The Case of the Collapsing ColiseumThe Case of the Collapsing Coliseum

Twenty-some years ago 5,000 New England hockey fans gathered at a Connecticut coliseum for a match. Ultimately winners were congratulated, losers consoled, and all went home into a snowy night. The next morning the coliseum had a very different look: it was less than half its earlier height.

Scene of the Crime

The roof had collapsed, at many points touching the tops of the seats. Providentially the building was empty at the time and there was also no fire. Had the collapse and a fire occurred during the match, upwards of a thousand people could have died, making it the largest peace-time disaster in U.S. history.

The roof was supported by what is called a space frame, which rather resembles a gigantic spider web. The frame is a welded steel structure of plates, angles, tees, and other forms. One advantage of the space frame is that it gives unobstructed views-very unlike the old Boston Garden.

The collapse had twisted and torn this inch-thick steel like so much papier mache and, importantly, had torn out many of the welds. The figure shows some of the twisting and tearing that occurred. (The torch cuts made after the collapse.)

The coliseum was city-owned, and the city fathers were understandably perturbed when a prized landmark of their city transformed itself into a twisted pile of rubble. The city set about suing those involved in the collapse, and those being sued hired experts. I was hired by the insurer of the architect. Early analyses had claimed design flaws caused the collapse, so the architect and his insurer were in a difficult position.

The Investigation

The experts spent several days crawling over, under, and around the collapsed roof. The design experts with my client proposed that the collapse was due to the buckling of a member due to compressive stresses. But why would the member buckle? Yes, there was snow, but the roof was designed for a snow load. Could the cold weather have embrittled the steel? Could the welds have been inadequate? I explored these questions at Cambridge-area laboratories.

Toughness in steels is commonly measured by the Charpy V-notch test. A tough steel absorbs a lot of energy before fracturing. At low enough temperatures steel becomes brittle and absorbs little more energy than a ceramic. Steels are best used at temperatures where they absorb a lot of energy in fracturing.

The Smoking Gun

Toughness testing was done at 32F to match conditions at the time of the collapse. The steel specimens showed good toughness. I enlisted the help of a welding expert at MIT in examining the welds. The metallurgical quality of the welds was acceptable, but the sizes struck us as a bit small, considering the thickness of the steel. Welds are supposed to be stronger than the material welded.

I reported that the steel had adequate toughness. This was not good news, as the insurer hoped to shift blame to the steel maker and fabricator. I drew attention to what seemed to be light welds, but my client did not seem that interested. In any case my job concerned the quality of the welds. Whether or not the welds were heavy enough to serve their purpose is the realm of the design engineer. At that point I exited the case.

The city's insurer paid $14 million for the loss. The cost of a new structure with a conventional roof was $27 million. I never learned how the case came out, though I suspect all the involved parties chipped in a share.

A decade or so later I worked with the architect's attorney on a different case. He told me that a later expert had found evidence of fatigue on the fracture surfaces. Any such report was bogus. I examined numerous fracture surfaces and found none of the polishing and striations that go with fatigue failure. The fractures were all of the rough, fibrous sort characteristic of ductile failure by sudden overload. Fatigue, which only occurs under repetitive loading, simply was not a factor.