Discussion Forum

Hi

a sweep mesh might also work, anyhow it's not the shape but the density of the mesh w.r.t relevant feature sizes and more important, dense enough to sample correctly any gradients of your dependent variables

--
Good luck
Ivar

Is it necessary to make tetrahedral mesh in 3D model?

No

Will it work if I use free triangular mesh in one surface and sweep it across the other surface.

Sometimes it works, not always. Depends on the geometry, symmetry, etc. Tetrahedral mesh is more general approach

What may be the cause? is it the meshing?

Might be. Try tetrahedral mesh

Hi,

I have mesh one faces with free triangular element, and swept it across the other faces. When I start computing, in the message window it displays about the degrees of freedom to be solved. and then there is nothing. I see no convergence plot or any other default plot. I just get the message which says that the solution could not be converged.

HI Alexander,

Thanx for the wise reply. The problem with tetrahedral mesh is, whenever I build a tetrahedral mesh, there is an error shown in some boundaries. The error message is 'failed to respect a boundary edge'. And when I zoom into that boundary, I see a small line (or protude) that comes from no where.

May be there's a problem with the geometry!

Hi,

Try meshing the problematic boundaries with triangular mesh first.

Hi

Our geometrical models, if received from a CAD designer engineer, have mostly a lot of "small" mechanically useful details, but that have no FEM analysis use. Rather the contrary, most fillets, champfers, screw holes, threads etc are only a complication that can easily multiply the DoF by an order of magnitude.
Good practice is to remove these from the CAD BEFORE exporting the geometry. I do not know if this applies to your geometry, but take a closer look ;)

The difficulty is to decide what is of FEM importanceand what can be left out, this you gain only by experience, by trying out and studying intensely your results

Another issue is complex geoemtries that "loop around" as a thorus, this kind of topology often gives difficulties to the mesher as it starts somewhere then propagates as a wave, but then suddenly must match the present wave to a fixed meshe once closing the loop. Such geoemtries are often easier meshed by cutting them up, or by meshing from interiour to exteriour, or sweeping along the axis, hence some manual tweaking. Again, meshing is for me more an ar than a pure science, here again your gain time with increased experience. Often it's worth to remsh a couple of times a geometry, just to learn how it's done quickest, and what are the final quality, use the tooften overlooked "Mesh statistics" to help you "measure"

--
Good luck
Ivar

Hi Ivar

Thank you so much for the explanation. I have tried remeshing several times, but the problem when I compute, there is no progress at all, and after an hour or so, it says that it could not be converged.
However, I'll try to remesh the geometry again! ;)

Also, I will be thankful to you if you look at this another problem of mine. It is about 2D rotationary modelling.

www.comsol.com/community/forums/general/thread/24687/


Thank you Again! :)