Did Design Flaws Doom the Minneapolis Bridge?

I don't pretend to do breaking news in this blog, although timely events do make it in here on occasion. A little over a year ago, on Aug. 1, 2007, to be exact, a major bridge carrying I-35W over the Mississippi River in Minneapolis collapsed, killing 13 motorists and drawing attention to the generally poor state of highway infrastructure nationwide. At the time, there was speculation that neglected maintenance might have been a factor in the collapse. But last January, the National Transportation Safety Board issued an interim report which pointed to a design flaw as the possible culprit. The final accident report is due out in November, but the interim report sheds a lot of light on the issue already.

Here are some details. The bridge was an example of a truss design that was popular in the 1960s. The trusses used in the ill-fated bridge were made of steel beams that formed triangular shapes and converged at joints that were held together by gusset plates. Gusset plates are like the slices of bread on either side of a sandwich whose meat is the ends of the steel beams that you're tying together at the joints. The bridge builders fasten the beams to the gusset plates to make the joints secure. As you might imagine, the weight of the vehicular traffic plus the dead weight of the bridge itself and any wind forces, etc. all factored into the forces that these gusset plates had to deal with.

Figuring out all those forces is the job of civil engineers, and although the calculations are detailed and tedious, there is nothing involved that requires the genius of the proverbial rocket scientist. The NTSB people have laboriously recalculated the loads for all the gusset plates on the bridge, and using standards and assumptions current at the time the bridge was built, they found that most of the gusset plate designs (there are eight actual plates in each location) were done right. That is, the gusset plates were thick enough to have enough "capacity" to meet the "demand" that the bridge loads imposed. The report is written in terms of "demand/capacity ratios." If the calculated capacity of a plate to deal with loads is exactly equal to the demand placed on it, you have a demand/capacity ratio of one. Since all these calculations have some margin of error, it doesn't mean that the instant you go to a demand/capacity ratio of 1.1 the thing will bust. But generally, you want most of the items in your bridge to be capable of meeting more demand than they actually will have in practice—that is, a demand/capacity ratio of less than one is desirable. A 1.1 here and there won't be fatal, necessarily, but 1.5 or higher is not a good idea.

Well, somebody back in the sixties must have put some kind of thought like this into the design, because most of the gusset plates with higher demand were thicker, just like they should have been, and had reasonable demand/capacity ratios. To save money and fabrication time, the plates with less stress on them were made thinner, and most notably the set of eight plates designated "U10," near the middle of the bridge. The NTSB engineers calculated three different kinds of loads that were imposed on the plates: shear, principal tension, and principal compression. The U10 plates were the only ones that had a demand/capacity ratio of 1.5 or greater in all three kinds of loads. And the NTSB's study of the wreckage showed that all eight of those gusset plates had popped during, or more likely at the beginning of, the collapse.

You may recall that there was some heavy construction equipment on the bridge at the time. It begins to look like the I-35W bridge was a marginal design that somehow made it through forty years on the edge of safety, only to have a combination of factors topple it into the river last year.

We will have to wait for the full accident report to be released in November to know more. But in its conservative engineering way, the NTSB has shown that age or maintenance may not have had as much to do with the collapse as we thought at first. This shows how errors in engineering judgment can lie dormant for decades before bearing their bitter fruit, to borrow a phrase from the old "Shadow" radio program (no, I'm not quite that old, but I've heard recordings).

We may never know exactly what the designers were thinking when they designed this detail of the bridge, because evidently the Minnesota Department of Transportation doesn't have any records of the consulting firm's calculations for the gusset plates. It might have been a case of rushing a job, or a simple error of some kind. The engineers who actually performed the initial design may be dead, or retired, or otherwise unlocatable. And finding them or suing their firm is beside the point, in a way. Certainly, legal measures are one way to deal with this sort of thing. But a better use of resources, to my mind at least, is to include this story in civil engineering classes and textbooks as a lesson in how not to do it.

Sources: An article describing the news conference at which the NTSB issued its preliminary report can be found at http://www.designnews.com/article/1790-Design_Flaw_Identified_in_Minnesota_Bridge_Collapse.php. The NTSB's interim report no. 383930 can be downloaded at the NTSB docket website http://www.ntsb.gov/dockets/Highway/HWY07MH024/default.htm. My thanks to Michael Sherry for drawing my attention to this report.