Modeling beams and slabs to work together

[fare]

What do you think what is the best way to model RC slab on a beam, that would act together and be closest to reality?
Both elements draw as much of the moment as their stifness is...

 

[Maraveas]

I know that sometimes the plate elements and the beam
are not ccoperate succefully, because they have very different factions/equations. You can try use higher order shell elements for the plate.
But unfortunatelly you must try compare your results
with a closed results example and then use the same
way to do the calculations of a more complex case

 

[Jonaidi]

Closest to reality! If the reality is to have a stiff untracked connection between slab and column, you can model it with using 'rigid link' elements connecting the node of column to a few nods in the slab around the column's node. (you need, say, 4 nodes in the slab which represent the size of the column at its 4 corners)

 

[cdemir]

Dear Sir,
Beam element and shell element have to be same DOF. you mentioned moment transfer so as far as ý know six DOF is required for moment

 

[fare]

FEA programs treat slab and beam as separate elements connected in common joints. In reality they work together as a T-beam, where the width of the slab participating has to be selected.
There are several methods of modeling this connection. Does anyone has good experinece with some of them?

 

[corus]

You can integrate a beam element with a shell element or a 3D solid element so that they share common nodes. The beam centroid has to be offset however to get reasonablish results.
For your beam and slab I would use 3D solid elements if you wish to get th ebest results. One element through the thickness of the beam should be sufficient. You can either tie the nodes or have contact between the two surfaces. Similarly you could use shell elements for the beam and tie the translational freedoms to the slab.

 

[fare]

Corus,
do you mean defining beam and shell at the same coordinates and then to add the offset to the beam?
I prefer using shells than solids since I need moments more than stresses (for reinforced concrete).
The 'tying translational freedoms to slab' - you mean tying translational freedoms of some beam to some slab nodes (closest ones) if beam is not at the same coordinates as slab?

 

[corus]

You can have the beam along a line of the shell, ie. sharing a co-ordinate. Offset can be input as one of the beam properties so that although the beam may lie along a line, the offset gives results as if the centre was shifted away. I think this was implemented into FE code for such a purpose as reinforcing shells. I'd only use this method if you were wanting the overall behaviour of the structure however, rather than any localised effects.
Tying the beam and slab nodes in 3D can be done even if the nodes are at the same co-ordinates. I had in mind that the beam and slab were separate bodies. A simpler way of defining 3D geometry would be to have the beam and slab as a single body. You wouldn't need to tie anything together then.

 

[PXC]

I think that there are a number of ways to model a concrete slab with integral beam, and that the choice will depend on what you want for the output and what your software can do. I invariably want to get bending moments for design. If the "beam" can be described as a thickened slab, then I would use plate elements for both the slab and thickened slab. If the beam is relatively deep compared with its width, then I would most probably reluctantly use an offset beam element. I say reluctantly as the required bending moment and shear force distributions can not be obtained directly from either the slab or beam elements. The bending moment diagram from the beam elements will in fact be a sawtooth shape. An explanation of this and the theory for calculating the required bending moment diagram is given at the following site under "Web notes"/"Stiffened plates".

http://www.strand7.com/

I suspect that some software may automate this calculation for you.

The above reference does include some other methods of modelling stiffened plates.

One method that I have used for some unusual footings is to model them with brick elements. Strand will then calculate the bending moment on a section or slice through brick elements by integrating the normal stresses over the section. I do not consider this suitable for normal design as it is too slow.

 

[fare]

Thanks PXS,
it is good that some programs can integrate bending moment from stresses. Can you determine the area where to integrate (part of a slab)?
I usually use beam elements in the plane of the plate (with the moment of inertia of T beam). It seems (even according to this text you mentioned) it gives the smallest error.
I would use thickened plate to represent the beam but it will again be in plane of the plate, and would require intergration of bending moment (even if the thickened plate is beneath the slab). How do you use this method? Maybe you define part of slab that participates in T beam as separate plates, and then integrate? Do you include some offset for this thickened plate?

 

[PXC]

With Strand, the integration is over the area that is displayed. So for a stiffened slab, it would be necessary to display the width that the code allows.

I often model steel plates with stiffeners for materials handling chutes and bins etc. I always put the centroid of the platework on the centroid of the beam elements. I do not bother to take the increased bending strength from the plate into account at all. This is an appropriate conservative approximation for this application.

I will explain my use of plate elements for both the slab and thickened slab in a little more detail, as think you may have misunderstood me. I would only use one layer of plates for both the slab and one layer for the thickened slab. This means that I do not need to integrate to calculate moments. I simply offset the plate elements so that the top-of-concrete is the same for all elements. The offsets are a built in variable in both beam and plate elements, so that the nodes can all be in any single plane, whether the centroidal plane of the slab or the top-of-concrete. I suggest a little caution for the results at the change from 1 depth to another, but I would have confidance in the results away from the step change. I have not attemted to benchmark this method, but it seems to work well, the results look good.