Rocking the House

Seismic tests of a life-size, wood home

On Jan. 17, 1994, about 2 hours before sunrise, a magnitude-6.7 earthquake struck Southern California’s San Fernando Valley. The temblor, which originated beneath Northridge, occurred along a previously inactive and unknown fault zone. In a quake that generated an estimated $10 billion in damages—the United States’ costliest earthquake to date—dozens of bridges collapsed, hundreds of buildings were destroyed, and 57 people died.

MOTION PICTURE. The blur in this image results from the shaking of the townhouse, not the camera. Researchers simulated a magnitude-6.7 quake to better understand how wood-frame structures behave in such temblors. SUNY Buffalo

DORM DISASTER. The contents of this room, furnishings appropriate to a college dorm, were tossed about within the first seconds of the simulated quake. SUNY Buffalo

ALL CRACKED UP. Structural damage around the townhouse’s garage door was severe, which is typical for such wide, unsupported spans. SUNY Buffalo

Last month, that quake struck again—this time, it rumbled through a corner of a lab at the State University of New York at Buffalo. In a building the size of an airplane hangar, scientists used computer-controlled hydraulic equipment to recreate the ground motions of the Northridge quake. The team applied the tremors to a two-story wood-frame townhouse that members of a five-university consortium had constructed atop two large platforms called shake tables.

“This is the first time that a wooden building this size has been [seismically] tested anywhere in the world,” says Andre Filiatrault, a civil engineer at the university.

November’s experiment, the latest in a series of seismic shakedowns conducted by the researchers, will shed light on how wood-frame structures behave in earthquakes, says Filiatrault. Results of these tests and of an even larger test scheduled for 2009 may lead to improvements in construction techniques, building materials, and building codes—all of which could end up saving lives.

Model home

Of the 24 people who died in buildings during the Northridge quake, most perished in wood-frame structures. Sixteen of those fatalities resulted from the collapse of a single apartment building. In that structure, much of the floor space at ground level was occupied by garages. The walls, weakened by the garage’s large doorways and the open space within the structure, didn’t provide strong support for the living areas on the upper floors, permitting them to move back and forth and quickly collapse, flattening the building’s bottom level.

While that particular architectural design is no longer used for new construction in California anymore, plenty of older buildings have such open spaces at ground level. A garage, even with reinforced walls, is one of the weakest areas in a structure, says Filiatrault.

More than 80 percent of buildings and more than 90 percent of the residences in the United States have wood-frame structures, but engineering students are offered few courses about the design of such structures, says Filiatrault. Furthermore, “scientists don’t really understand how wood-frame buildings perform in a quake,” he says.

When researchers do test such structures, they typically use small-scale models, says John W. van de Lindt, a civil engineer at Colorado State University in Fort Collins. However, those models, which have been used in seismic simulations for decades, aren’t perfect stand-ins for the structures they’re meant to represent. For one thing, scaled-down versions of wooden buildings are in many ways stiffer than the real things, he notes. If researchers try to correct for that trait by loosening connections between structural components in a model, then the structure may bend and flex correctly but not vibrate at the appropriate frequencies.

It’s particularly difficult for scale-model tests to accurately depict the damage that a full-size wooden structure will experience. Although researchers can carve a miniature version of a wooden two-by-four, they can’t scale down the size of the wood cells in that board.

Van de Lindt, Filiatrault, and their colleagues avoided such problems by building the real thing: a 1,800-square-foot townhouse. They followed construction methods typical of those used in Southern California during the early 1990s.

The full-scale model had a skeleton of two-by-fours, interior walls clad with drywall, and external walls sheathed with large sheets of glued-together wood chips and covered by three layers of stucco. The three-bedroom, two-bath home, with a small alcove leading to a patio was designed to represent the central dwelling in a three-unit row of townhouses. Its construction followed the California building codes in place when the Northridge quake occurred. Many of the houses that people live in today were built during that era. Many of the houses damaged by the Northridge earthquake, by contrast, had been built at a time when building codes were less stringent.

The test townhouse didn’t have plumbing and had only a few runs of electrical wiring for lighting fixtures. “Neither of those [features] adds much structural integrity in a real home,” explains van de Lindt.

To ensure that the structure didn’t collapse completely and thereby endanger people or demolish the lab, engineers had wrapped broad straps around several major structural components. Thus restrained, the building could lean no more than 40 centimeters at a height of 2 meters. At that angle, the house probably would have collapsed.

By building their own townhouse, the scientists ensured the quality of their test object. Although researchers can dissect a rubble pile after a disaster, they usually can’t tell how well the structure had been built to begin with, says van de Lindt.

To gather quantitative data, the scientists installed a bevy of sensors in the test home, including 50 sensors to measure forces experienced by components within the structure, 75 to monitor accelerations caused by the shaking, and 125 to measure how far various parts of the structure moved back and forth. Eight video cameras inside the house and four outside recorded what happened to the house and its contents.

Beyond testing the townhouse’s structural response to the Northridge quake, the researchers wanted to examine the effectiveness of strapping down bulky items. So, the team—under the guidance of Louise Filiatrault, Andre Filiatrault’s wife—fully furnished the townhouse. She was in charge of “collecting the items donated by students and scouring yard sales for the rest,” she notes. The night before the seismic test, she and the Filiatraults’ children tidied the home, making sure that flower arrangements and lamps were placed just so. They also poured water into the glasses on the dining room table.

As for Andre Filiatrault, the afternoon before the seismic test, he expressed concern about the damage the townhouse might suffer. “I don’t think the house will fare too well,” he predicted.

What’s shaking?

The townhouse replicates a type of construction called slab-on-grade, in which a home is built on a concrete platform that rests on soil. The horizontal wood pieces, or sill plates, placed at the base of the structure’s walls are bolted to the foundation, and the rest of the structure is built on that base. (For time-lapse video of the townhouse’s construction in the lab, see

In the lab, the 40-ton, 17-m-by-6.7-m townhouse was attached to a concrete slab that was built atop and attached to the two side-by-side shake tables. Eight hydraulic pistons, each the diameter of an adult’s thigh, moved the tables back and forth and up and down.

The researchers directed the tables’ movements to simulate the Northridge quake, as described by seismometer data gathered about 6 kilometers from the quake’s epicenter. “These are some of the strongest ground motions recorded anywhere in the United States,” says van de Lindt. Horizontal accelerations during the Northridge quake reached 80 percent of the acceleration due to Earth’s gravity.

In the half-hour before the quake simulation, the researchers conducted two preliminary tests, each a 3-minute series of small vibrations at various frequencies. One set of these simulated ground motions shook the townhouse from front to back, and the other drove the structure from side to side. Data gathered during these tests will enable the scientists to identify the structure’s resonant frequencies—the frequencies at which small ground motions can cause abnormally large structural flexing. Architects and engineers try to avoid designing structures with resonant frequencies at or close to those of the strongest earthquake-generated vibrations.

Then came the big shake. After a few small tremors, the simulated quake struck hard, and anything that wasn’t fastened down was on the move. Flower boxes leaped from their perches beneath second-story windows. In the dining room, the video cameras revealed that the chandelier danced, the water glasses fell over, and the place settings slid back and forth on the table as if an incompetent magician was practicing the old tablecloth trick.

In the bedroom, which was furnished college-dorm style, the cinder block–and-board bookcase toppled to the floor. So did the dresser, its contents, and the lamp and computer monitor on the desk. The intense shaking hurled a television more than 3 m across the room. That and the other heavy items could have badly injured anyone in the room during a real quake.

While inspecting the damage after the test, a local radio reporter said, “This looks just like my teenager’s room.”

In another bedroom, the story was different. A tall bookcase, which had been attached to the wall with brackets, didn’t fall, although its contents of stuffed animals, children’s books, and other bric-a-brac tumbled out. A small television, chained to a table that had been bolted to the floor, likewise stayed in place.

In the garage, a station wagon hopped around like a lowrider. One water heater, strapped to the wall as has been required for more than a decade in new construction by California law, held its ground. Another heater—unconstrained, as those in many older homes are—fell over in the first 2 seconds of the quake. Leakage of water is the least of a home owner’s worries, says van de Lindt; rupture of the gas line leading to the device poses a much greater risk.

While most of the structure’s contents got tossed around, the building didn’t collapse. In fact, it withstood the quake better than expected, says Filiatrault. “I’m surprised the windows didn’t break,” he notes.

George Digman, director of research and development at Kolbe & Kolbe Millwork, the Wausau, Wis.–based company that provided all the windows installed in the test article, concludes: “The test was dramatic but anticlimactic.”

As anticipated from analyses and from real-world experience in homes, the walls and frame surrounding the garage door were severely damaged. The townhouse’s coating of stucco fractured in many places, and the video camera mounted in the first-floor home office showed that a thick cloud of white dust erupted when drywall in the room cracked.

Nevertheless, during the quake, the walls of the home flexed back and forth at an angle of about 3°. So, at a height of about 2 m, walls shifted about 10 cm one way and then the other. After the test, the resting structure leaned in one direction at an angle of about 1°.

That might not sound like much, but Filiatrault worries about the building’s integrity. “If I lived in this home in California, I think I’d camp out in the back yard until an engineer could look at it,” he says.

Product testing

A cursory look at the test townhouse just after the Northridge simulation revealed that the sill plates were cracked at many points where they attached the foundation to the rest of the building. Because of that fundamental damage, the researchers say that they won’t do any more shaking of the townhouse. It might detach from the shake tables if put in motion, they say.

Filiatrault and his colleagues are now taking the townhouse apart bit by bit, looking at structural components that normally are hidden, to see how they fared during the quake. Such information could lead to better designs, for instance, of connectors used to attach walls, beams, and sill plates to one another, says Steven E. Pryor, research and development manager at Simpson Strong-Tie, a Pleasanton, Calif.–based firm that donated materials for construction of the test structure. While such components are essential to a building’s structural integrity, he admits, “It’s tough to get customers to think about them when they’re thinking about granite countertops.”

Besides the information gathered during the simulated quake and the ongoing teardown, the researchers have reams of data that they collected during almost three dozen preliminary, mild shake tests conducted at several times during construction, says van de Lindt.

Results of those tests will enable researchers to isolate the structural contributions of the various materials. The data already reveal that drywall on interior surfaces of outside walls contributes to a house’s strength. The tests also settle a long-standing debate about whether stucco can provide some structural support for a building. The answer is yes, says Filiatrault.

Last month’s test is the culmination of the first year of a 4-year project, says van de Lindt. He and his colleagues will spend the next 2 years scrutinizing the data they’ve collected and developing computer software to enable engineers to better analyze and design wood-frame structures.

The team plans to put that software to the test by designing and building the major components of a six-story wood-frame structure that will be assembled and tested on a huge shake table in Japan in 2009.

Beyond improving wood-frame design-and-analysis techniques, the test results may inform engineers when they consider revising building codes, which typically are reviewed every 5 years or so.

“More than 100,000 people lost their lives in earthquakes in the 20th century,” Filiatrault notes. “Maybe this test will save some lives in the future.”

More Stories from Science News on Earth

From the Nature Index

Paid Content