3doorsdwn
Structural
- May 9, 2007
- 162
This thread isn’t [at least not initially] posing a question……just sharing some results of a study I did…….any comments welcome.
Lately I have had a number of beams with torsional loads with simple shear connections [i.e. a pair of bolted angles] at the ends. Most of the time I avoid this type of connection (for torsional loads), but the project that I was on (and the lightness of the loads) sort of dictated I went with this. Because of uncertainties with this in the past, I decided to do a complete study of this using classic and FEA results. Basically what I did was take a wide flange and model it in STAAD two different ways: as a stick/beam element and a model of it as plates. The beam was subjected to a torsional load only. I also calculated theoretical results using AISC’s Design Guide No.9.
My purpose was to evaluate the impact (if any) of having a [single] beam with this type of connection at the end. I wanted to know how the displacements compare to theoretical results, the stresses induced, and the mode of force transfer at the ends.
Overall the results tended to match up. As far as stresses go the overall magnitudes [i.e. the highest predicted] approximately matched (with a notable exception which I will discuss in a moment). The (rotational) displacements were close with the plate element “beam” and the theoretical results predicted by the charts in design guide No.9 [for torsionally “pinned-pinned’ beams]. One thing that did sort of surprise me in this regard: the rotation predicted by the stick element model. In order to model the end conditions there, I released 90% of its rotational capacity [through the partial moment release command in STAAD]. The rotational stiffness obtained by other methods only got close to the stick element model when I restrained all its rotational capability. This presents sort of a problem because if you do that you are [in effect] saying warping is restrained at the ends…..and that is not the case.
The highest stresses [in fact about 2 times more than theoretical results predicted for this area] were obtained at the locations where the angle connections were modeled. (Represented by thicker plate elements at that location and pinned [in all directions] supports.) As I figured it would: the torsion hit the beam support as a couple created by forces acting on the angles as if they were seated beams (i.e. shear forces that acted perpendicular to the direction beam [vertical] shear forces normally act on angles). As a result, unanticipated bending stresses are showing up in the web. (It’s almost like the web is being bent over the angles to transfer the force to them [and when you think about it: that’s probably how it would have to happen].) So apparently this is the weak link. (And considering how thin most webs are: it could be a fatal weakness.)
Finally, [after getting these results] I modeled a framing system where I had the same beam [with the same load] but it was part of a 5 beam system that consisted of an identical beam 4’ away and 3 [smaller] beams spanning between the 2 larger beams. All the connections had 95% release of moments/torsions [again using STAAD’s partial moment release command and assuming all clip angle connections]. This model produced much better results, as no high torques were attempted to be delivered to the angles [as torques].
Given these results it seems the logical thing to do is always have some framing members present [as discussed in the last paragraph] so that these forces are minimized. (Or have the proper connection for this type of force transfer [despite who may protest].)
Lately I have had a number of beams with torsional loads with simple shear connections [i.e. a pair of bolted angles] at the ends. Most of the time I avoid this type of connection (for torsional loads), but the project that I was on (and the lightness of the loads) sort of dictated I went with this. Because of uncertainties with this in the past, I decided to do a complete study of this using classic and FEA results. Basically what I did was take a wide flange and model it in STAAD two different ways: as a stick/beam element and a model of it as plates. The beam was subjected to a torsional load only. I also calculated theoretical results using AISC’s Design Guide No.9.
My purpose was to evaluate the impact (if any) of having a [single] beam with this type of connection at the end. I wanted to know how the displacements compare to theoretical results, the stresses induced, and the mode of force transfer at the ends.
Overall the results tended to match up. As far as stresses go the overall magnitudes [i.e. the highest predicted] approximately matched (with a notable exception which I will discuss in a moment). The (rotational) displacements were close with the plate element “beam” and the theoretical results predicted by the charts in design guide No.9 [for torsionally “pinned-pinned’ beams]. One thing that did sort of surprise me in this regard: the rotation predicted by the stick element model. In order to model the end conditions there, I released 90% of its rotational capacity [through the partial moment release command in STAAD]. The rotational stiffness obtained by other methods only got close to the stick element model when I restrained all its rotational capability. This presents sort of a problem because if you do that you are [in effect] saying warping is restrained at the ends…..and that is not the case.
The highest stresses [in fact about 2 times more than theoretical results predicted for this area] were obtained at the locations where the angle connections were modeled. (Represented by thicker plate elements at that location and pinned [in all directions] supports.) As I figured it would: the torsion hit the beam support as a couple created by forces acting on the angles as if they were seated beams (i.e. shear forces that acted perpendicular to the direction beam [vertical] shear forces normally act on angles). As a result, unanticipated bending stresses are showing up in the web. (It’s almost like the web is being bent over the angles to transfer the force to them [and when you think about it: that’s probably how it would have to happen].) So apparently this is the weak link. (And considering how thin most webs are: it could be a fatal weakness.)
Finally, [after getting these results] I modeled a framing system where I had the same beam [with the same load] but it was part of a 5 beam system that consisted of an identical beam 4’ away and 3 [smaller] beams spanning between the 2 larger beams. All the connections had 95% release of moments/torsions [again using STAAD’s partial moment release command and assuming all clip angle connections]. This model produced much better results, as no high torques were attempted to be delivered to the angles [as torques].
Given these results it seems the logical thing to do is always have some framing members present [as discussed in the last paragraph] so that these forces are minimized. (Or have the proper connection for this type of force transfer [despite who may protest].)
