Timoshenko Beam Solver and Conforming Boundary Conditions

[Stefan Hiemer]

Hi everyone,

I am no sure whether I have spotted a mistake within the TimoshenkoBeam solver (unlikely) or whether I do not understand the conforming BC (CBC) correctly by reading the Elmer documentation (likely). In my current project we need to create a number of pure beam networks with periodic boundary conditions. An example beam network is attached where the gray line shows the simulations box and dotted lines indicate beams crossing the periodic boundaries. The default PBC in Elmer should not be used as we have an exact node-to-node correspondence therefor I turned to CBC.

1. I overtook the simple cantilever test (elmerfem/fem/tests/Beam_3D_Cantilever/) and applied CBC to the free end. However my results are identical to the test case without CBC, ergo CBC does not do anything. Do I need to put additional input there? I would expect not as CBC should just replace the dofs with the corresponding node and checked a test . I attached the sif file (simple-example.sif) and the geometry (periodic-beams.msh).

2. In the case for beam networks (example-beam-network.png) my starting point is the adjacency matrix of the network whose connections are beams. In Gmsh (gmsh-representation.png) I create artifical nodes (the ones with just one connection) that I intend to reassign via CBC to the opposing side and then encase them in physical groups of Gmsh to make it Elmer compliant. For the example, it is trivial. I will however later encounter cases where the bounding box is not rectangular and where I may need to create a large number of physical groups to do it appropriately. I just wonder whether it would not just be more efficient to give the adjacency matrix / pairs directly to elmer to make the connections.

Thanks and have a nice weekend
Stefan

Edit: for some reason I cannot put the sif file. I will put it into the next post.

[Stefan Hiemer]

sif file

[kevinarden]

The intent is that the nodes are coincident and you are using conforming to glue them together. They do not have be to coincident but then you need to define the mortar conditions so that the non-conforming mesh nodes can be determined for the mortared boundary. It is not intended to be used as in your example, to simply set one BC equal to another.

[mika]

Do I need to put additional input there? I would expect not as CBC should just replace the dofs with the corresponding node and checked a test .

I think there is a problem in the library subroutines that prevents the creation of constraints when the boundary is defined in terms of point (101) elements. The both commands "Periodic BC" and "Conforming BC" cause the Elmer solver to call the subroutine CreateInterfaceMeshes which denies to create a list of connected boundary elements in the case of 101 elements. This should be revised in order to enable this functionality for 1-D models.

-- Mika

[Stefan Hiemer]

Hi Mika,


thanks for your answer. Two questions: 1. Constraints are only needed for PBC, correct? CBC should work by replacing indices if I am not mistaken, so the requirement of conformity will always be fulfilled. 2. Do you know whether this change in the routines will happen soon?

Thanks and regards
Stefan

[mika]

Do you know whether this change in the routines will happen soon?

If you update to obtain the change

https://github.com/ElmerCSC/elmerfem/co ... 1f1aa44250

it might now be possible to apply Conforming BC in connection with a 1-D model. I also added a simple test

https://github.com/ElmerCSC/elmerfem/co ... 6da044a8d8

to check that a sensible looking periodic solution under a periodic load can now be defined by using the keywords Apply Conforming BCs and Conforming BC. If you try to run this example, note that the file mesh.boundary was modified so that the boundaries can be identified by using the command Target Boundaries... The reason for this is that here the command Target Nodes cannot yet be combined with Conforming BC, so I needed to update the identifiers for the boundaries.

-- Mika