Unusual models without any constraints

[FEA way]

Hi,

recently in Abaqus I've seen something that people interested in FEA will probably find very unusual. There are two YouTube videos where structural mechanics model is solved in Abaqus without any constraints. There are no BCs, inertia relief nor other methods restricting Rigid Body Modes in these simulations.
Here are these two videos:

https://www.youtube.com/watch?v=uRZZeOxOGRQ

https://www.youtube.com/watch?v=Y7VkY0qkxZw

In both cases loads are self-equilibraiting (stable models) but in any other CAE program I know about such models would require constraints anyway.

I performed these simulations to make sure it really works and found out that it's true.

Does anyone have an idea how it is possible that such simulations run in Abaqus ? What is different in Abaqus solver that it obeys unwritten rules of FEA.

 

[Erik Panos Kostson]

Like you say it is self equlibrating.

I do not know what you mean with any other software you know (have you actually tried this in all other software). Say in ansys, one can solve this (no restraints, see image below). Since some pivot ratios are zero (FEA does not like that), the pivot checking needs to be turned off and then it will solve without any problems (see below). This was also done for linear static (K*u=F), where the singularity in the stiffness matrix can be severe.


In any case this could be solved in a FEA software that has a dynamic solver (full equation of motion).

The same problem solved (both sparse and iterative solvers) in Strand7 (no restraints again), again no major issues (second image below).

 

[FEA way]

I tried these cases in COMSOL and SolidWorks Simulation as these two programs are available at my institute where I perform FEA researh. It didn’t work there so I thought it won’t run in most types of software.

Apparently some FEA programs check for constraints before running the analysis and show errors if there are no boundary conditions while others ignore the lack of BCs and solve the problem without error as RBMs are eliminated by the loads themselves.

 

[Erik Panos Kostson]

Yes, say Strand7, one would get a warning about the constraints, if sparse solver is used the pivot ratios will be close to zero and reported, but it will solve (in ANSYS, one would need to tell it that it is OK to have a small pivot, so switch the check off as I said). For iterative solver the pivot is not such a problem (solves differently than the sparse).

So it is all how different software treat these things (some move on and solve and give a warning, some give an error, and stop, and so on...)