My intention was to use simple rigid links and handle the rotations using a constraint. It works at one of the beam end, not the other which I find strange.
I seem to remember that this wouldn't work. I checked and tried using two advanced RL with all DOF locked and then unlocked the rotational DOF via the constraint. The analysis failed to solve due to a singular stiffness matrix. However, if I locked all DOF on the constraint and then released the rotational DOF on the advanced RL, it will solve. I'm thinking it might have something to do with how rotational for an advanced RL are handled and how rotations at a constraint are handled.
If you have time, check out the constraints in this model:
So I played around with your model and found that rrotational DOF for a constraint only become available for a simple RL between that point and some other reference geometry AND if that points coincides with 0,0,0 of the WCS. I seem to vaguely recall my professor for FEM talking about how rotations are with respect to the WCS or something along those lines (he use to be the R&D manager for Mechanica/Simulate). He's a member on my thesis committee, so next time I see him I'll ask him about this.
I did try creating a local CS at the second datum point and defined a new constraint with respect to that CS, but the rotational DOF on the constraint were still grayed out. However, if I created a beam between the datum point and another point (like a vertex of an edge), then the rotational DOF of the constraint un-gray and I can lock them. That being side, I think the default (grayed out) state for the constraint is the rotational DOF being free. I'm not sure if after doing this beam trick and locking the rotational DOF whether they will stay locked.
I can solve it using a workaround, as you mentioned using an advanced RL, or a short stiff dummy beam where I can add beam release at one endpoint. I had planned to use it for educational purpose, in situations like that one shouldn't have to struggle with workarounds...
A short beam (or a zero-length spring) was my next suggestion. I seem to recall my professor telling me how PTC changed the rotational DOF control on a constraint because too many users were not using it incorrectly, and that it was just easier to take the option away from them except for under a few circumstances.
However, if you were going to use this in an example for educational purposes, then maybe this will be a good thing. With this example, you can show how sometimes modeling strategies that an analyst planned on using don't work out, and that workarounds need to be developed that result in the sample (or very similar) stiffness characteristics.
