Interference-fit analysis

Hi forum,
I'm interested in the estimate of the stress distribution in interference fitted shaft-hub assembly. As a starter I'm investigating a simple case, easy to evaluate with conventional equations.
In order to impose the interference I've used the keyword DIST_MAIT=radial_interference
The results I've obtained are “quite corrects”. I'm not sure everything has worked properly because:
1) average stress on contact zone is correct but rather nonuniform. The mesh I've use isn't so coarse, and with quadratic elements I expected a better result.
2) Looking at the displacement representation of the contact elements on post-pro seems that only few elements apply the dist_mait parameter (in the attachment the surface “esclave” is visible)
3) I can't use the same scale factor plotting parameter from field DEPL or from field SIGM. In the attached image the upper plot is the contact pressure (LAGS_C from field DEPL) with scale factor 70, the lower plot is the equivalent stress (VMIS from field EQUI_ELNO_SIGM) with scale factor 1e-4. The plot of VMIS with the scale factor 70 lead to an exploded model that don't respect even the boundary conditions applied with DDL_IMPO.
Any help or pointers will be appreciated.<br /><br />Post edited by: theCapitain, at: 2010/08/13 17:22

the image I was talking about

Hi,

It would help if you could post the *.comm file so we can see details of the methodology, as well as the mesh file to run it and see the corresponding *.mess file.

Regards,
JMB

Tanks for your help JMB,
I've attached the .med file (nothing special, a mesh generated rapidly just for test purposes), the command file and my .mess file.
I'm looking forward to read your considerations.

I had hoped to have had the chance to run your analysis before commenting but I haven't had the time. However a couple of quick pointers.
1. Most of the contact methods only check when a slave node penetrates a master surface and then apply a correcting factor to remedy this*. Which means you will have a wide range of stresses on the the nodes. Code Aster has some alternatives but I forget what they are. You can either consult the documentation or try them all.

2. Quadratic elements are not always better than linear elements when contact is concerned different methods will treat quadratic elements differently. Sometimes linear elements will do better.

I'm sorry this wasn't more informative. I hope this helps