We structural engineers are inveterate ambulance chasers; the only thing we enjoy more than designing bridges is trying to figure out why some of them fail. Last week’s collapse of the Fern Hollow bridge provided us with an outstanding opportunity to pretend we were forensic investigators; fortunately, there were no fatalities associated with this catastrophe.
When I got up that morning, the first message in my email mailbox was one from Don Toney, “Turn on your TV! There has been a bridge collapse in Pittsburgh!” I immediately complied and found myself engulfed in an intriguing drama. I spent the rest of the day splitting time between being glued to the television and trading email dialogue with other would-be structural experts, as each new bit of information became available.
My first reaction, of course, was “Near zero temperature – brittle fracture!”. An important phenomenon in material science is the fact that steel loses its ductility (its ability to deform significantly without fracturing) at lower temperatures, precipitating brittle fracture. This abrupt failure requires the combination of three factors – a temperature below the material’s Nil Ductility Transformation Temperature (NDTT), a very high stress, and the presence of a reasonably large crack. We presume the National Transportation Safety Board (NSTB) will test samples of material in key components to confirm their NDTT is below the extreme temperature anticipated in the bridge.
The Fernwood Hollow bridge was designed as a batter-post rigid frame structure – essentially a continuous girder 447 feet long, supported by sloping legs at its quarter points and abutments at each end. A good example of this design is the Canon-McMillan Alumni Bridge on I-79 south of the Canonsburg Exit. PennDOT refers to this design as a “K-Frame”.
Based on aerial photographs of the collapsed bridge, it appears that the failure probably was initiated at one of the sloping legs. The deck appears to be have come down as a unit, then broken into four distinct pieces. The fact that the Frick Park end is much closer to its abutment than the Regent Square end suggests that the legs at the Frick Park end failed first. An NSTB report on February 6 reinforces this indication. We hope that part of the investigation by the NSTB includes a mathematical model simulating possible failure modes.
We assume each of the supports was designed to provide an appropriate degree of freedom – translation and/or rotation. The bases of the sloping legs were probably pins, rollers, or rockers, each permitting free rotation of the end of the leg. Based on photographs and visual reports of massive corrosion on the legs, it is possible that one or more of these components became jammed and induced a clamping moment for which the leg was not designed.
Early in the news coverage we learned that several persons on the Frick Park trail under the bridge had noted apparent problems with the bracing between each pair of legs and reported them to “the City”. Although the city does own the bridge, it is not clear which department has jurisdiction over its maintenance. It is clear that the response to these reports was questionable.
As designed, each pair of legs had two sets of “X-Braces” tying them together at mid-height. The primary function of the braces is to transfer lateral (wind, for example) loads from the deck level to the supports. Although such loads are small (10,000 pounds), they would be too great for unbraced legs to accept without failure.
A secondary function is to reduce the unbraced length of the legs by one half. Each leg acts as a sloping column carrying about one fourth of the dead load of the bridge. The ability of a column to carry a load is inversely proportional to its unbraced length – in this case, eliminating the bracing would reduce the leg’s allowable load by a factor of four.
Apparently the original bracing had deteriorated so much by 2014 that it was removed and replaced by tensioning cables between the deck and the bases of the legs. Properly sized and tensioned, such cables should be able to transfer anticipated wind loads; they do not however satisfy the secondary function. It is easy to suspect that buckling of the unbraced legs caused the collapse.
My colleagues and I have spent a week speculating on possible causes of the failure and wishing we could be part of the NTSB investigative team. Once the deck is picked up, they may find “a smoking gun” that clearly explains the cause of the failure; it is equally possible that this cause is more complicated and will require extensive analysis. Their February 6 report did state that “Although certain areas of the welded steel girders were identified as fracture critical, no primary fractures were found in these areas”.
The media has focused on inspection and the PennDOT classification of this bridge as “poor”. Does “poor” mean a bridge should be closed? No, “poor” is an indication that the bridge is in the future repair queue, waiting its turn to be rehabilitated when funds are available. Does passing inspection imply a bridge is guaranteed not to fail? In reality, the system works well, but is not foolproof.
We are pleased that PennDOT District 11 has been assigned the task of replacing the bridge, on an expedited basis; their recent performance on remediation of the Route 30 landslide was exemplary. The choices of HDR, Inc. to design the new bridge and Swank Construction to build it ensure the replacement will be in operation in a minimum time.
The American Society of Civil Engineers has been campaigning for years for a bigger investment in our infrastructure; incidents like this one highlight the fact that the consequence of deferred maintenance can be massive, socially and economically. The recent “Bipartisan Infrastructure Bill” is a step in the right direction – many additional steps will be required to reach an adequate level of service.