Logo E-Lab
April-June 1997 Issue

Earthquakes: How Can We Design Structures Less Likely to Fail ?

Recent earthquakes in Japan and in California have been catastrophic. In 1995, an earthquake in Kobe, Japan, killed 5,500 people, injured thousands more, destroyed buildings and other structures, and caused about $200 billion in direct economic losses. In 1989 and 1994, earthquakes in California also caused widespread damage, some of it at locations 150 km from the epicenter. In all three cases, destruction far exceeded that predicted for earthquakes of such magnitudes. Damaged buildings included many that met the strictest earthquake-resistant building codes in the world. Indeed, observations of the damage showed an interesting pattern: freeways, bridges, shopping malls, and parking garages-most of them built according to code-collapsed, while nearby skyscrapers and even some older, smaller buildings sustained little or no damage. Today's building codes are clearly inadequate. Revising those codes and developing techniques for retrofitting existing structures are critical before the next destructive earthquake strikes.

The failure of today's building codes shows that--despite decades of research--we still do not fully understand how earthquakes affect structures. Developing such an understanding is difficult. Theoretical techniques can predict how a structure will behave during an earthquake, but only initially. Once the structure begins to deform substantially, the relationship between the imposed forces and the deformation changes, and theory fails. To gain insights, investigators have tried monitoring the behavior of a scaled-down structure placed on a shake table that replicates the ground motions of an earthquake. However, it is difficult to extrapolate findings from such tests to full scale because it is impossible to accurately scale down simultaneously all the necessary parameters (material properties, stresses, weight, and so on). Only tests with full-scale or near-full-scale structures can yield meaningful results.

For the past two years, MIT researchers have been examining the feasibility of building a shake table that could subject a full-scale building or other structure to the forces of a real earthquake. Performing the research are collaborating teams of experts in seismology; civil, electrical, and mechanical engineering; and electromagnetic energy systems. Leading the teams are M. Nafi Toksoz, professor of geophysics and director of the Earth Resources Laboratory; Professor Eduardo Kausel, professor of civil and environmental engineering; and Emmanouil A. Chaniotakis, research scientist in the Plasma Science and Fusion Center. The researchers also work with collaborators at the Idaho National Engineering and Environmental Laboratory (INEEL), the proposed location of the full-scale shake table. The table is intended to be part of INEEL's proposed "Advanced Combined Environmental Test Station" (ACETS), a grouping of facilities that will test the response of full-scale structures to three types of natural threats: earthquakes, winds (hurricanes and tornadoes), and aging due to environmental factors such as humidity, salt spray, and solar radiation.

The first step in designing a full-scale shake table is to understand the ground motions that it must replicate. Professor Toksoz and his coworkers have analyzed data on peak ground accelerations (PGAs) measured during 15 significant earthquakes worldwide. (Strong-ground-motion instruments generally measure accelerations; velocities and displacements must be calculated from the accelerations.) Their analysis led to some unexpected observations. First, the maximum PGAs in recent earthquakes were much larger than traditional earthquake models predict. Moreover, earthquakes with similar magnitudes did not necessarily yield the same ground motions. In fact, for earthquakes with similar magnitudes at similar locations, the maximum PGA varied by factors of 2 to 5. Further analysis showed an interesting trend: as the number of measurements increases for a given size earthquake, the maximum PGAs tend to be higher. There must be isolated high local motions that are identified only when sensors are highly concentrated. The components of motion were also unexpected. At some stations the vertical accelerations--usually thought to be less important--were about the same as and occasionally greater than the horizontal.

Perhaps most surprising was the dramatic variability of ground motions over small distances. Seismologists have long recognized that waves from an earthquake can get absorbed, reflected, and redirected by geologic structures within the earth, leading to a variation in surface ground motions from place to place. Examination of recent data from closely spaced sensors shows that variations up to fivefold can occur at sites as little as 100 m apart, even for locations as much as 100 km from the epicenter. Data from the 1994 earthquake in Northridge, California, shown in the figure below, demonstrate some of those findings.

1994 Earthquake in Northridge, California: Predicted Versus Measured Peak Ground Accelerations

Such observations explain the shortcomings of current building codes during recent earthquakes. First, the ground motions were greater than expected, thus greater than the building codes are designed for. The vertical motion was significant, yet the codes focus on withstanding horizontal motions such as sliding. And most important, the measured ground motions varied over distances comparable to the "footprints" of large structures such as freeway spans, parking garages, and shopping malls. That finding could explain the widespread damage to such structures. If, for example, all the supports of a bridge move in the same direction, damage may be minimal. However, if one corner of the bridge twists or slides in a different direction from another, the likelihood of damage or collapse greatly increases.

Examining how such motions affect structures requires a shake table unlike those now available. It must be able to carry larger loads--preferably a full-scale structure--and impose stronger forces and larger accelerations with precision. Moreover, it must be able to create complicated motions, with different forces imposed at different points on the structure being tested.

Dr. Chaniotakis and his coworkers have developed a conceptual design for an electromagnetic seismic simulator (EMSS) that can meet those requirements. The EMSS is 30 x 30 meters in surface area and can hold a 10-story building weighing 1000 tons. It can operate for 30 sec and achieve an acceleration of 1 g, a velocity of 3 m/sec, and a displacement of 0.5 m. To provide spatial variability, the EMSS is not a single table but a platform made up of multiple panels that operate independently. Each panel can shift up and down and sideways, rock front to back and side to side, and twist. The basic design includes nine panels in a three-by-three array. However, the system is modular so the panels can be arranged in any pattern. For example, to test a bridge, the panels can be placed side by side to create a long platform.

A major challenge was finding a means of moving the panels. Existing shake tables are generally driven by hydraulics--a method that does not yield the precision and flexibility needed for the large, "articulated" shake table. The MIT researchers have come up with a novel solution: they use "actuators" driven by electromagnets. Drawing on their experience with using very large magnets in fusion energy experiments, Dr. Chaniotakis and his colleagues in the Plasma Science and Fusion Center designed an actuator for the EMSS that consists of two magnets. One is a stationary hollow cylinder; the other is a rod inside the cylinder that moves up and down like a piston. Current passing through the outer cylinder generates a magnetic field. Current flowing through the central rod interacts with the magnetic field, generating a force that causes the rod to move. By altering the polarity, the researchers can make the rod go up and down. The result is a large, well-controlled force capable of moving a heavy weight. In the EMSS design, several actuators are placed on each panel, oriented in different directions. By controlling the magnetic fields in the individual actuators, the researchers can make a given panel move in all directions. Taken together, the panels can simulate the complex ground motions measured during earthquakes. Altering the pattern of motion requires changing the way electric power flows to the system--a far simpler task than changing hydraulic fluid flows.

Computer simulations prepared by Professor Toksoz and his seismology team provide extensive details on the motions resulting from earthquakes. However, performing a shake-table test requires knowing how those relatively large-scale motions translate into forces exerted on a single structure. Professor Kausel and his coworkers have therefore performed simulations that describe the propagation of seismic waves through the ground and the differential motions they create at points on the surface--points that may be only a few meters apart, thus within the footprint of a sample structure being tested on the EMSS. Other analyses examine further refinements and adjustments that are needed. For instance, as an experiment begins, a massive structure being tested will initially resist moving. Once moving, it will have a tendency to vibrate at a certain frequency or even to tip over. Indeed, the structure could begin to move the table rather than vice versa. Professor Kausel has calculated the components of such "feedback effects" that would not occur in natural systems and has defined the forces needed to offset them.

The researchers have now performed several computer simulations to evaluate the performance of the EMSS. In one simulation they modeled a full-scale 10-story building using the real signal of the Kobe earthquake. In the simulation, the EMSS consists of nine panels arranged in a square configuration. Each panel is 10 x 10 meters and is driven by three electromagnetic actuators, one vertical and two horizontal. Using their computer models, the researchers calculated the forces that the simulator must provide at the base of the building to reproduce the Kobe earthquake signal. The simulation showed that the EMSS is capable of delivering the needed forces. The calculated power and energy requirements are comparable to those used in today's fusion experiments. In fact, as designed, the power system for the EMSS can satisfy the requirements of other facilities at the ACETS site, in particular, the facilities to test the response of structures to high winds like those during hurricanes.

Because of the novel nature of the EMSS, Dr. Chaniotakis and his team have designed and built a table-top seismic simulator to demonstrate certain principles inherent in the full-scale concept. The table-top simulator is 1 meter by 1 meter and has four rectangular panels, each driven by two electromagnetic actuators. The panels move independently, though thus far only in the horizontal direction. The size of the simulator and the forces it creates are relatively small, yet the simulator demonstrates the general design of the full-scale EMSS and the spatial variability and fidelity that can be attained. Because of its modular nature, the table-top model can be disassembled and moved for demonstration purposes. Indeed, during the summer of 1996 the researchers took it to INEEL, where it operated well with a scale-model structure placed on its platform.

An important practical question is how the EMSS will affect the region around it. The researchers estimate that operating the table will produce forces equivalent to those of a magnitude 4.5 earthquake. The geology at INEEL would seem capable of handling such a disturbance: layers of hard basalt are interspersed with layers of soft volcanic ash that are ideal for damping seismic motion. Using two numerical methods and descriptions of the INEEL site geology, the MIT researchers determined that most of the ground motion caused by the shaking of the table would be confined to a small region close to the EMSS.

In the coming months, the MIT teams will continue working on various aspects of the EMSS. Tasks include adding vertical motion to the table-top model and developing a half-scale prototype of the actuator included in the conceptual design. The teams will continue to refine their computer simulations of the operation of the EMSS and the interactions between a sample structure, the table, and the ground beneath. They will investigate the feasibility of scaled-down experiments to examine certain aspects of the shake-table operation. They are also considering a "hybrid" method of seismic testing potentially applicable to the design of the EMSS. The method calls for using not only the shake table but also independent force actuators placed directly on the structure to provide further forces, perhaps in the vertical direction. The two methods would create motions of different frequencies, further increasing the accuracy with which the EMSS could replicate actual seismic signals.

[e-lab Home Page] [Energy Lab Home Page] [MIT Home Page] [Up]
Last updated:

Copyright © Massachusetts Institute of Technology 1997. Material in this bulletin may be reproduced if credited to e-lab.