Because you’re a very responsible person who doesn’t text and drive, when you roll over a bridge your smartphone is stuck to the dash, where it is perhaps giving you directions while streaming a WIRED podcast. But in the background, your device is also gathering reams of accelerometer data. One day, this could help diagnose problems with the very bridge you’re speeding across.
Every bridge has its own “modal frequency,” or the way that vibrations propagate through it—then subsequently into your car and phone. (Tall buildings, which sway in the wind or during an earthquake, have modal frequencies too.) “Stiffness, mass, length—all these pieces of information are going to influence the modal frequency,” says Thomas Matarazzo, a structural and civil engineer at MIT and the United States Military Academy. “If we see a significant change in the physical properties of the bridge, then the modal frequencies will change.” Think of it like taking a bridge’s temperature—a change could be a symptom of some underlying disease.
In the US, much of the bridge infrastructure was built to support car culture after World War II, and it’s getting old and unsound. Irony among ironies: Earlier this year, a bridge in Pittsburgh collapsed hours before President Joe Biden was scheduled to visit the city to talk about infrastructure. A 2007 collapse in Minneapolis killed 13 and injured 145, and the 1993 failure of a railroad bridge near Mobile, Alabama, killed 47 and injured over 100.
To monitor for cracks, corrosion, and other defects, some bridges have expensive sensors that detect how their modal frequency changes. But the vast majority of spans around the world—there are some 600,000 highway bridges in the US alone—lack these sensors. (They’re not set-it-and-forget-it: It takes hundreds of sensors to cover a particularly long bridge, and you’ve got to swap out their batteries and download data every few months.) Instead, bridge operators rely on slow, labor-intensive visual inspections.
Engineers, then, need a better way of monitoring modal frequencies, ideally cheaply and in real time. In a new paper in the journal Nature Communications Engineering, Matarazzo and his colleagues describe how they used ordinary smartphones in passing cars to accurately estimate the modal frequency of the Golden Gate Bridge. That could pave the way (sorry) for a future in which thousands of phones going back and forth across a bridge could collectively measure the span’s health, alerting inspectors to problems before they’re visible to the human eye.
The researchers began with a controlled experiment, in which they collected data by driving across the Golden Gate Bridge with smartphones on their dash. They knew all the variables: What kind of car they were in, their speed, their location at any given time, and where exactly the phones were in the car. As they drove, the phones collected data from their accelerometers, which measure movement—in this case the car’s vibrations. This allowed the researchers to accurately measure the modal frequency of the bridge, which matched data from traditional sensors that had already been deployed along the span.
Next, to get more loosey-goosey with it, the scientists asked the rideshare app Uber for accelerometer data from its drivers as they crossed the Golden Gate. “I don't know what type of phone they have, I don't know what type of car they have, I have no idea how fast they're going,” says Matarazzo. Still, that data was also able to accurately estimate the modal frequency, compared to the bridge sensor data. “Even if you cannot really precisely control the aspects of the trips for the data being collected, you can still get this information, which is really remarkable,” he says.
That’s because, while the Uber data isn’t necessarily consistent—the drivers are using many different vehicles and phones—the Golden Gate Bridge is. Its modal frequency breaks through the noise of those other variables. “The diversity of the data can actually be hugely beneficial in making those individual effects a little bit less pronounced,” says Matarazzo. “If we're thinking about collecting hundreds of thousands of data sets, we're going to have a huge distribution of vehicles, and the underlying characteristic between all those data sets would be the vibrations from one common bridge.”
But if measuring a change in the modal frequency is like figuring out that the bridge is running a fever, engineers still have to figure out what’s causing it. That may be tricky, because the natural environment can also influence the frequency by changing the state of steel and concrete—for instance, when outdoor temperatures rise and fall. But Matarazzo notes that previous research has shown that it’s possible to isolate and account for that signal.
With every passing year, the need to properly monitor these thousands of aging bridges grows. Some are getting well beyond their predicted lifespans, and all of them are now burdened with more—and bigger—vehicles, like the legions of trucks delivering our online orders. “It's like you're asking an 80-year-old person to carry more load than he used to carry when he was young,” says University of Alabama at Birmingham engineer Nasim Uddin, who researches the use of smartphones to detect the modal frequencies of bridges. “That's why bridges are collapsing everywhere. Unless you have a system like this, I think we're not going to be able to handle it.”
But, says Matarazzo, if you’re concerned that municipalities are going to start harvesting your smartphone’s location and accelerometer data to monitor their bridges, worry not. His team envisions using city vehicles, like police cars, to do the data collection. Rideshare companies like Uber and Lyft might provide data from their drivers, while logistics companies might allow researchers to tap into information from semi trucks and other large vehicles.
That said, it’s also feasible to analyze information from passenger cars. Vehicles are increasingly loaded with accelerometers, which feed data into active suspension systems, for example. “The vehicle is becoming ever smarter,” says University College Dublin structural engineer Eugene O'Brien, who studies bridge monitoring but wasn’t involved in the new paper. But, he points out, that raises data privacy and ownership questions. “There's issues about how to share the data—how do you get that data back to the road owner to tell them about the condition of their infrastructure?” It might require an agreement between bridge owners and vehicle manufactures—who’d ideally ask for the consent of drivers first.
Cities might even develop systems to compensate drivers who opt in to providing data: If you’re helping with the maintenance of spans, and therefore saving a government money, maybe you’d get a discount on bridge tolls. “So that gives them this extra stake: Hey, I rely on this bridge, but I'm also collecting data that's going to support this bridge,” says Matarazzo. “That's going to help it work better today, tomorrow, and for future generations.”
