The “Wall of Wind” at Florida International  University in  Miami can blow as fiercely as a category-5 hurricane.

The “Wall of Wind” at Florida International University in Miami can blow as fiercely as a category-5 hurricane.

Robert Sullivan

Read our COVID-19 research and news.

These disaster machines could help humanity prepare for cataclysms

For the past year, Tara Hutchinson has been trying to figure out what will happen to a tall building made from thin steel beams when “the big one” hits.

To do that, she has erected a six-story tower that rises like a lime-green finger from atop a shrub-covered hill on the outskirts of San Diego, California. Hundreds of strain gauges and accelerometers fill the building, so sensitive they can detect wind gusts pressing against the walls. Now, Hutchinson just needs an earthquake.

In most of the world, this would be a problem. Even here, where a major fault runs right through downtown, the last quake of any note struck 6 years ago and was centered in nearby Mexico. But Hutchinson, a structural engineering professor at the University of California (UC), San Diego, doesn’t need plate tectonics to cooperate. This summer she has an appointment at one of the world’s biggest earthquake machines.

This device—a sort of bull ride for buildings—is one in a network built around the United States over the past 15 years to advance natural disaster science with more realistic and sophisticated tests. Costing more than $280 million, the National Science Foundation (NSF) initiative has enabled scientists to better imitate some of the most powerful and destructive forces on Earth, including earthquakes, tsunamis, and landslides. 

The work has led to new building standards and better ways to build or retrofit everything from wharves to older concrete buildings. Scientists have gained insights into how quakes damage pipes in walls and ceilings and how to help quake-proof highway ramps, tall steel buildings, parking garages, wooden homes, and brick walls, to name a few.

That expansion continues today. In a new $62 million, 5-year program, the network of doomsday machines is expanding to simulate hurricanes and tornadoes and is joining forces with computer modeling to study how things too big for a physical test, such as nuclear reactors or an entire city, will weather what Mother Nature throws at them.

Scaling down disasters

Credit California's Northridge earthquake for helping set this in motion. The 1994 quake, centered near Los Angeles, killed 72 and cost an estimated $25 billion in damages. In its aftermath, a report commissioned by Congress warned that the country needed a more systematic approach to studying how to reduce damage from earthquakes. NSF responded with the $82 million Network for Earthquake Engineering Simulation. The money funded a construction spree at 14 sites around the country. Another $200 million paid for operating the sites through 2014. That included UC San Diego, which unveiled the world’s largest outdoor shake table in 2004.

Researchers at Oregon State University, Corvallis, unleash tsunamis in a wave basin.

Researchers at Oregon State University, Corvallis, unleash tsunamis in a wave basin.

© Aurora Photos/Alamy Stock Photo

Descriptions of these disaster labs are often couched in superlatives: the biggest, the longest, the most powerful. In addition to the San Diego facility, the projects funded under the original program and its successor, the Natural Hazards Engineering Research Infrastructure (NHERI), include North America’s largest wave flume for studying tsunamis at Oregon State University, Corvallis; the world’s largest university-based hurricane simulator at Florida International University in Miami; and, at UC Davis, the world’s biggest centrifuge for making scale models mimic the stresses on tons of buildings, rock, and dirt—crucial information for assessing how structures will weather earthquakes and landslides.

More than bragging rights is at stake. When it comes to learning how buildings cope with the forces generated in a natural disaster, size often does matter. For example, the way soil particles stick together, an important factor in landslide risks, depends on how much mass is pushing down on them. Similarly, it’s nearly impossible to build accurate, tiny versions of rebar: steel rods embedded in concrete structures that are critical to building performance. Similar difficulties arise with measuring how hurricane-force winds interact with a building.

“You can’t take a real building and scale it down to one-tenth and put it in a wind tunnel. The physics doesn’t work,” says Forrest Masters, a wind engineer at the University of Florida in Gainesville who directs his university’s share of NHERI. That includes a machine capable of subjecting 5-meter-tall walls to the air pressures found in a 320-kilometer-per-hour hurricane, and a wind tunnel whose floor can be modified to see how different terrain influences the way wind interacts with structures.

Computer models too can fall short in accurately reproducing all the forces at play as, say, a bridge twists and sways in an earthquake. So many different pieces in the bridge are pulled in so many directions at once that it can fail in unpredictable ways, causing models to misrepresent reality. In 2010, a contest at the San Diego shake table pitted 41 teams of experts running models against a real-life test of a 7-meter-tall bridge column topped with 236 metric tons of concrete blocks. The computer results were all over the place, says Stephen Mahin, a structural engineer at UC Berkeley who helped orchestrate the event. On average, they underestimated how much the column would sway by 25%. “You can’t quite trust the computer results yet,” Mahin says.

A building awaits its ordeal on the shake table at the University of California, San Diego.

A building awaits its ordeal on the shake table at the University of California, San Diego.

Erik Jepsen/UC San Diego

One morning in mid-May, Hutchinson inspects her building in the final stages of preparation for the test. She points to tiny gaps that have sprung open where metal ceiling joists meet the wall in a first-floor room. That happened during a minor, preliminary shake her team delivered to the building a day earlier. It’s the kind of thing that could make a difference in how load is shared between pieces of the building, and how much damage the building suffers in the next temblor. And it wouldn’t show up in a computer model.

“You’re not going to account for every screw,” she says. “Look at how subtle this damage is.”

Shake, rattle, and roll

Devising a machine that can pack the same wallop as a magnitude-8.0 earthquake or a category-5 hurricane isn’t easy, or cheap. A look under the hood of San Diego’s shake table illustrates the kind of mechanical muscle needed. Joel Conte, an engineering professor who oversees the shake table operations, leads the way into a cavernous under-ground room filled with machinery. A 20,000- liter metal tank holds the hydraulic fluid that drives the entire system. Two pumps slurp the fluid from there into a bank of 50 slender black cylinders reminiscent of street light poles at pressures reaching 34,000 kilopascals (5000 pounds per square inch). That high pressure is crucial, generating enough force to swiftly move an entire building.

Conte turns down a passageway, tracing the path of the fluid through steel pipes 30 centimeters across, and into a room dominated by a mass of steel resembling the hull of a flat-bottomed boat. This is the epicenter. A metal plate 5 centimeters thick, 12 meters long, and nearly 8 meters wide sits overhead, bolted to the steel underbelly. At either end, an actuator that looks something like a car’s shock absorber, but is as thick as a man’s torso, extends from this structure to the concrete wall. When the commands come from computers in a nearby building, the actuators will jerk to life, the hydraulic fluid driving them back and forth. The plate, pushed and pulled between them, will slide across metal sheets polished mirror-smooth at speeds of up to 1.8 meters per second. Voilà! Instant quake.

“The real world, you cannot count on it,” Conte says. “You cannot say, ‘Oh, I’m going to sit and wait for the next earthquake in front of this big building, and I’m going to invest a lot in sensors.’ You may have to wait 30, 40, 50 years. So you produce an earthquake.”

Since its construction for $10 million, the shake table has tested a four-story concrete parking garage, a wind turbine, and a five-story concrete building complete with elevator and stairs, among other things. The tests have shown that special inserts can increase resilience by allowing a building to move over its foundation and that modular concrete floors can behave erratically unless they have additional reinforcement. They have also revealed how tall, wood-framed buildings fail and how reinforcements can strengthen old brick buildings.

Back in his office, Conte gleefully clicks through the “best of” video highlights. A four-story wood building twists and splinters to the ground. A parking garage teeters back and forth like a rocking chair. A split screen shows two identical rooms filled with hospital beds and medical equipment. One is in a building outfitted with padded foundations that help it absorb an earthquake’s shock; the other isn’t. As the video runs, beds in the regular building suddenly lurch back and forth before toppling over. In the other, they barely move.

In the current test, Hutchinson wants to see how a building six stories tall made from lightweight steel performs during and after an earthquake. She thinks it could do well, partly because it’s lighter than a concrete building of the same height, giving it less mass to generate damaging forces during a quake. Today, building codes allow this type of construction to be just shy of 20 meters tall. But the tallest building really put to the test was only two stories high.

The structure, modeled after an apartment building, is destined for a multistage torture test. Hutchinson and her colleagues will first put it through a simulation of several quakes, including Northridge and a 2010 magnitude-8.8 in Chile. Then they will set fires in parts of the building to see how it holds up in a blaze triggered by quake damage. Then they will shake the building again in a mock aftershock, hard enough that it might collapse.

The results aren’t just of academic interest. Sponsors of the test include manufacturers of the steel construction parts, the insurance industry, and state government. “There’s nothing like a full-scale test,” says Richard McCarthy, executive director for the Cali--fornia Seismic Safety Commission in Sacramento, a government commission that advises policymakers. It contributed $100,000 to the event, he says, partly with an eye toward potential changes to building codes governing construction using these materials.

Conte is now lobbying state officials for a $14 million upgrade that would allow the machine to run even more realistic tests. Right now it can move only back and forth in two directions; new hardware would add up-and-down, side-to-side, and diagonal motions, enabling it to move in every direction—like the world’s biggest shake table, an indoor facility in Miki, Japan.

Up next: Hybrid simulations

Scientists are trying to go even bigger by marrying such physical tests with computer models. The resulting “hybrid” simulations can test massive structures too big to fit inside any test facility, says James Ricles, a civil engineer at Lehigh University in Bethlehem, Pennsylvania. His lab, which is part of the NSF network, tests well-understood parts of a structure with computer models but stages physical tests for parts that the models can’t handle. In a feedback loop measured in milliseconds, sensors from the physical test send data to the model, which adapts and sends new signals that tell the machines driving the physical test how to tweak their next moves.

Ricles’s lab simulated the behavior of an elevated highway during an earthquake by physically testing the concrete columns while testing a virtual model of the bridge deck in a computer. He recently applied the same strategy to testing a design meant to allow a steel building to rock back and forth rather than bend during a quake. A four-story chunk of the building stood in the lab; the rest of it existed only in the microprocessors of a computer.

Destruction is a definite part of the work’s appeal, says Gilberto Mosqueda, an engineering professor who runs hybrid tests at UC San Diego: “You build these models, and essentially you shake them till you break them.” But the mountains of data generated by the tests also open the way to more sophisticated numerical models that could one day do some of the work of the doomsday machines.

Whereas the earlier NSF program focused on big testing platforms, the NHERI initiative is putting more money into the virtual side. The University of Texas, Austin, won $13.7 million to build a data repository and software platform to store information from years of field tests. In the future, engineers should be able to tap data in the digital repository to boost the accuracy of their computer models. And NSF will soon issue an $11 million award for a computational modeling and simulation center.

“Will we get to the point where we can just model everything and have confidence? That may still be a long way off,” says Joy Pauschke, a structural engineer and director of the NSF program that funds the testing work in Arlington, Virginia. “But hopefully as we test and improve models, we start moving towards having better capabilities with the computational modeling.”

Berkeley’s Mahin—whose 2010 contest exposed the shortcomings of models—now also foresees bright prospects for modeling. Advances in machine learning and cloud computing, he predicts, will lead to models capable of simulating not just single buildings but entire communities. Unleashing “virtual disasters” could then enable researchers and government officials to grasp the region-wide effects of a major quake or storm and decide which measures today would prevent the most damage.

“In 20 years, you can model a whole city in a very complicated way, I think,” Mahin says. “There’s a great hope this analysis can help mitigate the damage from future natural disasters.”


Additional articles in our Natural Hazards feature package: