Shell and Beam Interaction with Abaqus

[RicardoF]

Hello Everyone,

I am currently learning how to use abaqus, focusing mainly on composites to analyse our monocoque.

Can anyone tell me how to mix shell elements with beam elements (monocoque and hoops)?
I am used to Ansys and, with it, I can mix a beam with a shell simply by meshing a line that is common to both elements.
Does abaqus (6.11) follow the same principle? I noticed that I cannot select a line from an area while trying to assign a beam profile. Should I create stringers that directly use the edges from the areas or must I create points and lines?

I am trying to follow this approach because I am really used to it (thanks to ANSYS) and I don't realy know how to create assemblys and contacts between several parts (assuming that is what I need to do to assemble the hoops with the shell)

Thank you,

Ricardo Ferreira
Technical Director
Projecto FST Novabase

TU Lisbon - Instituto Superior Técnico

[Francis Gagné]

Couldn't you use shells for the hoops as well? It should be fairly easy to integrate them to the shell of the monocoque. I have never used abaqus myself so I am not sure, but it is what I do under Ansys ACP. Might get you somewhere, and should not be much heavier on the calculation side.

[Moop]

I can't speak for how you'd get it setup in ABAQUS CAE(We pre-process in HyperWorks), but you should be able to mix beam and shells no problem in ABAQUS, as both shell and beam nodes have all 6 DoF.

For example, when I was helping the chassis guys get the IA plate setup in their model, I just made a ruled surface through the plane of the IA, trimmed it, added points on the surface edges close to the beam nodes so it was forced to put nodes there, threw a 2D mesh on the surface and equivalenced the nodes.

You can't mix and match nodes of solid elements with shells/beams, since the solid elements only have translational DoF. It's pretty standard to couple shells to solids with tripods of beam MPCs, as the tripod lets you react the moments in all three directions, as long as you don't care about the stresses right close to the connection. Also, I know ABAQUS had a shell to solid coupling element of some sort, but I haven't used it so I don't know how well it works.

[RicardoF]

Originally posted by Francis Gagné:
Couldn't you use shells for the hoops as well? It should be fairly easy to integrate them to the shell of the monocoque. I have never used abaqus myself so I am not sure, but it is what I do under Ansys ACP. Might get you somewhere, and should not be much heavier on the calculation side.

First off All, thanks for your Help guys,

Francis,Yes, I Already tried that and it works. I tried it with a simple model just for learning purposes.
The problem with this approach is that for the actual monocoque, I would probably have to take some steps back regarding modeling because we use structural members from solidworks. I also believe that the sweeps would imply a careful modeling in order for the areas to automatically interact. This obviously means that I try to avoid using couplings in order to make the analysis straightforward (import-> define sections-> mesh-> apply loads-> results). Couplings take a while to define and if we make changes to geometry we have to do things all over again. That's why I am trying to understand how to mix both elements like I do in ANSYS.
I can certainly work it in other way, but I am curious to investigate how can I do it this way because I want to learn abaqus, not only monocoques in Abaqus. And someday I may want to add a beam to my analysis that it is not a tube with small wall thickness.

I will investigate this with a simpler model. Then I will publish my conclusions on this forum.

Thank you for you help,

Ricardo Ferreira
Technical Director
Projecto FST Novabase

TU Lisbon - Instituto Superior Técnico

[Giova2k11]

I have never used Abaqus so I can't be on help on that, but I am wondering, if you know how to model this problem with Ansys and are used to use it why don't you just use that for the analysis? It can save you the time to learn a new program and it will still give you the results you need.

[RicardoF]

Originally posted by Giova2k11:
I have never used Abaqus so I can't be on help on that, but I am wondering, if you know how to model this problem with Ansys and are used to use it why don't you just use that for the analysis? It can save you the time to learn a new program and it will still give you the results you need.

There are several reason:

We use Ansys APDL. It can get the work done but is way worst than abaqus in terms of geometry import and corrections. Our current design it is a bit tricky so some problems are appearing. We want to cross the results from both softwares in order to evaluate if our results make sense.

In industry there are more companies using abaqus. We are moving towards a direction were technical partnerships with companies will be more relevant. If we speak the same language we can easily help each other

Abaqus seems to have a nice optimization solver. We want to check it in order to open some doors for future teams.

Abaqus still releases new features. Ansys (apdl) stopped in time maybe before I was born!

Regards,

Ricardo Ferreira
Technical Director
Projecto FST Novabase

TU Lisbon - Instituto Superior Técnico

[Giova2k11]

I see, your reasons are all prefectly understandable. I am aware that more companies use Abaqus, and that's a great plus.

I'm also aware of its design optimization feature as I was at a symposium last year in which the whole suite was presented; I'm sure you have tried out Ansys' own optimization solver and reckoned that Abaqus was better under that aspect.

All in all I guess Abaqus does have some advantages, and yet our team used Ansys to simulate both the carbon fiber rims and the monocoque, so I know it can give some valuable answers as well, and if I were you I would have spared myself the hassle to learn a new program (I'm lazy, I know!! http://fsae.com/groupee_common/emoticons/icon_biggrin.gif )

Although, once you know how to work with FEA my guess is that moving from one solver to another can't be completely different...