Case Studies: Bridge Collapses

The Tacoma Narrows Bridge opened in July 1940, connecting the Kitsap Peninsula to Tacoma, Washington. Its plate girder deck was barely 12 feet deep for a main span of 853 metres. Engineers and drivers noticed it was unusually flexible from the first day of service. By November 7, 1940, four months after opening, a 64 km/h wind set up an oscillation that shifted from vertical undulation into a violent torsional mode. In about 45 minutes the deck ripped itself apart and fell into the Narrows. A camera crew happened to be filming, producing what became the most widely reproduced footage in the history of structural engineering education.

The immediate post-collapse explanation, repeated in textbooks for decades, attributed the failure to resonance: the wind happened to match the bridge's natural frequency, pumping energy in at the right moment like a child on a swing. This explanation was not merely incomplete. It was wrong. Wind does not blow at a steady frequency that matches structural modes. What actually happened was aeroelastic flutter: the bridge's torsional oscillation changed the aerodynamic loading acting on the deck, which in turn reinforced the torsional oscillation in a positive feedback loop. The driving force was regenerated by the bridge's own motion, not by any coincidental matching of frequencies.

The forensic engineering lesson from Tacoma Narrows is not the mechanism alone. It is the model-validation lesson. The designers had calculated static wind loads per the practice of the day. They had no framework for dynamic aeroelastic behaviour. After the collapse, bridge engineering acquired that framework: wind tunnel testing of deck cross-sections became standard for long-span bridges, and aerodynamically stable deck shapes, open truss sections or streamlined box girders, replaced solid-plate girders. The failure of one bridge improved the safety of every long-span bridge built since.

The Hyatt Regency hotel in Kansas City opened in 1980 with a spectacular atrium lobby crossed by three suspended walkways at the second, third, and fourth floors. The walkways hung from hanger rods anchored to the roof structure. The original design showed the second-floor and fourth-floor walkways sharing a single continuous rod running from the roof through a nut-and-washer bearing at the fourth-floor walkway beam and then down to a similar bearing at the second-floor beam.

During the shop-drawing phase, the fabricator proposed a change: instead of one long continuous rod, use two shorter rods, one from roof to the fourth-floor beam and a second from the fourth-floor beam to the second-floor beam. This looked like a practical fabrication improvement. What no one recalculated was the effect on the fourth-floor connection. With the original single-rod design, the nut and washer at the fourth-floor beam carried only that floor's dead and live load. With the two-rod change, the fourth-floor nut had to carry the combined load of both the fourth-floor and second-floor walkways, because the upper rod now transferred the second-floor walkway's load upward through the fourth-floor assembly.

On July 17, 1981, the atrium was crowded with people at a tea-dance on the second floor and watching from the walkways above. The fourth-floor walkway connection failed in punching shear: the nut and washer pulled through the box beam flange. The fourth-floor walkway fell onto the second-floor walkway, which then also collapsed. The investigation by the National Bureau of Standards (now NIST) showed that the as-built connection had a capacity of approximately 90 kN. The load on the night of the collapse was approximately 175 kN. The connection was below code capacity even for normal occupancy, not just for the night of the collapse.

The professional responsibility dimension is significant. The shop drawing change required the engineer of record's approval. The engineer's stamped review was interpreted by the Missouri licensing board as an approval of the change. Whether the engineer actually checked the load calculation for the new configuration was the contested factual question in the subsequent licensing hearings. The board ultimately found that the standard of care required the engineer to have checked it, and revoked the licences of both the principal engineer and a project engineer.

The Ponte Morandi, designed by Riccardo Morandi and opened in 1967, carried the A10 motorway across the Polcevera valley in Genoa. Its cable-stayed towers used stays that were not conventional steel cables but prestressed concrete beams encasing wire tendons under tension. Morandi believed this design was more durable than exposed steel cables. In practice the opposite was true: the concrete encasement made the tendons largely inaccessible for inspection and created a closed environment where any water ingress would corrode the high-strength steel wires without visible surface indication.

By 1990 visible deterioration prompted the first major repair programme, in which two stays were encased in a second layer of concrete containing post-tensioned external tendons. Subsequent inspections noted ongoing deterioration and prompted further discussions about repair or replacement. On August 14, 2018, stay 9 of tower 9, the 80-metre deck section on the western approach span, failed. The deck fell 45 metres, killing 43 people. Vehicles were on the bridge and in the buildings below at the time of the collapse.

The Italian investigative committee and independent engineering analyses converged on corrosion of the stay tendons as the primary cause. The high-strength steel wires, once their protective oxide layer was destroyed by the carbonation of the concrete, were susceptible to stress corrosion cracking and hydrogen embrittlement under sustained tension. The cross-section of functioning tendon area diminished over time until the remaining section could not sustain the imposed load. Because the geometry of the stays gave limited redundancy, the loss of one stay was sufficient to cause catastrophic collapse of the associated deck section.

Tacoma Narrows, Hyatt Regency, and Ponte Morandi involve distinct physical mechanisms: wind dynamics, a connection load-path error, and hidden corrosion. Each mechanism is genuinely distinct. But looking at the investigation process rather than the physics, the three cases share a set of recurring root cause conditions that appear in most major structural failures.

• Knowledge boundary: in each case engineers operated at or beyond the boundary of what was well understood at the time. Aeroelastic flutter was not in the 1940 design toolkit. The load-path consequence of the Hyatt Regency rod change was not recognised. The long-term durability of the Morandi stay type proved worse than expected.
• Review failure: none of the three failures involved adequate independent review of the specific detail that caused the collapse. Dynamic wind behaviour was not checked at Tacoma Narrows. The load calculation was not re-run at Hyatt Regency. The tendon condition could not be assessed without destructive testing at Ponte Morandi.
• Warning signs missed: Tacoma Narrows had been oscillating since day one. Hyatt Regency had a documented shop drawing change. Ponte Morandi had been repaired once already. In each case the warning existed but was not read as a precursor to imminent collapse.

Bridge collapses present specific evidence-collection challenges. The collapsed span often falls into water or onto occupied ground, so recovery of structural components is complicated by rescue operations, environmental conditions, and jurisdictional questions about salvage rights. At Ponte Morandi, the fallen deck section was partly buried under its own debris in a residential area, and forensic access had to be coordinated with emergency demolition of the surviving towers.

In the Hyatt Regency collapse, the four hanger rod assemblies at the fourth-floor walkway connection were the critical physical evidence. Investigators recovered these, measured the box beam flange punch-through deformation, and used that deformation pattern to determine the direction and magnitude of the applied load at failure. The hardware was retained as court exhibits and matched precisely to the load analysis: the capacity was below demand before the event, not just marginally below it on the night.

For Ponte Morandi, the investigation included metallurgical examination of tendon wire samples recovered from the fallen stay, measuring cross-section loss, fracture surface morphology consistent with stress corrosion cracking, and chloride penetration profiles in the concrete. These measurements directly supported the corrosion mechanism and also established the timeline: the degree of degradation was consistent with decades of progressive corrosion, not a sudden overload event on the day of the collapse.