trouble with convergence [solved]

[kevinarden]

Nice work.

for trouble getting meshes to connect with each other across 3d boundaries, generally the coherence command right before meshing will resolve it.

[Rich_B]

Here is an alternate example, this time using elmergrid as the mesh generator.

Rich.

panel-magnified.png (208.83 KiB) Viewed 599 times

potential.png (76.79 KiB) Viewed 599 times

[greyltc]

Hey thanks for that! The potential map looks right on!

The fine geometry details are important to me. What I posted here is just kind of a baseline in my study. Other items in the study have much more complicated geometry. Being able to mesh the geometry as it appears in the step files is an important requirement. I'm still a bit unhappy with how I ended up meshing this. Being able to only get the meshes working with simple extrusions is a bit limiting. I'd love to be able to mesh arbitrary geometry with these aspect ratios. Somehow the anisotropic meshing capabilities of the meshing tools I've been trying seem to be pretty limited though.

The other weird thing that's got me scratching my head (as in the post title) is that the iterative solvers don't work on this; I'm only able to use the direct solvers.

[kevinarden]

The units in Elmer are arbitrary but required to be consistent. Therefore scaling is an option, but instead of scaling one direction by 10000, you have to scale all directions by 10000. This takes numerical thickness out of the problem. However you have to following scaling laws for all of the other constants in the problem for the units to remain consistent. For example density would have to scaled by 1/10000 so that the mass remains correct. There is a set of scaling laws for other things like force, stress, velocity, acceleration, etc. Usually these scaling laws are used to build small scale models of large structures for physical testing, however they can work in the other direction. Scaling up gets you out of the numerically very small numbers, but everything has to be scaled consistently.

[kevinarden]

scaling.PNG

(192.4 KiB) Not downloaded yet