combination of solvers for Shell elements and and solvers for 3D

[Wil]

Hi,

I know that Elmer supports solvers for shell-elements and 'usual' solvers for 3D-domains.

Is it possible to combine both sort of solvers in one model? Consider e.g. a model of a cube where the top 'face' or its mesh elements is used for a simulation with shell elements introducing eg. thermal influx. A second example would be a cylinder where the outer face (lateral surface) would define the shell-elements of a thin-walled pipe for heat transfer while the 3d-elements themself would be used for simulating a fluid flow.

BR
HoWil

[raback]

Hi

Yes, at least to some point, see:
https://rakenteidenmekaniikka.journal.f ... iew/120470

Code: Select all

elmeruser@elmeruser-VirtualBox:~/elmerfem/fem/tests$ ls | grep Shell_with_Solid
Shell_with_Solid_BenchmarkCase1
Shell_with_Solid_BenchmarkCase1b
Shell_with_Solid_BenchmarkCase2
Shell_with_Solid_Eigenanalysis
Shell_with_Solid_EigenanalysisPar

Edit: read too quickly. This is probably not what you're looking for. Different equations may occupy different bodies as you wish. The key is to introduce "Body Id = Integer" to some BC and then you can deal with that as a body. This allows you to solve equations of different dimensions. This being said you may need to tinker some parameters to take the depth into consideration in a proper way. Devil always lies in the detail.

-Peter

[Wil]

Thanks Rabak,
I try to summarize:
There is the possibility to combine shell-elements (2D-Faces) and 3D-elements in one simulation.
There are until now (only) solvers for mechanics implemented.

Is the coupling between shell and bulk not only on the boundary edges of the shell possible but also 'through' the shell?
Is the implementation of solvers for thermal (perpendicular to the shell-surface) or fluid-flow (inside the shell-face) problems in shell elements in general in Elmer possible.
I know that problems can always arise, but how much effort do you estimate to implement a new solver for shell elements for thermal problems for an inexperienced scientist with a little background in FEM and programming (no Fortran)? If the analytical equations are sorted out does the implementation take weeks, month or years?

Thanks in advance
Wil

[kevinarden]

This just sound like boundary elements to me. Not shell elements coupled to solid elements. You should be able to turn on the heat solver with flow solver and input heat Flux into the boundary elements.

[raback]

Hi,

Shell & solid is the toughest case since the formulation for shells is notoriously different than for solids. For heat and flow equations there is no such drastic difference. For example, heat equation in the solid and shell is the same when you multiply certain material parameters by thickness. For fluid flow there exists a reduced dimension ReynoldsSolver (for lubrication type of problems) and preliminary version of reduced dimensional full Navier-Stokes.

So how much work there is really depends. If your equation is a variation of advection-reaction-diffusion equation (parallel and arbitrary order p-elements) here is a good starting point:
https://github.com/ElmerCSC/elmerfem/bl ... handle.F90

-Peter

[kevinarden]

From your original description you want to solve this problem. Shell elements are not needed, unless you want to calculate responses in the shell wall. If you just want inputs to the fluid than boundary elements are enough, and the shell elements/solver is not needed.

problem.PNG

(13.98 KiB) Not downloaded yet

[raback]

Kevin is right. You can actually model a finite thickness pipe with known external temperature using k/d as the heat transfer coefficient. -Peter

[Wil]

From your original description you want to solve this problem. Shell elements are not needed, unless you want to calculate responses in the shell wall. If you just want inputs to the fluid than boundary elements are enough, and the shell elements/solver is not needed.

problem.PNG

Thanks Kevin for your input. Nevertheless, I need to model and simulate the thin domain surrounded with bulk material. The model I gave with the pipe was just an example in which I would need the temp. distribution, flow-parameters, magnetic-properties inside the 'pipe'/shell domain.

[Wil]

Shell & solid is the toughest case since the formulation for shells is notoriously different than for solids.

With "Shell & solid is the toughest" you mean 'mechanical' problems?

For heat and flow equations there is no such drastic difference.

And how about magnetics compared to mechanics, heat and flow? All for static simulations?

[raback]

Hi,

If you read Elmer Models Manual you can see why Shell equations are notorious, see Ch. 8 in:
https://www.nic.funet.fi/pub/sci/physic ... Manual.pdf

One could perhaps fuse the 3D solid equation to shell equations on matrix level. However, this is not totally trivial task as shell equations have different dofs: 3 displacements + 3 deflections.

For magnetics 2D and 3D are also very different mainly due to the fact that the \curl\curl operator reduces nicely in 2D wheres in 3D it doesn't. The 2D in magnetics means some symmetry since magnetic fields are not contained within shells. There are also thin sheet BCs for 2D regions of 3D magnetics.

-Peter