Generative Structural Analysis - Isostatic constraint

[kdx250]

I am trying to simulate a free cylinder (closed, with thickness) with a certain pressure applied to the inside surface (towards the outside) to inflate it. I am using to methods to do that. 1) Using only isostatic constraint and 2) Using isostatic constraint and self balancing (from the loads icon). On both occasions the result is not symmetric as expected. Instead the translation on one side is different than the other. Does anyone have any idea why is this happening or maybe a solution?

 

[flambter]

Well, I think it depends on what you want. The difference between the two displacements is about 1%, which is not that bad after all. I doubt that you can actually get a perfect match.

When using an isostatic restraint, Catia actually chooses 3 nodes and restraints them in order to avoid rigid body motion. (system 3-2-1). Those restraints block translation on some nodes, so maybe those blocked translations have a small influence (1%) over the overall displacements. For example, one node on your cylinder has a zero displacement value along the axis of the cylinder and therefore, all the nodes moves relative to that particular point. If this point is located on one side of the cylinder, it probably be the side where you have the less displacement, because the other side of the cylinder has a cumulative (small) axis displacement from all the nodes between that zero displacement node and the side of the cylinder. (Do I make sense??) This is my theory anyway, not an actual definitive answer.

Doing that kind of problem, I get about the same "error".

So maybe what you would like to do is try to make that zero translation node along the axis of the cylinder in the middle of the cylinder so both side of it would move the same amout.

FEM is always a compromise between what the software gives you and how you interpret the results. That second part is the most important, because after all, the software is just a machine! And to be honnest, this situation will never happen in real life, because it is always gonna be fixed.

 

[kdx250]

Thanks for your answer.
It is like you explain.
It is exactly this result (even the small difference) that I want to avoid. I am already aware of the 3-2-1 nodes system for constraining an object, either only with isostatic or isostatic+self-balancing.
At the first case (isostatic) there is no constraint visible whatsoever (it only shows the anchor icon) so I can not set the node to be constructed.
At the second case, yes you can see the nodes where it has selected for constraining the part.
The goal is to simulate a free body with internal pressure (imagine a body in space i.e.) and then see the deformations, stresses etc. I need that precision to decide for the material to be used. After all if you can not have control over the constraints and if you get these kind of errors on simpler modes how can you trust the results for a more complicated one (i.e. a part which is not symmetric)?

 

[flambter]

For my part, I used isostatic restraint and then right click on it in the tree and you can select something like visualize on mesh. Then you can see (really small but its there) which nodes are constraints.

If you want to specific exactly the node to restrain, then you have to put points on your cylinder where you want the constraints to be. Then use a local mesh - imposed points (so it creates a node at that point in the mesh) and then if you put a constraint on the point, I think it should work. That way, you can have a full control on your restraints.

The thing is you cannot just use some arguments that both side have the same displacement and therefore the model is valid. It is more a question of is the diplacement is valide in itself. Then only a hand calculation based on theory (close cylinder should be easy to find) could tell you if you model is valid or not or how good it is. Maybe the "real" value is something in between the two values or more, or less, making the fact that the two displacements are the same irrelevant. I still think that FEM is not a perfect science and getting the exact answer is something you can come close to, but not be perfectly spot on!

You can also create an applied load sensor to see if the resultant is really zero. If you have a small resultant, than maybe that is what makes the difference for both ends.

You should also be carefull with special features as isostatic constrain and self-balance. I saw models where using those features acutally really influenced all the displacements and stress plots.

But your problem is interesting. I am wondering if it is really possible to make a "free body in space" problem since the actual base of FEM is "no free body" motion, everything should be fully constrain.