Thank you for the intelligent discussion. I don't want to be contentious, but I didn't see any indication of this "roller nest" business in FATIGUE EVALUATION AND REDUNDANCY ANALYSIS, a 100 page pdf linked as ref 23 from the Wikipedia page, I-35W Mississippi River bridge It has grainy b/w pictures of roller bearings at Piers 5,6, and 8 which show one set of rollers on a transverse axis. The paper has a discussion of the expected, or "free", movement versus observed movement in the longitudinal direction, i.e. along the length of the bridge "for those of you in Rio Linda".
I notice that the pier 6 photo is from the west side, as I can make out the infamous drain pipe on the other side under the bridge.
The roller analysis involves movements mostly under 1", but some slightly more, but the point would be the stresses sustained due to the frozen bearings, rather than the amount of movement.
The authors describe a finite element stress analysis, and the frozen bearings were accounted for as boundary conditions, so I don't think the shear stress at the bearing was calculated, although it could have been, on physical principle. Anyway, it didn't seem to be a concern of the authors.
I don't see why a shear failure at the east pier 6 bearing isn't an attractive option. The way the road surface toppled to that side, with the west side truss collapsed to the east, and the apparent violent disruption of the east bearing assembly, it seems like it's obvious that's what happened, from a layman's perspective.
Better pre-collapse imagery, obtained since the post you refer to, show the roller nests to be one layer instead of two, confining permitted movement to the longitudinal axis.
The term “roller nest” comes from the June 2006 MnDOT Inspection report, found here:
http://www.dot.state.mn.us/i35wbridge/pdfs/06fracture-critical-bridge-inspection_june-2006.pdf
...to wit:
“....Truss Bearing Assemblies: The truss spans have six “geared roller-nest” bearing
assemblies, and two fixed bearing assemblies. The truss bearings have section loss, flaking &
surface rust; moderate corrosion, the bearings at piers #5 & 8 are functioning properly. They
are checked during each annual inspection. The bearings at pier #6 show no obvious signs
of movement, difficult to reach with snooper.”
...from page 12. Note that this section refers only to the east truss, with west truss inspection results reported later in the text.
I can’t rule out shear failure at east pier 6.
I can’t rule out shear, tension, compression, crack, bolt failure, or gusset failure a few panels east of there, as the main span obviously separated from the pier 6 truss panels before dropping straight down into the river, while the pier 6 elements leaned over and collapsed to the east. Those sections were separated prior to falling.
I can’t rule out failure at or about east pier 5, where the crossbeam was severely damaged in 1986, necessitating closing the bridge, jacking up the superstructure, and replacing rocker bearings and shoring up a badly damaged crossbeam assembly.
I can’t rule out a failure at west pier 5 where the 2006 report indicates a frozen rocker bearing similar to that which caused the 1896 damages and failures. Failures at either of these last two locations could have negated counterbalance cantilever weights, caused significant deflection midspan on span 6, negation of cantilever counterbalance south of pier 6, and ruptured members north of pier 6 where the truss assemblies transition from cantilevers to simple trusses, OR, they could have initiated the collapse sequence at or around pier 5, with failures at subsequence points following them in the failure sequence.
I can’t rule out a tension failure in any other members in this area, stretching from mid span 5 to mid span 7, and unrelated to frozen bearing assemblies.
I can’t rule out resonant vibration from cement trucks exacerbating any or all of the above problems accumulating in the bridge before collapse.
I feel reasonably comfortable ruling out a triggering failure north of the span 7 midpoint, or south of midspan on span 5.
I feel highly confident ruling out scour, at any pier, as a major or contributory factor in the collapse, because none of the piers in the arae which failed first was subject to scour effect, and all are still standing.
I feel highly confident in ruling out overloading from construction materials or traffic, as a major or contributory factor in the collapse, because the bridge was designed to carry bumper to bumper traffic loads, and these loads are a small fraction of the dead loads (weight of the bridge itself) the structure carried successfully for decades.
Just between you, me, and the lamppost, I’m pretty sure that the triggering failure occurred in a tension member (top chord, diagonal brace, or gusset connecting same) two struts north of pier 6, or in the sway bracing between the east and west pier 6 kingposts, as a direct result of greater than design loads, induced by differential expansion due to frozen roller bearings at pier 6, frozen rocker bearings at the southwest crossbeam, and the frozen hinge pin at span 2.
The evidence necessary to prove this is currently underwater at the south end of the center span. Unless any of the Navy divers publish pictures of the debris in this area, as it lies after the collapse, and before it is moved, there’s not much I can do to substantiate my suppositions, other than what I’ve posted so far.
Photos of the east and west endbeam/crossbeam/rocker bearing assemblies at pier 5 might be helpful, but in the imagery available so far, these members are at the bottom of the debris pile, with the south end of pier six on top of them, and at best, that data could only disprove crossbeam/rocker bearing failure as a trigger event, without confirming the span 7 main truss or pier 6 sway bracing failures I strongly suspect.
There a limit to how well one can analyze construction failures at a distance, and if we’re not at that limit now, we approaching it quickly. Intellectually unsatisfying, but that’s the way it is. NTSB will issue a report in a year or so, and whether it gets the full failure sequence right or not, it will contain enough data to ascertain the cause. For now, that’s about the best we can do.