Hyatt Regency Walkway Collapse

http://en.wikipedia.org/wiki/Hyatt_Regency_walkway_collapse

Here we have a walkway suspended from the ceiling and two connection details. Both of them are bad for different reasons. However one of them theoretically works, and the other killed 111 people. Which is the one that killed over a hundred people and left over 200 injured?




I think most people have guessed that Detail 2 is the faulty one, which is right. It was Detail 2 which failed. Most people are unsure of exactly why one detail works and the other doesn't which is understandable because without further information it would be impossible to make an accurate judgement. Having said that, if you were a design engineer looking at 1 & 2, and you had to make a call, what would you do?


In engineering design analysis the objective is to visualize the methods in which a structure will fail and design the structure to withstand those potential failures. Detail 2 would work if it were re-designed with different beams/rods/etc, but it was a modified version of Detail 1 with a different load path. Because the load path was different, the mode of potential failures was different. If you have the same elements (the same beam, the same rod, etc), yet the design is completely different, you may have problems.

The chief difference between the two connection details is that the second detail doubles the load on the beam itself. While in both cases the rod takes the load of both walkways, only in detail 2 does the beam take the load of the walkway below. In Detail 1 the lower walkway is supported directly by the rod and places no additional load on the upper walkway. In Detail 2 the lower walkway is actually supported by the upper walkway, which is in turn supported by the rod from the ceiling.

This adds a new mode of failure to the problem. Where previously the lower walkway did not interact with the upper walkway at all (structurally), there is now the possibility of the upper walkway beams failing where the lower walkway is connected to it. This is not something that was accounted for, and the beam failing was exactly what happened.

Secondary factors for the failure are the design of the beam itself as well as the method of attaching the rod to the beam. The beam was produced by welding two tapered flange beams together to form a hollow box section. The contributed to failure because the tapered flanges are narrowest in the middle of the box where the flanges come together, this also was where the weld was. The weld itself is not necessarily any weaker than the surrounding metal as such, however the high temperature of welding can affect the heat treatment of the surrounding metal.

The other secondary factor was how the rod attaches to the beam. The rod passes directly through the box section and the beam is suspended by a nut with a washer, right through the middle of the box where the narrowest sections of the tapered flanges came together. Given the relative weakness of the beam at this point, a better detail may have used a metal bracket that supported the beam rather than pass straight through the weakest part. This would be a cost increase, but a tiny one considering the scope of the project. These secondary factors did not cause the beam to fail, but they added to the catastrophe as I will explain later.

The fact that the detail looks so innocuous but has such massive structural implications emphasizes why experienced engineers need to carefully examine details like this. What happened in this case was a miscommunication/mismanagement problem between the design engineer and the *shop drafter. Specifically, the engineer produced Detail 1. The shop drafter/builder noticed that this design was actually not possible to actually build - a nut would have to be threaded over several yards of rod, and the threads would have to be protected against transport damage - all of this is highly impractical. Thus the shop drafter came up with their own detail which they submitted to the engineer for approval. The engineer did not properly review the shop drawings and thus over 300 people were killed or maimed. I myself have reviewed shop drawings and have noticed shop drafters making up their own details when they should be calling the engineer to ask questions. The fact that shop drafters sometimes send the drawings to us for review shortly before the steel is to be fabricated makes things even more difficult. A recent 3-level building I worked on had well over a hundred pre-cast panel and steel shop drawings to be reviewed, fortunately they sent us the drawings well before the stuff was to be fabricated though.

(*A shop drafter produces working drawings that a steel fabricator will use)



After all that we know that a faulty connection was to blame. So in what manner did the connection actually fail?

Firstly, in a walkway such as this, the beams span continuously over the support - this means there is a high negative bending moment where the beam is supported by the rod. In fact for a continuously supported beam, the negative moment where it is supported by the rod will be the point where the bending stresses will be the highest. In addition to this, when the bending stresses are in the negative direction, it means that the bottom face is the compression face.

When a beam is loaded in bending, one face of the beam goes into tension, and the other into compression. Imagine a ruler from your pencil case spanning between two tables. As you place a load on it, the bottom of the ruler stretches while the top is compressed. For most steel sections, the compression face is the critical face that will fail first because it can buckle.

Again consider your ruler and imagine you're holding both ends and trying to crush it. If the ruler remains perfectly straight you'll find it basically impossible, if someone else comes along and pokes the ruler from the side it quickly bends and you'll snap it.

Back to our walkway. Note that the bottom flanges have folded upwards. If it looks to you like the rod with the nut pulled through the beam and folded the flanges upwards, you would be absolutely right. When we remember that the bottom flange was the compression flange (the critical one), now imagine what was happening to your ruler when compressed happening to the bottom flange of the beam...

Once this initial failure occurred, the weld on the bottom of the beam "unzipped" because welds tend to be relatively brittle and so cracks propagate through them. This meant that if any of the other rod connections were still ok, they weren't for much longer when the crack reached them.