Five unanswered questions on the FIU pedestrian bridge collapse

The flood of news that followed the collapse of the FIU Bridge in Miami on 15th March has slowed to a trickle.

The US National Transportation Safety Board (NTSB) have been crawling all over the bridge wreckage (see videos below). It has been reported that they have asked project participants not to share anything with the media. Their preliminary report should be published this month.

A couple of stories note that the project was running over budget and behind programme, caused in part by design changes instigated by Florida Department of Transportation (FDOT). Is this relevant? It is hard to tell. Design-and-build projects often proceed to a difficult programme, never with a positive impact on quality and safety, but most are, of course, built safely.

Anonymous Canadian YouTuber AvE is said to have found the "smoking gun", and offers up a useful analysis of some of the evidence publicly available (I've embedded the video further down this post). His explanation suggests that stressing rods in truss member 11 (see diagram above, taken from the preliminary design drawings) were over-tightened, causing the rods to fail suddenly - but this was a compression member at the time of collapse, and the rods in it should not have been relevant to its load capacity.

Engineers have pointed to the lack of structural redundancy in the design, with its single truss carrying all the load. Catastrophic failure of any individual truss member would therefore inevitably result in collapse of the bridge. However, there's nothing wrong with "fracture-critical" design so long as members (and their connections) are designed to be invulnerable to fracture.

The most comprehensive discussion of the failure that I've seen can be found on the eng-tips forum, currently extending over five separate discussion threads: 1, 2, 3, 4, 5. Be prepared to give up several hours if you want to dive into those in any serious depth.

I think the cause of the collapse will be found to be multi-dimensional. There must be an immediate physical cause of failure: the structure was not adequate for the loads applied to it (at the time of collapse, the only loads of significance were self-weight and prestress). That may relate to defects in construction and/or design, and it may relate to failures of process (doing the wrong thing). That in turn may have been caused by human failures: miscommunication, or plain irresponsibility. Behind this, there will be a wider context of budget, programme, regulatory, political, commercial factors and the like. Some of this is captured in Alfred Pugsley's enduring phrase, the "engineering climatology", the cultural environment within which engineers operate.

I have some questions I would want answered before hazarding a clear speculation as to what happened, and why.

1. Who was responsible for what?
FDOT have used every opportunity to disassociate themselves from the bridge failure, issuing press releases to make clear that their role was only budgetary (channelling funding), or administrative (monitoring progress in use of funding). FDOT are clear that FIU, the contractor MCM and the designer Figg were entirely responsible for the safety of the structure and its construction.

But this is not clear at all.

FDOT have acknowledged that part of their role was to "authorize utilization of aerial space above the state road". They also attended meetings with the design-build team, including one just a few hours prior to the collapse to review cracks found in the concrete. Their representative at that meeting was an engineer, not an accountant.

It seems to me that the designer, Figg, was responsible for the safety of their design, and any amendments made to the design that they had knowledge of. The contractor, MCM, was responsible for following the design and any standard specifications. Both clearly have a duty to the public to ensure the works are safe.

However, FDOT also have a duty to the public. If they had any reason to suspect the works were not safe, presumably they would not have authorised use of the space above the road.

In the United Kingdom, they would have considered the competence (and available insurance) of the project participants. They would also have reviewed the technical proposals for the design to ensure they were appropriate and in line with good practice, and they would have accepted a certificate from the designer confirming the design had been prepared in accordance with what had been agreed. They may also accept a certificate from the contractor confirming the structure had been built in accordance with the approved design. In the UK, they would have required the appointment of an independent design checker, with further check certification.

A particularly prudent public authority might also consider that before reopening a road to traffic passing below a partially complete structure, they might seek specific assurances regarding the safety of the structure in its interim state, to confirm that the design covered the state the structure was being left in temporarily, and to confirm that the construction completed to that point was compliant.

I don't think FDOT can have expected anything to go wrong. The question, however, is whether their technical assurance procedures were sufficient for them to reasonably judge that it was safe to open the road below an incomplete bridge. A "hands-off" approach is clearly a nonsense, otherwise they would be obliged to let all kinds of dangerous work take place without regard to highway safety. The highway authority should, in my view, only be relying on the word of the design-build team if they have a process in place to ensure that word is trustworthy.

2. Why prestressed concrete?

This really does need explaining. Concrete truss bridges are pretty rare, and those that do exist are generally historic.

The reasons for this are not primarily safety-related. A steel truss will be lighter than a concrete truss, making foundations and temporary works less expensive. Parts can be largely prefabricated and assembled, rather than requiring complex cast in-situ works. Temporary construction arrangements are made easier due to the material's better ability to deal with reversal of load.

In some countries, steel will be preferred because there are fewer hidden critical details, a nervousness born out of a past history of failures in post-tensioned bridges when hidden prestressing tendons corrode. That is presumably less of an issue in a warm-weather climate such as Florida.

The positive side-benefit of selecting steel is that it is not normally prone to sudden, brittle failure. It will tolerate overstress by undergoing plastic deformation; yielding and sagging, and giving forewarning before failure.

The same is not true of prestressed concrete, and especially where it is subject to high shear stresses. Failure of a prestressing tendon can be sudden and explosive. Both compressive and shear failure of concrete can be sudden, with little prior warning, especially if there is a lack of conventional reinforcement.

Photographs of the FIU bridge do not reveal large quantities of conventional reinforcement, indeed they seem to show the opposite. The bridge may therefore have been highly dependent on the integrity of the prestressing rods for its load capacity. The interaction of forces at the truss nodes will have been especially complex, given the proximity of the prestress anchorages to these nodes.

With all this in mind, the choice of prestressed concrete seems likely to have contributed to the suddenness of the bridge collapse. So: why was prestressed concrete chosen?

3. What was the nature and location of the reported crack?
It's known that there was a crack at the north end of the bridge, the end which failed. The project's design engineer had phoned FDOT in the days before collapse to report the crack. FDOT had joined the project team for a site meeting to discuss the crack on the morning just a couple of hours before the bridge collapsed.

After the meeting, work was undertaken on the bridge to adjust prestressing rods. It's not entirely clear whether this work was intended to address the cracking, although a link is clearly possible.

It's not clear at this stage whether the crack is actually relevant. It is evidence of a problem, but not necessarily the same problem as was being dealt with at the time of collapse, and not necessarily the same problem which caused failure.

4. Why was work being undertaken on the stressing system immediately prior to collapse, what was this work, and who instructed it?
According to the NTSB:

The investigative team has confirmed that workers were adjusting tension on the two tensioning rods located in the diagonal member at the north end of the span when the bridge collapsed. They had done this same work earlier at the south end, moved to the north side, and had adjusted one rod. They were working on the second rod when the span failed and collapsed.  The roadway was not closed while this work was being performed.

This refers to member 11. Attentive readers will note from the truss diagram above that member 11 was shown (in the preliminary design) with no prestressing. In the permanent load case, it does not require prestressing, as it is under compression under all permanent and imposed loads. However, the design was evidently changed to suit the construction arrangement, which required the span to sit temporarily on a self-propelled modular transporter during installation, supported at the truss node below members 9 and 10. The end part of the truss cantilevered beyond this during transportation, which will have induced tension in member 11.

The prestressing bars in member 11 were therefore required only as a temporary measure during transportation. You would expect them to have been de-stressed (and possibly removed) once the bridge was sat on its permanent supports.

Indeed, that's precisely what a construction representative appeared to say would happen in the "smoking gun" video (starting at 8 minutes in):

There is an obvious discrepancy here. It makes sense that rods in member 11 would be de-tensioned before traffic was allowed back under the bridge, simply because it was convenient to do so while the highway remained a construction site. It does not make sense that any further adjustments were required afterwards; that implies that the bars had not been de-stressed at the intended time.

As well as knowing what was done, a key question is who instructed it, who agreed to it, and why they considered it to be a safe operation to perform above live traffic. There can have been no consideration that the de-stressing work could endanger the bridge.

5. Why was the end truss diagonal (member 11) insufficiently robust to accommodate whatever change in load effect occurred during the re-stressing operation?

Indeed simple calculations should show that the adjustment of stress in member 11 should have been minimal: the compression due to the bridge's self-weight should have been far greater than any stress induced by the prestressing rods. Follow the earlier link to the eng-tips forum for calculations which set this out.

Even in a temporary condition, where reduced factors of safety are sometimes accepted, the concrete truss member and its end nodes should have been robust enough to accommodate any small variations in load caused during construction operations. This should be true even for unexpected changes in load.

The prestress in the stressing bars was being adjusted by means of a hydraulic jack. According to the NTSB statement, one of the two bars had been adjusted, and the second was being worked on when the bridge failed. This will have created an eccentric load effect in member 11, but I doubt that on its own is sufficient to cause failure, and it can be checked beforehand.

There are other issues with hydraulic jacking: in order to loosen the nuts securing the stressing rod, a greater prestress has to be applied initially to allow the nut to be freed. There are risks of hydraulic failure in the jack. The possibility of some sort of failure in the jack, the rod, or the rod anchors, could result in a dynamic shock load being applied to the concrete, but it should have been designed to be robust enough to accommodate any foreseeable range of loading, especially considering that member 11 would be required to carry significantly greater loads once the bridge opened to the public.

I've read a lot of speculation about whether member 11 failed at its upper or lower end, or along its length. There isn't yet sufficient evidence available to do more than speculate. However, the general question remains: why was this part of the bridge not sufficiently robust? This is not a question about the load, or about material defects, it's a question about general good practice in design and detailing, especially for one critical member and two critical nodes on which the entire capacity of the bridge depended.