Symmetry BC for thick wall cylinder

[RaduP]

Geometry: concentric thick wall tubes, with differential thermal expansion. Software: ANSYS 10.0.
If I use plane stress elements, the solution for a 1/4 cross-section model with symmetry BC coincides with that for a full model which recovers exactly the analytical solution.
If I use 1/4th model with generalized plane strain (long cylinders) elements and same symmetry BC, the solution is no longer axisymmetric, with the contact pressure being non-uniform. Same (bad) solution if using contact elements or coupled DOFs between the 2 cylinders.
Question: why wouldn't the symmetry BC work with generalized plane strain?

 

[rb1957]

I'd say the answer is more to do what what is working (than waht isn't) ... i think that your plane strain plates allow a bending reaction that the plane stress ones don't have, so this should change the result; no?

 

[EdR]

RaduP:

You should still get results from the plane strain solution that are symmetric. If you don't then I would look carefully at your model as you have introduced some kind of asymmetry in the plane strain model.....

You should not expect to get the same absolute results from the two models as the end conditions (in the z direction are different) but I would expect to retain the symmetric aspects of the solution...i.e. just changing from plane stress to plane strain should not cause the general form of the solution to be different....just the absolute values...

Ed.R.

 

[corus]

You can't have the same solution as the analytical one with plane stress elements as your axial stress will be zero.

There's no reason why the generalised plane strain element wouldn't have worked. I'd check your input temperatures as a non-uniform contact pressure would indicate you have non-uniform thermal expansion or an error in your contact surfaces.

Surely 2D axis-ymmetric elements would have been much better to use?

corus

 

[RaduP]

EdR and corus, thank you for your input.
Of course I don't expect the same solution between plain stress and plain strain. With regard to model asymmetries, the generalized plane strain model is exactly the same as the plane stress model, except in the batch file I change plane strain to plain stress (KEYOPT...).
Corus,
I've chosen to use 2-D solid as opposed to axisymmetric, because I wanted to be able to later extend the model to some other non-axisymmetric geoemtries and/or loading.