All questions to: meshing, materials, boundary conditions and element properties
Thank you very much for your answers that, as usual, produce both certitudes and new questions. What do you think about initiate a new thread with your last post, in order to clarify the subject of this interchange of opinions?
On the other hand, at the report that you have read, there are some conceptual errors that I would like to correct. Tell me please if I can send you the new one (better if it's not to a moody email box )
Tet 16 with 10 nodes it what we call "Tet10"
and Tet 17 with 4 nodes is what we call "Tet4"
While I´m more or less fresh in the FEM community all I can tell you is that we found out
that usually you´re doing fine with Tet4 and higher resolution regarding result and time.
We usually mesh the parts finer than necessary at least (we do think so!).
It´s easier to judge for us to get at least to get a decent number of elements across a thin wall,
then to judge if a smaller number is o.k if made of Tet10 elements.
But as always: That depends on the part and the task.
(Hey - still waiting for the dl for V2 )
After reading your comments I realize that I had a kind of fantasy about quadratic tetrahedrons: that that they were not as stiff as linear ones. If I understood you well, the advantage of T16 is just the greater quantity of nodes for element. In other terms that one T16 is so rigid as an array of eight T17, but with fewer lines at the screen... what a disappointment . If so, is this because the meshers place the 5/10 nodes just in the middle of the edges and not in a serendipity mode as described the theory of quadratics tetrahedrons? Anyway... things are as they are.
But... (always there's one) I find differences between stress outputs of T17 / T16 meshes. To check this I tested two models (the same of the report, with the same BCs) with very close number of K:
Model 1: stl meshed directly in AURORA's NETGEN, T17, K = 85613, E = 453243, DOF = 256839
Model 2: stl coming from NETGEN-GUI, meshed in AURORA's TETGEN, T16, K = 85112, E = 60212, DOF = 255336
These are the von Mises results:
Model 1 :
1-3.006E+001-0.00% / 2-6.013E+001-0.00% / 3-9.019E+001-7.34% /
4-1.203E+002-28.88 % / 5-1.503E+002-45.51% / 6-1.804E+002-16.73% /
7-2.104E+002-0.93% / 8-2.405E+002-0.26 %/ 9-2.706E+002-0.20% /
10-3.006E+002-0.04% / 11-3.307E+002-0.12%
1-1.671E+001-0.00% / 2-3.342E+001-0.00% / 3-5.013E+001-0.80% /
4-6.684E+001-13.77% / 5-8.355E+001-26.54% / 6-1.003E+002-35.59% /
7-1.170E+002-16.76% / 8-1.337E+002-4.39% / 9-1.504E+002-1.30% /
10-1.671E+002-0.49% / 11-1.838E+002-0.36%
Discarding the non credible stress peaks of Model 1, and taking, say a 97 percentile, the outputs are: Model 1 = 60 / 200 MPa, Model 2 = 40 / 140 MPa. In my opinion, this is too much difference for that that seems to be a densely meshed model (remember please that it’s a kind of tube of Øs = 100 / 60 x h = 100 MM). At this point I wonder which of both results is better. Actually I can't determine this but... sincerely, I don't like too much those stress peaks of Model 1. I'm starting also to wonder if T17 meshes are not better for Rankine stresses (say brittle materials), and T16 for von Mises (say more ductile ones).
I have some additional comments for all this subjects, but this post is already long enough. Here is still Thursday and tomorrow is the Big Day, so, surely we'll be resuming this thread on next week. Receive please my best wishes of success for AURORA V2 and for all of you.