At the height of rush hour, on the evening of Aug. 1, 2007, an eight-lane steel truss bridge over the Mississippi River in Minneapolis suddenly collapsed. Dozens of vehicles plunged into the water, and 13 people died. Eric Hernandez, an expert on structural engineering at the University of Vermont, wants to make sure this doesn’t happen on another bridge.
Combining novel algorithms with existing sensor technologies, he’s developing new, lower-cost techniques to interpret the vibrations in bridges and buildings. His goal is to create affordable tools for engineers and regulators to more accurately forecast the remaining life of a structure — whether it’s a decades-old bridge or an earthquake-shaken building.
To support his research, the National Science Foundation granted Hernandez, an assistant professor in the College of Engineering and Mathematical Sciences, a five-year, $500,000 Faculty Early Career Development Award (CAREER) that will begin June 1.
Health tests
“We want to know: can we accurately estimate when a structure will fail? Then we’d be able to step back and say, ‘it has between 12 and 14 years left of service. We should plan now so that in 10 years it has been replaced,’” Hernandez says. “ Right now that information is often not known. And you can have cases like the Minneapolis bridge which collapsed dramatically without warning.”
Engineers had inspected that bridge annually and were concerned about its condition — it was slated to be replaced in 2020. But they didn’t expect — nor were they able to forecast — its catastrophic failure.
“Instead of just relying on visual inspections, we could be using sensors,” Hernandez says, to assess the structural health of “buildings, bridges, tunnels, and wind turbines — early, before there is trouble — like people get tests when we go to the doctor."
Hernandez and his students have been working in a UVM lab, and — in partnership with the the Vermont Agency of Transportation — studying a (healthy!) bridge in Vermont to test these ideas. “We're looking at how structures vibrate as different kinds of loads are acting on them,” Hernandez says, “and from those vibrations we can estimate their level of safety or reliability.”
Vibration-sensing technologies on the market today are very good, Hernandez says. “There are sensors with enough accuracy to do the work we need to do,” he says. But they’re not cheap. “The problem is that the number of sensors that we can put on structures is limited because of cost,” he says. That's why he’s focused on what’s called a “minimally instrumented” approach to measuring a structure’s remaining life. “We’re proposing to determine what is the minimum number of instruments that you need,” he says, using more sophisticated computational techniques for interpreting the sensor data that is collected. “It's an algorithmic contribution,” Hernandez says.
Fatigue life
Hernandez is exploring his new approach at both ends of the wear-and-tear spectrum. On one end are seismic loads — think earthquakes — on buildings. “The number of load cycles that it takes to break buildings is often low, on the order of 10 or 20,” Hernandez says. On the other extreme are the loads that car and truck traffic make on bridges. “Imagine a paperclip you bent — it's not evident when it will break due to fatigue. It's the same thing with bridges, but instead of three or four bends, it's millions of cycles. But at some point the fatigue life of the bridge will be exhausted,” Hernandez says, “and we want to be able to track that fatigue as the structure is operating.”
Eventually, Hernandez’s new approach could be packaged into software “that could be coupled with different kinds of sensors that people would put onto structures," whether built into a new wind turbine or attached to a hundred-year-old bridge. "Then the software will do all the analysis,” he says. “Fatigue is responsible for about 75 percent of all structural failures. That's why this is an important problem."
