Repository navigation
New Ferrite-based materials/assembler package: overlap, coordination, and naming #1377
Replies: 3 comments 5 replies
|
Very cool to hear that you are building so much on top of Ferrite and the other packages! Regarding the naming, at least I've refrained from using My motivation for creating I'm happy to see if your MPI support can fit into If there are specific parts of e.g. |
|
Cool to see more Ferrite use! When suitable, we should put your packages into the downstream testing https://github.com/Ferrite-FEM/Ferrite.jl/blob/master/.github/workflows/Downstream.yml to make sure we do not accidentally break them. |
|
Really awesome to see some coupled FE-MD code coming to Julia! As a side-note, the FerriteCon will be in Braunschweig this year. I agree with Knut here on the naming. Packages extending Ferrite specifically can (and probably should?) have the prefix. Packages consuming Ferrite should not have them. See e.g. FerriteOperators (the "assembly layer") vs Thunderbolt (the "modeling layer"). Do you mind briefly explaining what you mean by Also
does not really exist. Specific applications need different interfaces. With FerriteOperators I try to provide (among other things) building blocks, so interfaces can be put together quickly in the form of operators. I also think that the success of Ferrite being adapted by quite a few groups now is that we essentially provide a way to quickly put together a tailored API for specific application areas, in contrast to existing frameworks where users try to bend existing APIs to incorporate their special elements. |
Uh oh!
There was an error while loading. Please reload this page.
Uh oh!
There was an error while loading. Please reload this page.
Hello everyone,
I am Lukas, a doctoral researcher at FAU Erlangen-Nürnberg in Germany, working mostly on coupled FE-MD (Molecular Dynamics) problems. After @Joroks re-implemented our Matlab-based coupling code base ("Capriccio") in Julia as part of his Bachelor thesis, using Ferrite.jl for the FE part, parts of our research group started moving from Matlab towards Julia and Ferrite.jl.
From there on, re-establishing our previous FE capabilities on the new Julia-based Capriccio code became one of the main pursuits of my work: it started with the translation of Wuyang Zhao's VEVP material model, extended to several more (partly rather complex) polymer models, and the FE kit around them grew into a generic assembler (assembly, threading/MPI reduction, quadrature-point coupling weights for the FE-MD overlap region, state handling, and adaptive time stepping), which I separated into a standalone package. The original authors of the implemented material models agreed to publish their models within it, and we jointly plan to complete this work with a SoftwareX publication covering the full package, possibly after inviting further material model authors to contribute their models as well. Only while writing the documentation did I realize that related projects already exist, in particular @KnutAM's and @kimauth's repos with material models and generic assembler code.
I plan to release the package on GitHub within the coming weeks, and I wanted you to hear about it from me to discuss a few aspects beforehand. The overlap with existing work is layered: the assembler covers similar ground as FerriteAssembly.jl, though from a different starting point (FE-MD coupling weights, MPI reduction, recoverable local failures for adaptive stepping), the constitutive interface follows rather similar conventions as MaterialModelsBase.jl (after discovering it, I added an extension through which MaterialModelsBase materials run directly in the assembler), while the included model library itself does not duplicate anything existing (apart from basic models adapted from the Ferrite.jl tutorials). Given the overlap, mid/long-term, I am very open to upstreaming parts (e.g., polymer models into MechanicalMaterialModels.jl or MPI assembler features), to merging developments where sensible, or to whatever coordination you would consider most useful.
One question I would like your opinion on before proceeding is naming: is it okay to use a name that prepends "Ferrite" as someone currently outside the Ferrite.jl development team? I thought FerriteMaterials.jl would be a concise fit (even though it is specifically about "mechanical" material models and includes notable assembly logic), but I am also happy to use a different name if using the prefix is not desired. Please don't shy away from sharing your direct opinion; alternative suggestions are very welcome.
I am very much looking forward to your replies! If you feel like discussing this in more detail or want to have a look into the implementation before publication, I would also be glad to set up an online meeting.
Best,
Lukas
TL;DR
(a) A Ferrite-based materials/assembler package (grown from FE-MD coupling work, with partial overlap to FerriteAssembly.jl and MaterialModelsBase.jl) will be released in the coming weeks; open to upstreaming, merging, or other coordination mid/long-term.
(b) Is using the "Ferrite" prefix in the package name okay for developments from outside the Ferrite team, and specifically FerriteMaterials.jl?
(c) Anything else: happy to discuss, also via online meeting if desired.
All reactions