FIU 2018: cracks above an open road

The permanent structure needs checking during construction as well as in its finished state. At FIU, the main span was cast beside the road, moved onto its piers, and then subjected to remedial work, and each of those conditions changed what had to be checked. The NTSB examined the design calculations and the construction-stage review separately. [1]

At 1:46 in the afternoon on Thursday 15 March 2018, that 174-foot concrete span, still under construction, dropped about 18.5 feet onto SW 8th Street in Miami; the NTSB report states that a crew on it had been re-tensioning the post-tensioning rods inside truss member 11 and that the collapse occurred immediately after this activity. Traffic was still using the street below: of its eight lanes, two westbound lanes were closed, but one westbound lane and all five eastbound lanes were open. Eight vehicles were fully or partially crushed, seven of them occupied. Six people died: one worker on the bridge and five people in the cars beneath it. Ten more were injured. [1, pp. xiv, 38]

What the bridge was doing

The crossing was built for Florida International University using accelerated bridge construction: crews cast the main span beside the road, then on 10 March 2018 moved it onto its piers in an operation that ran from 4:30 a.m. to 12:30 p.m. [1, pp29-31] FIU owned the project, with MCM as the design-build contractor, FIGG as the designer and Louis Berger as the independent peer reviewer; construction engineering and inspection came from Bolton, Perez and Associates, and oversight responsibilities sat with FDOT. [1]

The National Transportation Safety Board investigated, and its probable cause is blunt: load and capacity calculation errors by FIGG in the design of the node where truss members 11 and 12 meet the deck. Contributing were an inadequate peer review by Louis Berger, which failed to detect the errors, and the FIGG engineer of record's failure to grasp the significance of the cracking in that node and to get an independent peer review of his remedial plan. On severity, the board names five parties (MCM, FIGG, Bolton Perez and Associates, FIU and FDOT) for failing to stop bridge work when the structural cracking reached unacceptable levels and to take appropriate action to close SW 8th Street as necessary to protect public safety. [1, p. 103]

Reading it through PRISM

My way into the decisions recorded in the report is PRISM: Problem Framing, Research, Interpretation and Synthesis in sequence, with Metacognition checking assumptions and confidence throughout. The report documents calculations and communications, including a month of crack photographs; it cannot establish everything anyone believed, and what follows is my application of the method, not evidence that the case validates it. [1, pp34-38]

Problem framing: are we solving the right problem? Starting on 13 March, the remedial measures included an additional shim under the diaphragm and retensioning member 11. The NTSB says these were outside the FIGG design and planned construction stages. [1, p32] I would also want the decision to address whether work and traffic could safely continue. FIGG's minutes of the 9:00 a.m. meeting on 15 March say no one expressed concern about the span's safety over the road. That records a lack of expressed concern, not proof that no one considered the question. [1, p37]

Research: what would prove me wrong? The checks that could have caught this were weakened one by one. The independent peer review never analysed the construction stages, although AASHTO and FDOT requirements called for analysis during the various construction phases. [1, p. 89] Louis Berger's engineer told investigators after the collapse that construction-stage analysis "was not part of our scope". [1, p. 50] In an email, FIGG had told Louis Berger that the original scope of work remained unchanged, and Louis Berger confirmed it; what changed was the fee, from the $110,000 bid to $61,000, and the timeframe, from ten weeks to seven, to meet FIGG's requirements. [1, pp. 49-50] The crack locations in the 28 February report were mislabelled, so its photographs pointed to the wrong places. [1, p. 34] On the morning of the collapse, a second FIGG employee took seven photographs of the cracks; the engineer of record never viewed those seven before, during or after the 9:00 a.m. meeting, though he had visited the cracks in person and received other site photographs. [1, p. 37] And the remedial plan itself got no independent review. [1, p. 103]

Interpretation: what else could this mean? FIGG used a dead-load factor of 1.25 when calculating the clamping force that increased interface shear resistance. FHWA strongly recommended 0.90 to minimise the assumed capacity, but the NTSB notes that 0.90 was not an explicit requirement for this calculation. FHWA used 1.0 in its post-collapse check and considered that the highest defensible factor here. [1, p46] The NTSB explains the effect of the 1.25 factor as an approximately 25 percent capacity overestimate; using 0.90 would instead have reduced capacity by approximately 10 percent. [1, p73] Separately, FIGG's capacity calculations for nodes 1/2 and 11/12 were about 20 percent above FHWA's checks. The NTSB found that even a joint roughened as assumed in the design would have been under-designed and could have failed. [1, p72]

Synthesis: what do all the pieces say together? Demand underestimated, capacity overestimated, at the one node that failed: against FHWA's models, FIGG underestimated the interface shear demand at node 11/12 by about 850 kips, or 46 percent. [1, p. 78] Cracks were documented from 13 February and photographed through the morning of the collapse, thirty days in all. [1, p. 34] The NTSB documented significant crack progression after the post-tensioning rods were detensioned on 10 March. [1, p. 35] The response was to re-tension member 11: the NTSB recounts that FIGG recommended restressing its rods to return member 11 to its previous state, when the cracks were known by MCM to have been smaller. [1, p. 35] The NTSB report states that the crew was carrying out that re-tensioning work and that the span fell immediately after it. [1, pp. xiv, 38] The board names five organisations whose failure to stop work and take appropriate action to close the road as necessary contributed to the severity of the collapse. The road stayed open. [1, p. 103]

Metacognition: how confident should I be, and what am I missing? The NTSB separates its probable cause from factors contributing to the collapse and to its severity. [1, p103] The design findings draw on post-collapse calculations by FHWA and the NTSB. I can follow that analysis without claiming to know what everyone at the 9:00 a.m. meeting believed. The minutes establish what was recorded.

What to take from this

The peer review needed to cover the construction phases. The NTSB found that it did not. [1, p89] Its finding on severity also addresses the open road: work should have stopped when the cracking reached unacceptable levels, with appropriate action to close SW 8th Street as necessary to protect the public. [1, p103]

Closing

Six people died: one worker on the span and five vehicle occupants below it. [1, p. xiv] The NTSB report states that the collapse occurred immediately after the remedial retensioning work, while traffic was still using the road. [1, pp. xiv, 38]

The report's own record is hard to read past. Cracks were photographed for a month. On the morning of the collapse, in his presentation to the 9:00 a.m. meeting, the engineer of record said "there is no safety concern relative to the observed cracks and minor spalls". FIGG's minutes of the meeting record that, based on the discussions, no one expressed concern with safety of the span suspended over the road: a lack of expressed concern, not a finding that the span was safe. Less than five hours after the meeting began, the span fell.

I examine decisions like this in What Are We Missing? How to Think More Clearly Before You Decide. The book has fifteen chapters, including the Deepwater Horizon case, and sets out PRISM for decisions of our own.

Publication is planned for 13 February 2027. Read more about it at zafaralam.co.uk/what-are-we-missing/. If you have a call coming that you can't afford to get wrong, sign up to hear when it's out at zafaralam.co.uk/#launch.