Convergence issues when changing forces.

[JGoeransson]

Hi!

I recently got a task at work to update a few already computed ANSYS-scripts and to evaluate the results.

One of the scripts did however present me with quite a wierd issue of convergence. As the only things to be updated in the scripts were the loads I went ahead and did this (note: no force have changed direction and they were all smaller than the old forces).

I quite quickly get the problem of convergence and the solution stops.

But the script does converge without a problem with the older larger loads, which I find very odd.

Might be an ANSYS issue, might be that my knowledge of the FE-theory is lacking and therefore I can't see the reason why I'd get problems with convergence.

Thanks in advance for any input!

P.S Below are the errors from the log.

*** WARNING *** CP = 2399.266 TIME= 14:20:03
The default values that can be modified on the CNVTOL command have been
internally reset to TOLER = 1.0E-2 with VALUE = 1.0 due to convergence
difficulties and the analysis is continued. Please verify the
analysis with this adjusted value of TOLER. You may need to specify
the NSUBST or CNVTOL commands for convergence and accuracy of the
solution. This automatic resetting of TOLER and VALUE can be turned
off if you include the CNVTOL command in the input.

*** ERROR *** CP = 2418.859 TIME= 14:20:14
Solution not converged at time 5.E-03 (load step 1 substep 1).

 

[kan123]

It depends what kind of nonlinearity you are modeling. For example, a large load might cause bilinear contacts to close while a smaller load might cause some of the contacts to oscillate between a closed and open condition. You might try an infinitesimal load to see if the solution converges. If it does, you can use load steps with progressively higher loads to see where the problem develops.