Accuracy of the solution with very thick plate elements

[giovap]

Hi, I just started a new job where we perform linear analyses on steel structures of cargo ships. Near doors and windows we need to increase the mesh density so to capture local stresses. Currently the accepted maximum thickness/side length ratio is 1x1, meaning that we often have 45mm plates meshed with 50mmx50mm linear quads. I also noticed that the software uses the thin plate formulation. My questions are:
1- Is an element so stocky (it is, in fact, a cube more than a plate) approximated well enough by the thin formulation?
2- If not, would a thick-plate formulation solve the problem?

 

[FEA way]

Which software are you using ? Doesn't it offer thick plate or special solid elements (like e.g. Abaqus does) ? Did you try solving some test models with different element types ?

 

[Denial]

Question 2 first.[ ] You need to distinguish between two separate issues/questions:[ ] element type and element size.[ ] Whether to use thick or thin plates depends on the structure's thickness in relation to its overall size and shape.[ ] Having made that decision, you then decide on the required mesh density.

Question 1 is largely irrelevant.[ ] You will never reduce your solution's accuracy by using too many elements.[ ] You just increase the time and cost of your computations.


[sub][ ]—————————————————————————————————[/sub]
[sup]Engineering mathematician / analyst.[ ] See my profile for more details.[/sup]

 

[rb1957]

I would do some single element tests to see the effect of thickness as a ratio of side length on the element results.

What does the element quality checks allow ?

If you really want to know the Kt, are you using the nodal stress ?

"Hoffen wir mal, dass alles gut geht !"
General Paulus, Nov 1942, outside Stalingrad after the launch of Operation Uranus.

 

[BlasMolero]

Dear Giovap,
In the blog of Simcenter FEMAP & NASTRAN I explain the differences between the formulations of Kirchhoff-Love vs. Mindlin-Reissner shell elements for thin & thick plates.

https://iberisa.wordpress.com/2021/01/05/ejemplos-de-mallado-con-femap/

In your case of plates with thickness = 45 mm you can mesh perfectly with 2-D shell/plate elements of element size = 50 x 50 mm.

In case of using Simcenter FEMAP with NASTRAN, the following image shows the definition of the Property of a Shell element in Simcenter FEMAP: the fields with MID2 and MID3 are always active by default, the material is assigned automatically, therefore the Shell elements defined in FEMAP include by default membrane rigidity, bending and shear stiffness term, and are ready to work with thin and thick plates, not problem.

Best regards,
Blas.

~~~~~~~~~~~~~~~~~~~~~~
Blas Molero Hidalgo
Ingeniero Industrial
Director

IBERISA
48004 BILBAO (SPAIN)
WEB:

http://www.iberisa.com

Blog de FEMAP & NX Nastran:

http://iberisa.wordpress.com/

 

[giovap]

Hello and thanks everybody for your answers.
Following Denial answer: what I understand then is that my error is in thinking that the hypotheses of thin plates do not have to be verified on a single element but on the overall structure that the mesh is representing. Correct?

You will never reduce your solution's accuracy by using too many elements.

That is so obvious that I am a bit ashamed that I did not look at the problem this way, Thanks!

Which software are you using ? Doesn't it offer thick plate or special solid elements (like e.g. Abaqus does) ?

Yes it does have indeed thick formulations and solid elements. My question was more a thought experiment.

Dear Blas, I remember that you helped me some months ago with my Femap-related questions on the Simcenter blog, but unfortunately now I work with Patran! I miss Femap...

 

[rb1957]

I still think you should do single element test to understand how well the element is working.

"Hoffen wir mal, dass alles gut geht !"
General Paulus, Nov 1942, outside Stalingrad after the launch of Operation Uranus.