Road Trip: Girders Underground

by Don Friedman on May 16, 2016

How can structure be simultaneous right and wrong? It’s easy: design models change over time and so do our expectations of what makes a design “correct.” To use an overworked phrase, paradigms shift. Here’s the station roof from the piece of the Circle Line that I previously wrote about:


That’s a series of very big built-up girders carrying street and buildings over the tracks. The girders are tapered on the ends, which seem almost unnecessary since it did not reduce the amount of excavation or construction required, but is both good-looking and a physical manifestation of the moment diagram for the girders.

The detail that I want to talk about concerns how load is transferred. The structural deck for the roof overruns the filler beams (that is, sits entirely on top of them) and then the filler beams overrun the big girders. All of that is correct in that it works exactly as intended to transfer gravity loads down from the deck to the beam to the girders. There are two reasons that, using today’s standards, it represents less than optimal design: load direction and buckling.

London is not subject to earthquakes of any meaningful size, so there is no real possibility that the load direction could reverse from down to up because of a seismic event. On the other hand, UK design has incorporated the possibility of load reversal for some time based on making buildings more resilient to the effects of blast. (It was definitely true in the late 1980s, when I worked briefly on the design of a high-rise in the Docklands development.) The problem with overrunning connections is that a reversal in the direction of load causes a full change in the type of connection stress, in this case from compression to tension. The connection may not have the same capacity for tension that it has for compression and therefore load reversal may be dangerous. (If the beams were connected to one another in the more complicated but more common manner of having web-t0-web connections, load reversal would cause a change in the direction of stress but not the type of stress. Web-to-web connections are governed by shear, regardless of load direction.)

Because steel is so strong, a lot of steel design does not concern gross stress but rather making sure that the relatively thin steel members can actually carry something like their full capacity. Thin structural members subjected to any load other than tension have a tendency to buckle sideways. In steel beams this tendency manifests as lateral-torsional bucking, where the compression flange (the top flange in a simply-supported beam) buckles sideways, subjecting the beam as a whole to a twisting motion along its length. In this case the deck is providing buckling restraint for the filler beams and the filler beams are providing (through the little stool-like connections) restraint for the girders, but not in a particularly efficient way. Direct web connections provides restraint that is directly in line with the theoretical buckling movement, so they simply work better.

So, this roof works as intended, and it works by code. But if were being designed today, it would likely look different in order to more efficiently address two issues that are more important to modern design than they were in the era this was built. Simultaneously right and wrong…like a lot of engineering design.


Previous post:

Next post: