Weird behavior with solid idealization

[luis_andueza]

Hello everyone,

I have an unexpected behavior with the solid idealization in a file (attached below). 

I am simulating a liquid deposit and on the first attempt, I did it with solid elements and the second with shell elements. The results were supposed to be very similar but they weren´t. I checked everything to look for a reason why they were different but I couldn´t find it. So, I did a third and a fourth attempt with a new solid idealization, and now,  the results between solid and shell elements are very similar. 

My concern is why there is a difference between these two solid idealizations. How could I trust the results if these keep happening? what did I do wrong?

Here is the file attached (Inv2024) if someone wants to take a look. The first analysis is the different one. 

Here is the link for the file. Cuba solida.rar

Regards

[John_Holtz]

Hi @luis_andueza 

Thanks for providing the model and results.

My understanding of your description is the results of the first analysis (named "Analysis 1") are different than the third analysis (named "Solido lineal"). I expected to see a difference, but the results in the file you provided are identical (to 16 significant digits). Also, the von Mises stress, displacement, and reaction forces of the solid idealization and shell idealization agree reasonably well. (The shell shows a larger displacement. Shells are more accurate for thin structures.)

Solid element results

Shell element results

Please load the results of the first analysis and check them again. (Right-click on "Results" in the Model Tree and choose "Load".) Make a screen capture of the results that are different from the shell analysis or the third analysis.

John

[luis_andueza]

Hello @John_Holtz , thank you for your help

Here are the files again, I did a clean-up of the files. ARCHIVOS CAD_2

The behavior is in the images below. Each image shows the analysis and the file name. But as you can see in the first image the stress is very different from the other two. 

Regards

[John_Holtz · Accepted solution]

Hi Luis.

The mesh in the two solid analyses is different:

• Analysis 1; 26355 nodes; 78337 elements (linear mesh).
• Analysis 2; 164114 nodes; 83638 elements (parabolic mesh).

Parabolic elements are more accurate than linear elements, especially when the model is a thin wall with bending and one element through the thickness.

John

[luis_andueza]

Hello John,

I totally missed that. Now makes sense.

Thank you for your help.