Mesh size in failure crash analysis

[Malik27]

Mesh size could be very important in crash analysis;
The most common mesh size: 5-15 mm is too coarse sometimes to obtain good correlation results:
Have you got past experience in terms of material law and modelization ?

 

[EricZhao]

What is your modal? Mesh size is important for any FEM analysis. And shouldn’t the mesh size be governed by element type, model, load,… instead of the just absolute dimension?

 

[Drej]

Seconding Eric's comments, mesh size is important in every FE analysis. Saying "The most common mesh size: 5-15 mm is too coarse sometimes to obtain good correlation results" is generalising, which is meaningless, relatively speaking. I guess you're probably talking about an explicit FE approach to modelling if you're dealing with crash analyses (DYNA, ABAQUS/Explicit &c.) , where element size and geometry is critical, since the time step is based on the smallest element cross-diagonal in the model. The smaller the element, the smaller the time step, the longer the analysis takes to run, etc.


------------
See faq569-1083 for details on how to make best use of Eng-Tips.com

 

[ghouri]

If you are talking about automotive crash(which i assume you are) then the ideal mesh depends on various factors such as thickness, load case, resources, and geometric details etc..

1. thickness is important becoz you dont want to have 5mm elements with 4.0 thickness(violates the shell theory)
2 load case is important coz you dont want 8 mm element on the rear bumper if you are doing front-impact.
3. resources are important, they dictate your model and mesh size.
4. If you want to capture beads, crush-initiators and etc then the geometric details will dictate your mesh size.

And in crash-analysis "the-more-the-elements-the-better-the-result" is not necessarily true. Infact in some case it can hurt the prediction by softening the model a bit too much.

So to sum it up i would say, its experience which determines the right size for the right problem and i guess thats the reason why companies pay big bucks to crash analysts...

Ghouri

 

[corus]

I've never heard of having the element size greater than the thickness because it violates shell theory. Shell theory generally assumes that the stress distribution through the thickness will be linear and you generally use them where the thickness to overall dimensions is small. That doesn't involve the element size in any way.

In general, the fewer elements you have then the stiffer the overall model. The more elements you have then the better the representation of the model stiffness. If more elements hurts the prediction then my guess is that some other factor in the model is not represented properly and using fewer elements is a random 'fiddle factor'.

corus