This article was originally published in Science Progress.
In August, an eight-lane interstate bridge in Minneapolis collapsed during evening rush hour, killing 13 people and injuring 144. This collapse, and the failure to anticipate it, calls into question the adequacy of current bridge inspection methods. Why were problems with the bridge not identified? And if problems were missed in Minneapolis, could they be missed elsewhere? Could this happen again?
There is good reason to worry. Before it collapsed, the Minneapolis bridge was one of more than 70,000 bridges nationwide declared by the Department of Transportation to be structurally deficient. One in three urban bridges fall into this category.
Such bridges may be safe for travel so long as they are carefully monitored. Recent advancements in sensor technology provide the opportunity to collect detailed, real-time data on bridge performance. But this technology is being used on less than a handful of bridges nationwide. Current inspection methods, unfortunately, cannot be relied on to catch a bridge on the brink of collapse.
“We do not know which bridges should be taken out of the system, and which should be maintained,” said A. Emin Aktan, a professor of civil engineering at Drexel University and director of the Intelligent Infrastructure and Transportation Safety Institute.
Every two years, each government-owned bridge is required to receive a “routine” inspection, in which technicians or engineers observe the bridge and take measurements of its physical condition. Underwater structures, meanwhile, must be inspected by divers every five years. There are guidelines but no requirements for “in-depth” inspections, which can include things like probing of the bridge, laboratory analysis of bridge material, and testing of surrounding environmental and water conditions.
This heavy reliance on visual inspection is inadequate for three major reasons. First, inspections are susceptible to human error. Indeed, a 2001 study by the Federal Highway Administration found that inspectors regularly missed problems and inconsistently rated bridge conditions. Second, there are long intervals between required inspections, during which time serious problems may emerge. And third, inspections may be superficial and might not produce the detail necessary to spot deficiencies.
This is not to say that visual inspection is unimportant—visual inspection is crucial to assess bridge conditions, in particular cracks and corrosion. But more is needed to assure the safety of the nation’s bridges.
That’s where sensor technology comes in. Instead of relying on sporadic and error-prone observations, matchbox-sized wireless sensors can be attached or embedded on bridges to take precise, continuous measurements of virtually anything relevant to a bridge’s condition, including strain, tilt, vibrations, temperature, and seismic activity. This sort of data is particularly important as the nation’s bridge population ages—the mean bridge age is now 40 years old—and traffic and truck loads continue to increase, causing more rapid deterioration.
The Minneapolis collapse has created a political opportunity to modernize bridge monitoring. In its aftermath, Secretary of Transportation Mary E. Peters initiated an ongoing review of the agency’s bridge inspection program to, in her words, “make sure that everything is being done to keep this kind of tragedy from occurring again.”
Congress, meanwhile, is also engaged in finding solutions. Rep. James Oberstar (D-MN), chairman of the House Transportation Committee, is developing legislation to significantly improve bridge inspection requirements as part of “a data-driven performance-based approach to systematically address structurally deficient bridges on our nation’s core highway network.”
Sensor technology can help meet the goals expressed by Peters and Oberstar. What’s needed now is a plan to move forward.
First Steps for Sensor Technology
Recently, the Federal Highway Administration awarded funding to the Connecticut Department of Transportation and the University of Connecticut to deploy and study different types of sensor systems for long-term bridge monitoring.
“The goal is to generate information between inspections, so that if there’s a major change, we can take action to prevent something catastrophic from happening,” said project leader John DeWolf, a professor of civil engineering, who became involved in bridge monitoring following the 1983 collapse of the Mianus River Bridge on Interstate 95 in Greenwich, Connecticut.
Over the last several years, six bridges in Connecticut have been outfitted with unique sensor systems. Five of these are wired systems, in which cables connect the sensors to a computer. The sixth relies on solar-powered wireless sensors. This wireless system is particularly exciting because it holds great promise to be more widely replicated.
It can take a great deal of labor and expense to run cables over a bridge—especially one that is large and difficult to access. For a wireless system, however, cables are not an issue. Sensors merely need to be placed in desired locations on the bridge. Installation typically takes no more than a few hours, at a cost less than half that of a wired system.
Because of these advantages, DeWolf decided to go wireless for Connecticut’s longest bridge, the Goldstar Bridge, which crosses the Thames River on Interstate 95 in New London. Like all new technologies, wireless sensors are expected to get much cheaper over time. But even now they are affordable. Installation of 12 sensors at the Goldstar cost about $30,000.
Over the long run, sensors may even pay for themselves by more precisely identifying when and where repairs are needed. Ten wireless sensors were recently used to test stress levels from passenger trains on the Ben Franklin Bridge, which crosses the Delaware River from Philadelphia to Camden, New Jersey. The state believed the bridge was in need of major repairs based on advice it received from an engineering consultant. But data gathered by the sensors showed the bridge was in fact secure, saving tens of thousands of dollars in unnecessary repairs.
Sensors can also reveal problems as they emerge—before there is visual evidence such as cracking. This allows remedial action to be taken in time to head off serious structural damage, which can be very expensive to repair. “If you get to it quickly and fix it, it’s not going to be a major problem,” said Mike Robinson, vice president for sales and marketing at MicroStrain Inc., which developed the sensors for the Ben Franklin Bridge. “You can reduce the overall life-cycle cost of the bridge.”
DeWolf approached MicroStrain to develop the solar-powered sensors specifically for the Goldstar. The sensors used on the Ben Franklin were powered by batteries—fine for short-term testing, but not long-term monitoring. Batteries eventually run out of power and then need to be changed or recharged, which is a difficult task on a bridge like the Goldstar, where sensors are in hard-to-reach locations.
The solar-powered system relies on photovoltaic panels to harvest energy from the sun. These panels are connected to the sensors to supply power for daytime monitoring and recharge batteries for overnight observation. This system is expected to generate power for years with little or no maintenance. MicroStrain is also developing other solutions for long-term power, including mini wind turbines and super efficient battery-powered sensors, according to Robinson.
MicroStrain first installed its solar-powered system on the Corinth Canal Bridge in Greece to monitor seismic activity. There, the sun is strong enough for continuous monitoring, which is crucial given the unpredictability of seismic activity. At the Goldstar, where the sun is not as bright, data are gathered for 5 to 10 minutes every hour to conserve energy. For what’s being measured, strain and vibrations, this is considered plenty sufficient.
The data collected are temporarily stored on the sensors and then downloaded daily to an onsite laptop computer. From there, the data can be remotely accessed through a DSL connection. Of course, it is not possible to manually analyze the voluminous amounts of data generated. Instead, automated systems are programmed to comb through and pick out relevant information for DeWolf and his team to review.
Ultimately, this information can help confirm whether the bridge is safe. Vibrations, for example, can be monitored to ensure that they do not exceed potentially dangerous thresholds. For the vast majority of the nation’s bridges, this sort of information is not available. Indeed, Connecticut is now the only state using sensors for long-term monitoring of multiple bridges. Other states rely on the same visual inspection methods that failed in Minneapolis.
“Let’s not debate that visual inspection has proven insufficient,” Aktan said. “Instead, we should focus on strengthening bridge monitoring, so that one day there will be little worry about another bridge collapsing.” Wider use of wireless sensor technology is an important part of the solution.
Building a Nationwide Sensor System
In the aftermath of Minneapolis, public attention is now appropriately focused on detecting an imminent collapse. Thousands of the nation’s bridges are badly in need of repair. The possibility that one might collapse is very real.
Installing sensors on all of the nation’s 70,000 structurally deficient bridges, however, is not practical or even desirable. Within the “structurally deficient” category, there can be vast differences among bridges. Some bridges may be quite safe, in need of relatively minor repairs, while others may have major problems that should be addressed immediately.
The Federal Highway Administration, unfortunately, does not systematically identify priorities among these bridges. Nor are bridges of greatest concern necessarily given more attention. Rather, each bridge is subject to the same biannual requirement for visual inspection regardless of physical condition.
It is thus paramount that more detailed categories be developed that group bridges by degree of concern. High priority bridges, of course, should be repaired as quickly as possible. But repairs may take time to complete, or funding may not be immediately forthcoming. In the meantime, sensors could be deployed to provide more careful monitoring and help further refine priorities for repairs.
Sensors, however, should not only be installed on the very worst bridges. Ideally, they should be used to assist routine maintenance, so that bridges never get to the point of imminent collapse. This requires a system to smartly and economically deploy sensors to monitor the nation’s entire bridge population.
The first step in this process is to classify bridges according to type. A suspension bridge like the Brooklyn Bridge obviously has different characteristics than a truss bridge like the Goldstar and the I-35 bridge that collapsed in Minneapolis. But even bridges of the same general type can have critical differences. Truss bridges, for example, employ a variety of bracing designs, may or may not use pins to connect joints, and may carry traffic on the top, middle, or bottom of the structure.
Bridges will deteriorate in different ways and at different rates depending on such variables. Currently, however, the nation’s bridges are not carefully categorized by similar design features. This information is needed to determine which bridges to outfit with sensors.
Because similar bridges can be expected to perform alike, it is necessary to install sensors only on a sample from each category. Again, this sort of sampling is not part of the current monitoring system—each bridge is subject to the same biannual inspection. “Looking at each bridge as an individual is ridiculous,” Aktan said. “There are tremendous similarities between certain types of bridges, but we don’t leverage knowledge about those similarities.”
The Federal Highway Administration recently launched an initiative—the Long-Term Bridge Performance Program—that begins to move in this direction. The goal of the program is to generate “high-quality, quantitative performance data” based on a representative sample of the nation’s bridges, likely numbering 500 to 1,000 bridges representing the majority of structure types. This includes data on deterioration and its causes—traffic load, corrosion, fatigue, and weather, among others—as well as the effectiveness of maintenance strategies.
As part of its data-gathering efforts, FHWA intends to subject the bridges in its sample to detailed periodic evaluations, over at least a 20-year period, using sensor technology and other state-of-the-art monitoring tools. In addition, a subset of bridges in the sample will be instrumented to permit continuous monitoring, while decommissioned bridges will undergo forensic autopsies.
Congress created this program under legislation enacted in August 2005, with funding authorized through FY 2009. FHWA requested $20 million a year, but will have to operate with only about $5.5 million a year over the first four years. Thus, decisions must be made over which parts of the program to launch immediately and which to postpone pending higher levels of funding.
The initiative will be especially valuable in determining what data to collect and what the data means. In particular, it is not always clear what and where to measure. If sensors measure the wrong things or are placed in the wrong spot, they may miss critical deficiencies. FHWA’s research will begin to identify key factors and pressure points in the deterioration of different types of bridges. “A doctor knows where to take a patient’s pulse,” Aktan said. “We need to know where to take the bridge’s pulse.”
A critical part of this process is knowing how to interpret the pulse, so that sick bridges are diagnosed and treated. The vast majority of bridges lack baseline performance data—that is, data collected at the time they were built—from which to judge deterioration over time. Without this information, there is uncertainty about a bridge’s optimal performance and exactly what constitutes poor performance.
FHWA intends to address this problem by comparing newer and older bridges of similar type to identify and predict life-cycle changes. This should bring into sharper focus the large amounts of data generated by sensors. “The problem we have now is making sense of this data,” said an FHWA engineer involved in the Long-Term Bridge Performance program. “That’s what we are trying to address. Determining the sample of bridges is the most critical step.” FHWA has already developed methodology to identify bridges for the sample. But the final selection will not be made until after a prime contractor is chosen, expected within the next couple of months, to oversee the program’s day-to-day operations.
FHWA’s research deserves the full support of Congress and the administration. The amount currently appropriated is barely enough to get off the ground. One enduring problem, unfortunately, is the tendency of Congress to fund transportation research through earmarks to specific universities or private contractors. These earmarks sometimes go to worthy projects, but frequently they are awarded according to political considerations rather than merit.
Moreover, because funding is disjointed and somewhat arbitrary, transportation research is not well integrated and coordinated. The Long-Term Bridge Performance Program can help add cohesion by drawing together information generated by disparate research efforts, including other FHWA-funded initiatives such as the sensor project in Connecticut. “We will try to piggyback on other research projects and make them fit into this national approach,” the FHWA engineer said.
Confronting the ‘Infrastructure Crisis’
For years, the president and Congress have repeatedly deferred needed maintenance of bridges in favor of other budgetary priorities. This shortsightedness will cost the nation far more in the end, as the scale and severity of needed repairs balloon and become impossible to ignore. For Minneapolis, the Department of Transportation released $55 million in emergency funds and Congress authorized $250 million for rebuilding.
Other critical infrastructure—including roads, dams, and levees—are similarly deteriorating and could also benefit from enhanced monitoring through sensors. Substandard road conditions contribute to 30 percent of all fatal highway accidents, according to the FHWA. More than 3,500 dams are unsafe or deficient, many of which may not hold during significant flooding or an earthquake, according to state inspectors. And nearly 150 of the nation’s levees pose a high risk of failing during major flooding, according to the U.S. Army Corps of Engineers. The American Society of Civil Engineers, which gathered these statistics, terms the current situation an “infrastructure crisis.”
The Minneapolis bridge collapse provided dramatic evidence of this crisis. But it was by no means an isolated event. Just last year, for example, an earthen dam in Kauai, Hawaii gave way and let loose nearly 300 million gallons of water, killing seven people. In late 2005, a 120-ton concrete beam fell from a bridge in Pennsylvania onto Interstate 70. And of course, the levees in New Orleans were not only breached during Hurricane Katrina, but structurally failed.
It is crucial that investments are made to upgrade the nation’s crumbling infrastructure. In the meantime, however, more failures should be expected. The question now is whether we will be able to anticipate these failures in time to head off disaster. Sensor technology, if effectively implemented, would give reason for hope.