Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
3 EXECUTIVE SUMMARY The collapse of the I-35W Bridge in Minneapolis, MN, on August 1, 2007, was attributed to a design error that underspecified the thickness of steel gusset plates connecting the truss members at a particular joint. Certainly the heightened awareness of an Interstate bridge collapse prompted national attention to the design and load rating of gusset plates. Up until that point, the AASHTO documents were unspecific regarding the design and rating of gusset plates leaving considerable discretion to engineers. After the collapse, bridge owners were highly encouraged to include gusset plates as part of the normal bridge rating, something not done routinely prior to the collapse unless there was a change in condition (i.e., corrosion, impact, cracking, etc.). To unify the load rating of gusset plates, the Federal Highway Administration (FHWA) published a guidance document outlining the minimum number of resistance equations that must be checked to adequately load rate a gusset plate. There were certain criticisms of the document and AASHTO (through the National Academies) along with the FHWA sponsored an experimental program to further enhance the understanding of gusset plate failure mechanisms and create proper resistance equations that could predict the various failure modes of gusset plates. These equations were to be consistent with a load and resistance factor philosophy. The research conducted included both experimental and analytical modeling. Primarily the physical limitations and expense of experimental testing dictated that a small population of specimens would be tested to provide an adequate number of finite element model calibration points. Once a robust modeling philosophy could be established, then a much broader study of different connection geometries could be conducted analytically to encompass the types of gusset plates that are in the nationâs inventory of truss bridges. The experimental program specifically tested 13 full-scale gusset plate connections (though one was accidently destroyed). The members were reusable and each new specimen was only defined to be two new gusset plates and a set of chord splice plates. The configuration of the experimental connections used five separate members; two were collinear chords, one compression diagonal framing in at 45 degrees to the chord, one tension diagonal framing in at 45 degrees to the chord, and a vertical member that could be in either tension or compression framing in perpendicular to the chords. Six geometries of plates were tested that differed on how closely the compression diagonal was to the chords, how long the free edge was, and what type of fastener was used. In addition, four specimens were tested with simulated section loss and one specimen had edge stiffeners. One-half of the specimens failed by buckling of the gusset plate causing the compression diagonal to sway out-of-plane as a rigid body. The other half of the specimens failed by full width shear yielding along the horizontal plane just above the chord. Analytical models of the specimens were used to define the level of detail needed to predict the experimental failure with certainty. It was found that a three-dimensional shell model of the gusset plates and members was necessary to properly predict the failure of the connection. For the purposes of shear yielding and buckling, fastener holes did not need to be modeled. However,
4 nonlinear material and geometric properties of the gusset plate were necessary along with initial geometric imperfections. The robust finite element modeling technique was applied to a parametric study that increased the breadth and depth of studied connection geometries over the experimental study. In particular this included connections with diagonal framing angles other than 45 degrees, chords that were not collinear, corner joints, Warren and Pratt configurations, loading scenarios representing joints over a pier, at midspan, and near inflection points of trusses, gusset plates with edge stiffening, gusset plates with section loss, and multi-layered gusset plates. Due to the fidelity of the models, only buckling, shear yielding, and chord splice failure modes could be identified. The models did not have the fidelity to capture net section type failures. Data from both the experimental specimens and analytical models were used to determine the best resistance equations to predict a particular failure mode. Since many bridge owners had begun to load rate gussets according to the initial FHWA guidance while this research was being conducted, it was decided to largely validate that guidance and change as little as possible unless warranted. For shear yielding, it was found that a plastic shear stress distribution could be used to predict shear yielding. For buckling, it was found that the Whitmore buckling model was appropriate, but recognizing it uses a column analogy to predict plate behavior, required an equivalent length factor of 0.5 for all gusset plates. In addition, the length of the Whitmore column should only use the length from the middle of the Whitmore section to the nearest adjacent fastener line. However, it was found that many compact connections still failed in buckling that was unconservatively predicted using the new Whitmore buckling model. In these compact geometries where the members are spaced very close together, buckling only occurred after a significant amount of yielding that was dominated by shear. In these situations, the buckling was better predicted by determining the load in a member that would cause a partial plane to yield in shear. The variability of the resistance equations was used in Monte Carlo simulations to define resistance factors according to a load and resistance factor philosophy. The ï¦-factors were provided at a variety of dead-to-live load ratios and two different reliability indices. The resistance equations were found to be valid even for connections that had simulated corrosion. In this case, the average plate thicknesses remaining in a failure plane could be used reliably in the resistance equations. No correlation between edge slenderness and buckling resistance could be identified. However, stiffening a free edge was found to be an effective retrofit for connections that have low buckling resistance according to the Whitmore buckling theory. For connections where the buckling resistance is controlled by partial plane shear yielding, edge stiffening could not provide any enhancement in buckling resistance.