Plasticity convergence error

[Anonymous]

Hello,

When running a thermal and mechanical DED simulation on Netfabb 2023.0, I am getting a plasticity convergence error that I cannot get rid off.

*** WARNING: 19
Code 1014
Plasticity algorithm did not converge. Reducing time step.


Signal tag 32A2
ERROR
Code 331
The time step required to solve is less than the minimum allowable time step specified in the *TRAN card. Rerun with numerical relaxation or reduce the minimum allowable time step.

I have tried increasing the number of iterations in *SOLU as well as increasing the residual tolerance and executing through the commandline with "pan" but that didn't work.

As advised by the error message, I am trying to define the time steps parameters with *TRAN instead or in addition to *TAUT=1

But I always get the following error  (or similar) :

Starting proc

Reading binary file: pale_4_19_thermal.bin


forrtl: severe (157): Program Exception - access violation
Image PC Routine Line Source
pan2.exe 00007FF65BBDA056 Unknown Unknown Unknown
pan2.exe 00007FF65BAED2D8 Unknown Unknown Unknown
pan2.exe 00007FF65BAC3128 Unknown Unknown Unknown
pan2.exe 00007FF65BA3140F Unknown Unknown Unknown
pan2.exe 00007FF65BA310A7 Unknown Unknown Unknown
pan2.exe 00007FF65C7EABCE Unknown Unknown Unknown
pan2.exe 00007FF65C8D74BF Unknown Unknown Unknown
KERNEL32.DLL 00007FFF325F7034 Unknown Unknown Unknown
ntdll.dll 00007FFF330C2651 Unknown Unknown Unknown

In fact, I am not sure how to set all the *TRAN parameters and would like to use their default values or let them be automatically calculated by the solver except for the variable I want to change which is the minimum time step value. I have tried a few things but it seems I must input all the values.

1) Is there a way to only change one of the variables of a card ?

2) If not, where can I find the timesteps used by the simulations that were run with *TAUT=1 and without *TRAN ? I cannot find that information in the log files. This would help me set the TRAN card properly instead of trial and error.

3) Increasing the number of iterations in SOLU does not seem to be reflected in the solver output, i.e. even when setting 50 iterations, the convergence error is triggered at the 21st iterations if the residual is too high, e.g. :

*SOLU
50, 1.d-2, 1.d20

4) What else could be done to help plasticity convergence ? (generally, without plasticity the simulations runs until the end)

Thanks in advance for your help.

Best regards,

Stéphane

[michael.gouge]

Stephane,

You certainly getting into the deep end here. Let's see if I can help:

1. You can change just one variable but I think what you want is to be able to only specify one variable and that's not feasible. What I can do is offer the values I generally use:

*TRAN
0, [End of deposition time + 7200], 0.01, 100, 1e-6, 2.5, 20, 99999

2. Hopefully the above will suffice, but the TAUT=1 time step will equal laser radius/speed from the LSR file. Without using *TAUT you'd be using a value very similar to the initial time step, which using the above defaults would make for a very slow simulation. If you are going to drop *TAUT I'd put in the value it would calculate and use as your initial timestep instead of 0.01

3.  I think this confusion is due to us using the term 'solution iterations' for two different things. In the *SOLU card the value there applies to how many attempts each time step the solver will take to achieve convergence. Once it attempts to exceed this value we take a cutback - we step back slightly in time and try to solve again. The solver can take up to the number of cutbacks allowed by the *TRAN card (default of 20) before ending the simulation.

4. The most powerful tool to achieve convergence is the numerical relaxation card, *RELA. This can be used to break the newton rhapson update into multiple steps (1st argument) and apply relaxation (2nd value) The default values written by the GUI for the mechanical simulation is 1, 0.2, so we take 1 step and apply only 20% of the calculated derivative. To increase the odds of convergence increase the 1st number and decrease the 2nd, i.e. 2,0.1 solves most things, with 3,0.1 being about the limit I'd recommend attempting. If the solver does not converge it may be due to a very large non-linearity in the thermal results, or due to the material property model, which we can discuss if the above does not resolve your issue.

Best,

Michael

[Anonymous]

Hello Michael,

Thank you for your response and for the TRAN parameters. I used these parameters along with TAUT=1 and it is already much better in terms of plasticity convergence. I got a few warnings along the way but the simulation went further than before. I had another unrelated issue so I have to rerun the simulation, thus I don't know yet if it's going to reach the end but it's on a better path than before 🙂

I had used RELA beyond the acceptable limit so I could not really rely on it to further improve convergence.

I was also thinking that it could be due to my custom material properties, but hopefully it's correct.

I will let you know if the simulation runs into such convergence issues again.

Best regards,

Stéphane

[marcoslopeztrazo]

Micheal,

I've tried to follow your steps changing the *TRAN card as you have shown above and even using the relaxation values you gave (3, 0,1), however the programm gives me anyway the same error when I use a LSR file with low scann speed ( around 5 mm/s) and a laser power around 1500 W, although these error dissapear when I increase the scan speed to 100 mm/s and with the same value of laser power (around 100OW):

Code 1014

Plasticity algorithm did not converge. Reducing time step.

Signal tag 1C3

ERROR

Code 331

The time step required to solve is less than the minimum allowable time step specified in the *TRAN card. Rerun with numerical relaxation or reduce the minimum allowable time step.

Thanks in advance for your help.

Best regards,

Marcos