Boundary Conditions for A Structure With Bolted Attachments

[spongebob007]

Let's say I am designing a structure that is bolted to something else. I am interested in finding the stresses in my structure, but I don't care about the stresses in the something else it is attached to. Let's say for example I am designing a cantilever beam that has a flange at the end where it is bolted to a wall that can be assumed to be infinitely rigid. What I have always done in this case is to fix all the degrees of freedom on the circular edge of the bolting holes. Always done it this way, but recently I have had two customers come back with comments on a structural analysis report I submitted to them stating that fixed DOF is incorrect and that the boundary conditions should be pinned instead. Are my customers correct? Does it make a big difference? and is there some reference out there that some one could point me to that discusses this? Thanks.

 

[spongebob007]

In both cases where the boundary conditions were questioned, the structure was modeled with thin shell elements. The attachment points on my structure are through holes, and there are either blind tapped holes in the foundation the structure attaches to or through holes with a bolt and a nut. If it makes a difference at all, I have no knowledge of the specifics of the foundation design or how my structure is attached. All I have is an interface hole pattern. I fix all six degrees of freedom on the hole edges.

 

[bkal]

This might be a long shot, so my apologgies if I am missing a point. I have come across similar issues in the past and it is possible that it might be just a misunderstanding.

The main question is are you (or your clinet) interested and talking about local stresses (i.e. around your bolt holes) or a global behaviour / stresses. Let's assume a beam supported at both ends and subject to lateral loading. The end reactions (on a 'macro' scale) could either be forces only (i.e. global pinned connection) and forces and moments (built in / fixed connections). If you modelled the beam with line elements than you would restrain either translations only (for the pinned) or translations and rotations (for the fixed connection). However, if you model the beam using shell elements, then even if you pin (i.e. restrain tranlations only) all the nodes at the end section you are actually modelling a (global) fixed end connection.

 

[rickfischer51]

The best way to determine if it makes a difference is try it both ways on a simple model.

I don't like the idea of fixing a nodal moment at a bolt hole. I don't think you really get that unless you weld, and I think it just results in ficticously high stresses that you have to explain away as a modeling artifact. I usually pin it for analysis. The ability to sustain a moment comes from the multiplicity of bolts. Per St Venant, it make no difference to the rest of the model once you get away from the BC. The Steel Construction Manual identifies two types of bolted joints: slip-critical and bearing-type. If the bolts are properly torqued, then the joint is slip critical, and as long as the joint is maintained (i.e., no slip), then the proper boundary would probably be to restrain (X,Y,Z only, no rotations) the contact area around the bolts. If the bolts aren't torqued, you have a bearing-type connection, and that is exactly equivalent to a pinned joint. Typically what I do when analyzing a structure is to use a code such as the Steel Construction Manual, to analyze the bolted connection. I pin the bolts, sum the reactions (forces and moments) around the centroid of the bolt group, and take those reaction forces to a spreadsheet and the SCM.



Rick Fischer
Principal Engineer
Argonne National Laboratory

 

[rb1957]

IMHO fixing 6 dof is a "lazy" short-cut and shouldn't be done without a great deal of thought.

pinning a bolt is better, but still a little "lazy" in that the support structure and the fastener itself will always have a finite stiffness (as opposed to the infinite stiffness of a constraint).

best is to support the fastener with strings. i usually use three rods to support the three directions. I'll use an area of 1, 0.1, 0.01 ... something to let the structure have a little "give". sure you can use fastener stiffness calc (Huth seems to be preferred these days) and use a CBUSH type of string element.

Quando Omni Flunkus Moritati

 

[corus]

If you have a number of bolt holes then it wouldn't make any difference away from the joint. The multi-pinned restraints would provide rotational restraints if they're assumed to be infinitely stiff. What the customer may be suggesting, however, is that the whole of the joint is modelled as a single pinned restraint (assuming it's not a cantilever). In fact the true behaviour of the joint would be somewhere between pinned and fully restrained as the bolts would give some resistance to bending moments there. If you didn't actually model the bolt stiffness then the restraints you would choose would depend on which part of the beam you wanted to give more conservatism to. Probably the best option is to model the bolted joints as both pinned and fully restrained to see which gave the worst case.

 

[rb1957]

sure, you can replace a fastener group with a single contraint ... but too many fixed contraints over-constrain the model.

Quando Omni Flunkus Moritati

 

[rickfischer51]

I wasn't replacing the fastener group with a single constraint. I was exporting the reactions from the fastener group to a closed form solution based on an accepted code to analyze the fasteners.

Rick Fischer
Principal Engineer
Argonne National Laboratory

 

[rb1957]

yes, i was continuing corus' post ... you Can replace a tight grouping of bolts with a single constraint ... depending; separate pinned is "better" in my mind

Quando Omni Flunkus Moritati

 

[corus]

In situations where the customer has some doubts then it has been the practice to run 'sensitivity' studies just to test their hypothesis. This can be just of one case, for example, to illustrate that hopefully there's no or very little difference in the results if you change the boundary condition to fixed or pinned, as in this case. This sensitivity study generally doesn't need any detailed assessment as you're just doing a simple comparison, so can be done quickly. It keeps everyone happy and next time you won't have to do it.