Difference in results between models with shell and line elements

[arun.sankar]

Hi Sir

We are attaching two Nastran files( AutoCAD Inventor Professional 2021) that have been modelled using line elements and another one using shell elements. I have a couple of questions pertaining to the same:

a) Why is "Continuous meshing" not working for the case with all elements modeled as shell

b) Why is there a huge disparity in the results when they have been constrained and loaded identically? The outer brackets in the case with shell elements get stressed but when the same is checked for the model with line elements, the outer brackets are loaded only to a less extent. Moreoever, the deflection for the line element comes to 12 mm while when it is modeled as a shell, the deflection is 8 mm.

Please advise.

Load per beam - 27.5 tons

length of beam - 8500 mm

cross section of beam - square tube 400 x 400 x 12 mm

  

[John_Holtz]

Hi @arun.sankar 

1. Continuous meshing is not working because that is an issue with the software. You did read the forum post Use Continuous Meshing with caution, correct?
2. The loads and constraints may be the same, but the geometry is different! The shell model is a 412x388 mm tube, and the beam model is a 400x376 mm tube. The smaller tube deflects more. Please change the dimensions of the beam model and check if the displacements are closer.
3. The beam model transfers all of the vertical load to the middle of the shell bracket. The shell model transfers the load across the face of the bracket. Therefore, the displacement of the brackets is vastly different. Also, bonding the shell model over the entire face stiffens the bracket. 

Shell model displacements

Beam model displacements

[arun.sankar]

Hello Sir

We did try the dimensional difference you had suggested but the results are not at all in sync with one another. A structure irrespective of it being modelled using beam or shell elements should produce results that are close. Please advise on what needs to be done, whether I should opt for the line element design or the one with shell elements.?

[John_Holtz]

Hi @arun.sankar 

Just to be clear, what results are you comparing? What are the results that you have now?

I agree that a proper model will give similar results regardless of the element type (line, shell, solid). However, that also requires the models to actually represent the same thing! In your original model, the support brackets behave much differently (item c in my original reply) because the line and shell model do not represent the same thing. Therefore, you cannot expect the results in the bracket to be the same. Any result in the beam that is affected by the support bracket (such as the displacement) will also be different.

As another example, the results of a miter joint will be much different when modeled with line elements compared to shells. The line elements results are based on the calculated results at one location: the neutral axis of the beam. All of the other results are extrapolated from that one result. The shell element results are calculated at all points around the joint.

Example miter joint.

In summary, you should use the element type that more accurately duplicates the physical part and create the model to reflect the physical part.

[arun.sankar]

Hello Sir

Thank you for the elaborate explanation of the methodology to be adopted when analyzing shell elements. We did try splitting the surface so that the mesh elements match one another and used Solver "Offset bonded". I am attaching a snapshot of the same and the results look correct.

Sir, can you please explain the situations wherein we have to apply "continuous meshing" as there is quite an ambiguity on our part in its application.

[John_Holtz]

Hi @arun.sankar 

You asked "can you please explain the situations wherein we have to apply continuous meshing".

I cannot think of a situation where you have to use continuous meshing. Perhaps another reader will have an example. You should be able to use contact to bond the parts together. However, continuous meshing would give better results if it works. If continuous mesh does not work, the results will be inaccurate or wrong, and that is not better.