2
$\begingroup$

To get realistic results from a FEM calculation, it is common (or so I've been taught by someone who may not be up to date) to use at least four or five elements through the thickness of a beam or plate. I am now modelling a composite plate bending problem, with two or three materials.

Should I use four or five elements through the thickness of every material, or can I reduce my calculation time by using one or two elements per material, as long as the total adds up to four or five?

Improve this question
$\endgroup$
7
  • $\begingroup$ A better method is (1) use a FE package with elements that don't suffer from shear locking (the reasons for that problem were understood and fixed 30 years ago!) and (2) use a FE package that has proper "composite shell" elements. $\endgroup$
    – alephzero
    Sep 27, 2018 at 12:15
  • $\begingroup$ @alephzero Thanks. Perhaps (1) is true (I'm using NX NASTRAN 12); this is just a rule I've been taught by someone who is probably not up to date anymore. Regarding (2), I'm specifically interested in some of the behaviour caused by the nonzero thickness near the edge, so shell elements are not an option. $\endgroup$
    – Sanchises
    Sep 27, 2018 at 12:33
  • $\begingroup$ @alephzero Yup, a quick look into the docs shows that whoever warned me about (1) was probably mistaken. Are there any other reasons to use multiple elements through the thickness? $\endgroup$
    – Sanchises
    Sep 27, 2018 at 12:36
  • $\begingroup$ I would advise to run a convergence test, gradually increasing the number of elements through the thickness until your results dont change. $\endgroup$
    – user190081
    Sep 27, 2018 at 13:42
  • $\begingroup$ I don't know what is in NX Nastran 12 - I've only ever used MSC Nastran. In MSC Nastran, shear locking only affects a minority of elements (with old mathematical formulations that nobody wants to change, for whatever reason). Just avoid those element types for new models, and the problem doesn't exist. $\endgroup$
    – alephzero
    Sep 27, 2018 at 17:44

1 Answer 1

Reset to default
1
$\begingroup$

This kind of a rule of thumb is valid only for specific cases. I assume the person who shared with you this rule was investigating a very particular case using the same FEA tool for many times. The number one rule, however, is to always use your own brain. In your case, are the stresses or the deflections changing along the thickness direction? If not, you should not worry much about having a good resolution there. One fatty element will do the work. If you ARE interested in the phenomena going on along the thickness of the material, you have a lot of weapons in your arsenal. You might use 2nd order or more elements (this way the deflection behavior between the nodes of each element is assumed to be polynomial rather linear), refine the mesh to as many layers as needed, and most importantly - check the convergence of the results to realize there is no need to improve the mesh even more. This way you will obtain your own rules for this specific problem.

Improve this answer
$\endgroup$
5
  • $\begingroup$ Why down vote??? $\endgroup$
    – user14407
    Sep 27, 2018 at 17:49
  • 1
    $\begingroup$ There are many element types that will model stress changes through the thickness with a single element - including 1D (beam), 2D (shell) and 3D (solid) elements. $\endgroup$
    – alephzero
    Sep 27, 2018 at 17:49
  • $\begingroup$ @SamFarjamirad Too many technical errors. (I just pointed out one in a comment, while you were asking the question). Another one is the implication that higher order elements are "better", for modelling this type of structural behavior - they are not necessarily needed at all! $\endgroup$
    – alephzero
    Sep 27, 2018 at 17:50
  • $\begingroup$ Please explain what technical errors are you talking about. I mentioned he might get a picture of the changes along the thickness without enlarging the number of elements. If the behavior is complicated and predictable, he can use a higher polynomial order if necessary. Furthermore, there was no implication for anything. The guy did not explain what is the model so all we can do is guessing. $\endgroup$
    – Yaniv Ben David
    Sep 27, 2018 at 18:01
  • $\begingroup$ @alephzero, it's hard to decipher tone online, i was just curios to know what is wrong. Thank you. $\endgroup$
    – user14407
    Sep 27, 2018 at 18:25

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Not the answer you're looking for? Browse other questions tagged or ask your own question.