Hello,
I suddenly faced the following messages after normal calculation step messages.
Intermediate output for t = 8500.0 s
CPU-time: 630 s
Current phase-field solver time step = 5.56E-04 s
Average conc. of comp. 1 = 0.1249433, Variation = -0.0208729 wt%
Average conc. of comp. 2 = 9.1675329, Variation = -0.1675345 wt%
Temperature at the bottom = 923.00 K
Temperature gradient = 0.0000 K/cm
complete relinearisation!
complete relinearisation!
complete relinearisation!
complete relinearisation!
not converged! 2 1Error: 1 8831.71893535601
2nd trial not converged! 0
3rd trial not converged! 0
4th trial not converged! 1
not converged! 2 1Error: 1 8833.69322314905
2nd trial not converged! 0
3rd trial not converged! 0
4th trial not converged! 1
not converged! 2 1Error: 1 8835.67528401362
2nd trial not converged! 0
3rd trial not converged! 0
Convergence problems, increasing smallest sitefraction from 1.00E-30
to hardware precision 2.00E-12. You can restore using SET-NUMERICAL-LIMITS
4th trial not converged! 1
And, the calculation seemed to stop at the last error message.
I tried to avoid this difficult by modifing values of 4th parameter after automatic option in time-step, of Phase minimum parameter, of Interface thickness parameter, and of cell dimension parameter. I, however, bumped against the same error every time.
Please suggest precise operation for parameters to avoid this error.
---
original message from Sukeharu Nomoto
Output message "not converged!" several times
Answer
Hello,
If you get the message
>>>>2nd trial not converged!
this means that the Newton-Raphson-Routine was not able to find a proper Quasi-Equilibrium for a certain interface cell. There are four different Newton-Raphson routines, so there are four trials. Even if the fourth one is not converging this does not automatically mean, that the program run cannot recover and continue properly.
But if you get the message
>>>>Convergence problems, increasing smallest sitefraction from 1.00E-30
>>>>to hardware precision 2.00E-12. You can restore using SET-NUMERICAL-LIMITS
then there is no chance to recover because the Thermo-Calc subroutines got some extreme parameters like negative temperatures or concentrations. We will try to catch this error on the MICRESS level in future releases.
Anyway, the reason for this behaviour in most cases is that the simulation conditions (parameter) are not stable. Instabilities of all kinds typically show up first time in the thermodynamic subroutines.
Most probably this error messages will vanish if you reduce slightly the interface mobility of the corresponding interface.
If you want to try whether some smaller time-stepping could be helpful you should reduce the second parameter in the automatic time-stepping input, e.g.:
>>>>automatic 0.9 0.5 1.E-3 1.E-5
This is a prefactor to the segregation timestep criterion, which is typically the limiting one in alloy simulation and which is 1 by default. (Parameter 3 and four are then still optional).
Good luck!
Bernd
---
original message from Bernd
If you get the message
>>>>2nd trial not converged!
this means that the Newton-Raphson-Routine was not able to find a proper Quasi-Equilibrium for a certain interface cell. There are four different Newton-Raphson routines, so there are four trials. Even if the fourth one is not converging this does not automatically mean, that the program run cannot recover and continue properly.
But if you get the message
>>>>Convergence problems, increasing smallest sitefraction from 1.00E-30
>>>>to hardware precision 2.00E-12. You can restore using SET-NUMERICAL-LIMITS
then there is no chance to recover because the Thermo-Calc subroutines got some extreme parameters like negative temperatures or concentrations. We will try to catch this error on the MICRESS level in future releases.
Anyway, the reason for this behaviour in most cases is that the simulation conditions (parameter) are not stable. Instabilities of all kinds typically show up first time in the thermodynamic subroutines.
Most probably this error messages will vanish if you reduce slightly the interface mobility of the corresponding interface.
If you want to try whether some smaller time-stepping could be helpful you should reduce the second parameter in the automatic time-stepping input, e.g.:
>>>>automatic 0.9 0.5 1.E-3 1.E-5
This is a prefactor to the segregation timestep criterion, which is typically the limiting one in alloy simulation and which is 1 by default. (Parameter 3 and four are then still optional).
Good luck!
Bernd
---
original message from Bernd
Re: Output message "not converged!" several times
Thank you for your suggestions. I am going to try them.
---
original message from Sukeharu Nomoto
---
original message from Sukeharu Nomoto