A New System for Preventing Bridge Collapses

On August 1, 2007, the eight-lane Interstate 35W bridge in Minneapolis, Minnesota, collapsed catastrophically during rush hour, killing 13 people and severing a crucial connection across the Mississippi River. Before an investigation even had time to get started, the Minneapolis bridge failure had reignited an old debate about whether the United States was investing enough in overseeing and maintaining its infrastructure.

In general, the primary cause of bridge failure is not sloppy maintenance or deviation from design standards; bridges fall down because extreme events are too much for them to bear. A bridge designed to current standards and properly maintained can still fail. That is because we still do not know enough about how exceptional stresses—and especially the interaction of exceptional stresses—can compromise a bridge’s integrity. This ignorance prevents the United States from setting design and engineering standards that can cope with extraordinary “loadings” caused by weather and other extreme events to prevent them from bringing down a bridge.

We need to revisit current standards with the benefit of much more extensive research. We need to install monitoring devices to record stresses on bridges. Of course, we need to examine what happens during catastrophic failure such as the Minneapolis bridge collapse, but we also need to study and learn from incidents of lesser damage and close calls.

In the case of the Minneapolis bridge collapse, speculation in the initial weeks focused on fatigue cracking and corrosion (implying inadequacies of inspection and maintenance). There was also speculation about the adequacy of steel plates called gussets, which connect a truss’s structural members at a joint (implying problems in design, construction, or maintenance). We will not know for sure what the cause was until the National Transportation Safety Board reports its findings, which will not happen before summer 2008.

However, most catastrophic bridge failures occur because of, or are significantly exacerbated by, factors outside the bridge structure itself, especially extreme events. There are many kinds of extreme events, including: the force of a surging river current; the collision of fuel-carrying vehicles with a bridge support; large ground acceleration in an earthquake; a group of trucks carrying more weight than is allowed by law; and unprecedented events such as a terrorist attack through blast at piers or support cables.

An extreme event imposes load effects on the bridge; these add to loads exerted by the bridge’s own weight and by moving vehicles. Bridge designers seek to be sure that the bridge provides more than sufficient capacity to resist the loads likely to be applied to it during its life. When a bridge fails under loads that it should have resisted under accepted standards of professional practice, it is said to be a failure of its resistance capacity. Otherwise, the bridge fails because loads imposed by external events exceed those against which standard codes and practices are expected to provide protection. The latter are the more common so we focus on them here.

Bridge design and engineering advanced significantly in the latter half of the past century because bridge designers and engineers applied lessons learned from a series of bridge failures. Despite these advances, however, the United States has continued to experience major bridge collapses. Those failures happened because we don’t know enough about the dynamic effects that extreme loads have on structures. That is why the United States, in addition to needing to increase spending to maintain its infrastructure, also needs to invest in the research and information systems that will tell engineers how to avert future disasters.

Researchers need to examine both the failure of materials at the microscopic level and the progressive failure of entire bridge systems. They should further develop computer modeling of nonlinear dynamics during bridge failure. Working with economists and risk analysts, they should seek better understanding of methods for minimizing costs during the life of the bridge. But these studies would largely be extensions and developments of research that is already proceeding.

We argue that the United States needs to invest in new national R&D initiative, one that monitors and reports on bridge performance and bridge failures electronically through smart sensor networks incorporated into bridges. The nation needs to move its transportation systems into the digital age.

A history of failure

Historically, the most infamous bridge failures include the Ashtabula bridge (Ohio, 1876), Tay Bridge (Dundee, Scotland, 1879), Quebec Bridge (St. Lawrence River, 1905), and Tacoma Narrows Bridge (Tacoma, Washington, 1940), which achieved notoriety because the catastrophic collapse was captured on an amateur’s movie camera. These bridges all collapsed because of design or materials flaws that have been rectified by present-day practice.

However, better design and engineering have clearly not eliminated the risk of bridge collapse. Aside from the Minneapolis case, important examples in recent decades include the Schoharie Creek bridge on the New York Thruway in 1987 during intense rain, the San-Francisco-Oakland Bay Bridge in 1989 during an earthquake, and the Walnut Street Bridge in Harrisburg, Pennsylvania, in 1996, from scour due to flooding. Several other failures occurred during flooding on the Mississippi River in the 1990s, earthquakes at Northridge (1994) and Kobe, Japan, (1995), and the Gulf Coast during Hurricanes Katrina and Rita in 2005.

Data compiled for 1989-2000 by researchers Kumalasani Wardhana and Fabian Hadipriono and published in the Journal of Performance of Constructed Facilities shows a continuing record of U.S. bridge failures, even before the resurgence of concern in the wake of Katrina. They obtained their data from a national database prepared by the New York State Department of Transportation (NYSDOT) after the Schoharie Creek tragedy of 1987. Over the study period, the authors identify 503 cases of bridge collapse, with a peak of 112 in 1993.

The highest number of bridge failures (85) occurred in Iowa and the second highest (64) in New York. Iowa’s high figure may derive in part from the effects of the 1993 floods on the Mississippi and Missouri Rivers, and New York’s perhaps from more complete in-state data available to NYSDOT. As a proportion of the number of bridges in these two states, the failure rate for the 12-year study period was 0.33% for Iowa and 0.29% for New York. In the United States in general, there was an average of 42 bridge failures every year. If for no other reason than the aspiration for excellence in public works, bridge safety planning needs attention.

Going to extremes

In most cases, the causes of bridge failure are difficult to diagnose because they occur in complex interaction between the resistance and load. Failures attributable to resistance may occur because of flaws in the structure’s original design, flaws in detailing documents submitted by contractors with engineers’ approval, errors in construction practice, materials deficiencies, environmental deterioration over time, or inadequate inspection and maintenance. Failures are otherwise attributable to excess loads. For any particular case of failure, a forensic study would be needed to scientifically establish whether a failed bridge did or did not fulfill required codes, design specifications, and expectations of the profession and whether, by accepted standards, problems should have been recognized and the bridge closed or retrofitted. Without consistent forensic study, statements about the cause of failure are suspect.

Nevertheless, the information available from databases does provide a suggestive list. Table 1, obtained from Wardhana and Hadipriono, lists both internal (resistance-side) causes and external (load-side) causes. The failures that can be readily labeled external make up just over 82% of the total.

Bridge failures during the Gulf Coast hurricane disasters of 2005 serve as a further wake-up call. According to the National Institute of Standards and Technology, three out of four major bridges from New Orleans to areas north of Lake Pontchartrain underwent catastrophic failure, and the fourth was reduced to partial service. Major bridges also failed over Bay St. Louis, Biloxi Bay, Back Bay, and the Pascagoula River, all in Mississippi, and over Mobile Bay, Alabama. In addition, several movable bridges became inoperable because electrical or mechanical chambers were flooded. In fixed bridges that underwent structural failure, the causes included the lateral forces of waves on bridge superstructure and substructure along with buoyant uplift of inundated bridges. They also included impacts from barges and other debris, such as floating vehicles, shipping containers, logs, boats, and large appliances, plus the undermining of foundations through scour.

The Gulf Coast’s vulnerability may not be at all unique. Massive land-use changes have exacerbated the propensity for flooding in other places as well. Even New York City is among the coastal cities at risk from hurricane and coastal surge.

The threats from terrorism

The possibility of bridge failure from terrorism is also a concern in the aftermath of 9/11. Long-span signature bridges that have high symbolic significance may be especially at risk. In contrast with natural and accidental hazards, terrorist threats are likely to focus on high-visibility, high-service bridges of strategic economic importance

In some ways an attack against a bridge, any bridge, is less threatening to life and safety than an attack against a building because bridge users are traversing the structure and not (as in a building) occupying it. If the bridge and its approaches are not congested with traffic, then simply preventing additional vehicles from entering will quickly clear the bridge. Bridge landings provide easier and faster egress than do the doors of a high-rise building. With proper protocols, bridges can be closed early in response to suspicious activity.

In other respects, bridges are more vulnerable than buildings. Bridges are subject to constant flow of vehicles, which as a practical matter cannot be inspected for explosives. Because in a bridge (as compared to a building) the structural members are exposed, a malicious but trained observer may more easily estimate the most vulnerable points where explosives can undermine the structure’s integrity catastrophically. Long-span bridges generally have less structural redundancy than buildings do. Suspension bridges are especially vulnerable because the rupture of one main cable might cause rapid, progressive, and unstoppable catastrophic failure. Cable-supported bridges are also often architecturally dramatic and would therefore attract attackers searching for high-visibility targets.

Several news reports have suggested that bridges have been targeted by terrorists. According to news reports in 2004, al Qaeda documents captured in Pakistan indicated that the Brooklyn Bridge was reconnoitered as a possible terrorist target. Other news items reported the arrest of a suspect, wanted on other terrorism-related suspicions, for close-up filming of details of the Chesapeake Bay Bridge. In California, there was public debate reaching the gubernatorial level on the ability of the San Francisco–Oakland Bay Bridge to withstand an explosive attack.

Even in the absence of actual incidents, bomb threats and fears of terrorism can cause interruptions of service. The National Cooperative Highway Research Program estimates that there are 1,000 “critical bridges” at which special anti-terrorism precautions may have to be taken, even though there is no certainty that attackers would target these particular bridges.

Multihazard events

Bridge safety is receiving renewed attention because there is a growing appreciation of how separate load effects can combine to increase hazards. For example, surge and strong wind caused by Hurricane Katrina’s imposed lateral wave forces, buoyant uplift, debris impact, and scour. During a flood, the bridges that remain in service may be subjected to unusually heavy traffic loads. Such multihazard events may not even be exceptional; they may be at least as common, and perhaps more so, than the simplest scenario in which a single extreme event exerts a single type of load.

We have identified five categories of extreme events, listed in order of approximate likelihood:

• A simple event in which one extreme event exerts a single load.
• A combined multihazard event in which a single extreme event exerts multiple load effects, such as an earthquake that sets off ground shaking, ground faulting, and soil liquefaction.
• A consequent multihazard event, which is a single extreme event exerting one or more initial load effects that set off secondary events, each of which exerts additional forces on the structure—such as flood and wind that cause barge collisions or earthquakes that cause vehicles to collide and catch fire.
• A subsequent multihazard event, two unrelated extreme events separated in time, perhaps even years, with the second event affecting a subsystem weakened in the first event—for example, a barge colliding with a pier a few months after the pier is weakened by scour.
• A simultaneous multihazard event in which two unrelated extreme events coincide; for example a heavily loaded truck convoy crossing a bridge during a flood, or an earthquake in Alaska during the winter, when there is extreme cold or pressure from an ice jam.

Of these, combined and consequent events are the most likely multihazard events except for scour, which can cause long-term subsequent events. In practice, it may be difficult to distinguish between combined and consequent events. The critical lesson is that bridge design and planning must increasingly take into account such combinations of events.

Comprehensive monitoring

Bridge experts are becoming aware that bridges should be situated, designed, retrofitted, and maintained with a view to an ever-widening range of extreme loads and their combinations, with terrorism posing a new challenge. It is appropriate that, as the field of engineering progresses, its practitioners should refine their work to improve quality, cost-effectiveness, and public safety. If the United States is going to increase investment in infrastructure, it is essential that it have reliable information on what makes bridges safer.

We can make progress in bridge safety by modeling the reliability relationship between loads (including multihazard loads) and resistance and by developing improved structural modeling and materials. Both kinds of progress depend on better data, but obtaining better information is no easy task.

For building or bridge design (as compared to the design of ordinary commercial products), it is prohibitively expensive to put the structure through destructive testing. The National Science Foundation does establish facilities for earthquake study, with shake tables where experimental structures can be tested, but this testing is necessarily conducted on models that are far smaller than the structures of interest. It is also desirable to conduct research during the demolition of an obsolescent or otherwise unneeded bridge, but such research is almost unheard of in the United States, perhaps because of legal constraints.

Additional results may be obtained from forensic engineering studies of failed bridges. A well-developed approach is the post-disaster reconnaissance study, in which teams of engineers and other investigators examine the unique confluence of events that caused a structural failure. But such failures are rare and highly varied in the combinations of hazard types, severities, structure types, soils, and chains of causation that they exemplify. The studies simply do not provide enough evidence to enable generalizations about hazard effects. In addition, the fundamental flaw of forensic studies is that they take place only after a bridge has failed.

We believe that an especially useful supplemental source of information will be incident reporting. Its essential and distinctive feature is this: Reporting is on all incidents, including near-misses, minor mishaps, and significantly stressful events, not just spectacular accidents, disasters, or failures.

With respect to bridges, the challenge would be to measure and document events that severely stress a bridge but result in little or no damage. Data should be collected and coded in a rigorous, consistent manner. The data should encompass both the loads (and the single or multiple hazards that caused them) and the structural impacts. When such incidents, as well as full-fledged accidents are studied, the resulting database may be large enough to allow for statistical analysis. A practical advantage is that the organizations involved are more willing to share information about small incidents than major ones, where issues of liability and official responsibility arise.

Incident reporting systems are now required for airplane accidents and near-misses and are routinely recommended for the study of medical errors and construction site incidents. To implement such a system for bridges, the United States should develop and widely implement a structural health monitoring system on bridges. Such a system will track stresses, strains, and other conditions on the structure and its components and will identify, locate, and measure the effect. Such systems could have added value beyond data collection—for example, for real-time emergency management or to inform bridge operators of conditions during and after extreme events.

This system should also unify forms and reporting that are now collected in many different ways. At present, hazard data are sought by specialists in meteorology, hydraulics, seismology, highway accidents, fires, maritime accidents, volcanology, hazmat events, and security threats. Coordinated and consistent data collection is essential because even measures of severity are difficult to compare across hazard types. Fire intensity, for example, is not quantifiable in the same way as flow velocity or buoyant uplift from flooding.

Of course, it makes sense that data must be defined and screened by the pertinent scientific disciplines. However, excessively uncoordinated data collection and inconsistent reporting serve as an obstacle to policy insight on the probabilities of harm to various infrastructures and on ways of alleviating this harm.

In view of the resurgent concern about bridge safety, we propose infrastructure incident reporting through a National Bridge Health Monitoring System. To the extent that it focuses only on bridges, the system should be implemented under the auspices of the Federal Highway Administration, in cooperation with disaster preparedness agencies such as the Federal Emergency Management Agency and its state counterparts, state highway departments, and universities. Implementation should follow experimental testing by means of small-scale pilot projects in selected states. Should the tests prove successful, the system should be expanded to the United States as a whole.

While collecting data on structural damage, engineering researchers also must systematically develop damage models, so as to be able to predict failure processes for various kinds of structures and materials under extreme events. For long-term improvements in bridge safety, observations from destructive testing, forensic studies, reconnaissance studies, and bridge-health monitoring have to be integrated in predictive analytic models.

With such improvements, U.S. infrastructure investment would be more cost-effective on a risk-adjusted basis, because bridge decisionmakers would have more accurately accounted for the range of hazards and multihazard effects to which bridges are susceptible. In addition, although such a project may be initially developed for bridge safety, U.S. infrastructure as a whole would derive economies of scope from extending the system to meet broader needs for national protection.