Convergence criterion in staggered scheme

[Anmo]

Dear all,

I am using a staggered scheme for an user element in FEAP 8.6. The 4 degrees of freedom are split as:

PARTition
  1 1 1 0
  0 0 0 2

The displacement value for the second field stays constant at all nodes for certain time steps. This is a property of the used physical model. In those cases the residual norm is 1e-16 in the first solution step. For example the initial residual is

Residual 2 norm =    4.4330922E-16   1.0000000E+00

As I understand, FEAP is using a relative residual tolerance as convergence criterion. Because the residual norm is that small from the start, the relative residual norm is about 1e0 at all steps and the output file gives for the example:

*WARNING* NO CONVERGENCE: Time = 3.0000000E+00: Residual =  3.76986E-16

even though it is converged from the beginning. If I change the relative residual norm, it would have to be to 1e0. This would also effect all the other steps, such that I do not reach convergence, when the displacement values change. Is it possible to add the absolute residual norm as a convergence criteria? Or is there a different solution to the problem?

Thanks a lot in advance.

[Prof. S. Govindjee]

There is also an absolute tolerance that you can set (on the energy norm).

Code: [Select]

TOL,ENERgy,<value>See also Appendix D in the manual for further information on the TOL command.

[JStorm]

I wrote a user macro in order to set the reference values for the relative residuum and energy criterion to the value of the first iteration in the staggered iteration.
That works quite well.
The variables are in the headers rdata.h and rdat1.h.

Hence, I store three sets of this pair of reference values for a problem with two fields, one for each field (embedded Newton loop) and one for the monolithic problem which can be used as convergence criterion for the staggered loop.

[Anmo]

There is also an absolute tolerance that you can set (on the energy norm).

Code: [Select]

TOL,ENERgy,<value>See also Appendix D in the manual for further information on the TOL command.

Thank you very much for your help. This works for my example.