Discussion Forum

Hi

when working with "assemblies" you must understand how COMSOL is "analysing" the geometry before you can apply the boundary conditions to the "geomanalyzed" geometry (search for geomanalyze in your pdf doc).

Normally (without assembly) when you click on "point, edge, boundary or subdomain mode" (="Finsih in V 4) your geometry is passing theough an ananlysis function removing any redundant info, cutting up overlapping geometries by keeping (or not if "keep interiour boundaries" is not checked) the interiour boundaries.

If you use "assembly mode", none of these combinations are preformed and all common borders are defined as pairs and the geometrical items might overlap or have duplicate items.

This means BEFORE you select assembly mode, you should group your geometries with the "create composite object" function (draw menu). The assemly will then be performed only between the last remaining "composite objects" (there should be at least two remaining, otherwise it's not an assembly). Whatever is inside a "composite object" will then appear as before and your pairs are only defined on the common border of the highest hierarchial geometrical items.

Take care to check your "repair" tolerance w.r.t. your geometry, for small MEMS type objects or RF diffractive optics you might need to set it to something smaller than the default 10[um], otherwise your geoemtry might loose quite some details ;)

Try it out on a simple geoemtry, you will quickly catch the principle

Have fun comsoling
Ivar

Thank you very much Ivan for you reply, I really appreciated it.

Maybe I've understood the concept of "assemblies", still I have problems to modelize mine structure.
Is forbidden to have a sliding contact without using assemblies?

Starting from the exact same structure of "Sliding Wedge" I add 2 thin plates on the sides. One on the upper slab (right), the other on the lower one (left). So I realized a capacitor where the sliding slabs are dielectric materials. I made the union of each slab with the respective plate (so still 2 elements in the drawing).
Then I created the contact pair (without imprinting) with a tolerance 10e-16 (where the size of the slabs are in the order of 10e-6).

until here it seems ok. Then, I created a big rectangle covering everything, that represents the "air" all around, with a dielectric constant equal to 1. It is NOT active in the structural domain. So, in the structural domain, everything has a sense.

At the end, we have 3 elements: 2 sliding slabs, 1 big rectangle.

In the electrostatic domain, the label continuity is nowhere. Now, if I deactivate the "use assembly" I can correctly set the continuity boundaries, but I loose the contact pair.

The problem is that I cannot find the way to draw the surrounding "air" of the system (in the electrical domain I mean).

Thank you very much again


PS I was reading the "geomanalyze" and it seems that I need something like "Solidify" (Create subdomains from empty regions), am I wrong?

Then I tried in this way..

I used the "difference" to remove the "slabs" from the "air". Then I copied again the 2slabs into the empty space and I created "identity pairs" between the external boundaries of the slabs. The boundary between the 2slabs is still a contact pair.

Now I can set the "pairs" as "continuity" in the boundary settings, is the same at the end?

thank you again,

Riccardo

Hi

If you want a sliding contact you need two distinct overlapping boundaries, but that is exactly what assembly is all about.

When you have a single boundary between two different domains you usually have "continuity" and you do not need to differentiate "up" and "down". If you want to put a special layer or having a sliding boundary, then you need to distinguis the two "sides" of the boundary and that is what the assembly mode is for.

One of the difficulties is its not alsways obvious to define what is "up" and "down", in 2D you have the "arrow" giving the sign of the edge/boundary "s" parameter, then you can define the left and right with respect to the arrow direction, but I have forgotten now if left is up or down, check it out on a simple case its the quickest andthe most reliable way.
In 3D you need to define something that shows up on a postprocessing or a boundary marker image to identify "up" and "down", perhaps the colors of the pair view could tell, I havent checked (Hey COMSOL developpers: "it would be nice to have a visual helper here", sometimes it helps to state out your whishes, one never knows, they may be heared and suddenly you get a nice present, in one of the new releases).

I normaly have 2 "air" items in such cases one on the fixed and one on the moving side. If you search on the forum there are a few examples of linear motors with sliding boundaries, in the doc too, but you can always apply the principle of the rotating generator to a linear case

furtermore, identity pairs are normally continuous pairs, I mostly make my assemblies with default all identity pairs, then when I want a few contact pairs, I start to delete them from the identity pairs (by unselecting them and by noteing down thier id and default master - slave side) and then I manually create my contact pairs from these previously deleted identity pair items, in this way, I'm more or less sure thay are all defined, if you have just a few you can also decide to make them by "hand" one set at the time. Note that you should not have the same surface defined as identity pairs and contact pair, but I'm woundering if COMSOL is not chencking that, not sure.


What I have understood from geomanalyze (its mostly transparent for the user) is that it combines the graphics, assembling lines and closes boundaries, all depending on the way you have constructed your geometry and check the topology, before starting to number the items: (points edges, surfaces=boundaries and domains=volumes)


Have fun Comsoling
Ivar

Dear Ivan,


thank you very much. I was missing the way COMSOL converts the dawing into "phsyical plan" where you can apply all the subdomains and boundary conditions. Now everything is more clear and I'm able to draw what I would.

Thank you very much again

Riccardo

Dear Ivan,

sorry to bother you again. Right now it seems that I was able to get the right design (more or less).
The problem right now is to achieve the convergence. I read all the guides, and I'm trying to change the different parameters but without great results.
It's failing before reaching the limit (25iterations). I was trying to implement the "manual scaling" but the problem is to evaluate the magnitude order of Tn_cp1_smpn, Ttx_cp1_smpn, Tty_cp1_smpn. I got them from the "modeling guide". But when I used the reference to get more information about that, is still difficult to estimate their values. And varying randomly 3 parameters does not seem a smart solution.

so I would have 2 questions:
1) do you think there is a way to change the design in order to achieve an easier convergence?
2) "manual scaling" is always the right way when I cannot achieve the convergence in other way?


Thank you again

Sincerely,
Riccardo