-
Notifications
You must be signed in to change notification settings - Fork 41
Workshop XIV Presentations
Jonathan Guyer edited this page Oct 27, 2022
·
9 revisions
09:12:34 From Daniel Schwen (he/him) to Everyone:
FFT frequency filter and then integrate!
09:12:48 From Daniel Schwen (he/him) to Everyone:
counting peaks will be ill defined at the interface
09:15:17 From Daniel Schwen (he/him) to Everyone:
(cutoffs, even if well defined, can amplify small errors)
09:16:28 From Jim Warren, NIST to Everyone:
@Daniel, can you repost your last comment? I just logged into the Zoom
09:17:30 From Daniel Schwen (he/him) to Everyone:
(cutoffs, even if well defined, can amplify small errors)
09:17:38 From Pierre-Clément Simon (He/Him) to Everyone:
I agree with @Daniel here. Using a threshold in the metric would make small continuous microstructure variations lead to discontinuous variations in the metric, potentially making similar microstructures “appear” quite different.
09:18:27 From Daniel Schwen (he/him) to Everyone:
How do you prevent interfacial motion here?
09:18:30 From Jim Warren, NIST to Everyone:
Interesting. I was thinking about a cutoff in the post-processing analysis, so, perhaps amplification is not an issue?
09:20:20 From Nana Ofori-Opoku to Everyone:
it should be done in the postprocessing
09:22:48 From Ahmed Kaci BOUKELLAL to Everyone:
Could you give the reference of Plapp's work? Thank you
09:24:47 From Daniel Schwen (he/him) to Everyone:
I suppose with conservation of energy the motion will eventually stop
09:24:57 From Daniel Schwen (he/him) to Everyone:
and then what you get is part of teh benchmark
09:25:36 From Daniel Schwen (he/him) to Everyone:
the discrepancy from a 50/50 liquid/solid config is part of the benchmarked value
09:26:07 From Daniel Schwen (he/him) to Everyone:
@Nana and @JW, yes of course in postprocessing, but the issue is still there
09:27:09 From Nana Ofori-Opoku to Everyone:
Frequency filtering, then integrating is probably the better way to go
09:27:50 From Pierre-Clément Simon (He/Him) to Everyone:
@Jim, @Nana, @Daniel Maybe we are not talking about the same thing. But even in post processing, using a cutoff (blue), the red and yellow curves would be seen as the same, but the yellow and green curves would be seen as very different since the green peak would be seen as not existing. (see illustration above)
09:28:47 From Nikolas Provatas to Everyone:
See discussion in Section C, around Fig 2.
09:30:10 From Nikolas Provatas to Everyone:
More generally, the approach of Section III A-C is useful as methodology to be applied to interface energies. and Figure 2 on the extrapolation idea which can also be used for looking the infinit size limit.
09:30:39 From Jim Warren, NIST to Everyone:
Thx. Enlightening.
09:33:26 From Daniel Schwen (he/him) to Everyone:
How about a grain rotation benchmark?
09:33:46 From Daniel Schwen (he/him) to Everyone:
circular grain in a matrix with a lattice rotation
09:33:54 From Nana Ofori-Opoku to Everyone:
@Daniel, we want to keep students not scare them away
09:34:03 From Daniel Schwen (he/him) to Everyone:
hey!
09:34:05 From Nana Ofori-Opoku to Everyone:
😄
09:35:20 From Nikolas Provatas to Everyone:
Also, to Jim's point, this would be more like pedagogy more than benchmark I guess (although I like pedagogy too and think we should include that too (?)
09:36:32 From Daniel Schwen (he/him) to Everyone:
agree
09:40:08 From Daniel Schwen (he/him) to Everyone:
hard to get periodicity right
09:50:47 From David Montiel Taboada to Everyone:
I’d volunteer to try this in PRISMS-PF, which is FE
09:51:58 From Nana Ofori-Opoku to Everyone:
@David, I’ve attempted this in fences before and it should be doable solving this model
09:52:00 From Nikolas Provatas to Everyone:
We volunteer to do this with one of our spectral codes.
09:52:20 From Nana Ofori-Opoku to Everyone:
*fenics
09:53:58 From Daniel Schwen (he/him) to Everyone:
We have PFC objects in MOOSE, but frankly... they suck
09:54:53 From Daniel Schwen (he/him) to Everyone:
I'd still be willing to subject us to the shame though
09:54:55 From Nana Ofori-Opoku to Everyone:
@Daniel, I didn’t know this. I’ll take a look at it
09:55:07 From Nana Ofori-Opoku to Everyone:
Hahahah, that’s the spirit Daniel
09:55:16 From Daniel Schwen (he/him) to Everyone:
Oh, wait, that stuff might be in export controlled Marmot
09:56:07 From [email protected] to Everyone:
I got PFC working in FiPy (finite volume) recently, so will also volunteer
10:00:09 From Peter Voorhees to Everyone:
There's an export controlled version of Marmot?
10:03:25 From Jon Guyer to Everyone:
Rodent munitions
10:04:23 From Trevor to Everyone:
Going 3D implies greater computational resource requirements (memory, time).
-
What exactly to upload?
- Specification is all over the place.
- Are
8a/r=0.99r*,8a/r=1.02r*, and8a/r=1.01r*separate uploads or all one?
-
What do files get called?
-
"The entries should be labeled as
solid_fraction_{i},free_energy_{i}orparticle_count_{i}depending on the simulation number (1 to 5)"What's an "entry"?
-
Do I need a
meta.yaml? Is it in the schema of Manual Result Upload? -
How do I see Version 0 of the specification?
- I have one run, not 5 and I only ran to
t=100. Is there no way to upload that to reflect that it was requested at the time? Otherwise I need to rerun my simulations, which is alienating.
- I have one run, not 5 and I only ran to
-
Any way to get faster feedback from the CI?
-
-
Why does code need to be manually changed to account for a new upload?
Shouldn't notebook be able to render any properly validated list of DOIs with appropriate metadata?
- What's the phase-field-unique aspect here?
- Ron Boivert has thought a lot about mathematical software
- Chandler Becker (and probably many others) has already done a lot about describing materials software and attribution
- Bring back for XV
- How to lightning talk
- Make point quickly
- Objective is to provoke follow-ups
- Don't overprepare
- Home
- Benchmark Presentations
- Peter Voorhees Phase Field Lectures
- Phase-Field Method Recommended Practices
- Workshop Presentations
- Miscellaneous