 KB)

Dear Merijn,
the deviation between the beam and the shell results is mostly due to a slightly wrong dimension of the shell beam (see the attached definition: (87.6 KB)
): The cross section of the beam modelled via shell elements should be placed at the mid-line of the box cross section, not at the outer boundary. When doing so, one has to scale the applied surface load as well. For the beam cross section you defined a non-zero fillet radius - I am not sure whether that was intended.
Another effect is, that in order to simulate in-plane bending with shell elements one has to use a fine mesh in order to get accurate results.

1 Like

Hi Clemens,

Thanks for your reply! The fillet radius was indeed unintended. And I actually didn’t deem the small difference in geometry as significant, but obviously you are right! Thanks!
For this example the problem is fixed!