Modelling closely spaced supports in FEM package 1

[jaskamakkara]

Hi everyone,

I am checking a steel crane beam in an old theater. It is about 21m long and is supported in 3 intermediate points for 4 continuous spans of approx 5m. It is connected at these points to perpendicularly spanning concrete beams by way of channel sections, welded vertically to the top of the steel beam either side of the concrete beam, bolted to the concrete. The concrete beam is about 380mm wide so there is that much distance between the channel sections at these points.

Here you can see the channel on one side of a concrete beam.

To get the proper buckling capacity of this crane beam I decided to do a proper 2d element stability analysis since I wasn't sure how to get the buckling length from first principles given the nature of these connections and I wanted to capture the lack of proper torsional restraint at these locations. But this then made me realise that modelling these channels individually (rather than condensing each pair of channels to a single support) meant that I was getting some lever action and the support reactions are much higher than they would be if I condensed the supports to single points. The problem is now that the support capacity is driving the check and I'd prefer to get some more capacity out of these beams if possible. I suspect that if I were able to model the bolt connections as very stiff springs it might alleviate the issue a little bit, but I'm not sure how to put a number on it.

Snippets from my model

Difference in support reactions between modelling supports separately, and condensed. The 2d model agrees well with this 1d model.

Question is do you think it's reasonable to model the connections like this? I'm finding it hard to argue with the results since the FEM gives the "real" situation and these channels are spaced sufficiently far apart that I think they will behave more like separate supports. But, then again, when you have a seated connection with 2 pairs of bolts 200mm apart you would always treat that like a single support since otherwise the lever action would be huge and artifically increase your bolt forces. What do you think?

Cheers

 

[jayrod12]

I would honestly be designing it as a single point of support. There's likely enough slight give in the bolted connection to alleviate that prying/level action you're experiencing. Bolt plowing in the steel and just fit-up tolerances would provide a bit of flexbility.

 

[rb1957]

Don't really understand the pix, but ... modelling the two clips as one seems reasonable. But ... you may want to take some moment out (which'll drive it back to the duplicated model).

another day in paradise, or is paradise one day closer ?

 

[JoshPlumSE]

The problem I have with modeling two supports so close to each other is that they are usually modeled as "infinitely rigid". This is fine when a support is off on it's own because that support is so much more rigid than the beam it connects to. But, when it's so close to another support, the interaction of those two supports tends to over restrain the beam and produce unreasonable results.

If you're going to model it as two supports, I'd try to replace the closely spaced rigid supports with two spring supports where you have calculated a more realistic value for the spring constant.

 

[rb1957]

@Josh ... absolutely ... all too often I see redundant fixed supports over-constraining models. Using finite stiffness constraints is Much better. Of course then the discussion is "what stiffness?" and the answer is either ...
1) know or calculate something of your local structure, or
2) "plug and pray" with a range of stiffness (to see how small is the difference).

another day in paradise, or is paradise one day closer ?

 

[jaskamakkara]

Thanks for the replies, everyone. And yes, I think you have confirmed my suspicions that I should just condense these double supports into singles, as jayrod said there is likely enough give in the joints that the loads are shared. JoshPlum, I agree that if I could put in some springs instead of infinitely stiff supports that would also fix it, but I think that it's probably easier to put in single supports.

Thanks again.

 

[bones206]

Perhaps you could explicitly model the bolts. Just very short cantilever bar elements with length equal to the channel web thickness plus maybe another 0.5 x bolt diameter to account for localized concrete deformation. That should give you a spring value somewhat close to reality.

I’m guessing even if you do model the bolts, that the connection will still be quite stiff and constrained though... may not change your results by much.

 

[jaskamakkara]

@bones206 yeah I had that idea too, think I'll make that model just out of interest and see what happens.

 

[Ng2020]

There are numerous methods for computing bolted joint stiffness to be used in loads models. In my field the Huth formulation is commonly used. There's a good discussion about it here: thread181-27567