a member of another forum wrote, that nodes are not running on nodes, instead a contact point is established when a node of the source face is in contact with the element-face of the target face. (Link to the German Forum.)
But if I use the Mesh Mating Condition - Free-Coincident the source-node is initially in contact with the target-node, and not the target-element-face.
You guys (Karchun & JimB) wrote in this thread that this concentricity would be great for precision and convergence. But the other member wrote that nodes on nodes isn't the way it works.
So which case is better/worse or the "correct one"? I am a little bit confused because of this different statements. I would appreciate if someone could clarify this for me.
In general: I understand from Page 13 how contact elements are created:
"The solver projects vector normals from the source region to the target region. It then creates contact elements when these normals intersect elements in the target region and are within the search distance criteria for the contact pair."
This would make concentricity the winner.
Anyway, do you have any idea what the other member might have meant? Maybe another software? Another member even wrote that node-on-node is NOT what is used with most FE-software.
Yes, this is how contact algorithm works (contact can be symmetric). Check picture below and wee how selecting of contactor/target surface affect contact solution.
Coincident node still have contact with target elements. The problems may occur when target faces have large angle, like 90 deg. In this case, you have to break one contact surface to multiple flat/smooth parts.
In press-fit problem, coincident nodes remain coincident after solution and node-to-node mesh guarantee that contact have zero gap and you can set overlap value. For nonuniform mesh you can set initial gap/penetration to zero and overwrite it with desired overlap value.
If contact surfaces slide then no mater are nodes coincident or not, after first iteration they move and become non coincident.
Glued contact created to connection of nonconformal meshes (if you have coincident nodes then you can simply merge them).
Therefore, in conclusion Nastran can work both with conformal and nonconformal meshes. Some features like sharp edges/corners can lead to difficulties in contact convergence. For sliding contact mesh mating will be not so useful.
As I say before - creating of uniform mesh for contact depends from amount of free time and perfectionism of user. You can acheive good results on nonconformal mesh too.
Zero gap, it`s good for press-fit/gluing/acoustics. Zero gap is also good for starting of contact analysis. However, nowadays this is not necessary, nonuniform meshes work well too. You can make some test models and compare these two types of mesh.
To clarify, all of the information posted by @Karachun above is relevant for the Adina-based SOL 601 only. In this solution, a node to surface contact algorithm is used.
In SOLs 101, 103, 105, 111, 112 and 401, a surface to surface contact algorithm is used. Here the contact elements are created at the gauss point locations on element faces. Depending on geometry and discretization, there may be overlap at the nodes (particularly at the perimeter of the contact region), because the zero penetration constraint is not being enforced at the grids of either source or target region.