Shells-Solids Elements Connection in FE Model 1

[struclearner]

Dear All,

When a FE model has solids and shell elements, one way to make their DOF's compatible is to surafce coat the layer of solid elements next to connection of shells with the shell elements. What thickness is given to these embedded shell elements.

Thanks for your input.

 

[crisb]

Its not an ideal method, because as you have realised the localised thickness will be difficult to model well. Some codes have shell to solid transition elements, which may be a better way to go.

 

[EngAddict]

AFAIK it shouldn't matter, I normally make it the same thickness as the plate. More importantly IMO don't make the strip (tee piece) wider than the shell thickness or your strip thickness may then become an issue.

Some codes also allow you to intersect/overlap the plate with the solid but must be embedded enough to transfer sufficient rotation.

Some other simpler codes like solidworks automatically create transitional equations for the user so you don't need to model anything extra, just end the shell at the edge of the solid. Not ideal but has some productivity merits.

 

[rb1957]

so long as you recongnise that you'e modelling a pinned shell ... solid elements don't have moment freedoms at the nodes.

maybe RBE the 2D shell to pick up two sets on nodes on the 3D solid to give it some fixity ?

personally, i don't like overlapping elements trying to model the same structure ... either you're double counting or 1/2 counting, but neither sounds like a good thing, IMHO.

 

[EngAddict]

It isn't pinned that is the point of modelling the tee shape.. it transfers the rotation to the solid through the torque arm yeah? The shell tee junction has full fixed DOF.

I don't like that overlapping elements either that's why i use the other method. Generally speaking these connections will not be in a region of interest, it will only be used for continuity of the whole model, allowing the user to simplify the size of the simulation.

 

[rb1957]

"Tee shape", "torque arm" ??

i don't know when these terms entered the discussion.

i was commenting on the OP's OP ... joining solids to shells you have to be aware that solids only have translational freedoms, so connecting to a face of the solid with a shell won't connect the rotational freedoms of the shell (which cold be a problem in itself) and won't transfer the moment that the solid is reacting over it's depth. you need to connect to both faces of the solid.

RBEs, or beam elements, or constraint equations (maybe) ...

 

[EngAddict]

I said tee shape because it is what the OP is referring to with this method "surafce coat the layer of solid elements next to connection of shells". I am pretty sure he is aware of the DOF issue that's why he was asking about this particular method.

 

[rb1957]

i thought "surface coat" meant "to coat the surface" ... for example, by overlapping the solids with a bay of plates you'd give the plates time to develop moment at the edge of the solids (but it sure looks like you're double counting).

 

[AIUK]

To recover the rotational DoF at solid element's nodes coincident with that ones from a plate/shell element I normally make use of RBE3 with a 456 on the dependent (coincident) node and 123 on three independent nodes chosen from the solid's nodes in the face "normal" to the solid-plate joint.

 

[jean15paul]

At a previous job when we'd occastionally have solids and shells in the same FEM (Nastran), we would "tile" the surface of the solid mesh with shells, which is what I think you're referring to.

We'd give the tiled shell elements an extremely small thickness (e.g. 0.001 inch), but we'd also give them an extremely high "bending moment of inertia ratio" (e.g. 1e10). The rationale being that the high bending moment of inertia ratio would transfer the bending between the shell and solid elements without having much affect on the stress value due to the small thickness. It's still not an ideal approach, but it worked for us.

Anyone have any thoughts on this method?

I copied the discription of bending moment of inertia ratio from the Nastran Quick Reference Guide for everyone's benefit:
Bending moment of inertia ratio, 12I/T^3. Ratio of the actual bending moment inertia of the shell, I, to the bending moment of inertia of a homogeneous shell, T^3/12. The default value is for a homogeneous shell. (Real > 0.0; Default = 1.0)

 

[EngAddict]

Here is the page from the Algor user guide:

http://download.autodesk.com/us/alg...A5A-3413C8AA545-240.htm,topicNumber=d28e31141

It demonstrates both methods but does not give any recommended thickness or stiffness. I would suggest running a few test cases with different variables to ensure you are getting the expected results.

 

[BlasMolero]

Hello!,
The best approach I run is the use the GLUE "edge-to-surface" contact feature, this is a great resource we have when using the solver NX NASTRAN, here the user mesh the solid part with solid elements like CHEXA and the shell parts with plate elements like CQUAD4, without the need to create any intersection on geometry, it will be the solver who will account for connecting both meshed without increasing the DOF of the model or computation time, great!.

And also this feature is available for every solution SOL101, 103, 105, etc..

Best regards,
Blas.

~~~~~~~~~~~~~~~~~~~~~~
Blas Molero Hidalgo
Ingeniero Industrial
Director

IBERISA
48011 BILBAO (SPAIN)
WEB:

http://www.iberisa.com

Blog de FEMAP & NX Nastran:

http://iberisa.wordpress.com/