Negative time increments with AUTO and PART

[Anmo]

Hello,

I want to use AUTO TIME with PARTition in FEAP 8.6, but the combination leads to negative time increments. I am using a user element, but I could reconstruct the problem the input file example  in "PARTITION + AUTO TIME" (http://feap.berkeley.edu/forum/index.php?topic=1688.msg6484;topicseen#msg6484). The file is attached. I introduced a third partition, which consists of all degrees of freedom. This partition is solved to check, whether the complete residual converged as well. That is necessary due to the material model in my user element. Is there a possibility to prevent the negative time increments? Can the increment only be adjusted in partition 3?

In the case of only two partitions, the time increment is often adjusted in partition 1 and partition 2, such that dt-old and dt-new are not the same in both partitions. For example
partition 1:     
   dt-old =   9.9999999999999995E-007
    dt-new =   1.5848931924611141E-006
partition 2:     
   dt-old =   1.5848931924611141E-006
    dt-new =   2.5118864315095823E-006
Does that mean, that partition 1 and 2 are not calculated for the same?

Thanks a lot in advance.

[Prof. R.L. Taylor]

I have never considered using auto dt with partitions.  I would assume that it would not work with the current design since corrections to the dof's in other partitions would not get correct corrections to their increments, velocity, etc.

[Prof. S. Govindjee]

The way that AUTO is programed is going to break what you are trying to do.  AUTO is set up to ensure that THE inner loop's iteration count is controlled.  But in your case you have multiple loops within your LOOP INFInite.  It will thus be very confused as what to do.  Note that all the logic for
AUTO is in the code for the execution of the NEXT command (of the inner loop).

[Anmo]

Thank you very much for the fast answer. In that case I will stop using AUTO in my examples.

[Prof. S. Govindjee]

Your other option is copy program/pmacr2.f and modify the auto time stepping logic, but this would be a non-trivial exercise.