Difference between beam elements and plate elements to create beams

[Anonymous]

I was curious about if there is a difference between making a model out of just beam elements for I-beams and C-channels or if it would be more accurate to make them out of plates. To test I created two files, one was a curve and meshed into an I-beam, the other was the same shape I-beam made from plates with a point load in the center.  The results were surprising, the beam element had less stress but more deformation than the plate element.

I was wondering if anyone had more information or if maybe NASTRAN doesn't work well with plate elements to make beams. 

[Anonymous · Accepted solution]

The displacement and stress results between the two models should be very closely matched if your models are similar and made correctly.

There are a few things you need to make sure are the same if you are going to do a comparison between a beam and plate model representing the same geometry:

1.)  To get as close as an apple-to-apple comparison as possible, using element edge lengths (mesh size) that are exactly the same between two models.  (not a necessity as long as the mesh size is fine enough, but too readily easy to do to not do it for this purpose of benchmarking)

2.)  Use the same boundary condition between models.

3.)  Use the same loading pattern (magnitude, direction, and location).

4.)  Use the same materials properties and represent the exact same cross-section geometry (with beam {CBEAM} and shell {CQUAD4} elements).  DON'T use CTRIA3 elements.  Also represent the same length of the beam between both models.

5.)  Although you can pull beam {CBEAM} stresses directly from the model while post-processing...

You can also pull CBEAM forces from you model and post-process those forces for stresses to get a more exact/detail view of the variation of the stresses throughout the cross-section of the beam. 

One of the benefits of using shell elements to represent a beam is that post-processing beam forces for detailed stress variations from beam elements isn't a must.  However, you end up w/ more degrees of freedom (DoFs) in your model using shell elements.  Not an issue for small to medium size models/problems.

Using beam elements allows the user to rapidly and readily create different cross-section and type of beams by simply changing the property card.  Whereas, using shell elements requires remodeling (assuming you are meshing geometry) and re-meshing the discrete geometry and then also altering the property card (plate thickness).

Other than those high-level things, they should give similar results for deflections, stresses, and frequencies.  Similar doesn't mean exactly the same.  But they should be close enough to be well within the range of the differences being in the engineering insignificance range.