Here are some early gleanings from the hundreds of pages of inspection documentation MnDOT have released.
A draft report prepared by the engineering firm URS Corporation in June 2006 emphasizes that Bridge No. 9340 was designed in accordance with the 1961 American Association of State Highway and Transportation Officials (AASHTO) Standard Specifications for Highway Bridges , "which was based on a completely different fatigue design method that was revamped in the 1974 interim edition. The poor fatigue details on the truss spans, particularly those inside the main truss tension chords are difficult to inspect, have raised concerns on the consequence of a possible main truss member failure triggered by a fatigue crack." This concern prompted the Minnesota Department of Transportation to commission the study, which included detailed 3-D finite element analysis models of each of the structural members.
Using the 3-D finite analysis model and another model specified by AASHTO called a fatigue truck (which represents a typical truck in terms of dimension and load to simulate weight-in-motion over a surface) for live load stress analysis, URS engineers determined that the probability for fatigue crack development in crucial parts of the superstructure was very remote.
[T]he fatigue concern should not be completely discounted for the following reasons: (1) the access to the fatigue susceptible details inside the truss sections is very limited for crack inspection at the weld toes and therefore a timely discovery is unlikely to happen should a crack occur for some unusual causes; (2) the length of the welded tabs at the box section diaphrams was specified at 3.5" in length in the original contract plans, which is very close to the lower limit of 4" for the Category E detail. Should a fabrication error or the workmanship modify the detail to the extent that it has the fatigue resistance of a Category E detail, the infinite fatigue life requirement would not be satisfied per the AASHTO Fatigue Guide Specifications; (3) the traffic on the bridge is heavy compared with the average highway bridge and therefore the use of a single fatigue truck may be underestimating the repetitive load effects on the structure.
Just these two paragraphs (found on page 6 of a 299 page report--yep, no sleep for yours truly tonight) raise several crucial questions: Were the areas most likely to fail the hardest to inspect? Were certain structural details--the welds--vulnerable to potential fabrication errors or errors of workmanship that might ultimately contribute to a failure? And finally, was the data being fed into the model used to analyze truck loads on the bridge, which MNDOT's Dorgan pointed out at a press conference today as the biggest single contributor to structural fatigue, have been, well, off?
By the way, the Wikipedia entry on the collapse is pretty good just 30 hours after the event.