Interference-fit analysis

So it seems to work only with CA 10.1.
These days I'm away and I can only use the caelinux live DVD. None of the cases I have previously solved converge with CA 9.4 (automatic time stepping activated).
The difference is substantial: with CA10.1 I see the relative residual values rapidly decreasing (of 3 orders of magnitude at each Newton's iteration), while with CA 9.4 they remains almost constant.
Well, this could be explained since features are added to new versions but I didn't expect a regression in CA 10.2.

I'm not sure if there is a regression. I couldn't get it to converge on STA10.1 (Running CAELinux 2010 RC). Also comparing the two files they are not the same. The 10.2 file did not have initial penetration enabled. Also after looking at the file there are some improvements that I would make.

1.) The use of the timesteps in this problem are irrelevant as it is the same solution at each timestep. (i.e. the displacement at t=.1 ==t=1) because there is no applied displacement.

2.) not having a ramped loading causes the solver to need to solve it all in one step. Perhaps a solution that would converge more easily would be to start the two meshes so that there is no initial penetration and displace it down the necessary distance to cause the penetration. This gives the solver an easier method to solve i.e. if it cannot reach a solution at x time perhaps it can at x/2 and then x.

<br /><br />Post edited by: Peter Halverson, at: 2010/09/01 06:44

Actually, After looking at the mesh is there initial penetration?

Hi Peter,
there is no initial penetration of the mesh, I prefer to control the interference by adjusting the &quot;gap&quot; parameter (DIST_MAIT or DIST_ESCL in Code_Aster). This involves the introduction of the initial interference, not present on the model. To solve this problem the use of timesteps is very useful to achieve convergence: the gap do enforce displacements at the contact surface.
I see that the output you've got is very similar to that I obtained from STA 9.4 but completely different from mine for the same model on STA 10.1, how is it possible?
All model I posted converged, with more or less satisfactory results, running Salome-Meca 2010.1 (that contains SALOME 5.1.3 and STA 10.1 versions of Code_Aster) on Slackware 13.1 32 bit.

It gets even stranger.

On Salome-Meca 2010 it does not converge for me.
(Intel I7 processor)
Post edited by: Peter Halverson, at: 2010/09/01 15:35<br /><br />Post edited by: Peter Halverson, at: 2010/09/01 15:35