Cancel
Showing results for
Did you mean:
Highlighted

# SOL 601,106: force program to continue when convergence was not achieved at last iteration

Experimenter

Where is the "proceed when not converged"-button in SOL 601?

My contact-problem has difficulties achieving convergence at the beginning of the analysis.

When automatic time stepping has reached the smallest allowed time step size, execution of NX Nastran stops.

NXSTRAT      100    AUTO       1 LOADOPT       0  MAXITE      20        +
+        ATSSUBD    1000 ATSMXDT1.000000 ATSNEXT       1   NSUPP      10+
+         TNSLCF       1   IRINT      25

I don't care if a few increments do not converge until contacts have stabilized.

Is there a trick like entering MAXITE as a negative number?

6 REPLIES 6

# Re: SOL 601,106: force program to continue when convergence was not achieved at last iteration

Siemens Phenom

By definition, an increment needs to converge before the next increment can be attempted.

However, note that "converged" is an indefinite term. Converged simply means that residuals are less than the defined convergence criteria, so you can change the definition of converged by modifying the convergence criteria.

If you increase the appropriate convergence tolerances, then the solution will consider itself converged and move to the next increment.

Note that increasing convergence tolerances may (probably will) affect stability in future increments. Relaxing tolerances may allow one increment to "converge" , but then the next increment may not, etc.

# Re: SOL 601,106: force program to continue when convergence was not achieved at last iteration

Gears Phenom

Searct Autotimestepping or ATS in solver parameters. This option instruct solver to break timestep and try to solve with smallertime interval until solution is reached or timestep reached limit.

# Re: SOL 601,106: force program to continue when convergence was not achieved at last iteration

Experimenter

I agree with what you are saying. My question is how I can override the decision of the program to stop instead of continuing with the next time step.

I am sure, there exists a System Cell or a undocumented Parameter that does just this thing. In Nastran everything is possible ;-)

# Re: SOL 601,106: force program to continue when convergence was not achieved at last iteration

Gears Phenom

As JimB say solution is sequental, you can not calculate new timestep step until you not converged in previous. You can try to use restart and make some changes in loads and continue from last conwerged step.

# Re: SOL 601,106: force program to continue when convergence was not achieved at last iteration

Experimenter

I do so almost every day with MSC Nastran or MSC Marc and it works fine for me. As long as the final timesteps converge, results are o.k.

I cannot believe, that I am the first person who requests a "proceed when not converged"-feature in NX Nastran SOL 106.

NLPARM iteration strategy computes the best attainable solution and continues to the next load increment, when convergence cannot be achieved. Why should the same be impossible for NXSTRAT?

# Re: SOL 601,106: force program to continue when convergence was not achieved at last iteration

Siemens Phenom

"As long as the final timesteps converge, results are o.k." - this is a very dangerous and typically invalid statement, particularly for nonlinear anlaysis. More often than not, results are path dependent and not converging properly in early time steps leads to a different path being taken to arrive at a different set of "final" results. If it truly did not matter, the convergence tolerance defauts would be larger.

The answer to your question was contained in my first response: "convergence" is not an absolute. "convergence" simply means that residuals are less than specified tolerances. The parameters you can adjust to allow the solution to continue are the convergence tolerances. If you loosen the tolerances up, the program will achieve "convergence" earlier.

If you wish to elaborate exactly how you are telling other software to proceed (i.e. what parameters you are setting), you may receive a more satisfactory response.