EDIT Bernd: topic title
A Elmer Solver in Elmer terminology = FreeCAD FEM Equation in FreeCAD terminology.
I suggest to … bolt is what has to be implemented
start implementing the static-current-solver next. This is what I can need and it shows very nicely the multiphysics capabilities of Elmer. https://www.nic.funet.fi/index/elmer/doc/ElmerTutorials_nonGUI.pdf page 38
“Tutorial 9 Thermal actuator driven with electrostatic currents”
Directory: ThermalActuator
Solvers: StatCurrentSolve, HeatSolve, StressSolve
Tools: ElmerGrid, editor
Dimensions: 3D, Steady-state
Than implement the magnetostatic solver… this is what looks nice and should be a cliffhanger for all magnetos out there https://www.nic.funet.fi/index/elmer/doc/ElmerTutorials.pdf page 39
“Tutorial 8 Magnetostatics – Magnetic field resulting from a permanent magnet”
Directory: Horseshoe
Solvers: MagnetoDynamics2D
Tools: Gmsh, ElmerGUI
Dimensions: 2D, Steady-state
And implement the following example with the magnetostatic solver and a axi-symmetric problem https://www.nic.funet.fi/index/elmer/doc/ElmerTutorials_nonGUI.pdf page 34
“Tutorial 8 Induction heating of a graphite crucible”
Directory: InductionHeating
Solvers: StatMagSolve
Tools: ElmerGrid, editor
Dimensions: 2D, Axi-Symmetric
To solve the long wanted natural convection examples (https://forum.freecadweb.org/viewtopic.php?f=18&t=32583&hilit=elmer+heat+board)
the “Convection = Computed” has to be implemented and support for transient simulations has to be implemented (with use of paraview?)
“Tutorial 13 Transient flow and heat equations –Rayleigh-Benard instability”
Directory: RayleighBenardGUI
Solvers: HeatSolve, FlowSolve
Tools: ElmerGUI
Dimensions: 2D, Transient
Finally, yes this is somehow a direct ask for help … I will not implement all the listed solvers I have some need for static-currents but I know and see that support will only come up if others can use the FreeCAD-Elmer framework for their topics/areas.
The above list is more for gathering opinions and support . So, if you are interested in one of the above topics let me/us know it here … than we can decide how in which way it is best to help and learn people to implement there solvers on their own.
BTW: I’m also just a script kiddie so don’t be shy
I think convection computed and transients in general would be tremendous, as these are the most troublesome to obtain in FreeCAD and would benefit tremendously many applications including some of mine (vortex shedding would help me optimize a system).
Elmer has proven to be great for both 2D and 3D with the snap of a finger in FreeCAD and for transients, 2D cases slash the computing time radically. Until now I’ve only been using CfdOF and while the implementation of OpenFOAM is stellar, the whole meshing and sim are computationally very heavy at the moment.
I think paraview is the best at this point. While having a few cool vtk tools directly in freecad is great, the community seem to lament the lack of an active coder with the will and knowledge. Since Paraview is the state of the art anyway, making the link betweem paraview and freecad as automated as it is currently in CfdOF workbench seems like a great option to me.
Currently i would say that documentation is also a lacking aspect for Elmer’s integration. This would probably entice more support. I’m going to support that effort as best I can. The running GSoC project should help with that and should definitely advertise the results in other places of the community ( user showcases, tutorials, open discussion…).
I do indeed advocate electromagnetism because I see that there the open source world is not as crowded. This often means that people looking for decent 3D tool end up using Elmer. On the other hand, it is conseptually quite different from fluid or solid mechanics and there are other fields to push as well. In 3D electromagtics the MagnetoDynamics and VectorHelmholtz solvers are the most important ones. These involve \curl\curl opetaror and suitable element basis to deal with it.
Also the (scalar) Helmholtz equation for acoustics could be of interest. It naturally couples to elastic vibrations and is also not as crowded.
Acoustics and electromagnetics offer problems where the fields need to be solved outside the solid objects. I don’t know how keen you are in meshing the exterior volumes.
In elasticity the shell solver has been developed in recent years. Since very recently we can even solve coupled systems of solids and shells. This is still work in progress but if you take shells, then you may plan that the coupling will also be possible.
In heat transfer radiation from surfaces is well supported. I don’t know what the status is in FreeCAD but that is also something not supported by all FEM codes.
Elesticity and heat radiation do not usually require meshing of the exterior and might therefore fit more smoothly to CAD based workflows?
Meshing exterior domains should not be a terrible issue in freecad, the ability to create boolean fragments helps with defining exterior and interior as well as intersection volumes.
Regarding setting boundary conditions, user realthunder is currently deploying the ability to select hidden faces intuitively, so i for one would be all for encouraging multi-domain meshing in FEM, if only to see more interesting use cases but also because it would again show that FreeCAD is indeed a force to be reckoned with even in the powerful prorpietary ecosystems.
As an example, back at CfdOF workbench (cfd based on OpenFoam) on of the basic examples is the aerodynamics of a drone, which involves exterior domains . Closer to FEM, there’s the example of the two spheres in electrostatism.
My only issue with this in the current FEM status is the fact that Gmesh is not yet functionally integrated for boundary layers meshing.
Hello Maybe it is up to me… But i think that the name solver in (Elmer?) this topic is a bit weird. Shouldn’t that be a analysis? In my opinion it leads to confusion between Elmer, Calculix, Z88 etc compared to Elasticy, electrostatic, flow, etc…
I understand that when using plain elmer different naming confentions could exist. But compared to the way FC with the different solvers work it makes it fuzzy
Yes, the word “solver” has been suboptimal term for Elmer. To avoid confusion I sometimes try to use word “model” instead but “analysis” could of course be one too. Also that might not be ideal since same model/solver may be used to to create transient or eigenmode analysis, for example.
again i don’t have experience with elmer.
Here (http://feacluster.com/CalculiX/ccx_2.13/doc/ccx/node156.html) is a list of different analysistypes that exist in Calculix.
If a compatible brakedown can be made for Elmer than is it maybe an ide to breakdown mechanical calculations further(so take static, frequency, buckling etc apart ).
a short check here(https://www.nic.funet.fi/index/elmer/doc/ElmerModelsManual.pdf) it looks there some overlap but i haven’t dive in the hole manual(only table of content) so i’m not into the details!
@Bernd can you split the naming off from the original topic?
This is may be not a good idea. The terminology in CalculiX is rather bad. There we use analysis type but is is an attribute of the solver. This is for historical reason, because the analysis type used to be a analysis attribute and was moved to the solver years ago.
What is called Solver in Elmer terminology is a Equation in FreeCAD FEM. Means a Elmer Solver is a FreeCAD Equation. This topic is about extending these Equations. Equations are implemented in Elmer solver only. I like the concept. I would like to implement this for CalculiX too, and move all the solver attributes into the correct Equation. But this is not on hight priorty ATM. Because I did not want to change names twice I decided to stick with the name analysis type so far.
What we need to make a consens about is the Equation terminology. Is Equation a good word for this object even if we would start to use it for other solvers (CalculiX, Z88, OpenSees) as well.
I actually like “equation” a lot, it describes exactly what it will solve. The current implementation of Elmer in FreeCAD is crystal clear to me:
I create an Analysis (i.e. breaking down a problem into the sum of its components) to study the problem at hand
in that analysis I create a Solver which will be there to direct and perform calculations
in that solver I populate a list of Equations each describing a very physical problem (Navier-Stokes, Heat Transfer, Elasticity, Frequency…)
I know I am here only enumerating the current state, but written like this, it makes perfect sense to me. Each name in that nomenclature is very close to its real definition.
What I love about this architecture is how complex multi-physics problems are reduced to a simple stack of physical equations each with its priority.
and it shows very nicely the multiphysics capabilities of Elmer.
.