Structure Analysis with both Beam and Shell elments

[JamesMarine]

I noted there are two seperate shell and beam solvers in Elmer. i assume it should be possible to solve a structure with both shell and beam elements included, is that correct?

I can just add two solvers in the .sif file, is that correct? Is there any example projects that with both shell and beam elements?

[mika]

An ideal implementation of the coupled shell and beam analysis would be to assemble both models to a single system which would then be solved simultaneously. Nevertheless this type of approach has not yet been implemented. Since the models have compatible DOFs, it might be possible to implement this relatively easily. When time permits, I could look at this.

A hierarchic model could already be possible (solve a shell model first, then a beam model and iterate), but such a model is expected to have a limited applicability.

-- Mika

[JamesMarine]

Thanks Mika! It is very nice of you to help to look into this. When do you think you maybe able to find some time for this?

An ideal implementation of the coupled shell and beam analysis would be to assemble both models to a single system which would then be solved simultaneously. Nevertheless this type of approach has not yet been implemented. Since the models have compatible DOFs, it might be possible to implement this relatively easily. When time permits, I could look at this.

A hierarchic model could already be possible (solve a shell model first, then a beam model and iterate), but such a model is expected to have a limited applicability.

-- Mika

[mika]

I will try to make a first look at the necessary amount of coding within a couple of weeks, say. If the implementation turns out to be more complicated than expected, this may change to be a longer-term goal.

-- Mika

[JamesMarine]

Thanks Mika, it would be appreciated if you could kinldy me informed with the status.

[mika]

I used some hours for this and wrote changes to the shell solver so that the shell solver can also assemble and solve a pure beam model, obsoleting the separate solver for the beam equations basically. However, in my earlier comments I ignored an issue which makes the DOFs of the shell model and the beam model only partially compatible. The generalized nodal forces of the beam model are three forces and three moments, while one of the generalized nodal forces in the shell model is a reaction related to the elongation of the shell normal, which is not taken into account in the current beam model. This makes the started line of development not so straightforward as I thought, since at the moment a beam stiffness matrix cannot be assembled into the shell stiffness matrix simply without enforcing additional constraints somehow. It seems I need to reconsider the problem to find the easiest way to proceed.

-- Mika

[JamesMarine]

Thanks, Mika. Is there any latest update on this? I am working on marine FEM models, as you may understand, the structures that invovled both shell and beam is quite common. It would be appreciated if Elmer can be fitted with this kind of function.

I used some hours for this and wrote changes to the shell solver so that the shell solver can also assemble and solve a pure beam model, obsoleting the separate solver for the beam equations basically. However, in my earlier comments I ignored an issue which makes the DOFs of the shell model and the beam model only partially compatible. The generalized nodal forces of the beam model are three forces and three moments, while one of the generalized nodal forces in the shell model is a reaction related to the elongation of the shell normal, which is not taken into account in the current beam model. This makes the started line of development not so straightforward as I thought, since at the moment a beam stiffness matrix cannot be assembled into the shell stiffness matrix simply without enforcing additional constraints somehow. It seems I need to reconsider the problem to find the easiest way to proceed.

-- Mika

[mika]

Hi,

The devel version now contains the first version of an enhancement which allows a shell model to have one-dimensional beam sections. At the moment the combined model can be assembled in the case of linear analysis only (Large Deflection = False in the solver section of the shell model).

Parameters of the beam model are currently read from material sections, so elementwise organized data cannot yet be given. In particular one of the principal directions of the beam cross section (the y3-axis) can be given by specifying the value of the keyword "Director".

The shell model doesn't inherently know about a moment around the shell director (normal) of the mid-surface, while the beam formulation (see the chapter "One-dimensional Equations for Elastic Beams" in Elmer Models Manual) which has been used as a starting point employs the full resultant couple (moment) vector with three components. To cope with this discrepancy, the beam stiffness matrix is manipulated internally in such a way that the resultant couple vector cannot have a component with respect to the director vector of the beam. The beam and shell directors are now defined independently, but it is natural to construct the combined model so that the shell and beam directors agree in places where the domains of the two models intersect.

Testing has been minimal so far and solving some verification cases would be needed. At the moment there is just one example

https://github.com/ElmerCSC/elmerfem/tr ... m_Sections

to instruct writing a sif fie.

-- Mika

[kevinarden]

https://github.com/mrkearden/beam_solver

Files in link above. 12 Meter long, 3 Meter wide plate with a stiffener on the center line along.

problem.png

(106.93 KiB) Not downloaded yet

Elmer beamsolver was 6.97% of the theory solution

The shell solver was 5000 times the theory solution.

There may be errors in all 3 solutions, I am still checking.

[kevinarden]

Resolved the beamsolver case. There was an error in calculation of section properties.
beamsolver provides reasonable answer