Target Nodes vs Target Boundary

[Chris74]

hi all,
I've just a simple question about the boundary condition in the SIF file.
What's difference between "Target Nodes" and "Target Boundary" if the nodes of the "Target Nodes" match with the element's nodes of the "Target Boundary". In my opinion, I don't expect any difference, but the results are a bit differents when I compare the results files.I wonder if the "element=p:2" feature could be the cause.

Main features of my study :
- Simulation Type = Transient
- Equation = Nonlinear elasticity
- displacement + fixed conditions on the boundaries
- element=p:2

Best Regards.

[kevinarden]

Target boundary is per/length if a line or per area if a surface. Target node is that value applied to the node. They could be the same if you calculate the nodal loads that way. Also I do think target boundary would load mid-side nodes if present. However I think "element=p:2" increases the order of the element mathematically but does not physically add nodal points.

[raback]

Yes, the main purpose of "target nodes" was to be able to fix single Dirichlet nodes. The "target nodes" are not proper elements so the module specific boundary integrals are not applicable. The cumbersome part of "target nodes" is that they really require the knowledge of nodal index. The "target coordinates" keyword is more robust in that it activates searching of the nodes closest to the given coordinates. -Peter

[Chris74]

Thank for your replies, but it's not very clear for me, I wonder why the results are differents :
I have only Dirichlet conditions.
A short example to illustrate my issue :
a)At the beginning, with Elmer GUI, I was appling a condition on the edge. For example, on the edge n°114 (defined by nodes 119 and 121),
Target Boundary(1) = 114
Displacement 1 = Variable "time"
Real MATC "tx(0)*0.05"
b) Then, I replaced this previous condition by this one :
Target Nodes(2) = 119 121
Displacement 1 = Variable "time"
Real MATC "tx(0)*0.05"
And the both results are differents. Which is the best ?

Regards

[Chris74]

A simplify picture to illustrate the study

[raback]

Hi

As Kevin pointed out the difference may be the treatment of the 2nd order term in p:2 element. Are the results exactly the same if you use linear basis?

-Peter

[Chris74]

Hi all,
The results are exactly the same if I use linear basis.
So the treatment of the 2nd order term in p:2 element is the cause of this difference.
what kind of loading should I consider if I use p:2 element ?
Regards

[kevinarden]

I suppose it depends on the intent of your boundary condition. Target nodes is specific to those nodes. However, higher order elements (i,e, p:2) have more terms associated with element edges and boundaries. So if you intend to put a condition on an edge or boundary use target boundary, if your intention is to specifically apply conditions to the nodes use target nodes. The actual math happens at the integration points in the mesh, not necessarily at the nodal points, unless of course you want them to.

[raback]

Hi

We usually prefer defining boundary elements for setting BCs. The "target nodes" was initially just a way to define single nodes which you sometimes need to set. If you have p:2 elements you basically create a 2nd order bubble that is a product of the the nodal basic functions. However, when you use the "target nodes" logic there is no way to address it. For Dirichlet BCs using "target boundaries" this bubble is enforced to be zero. Also "target nodes" has the annoying feature that it requires additional knowledge of the mesh. It is cleaner to only refer to the mesh in abstract terms - not coordinates or node numbers.

-Peter

[Chris74]

Thank you Peter & Kevin for your replies.