Issues using periodic boundary conditions

[marandra]

I am trying to use periodic periodic boundary conditions via the apply_periodic_condition_process for the first time, and I am not getting the results that I expect.

I may have a problem with the setup of the case, so I have attached a minimal working example of a case that use the these conditions. The first figure shows the displacements obtained with the process, and in the second one the results with other routine written by @RiccardoRossi , which I think are correct.

In the example I am applying a displacement in the center nodes of the z-faces in x-direction. The results that I obtain in this (and all) cases is that the nodes of a set of x, y, and z edges just keep fixed.

Any ideas?

mwe_pbc_process.zip

[loumalouomega]

@marandra which is the routine which you mean?

[adityaghantasala]

@marandra I am having a look on the case you posted. I will let you know !

[marandra]

@adityaghantasala thanks! @loumalouomega , it is rve_periodicity_utility, for which I created an ad-hoc analysis which calls it similarly to how it is done in rve_analysis.py. The rve_periodicity_utility assumes the model part is a cuboid with parallel faces, computes the face-pairs from a given skin, and solve the circular dependencies of dofs that arise when imposing periodic bcs.

[marandra]

Hi @adityaghantasala Please remember to take a look at the case when you have a chance. Thanks.

[adityaghantasala]

@marandra sorry for late reply, can you post the case using rve_periodic_utility too ?! Thank you.

[marandra]

Sure, here you are: mwe_pbc_utility.zip (you'll need to copy periodic_bc_structural_analysys.py to python_scripts directory of StructuralMechanicsApplication.)

Note that in this case I need to specify a pinned node to avoid translation. If I do specify it in the previous case, the solution does not converge.

Let me know if you need anything else. Thanks!

[adityaghantasala]

perfect thank you !

[adityaghantasala]

I think I have found the reason. This is related to the corner nodes (which here are all except 2 nodes). The periodic boundary process has not a mechanism to handle them. In the next hours I will think of a solution and post here.

The RVE utility is taking care of these as it has the information about all the possible combinations of the periodic condition.

[adityaghantasala]

Consider the above node numbering from the example you posted. RVE utility connects the following nodes explicitly (in total 42) :

(slave and master) 17 and 9 18 and 9 7 and 15 1 and 9 2 and 9

which are not possible with the periodic bc process. On the other hand the Periodic bc process will create connections like this (because of the order of master and slave model parts, and the corner nodes ) with this process more than 42 : (slave to master) 9 to 1 1 to 10 1 to 2 14 to 2

in the above combinations 10 to 9 is established via 9 to 1 and 1 to 10. But with periodic bc process ,at the same time 1 is also connected to 2. This is because they are corner nodes. This is the reason why the process is not working for this case. I am having a more deep work in the rve utility to how the unique set of constraints (connections) are made to introduce them into the periodic bc process.

[marandra]

Hi @adityaghantasala , here rebumping the issue.

I've been doing fine with the utility so far, but I would like to give the process another try, as it is more flexible for my application. If I understand correctly, the issue is that the process is not automatically establishing some connections between corner nodes, and maybe also between side nodes. Does it requires some ad-hoc steps? Do you think it is possible to fix it?

[rubenzorrilla]

Closing as current version of the RVE process already takes care of the corner nodes. @adityaghantasala @marandra feel free to reopen as required.