Hi Anthony,

I am very interested in this. Who knows more about this?

I am doing some problems which are close to what you do. My problem is
contact (contact area changes with loading/unloading), elastic-plastic
deformation and non-linear material.

It always stopped due to low convergence rate. I do not got negative
eigenvalue message. This maybe due to my ultra fine mesh.

I use 3D 8 nodes element. I found this kind difficulty when the contact
surface goes in at diagonal direction.

Any suggestion is appreciated.



On Mon, 31 Jan 2000 16:16:12 -0500
"Anthony J. Petrella" <[log in to unmask]> wrote:

> Hello,
> 
> I am working on a difficult contact analysis in ABAQUS that contains 
> multiple contact interfaces and moderately complex geometry.  It is also a 
> large-sliding problem with nonlinear material properties.
> 
> I have had a great deal of trouble with a phenomenon I believe is widely 
> known as 'chatter.'  That is, in a given *STEP of the analysis certain 
> nodes in a contact interface will become closed (in contact) and others may 
> open.  In subsequent increments of the same step certain nodes will 
> repeatedly become closed and open and closed again - leading to either 
> poor/slow convergence of the increment or no convergence at all.  Failure 
> of the solution to converge is associated with multiple cutbacks due to 
> severe discontinuities (interference of the contacting surfaces) and is 
> usually accompanied by warning messages reporting one or more negative 
> eigenvalues in the system matrix.
> 
> I should mention that most of the contact interactions are 
> rigid-deformable, and I initially achieved some improvement in 
> stability/convergence by refining the rigid surface mesh.  However, I am 
> now pushing the limits of available resources (memory, CPU) and would like 
> more insight into the real cause of my problem.
> 
> If anyone has experience with this type of analysis, I would greatly 
> appreciate you sharing your experiences with me.  Specifically, what is the 
> meaning of the negative eigenvalues in this type of analysis, and what 
> general measures can be taken to prevent chatter?
> 
> Kind regards,
> --
> ----------------------------
> Anthony J. Petrella
> University of Pittsburgh
> [log in to unmask]
> ----------------------------
> 



%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%