What caused a bridge that had satisfactorily performed for 40 years to collapse so suddenly?
At 6:05 pm on 1 August 2007, the 40-year-old I-35W Highway Bridge in Minneapolis, Minnesota, suddenly collapsed. A 140 m-long section of the main truss, carrying 111 vehicles at the time, fell 33 m into the river below. Thirteen people were killed, and a further 145 were injured.
Two factors played a role: one was an error in the bridge’s design; the other was the significant loading on the structure at the time of the failure.
Design
The I-35W bridge was designed by consulting firm Sverdrup & Parcel and Associates, and it was opened to traffic in 1967. It was 580 m long, carried eight lanes of traffic (four in each direction), and was made up of 11 approach spans and three truss spans. The bridge was of steel construction, with a concrete deck. By 2004 it carried an average of 141,000 vehicles daily.
Following the bridge’s collapse, an investigation undertaken by the National Transportation Safety Board (NTSB) identified a design error in several of the bridge’s connections. These connections were comprised of flat plates, known as gusset plates, that connected bridge members together.
The NTSB discovered that the gusset plates at the locations U4, U10 and L11 were only half as thick as was required by the relevant design standard. They should have been 24 mm thick, but were only 12 mm thick, which meant they were significantly understrength.
Why was this design error made?
The NTSB investigation concluded that the designer did not perform all the necessary calculations to design the gusset plates correctly. The NTSB examined preliminary design calculation sheets for the main truss gusset plates, which included those at U4, U10 and L11. As the design progressed, the design details for some of the other gusset plates in the structure changed, but the gusset plates at U4, U10 and L11 remained the same thickness, 12 mm. This was because the designers did not appear to account for all the forces on these gusset plates and, therefore, did not thicken them accordingly.
(In technical terms, the design thickness of 12 mm appears to have been based on only considering the forces expected to pass across the splices between chord members, without considering the shearing forces introduced in the gusset plates by the diagonal and vertical members.)
The NTSB concluded that the design firm clearly knew how to design connections correctly, as evidenced by other correctly designed gusset plates in the structure, but they failed to do so at all locations. While this design error was not sufficient in and of itself to collapse the bridge, it produced latent weaknesses, with the trigger for the collapse coming 40 years later in the form of abnormal loading.
Loading
At the time of the failure, construction works were taking place on the bridge. These works involved removing approximately 50 mm of material from the bridge’s concrete wearing course, and replacing it with fresh concrete. By 1 August, the contractor had removed and replaced seven sections of the wearing course, and they were preparing for the eighth concrete pour that evening.
Their preparations for the pour involved stockpiling sand and gravel (to make concrete) on two of the bridge’s southbound lanes. The contractor’s reasoning for stockpiling material on the bridge, as opposed to mixing the concrete off the bridge and transporting it into position, was that the Minnesota Department of Transport specification required that only one hour elapse between initial concrete mixing, placement on the bridge and final screeding.
To comply with the one-hour window, the contractor decided to make the concrete on the structure, close to where it would be placed.
But the stockpiling added significant loading to the bridge. By 2:30 pm on the afternoon of the failure, there were 84 t of gravel, 90 t of sand, and 90 t of construction vehicles, equipment and personnel on the bridge for the pour. This loading totalled 264 t, with the Minnesota Department of the Transport estimating that the stockpiled materials alone was equivalent to four times the design load of the bridge.
How did the contractor get permission to stockpile the material?
The NTSB determined that they did not, in fact, obtain permission for the stockpiling. They also determined that the Minnesota Department of Transport did not have a policy that specifically required contractors to obtain such approval.
What appears to have occurred instead was that some of the contractor’s employees had, on a previous occasion, asked a Minnesota Department of Transport site representative about stockpiling on the bridge. The contractor’s employees interpreted the representative’s response as permission to do so, and as a result, stockpiling had taken place on a number of previous pours.
So, on 1 August, the contractor simply commenced stockpiling without permission from the Minnesota Department of Transport, and without any apparent concern for the significant weight they were adding to the structure.
Read next: Dr Sean Brady on what causes fatalities in the mining industry
Cause of failure
The NTSB investigation would conclude that the loading on the bridge was sufficient to result in the failure of the U10 gusset plates, with this failure culminating in the progressive collapse of the bridge.
But was it likely that the significant loading, even if the design error had been absent, would have been sufficient to collapse the structure? If this were the case, then the design error, despite being a deficiency, would not have been causative in the failure.
The NTSB found this was not the case. The cause of the failure was the significant loading on the bridge (in addition to other loading added to the bridge over its lifetime), in combination with the design error in the gusset plates. If the gusset plates had been designed and constructed in accordance with the relevant standard, then the loading, despite being significant, would have been unlikely to result in the failure of the bridge.
The I-35W is the story of how a critical design flaw lay dormant for four decades, missed by the original designers and undetected throughout the bridge’s lifetime. This latent weakness, combined with the significant loading placed on the bridge – a loading placed without any form of engineering due diligence – resulted in the collapse. The I-35W tragedy reminds us that it’s rarely a single factor that causes failures, but a combination of factors that come together to overwhelm the conservatism we build into our structures.
This article was originally published on LinkedIn. Read the original post.
About the author
Dr Sean Brady FIEAust CPEng is a forensic engineer and Managing Director of engineering consultancy Brady Heywood. In 2020, he completed the Brady Review, which was tabled in parliament and made 11 recommendations to the regulator and mining companies on how to improve safety. In 2024, he completed the technical and organisational investigation into the 25 May 2021 incident at Callide C Power Station in Queensland.
The impacts of disaster can be wide-ranging. At this on-demand webinar, explore the hazards of hydrogen explosions and the knock-on consequences on infrastructure.
Based on the report, I understand that 3 gussets were supplied as 12mm (U4, U10 & L11) and all other gussets were 24mm. If this is the case, then despite the clear error in the design and documentation process, I would also have expected that this should also have seen as an obvious inconsistency during the construction on site. Obviously if U9, U12, L13 etc. were all 24mm gussets, the reduction for U10 and U11 to 12mm would have been clear for everyone on site.
Clearly the site supervision was probably also inadequate, with inspectors not looking at the overall intent of the design, and only looking at joints in isolation ?
In any case, this is the type of mistake which we are all trying to avoid, with a view to making our communities a safer place for everyone.
How did the gussett plates fail ?
Did the gussett plates fail by shear or bearing ?
One would like to see photos of the
damaged gussett plates
Agree that gusset plates were a design flaws but putting loads 4times the design load is really the culprit. To expect an older bridge to have that much redundancy is not realistic.
could the effect of vibration and lateral forces due to construction activity imposed additional stresses to the 12mm gussett plates and the bolts (or rivets) thus weakening the whole connection?
As a student in engineering , this was a great article to read. Well written in simple language.