IN718 Heat Treatments - Time/Solver Steps

[dhicks]

Hello MICRESS Representative,

I am currently running some tests in MICRESS to simulate an additively manufactured (laser beam powder bed fusion, LB-PBF) IN718 specimen accompanied by multiple-step heat treatment processes. For reference, I am using these articles as a starting point: (https://doi.org/10.1016/j.addma.2018.11.024 and https://doi.org/10.1016/j.mtla.2020.100862). I've also been referring to Post #215 (https://board.micress.de/viewtopic.php?f=9&t=563S) for guidance as well.

I can successfully simulate the expected phases during the AM process (e.g., growth of C14_LAVES in the interdendritic regions and NI3TA_D0A near the grain boundaries; I am ignoring the smaller gamma' and gamma'' phases for now).
I'd like to simulate the dissolution of the C14_LAVES phase upon heat treatment; however, the simulations are prohibitively slow given my current parameter choices, and I have yet to have a completed simulation for a single AM cycle followed by a 6hr heat treatment.

I'm guessing that the update interval for molar volume, diffusion coefficients, and nucleation are some of the main reasons for an excessive runtime.
The current values are:
- molar volume TQ update interval: 1e-2
- diffusion coefficients update interval: 5e-3
- nucleation coefficients update interval (same for all nucleation types: C14_LAVES at FCC_L12/LIQUID interface, FCC_L12 at C14_LAVES/LIQUID interface, and NI3TA_D0A in FCC_L12 matrix): 2e-4
For reference, I've attached my current driving file.

The values I have yield reasonable results for the AM portion of the thermal profile, but they appear to be too frequent for the multi-hour heat treatment(s). I could partition the simulation into two: an AM segment and a heat treat segment. However, if possible, I see utility in being able to run a full thermal profile in one simulation. Nevertheless, I do not have a good sense of how to change the parameters for the heat treatment portion.
I am using the temp_limit_phase_TQ option to help speed things up, but the heat treatment sits right in this range.

1) Do you have any recommendations on how to speed up this simulation?
2) Is there a general rule of thumb - or recommendation - for time stepping/updates for long heat treatments? I'm guessing it is generally material-specific, but any insight for this materials system would also be appreciated.

Thank you in advance for your time!

Best regards,
David

[Bernd]

Hi David,

Welcome to the MICRESS Forum! It is also nice to hear that, as you say, you have overcome the numerical issues with the LAVES phase and already got reasonable results for solidification under SLM-conditions.

With performance issues, the first thing to do should be to have a look at the .TabP output. I had already summarized here, what can be typical reasons for bad performance, and how the output of the .TabP-file can be interpreted. But performance problems can always be good for surprises...

A specific difficulty of your setup arises from the strongly different time-scales which you have within one temperature cycle. For some of the parameters it is possible to adapt them by specifying them temperature or time dependent. This is the case for the diffusion data updating and the thermodynamic updating. For nucleation checking, it can be achieved by consequently restricting checking to the temperature range where nucleation is expected, and by using different seed types (with different checking intervals) for different temperature/process regions. Updating of molar volumes is uncritical, because this quantity does not change significantly and is only of minor importance, as long as you do not couple to stress, volume expansion, or latent heat - thus you can simply use a constant value here, or work with very little updating.

But still, it is not satisfactory to have many other numerical parameters identical under so different process conditions. Alternatively, you also should think about starting an extra simulation after solidification has finished (based on the restart output file as initial microstructure). This could still be automatized by a suitable Python script or Jupyter notebook.

Unfortunately, there is no general rule how to select updating intervals, even (or especially!) if they can be adapted to the respective parts of the process cycle. The rationale, anyway, should always be based on how much change of temperature and composition the system realizes in between the updates. So, e.g. during a solution heat treatment, the diffusion coefficient won't change a lot (especially when using global ("g") diffusion data).

If you could send me your .GES5-file and your thermal_profile.txt, I would try to have a look if there are other problems/chances for further optimization.

Best wishes
Bernd

[dhicks]

Hi Bernd,

Thanks for your prompt response and for your feedback/suggestions - I appreciate it.

Thank you for also directing me to the TabP file discussion. Upon an initial glance, I do not immediately see indications of bad performance after looking at the columns (I will take a more comprehensive look though).

I do like the idea of adapting the parameters to make them temperature- or time-dependent. But, I understand that this approach may still not be the most performant (good to know that these options are available though - thank you). I had a feeling that the best way was to speed this up was to partition the simulation based on the differing time scales. I will try to set this up and test.

For an isolated heat treatment simulation, do you recommend changing any other parameters outside of the diffusion update interval? Or do I just need to test/change different parameters (e.g., nucleation time intervals, min undercooling, etc.) to see if they yield expected results with an improved performance?

As for the GES and thermal profile, I've attached these files as well. If you notice any opportunities for optimization in these files, please let me know.

Best regards,
David