AMERICAN and Japanese structural engineers are having a crisis of confidence. Two devastating earthquakes 鈥� the first at Northridge in California in January last year and the second in Kobe, west of Osaka, 12 months later 鈥� have shaken their faith in what they regarded as the most resilient of their modern designs, the steel frame.
This rigid skeleton of vertical columns and horizontal beams, connected with welds, has become the most popular way of supporting buildings, especially those taller than five storeys. Although most of these frames remained standing, often without visible damage, this apparent resilience is dangerously illusory.
In the months after the Northridge disaster, engineers began to find cracks in the welds joining the columns and beams together 鈥� in the very connections designed to enable the structural frames to absorb the energy generated by an earthquake without toppling over. Left unattended, the cracks could trigger catastrophic failure in subsequent earthquakes of even modest power.
Advertisement
Escalating costs
By the end of last year, a state-funded survey in California had discovered cracked connections in more than 100 buildings thought to have escaped unscathed and, earlier this year, the city of Los Angeles ordered the owners of 400 more buildings to check for similar defects. Those found to have steel frames with cracked welds need up to $40 000 to repair each connection. As a result, the cost of damage from the Northridge quake, which killed 60 people, is expected to be double early estimates, rising to more than $20 billion.
Widespread problem
鈥淎 similar phenomenon is happening in Japan,鈥� says Makoto Watabe, a senior director of Shimizu, one of the country鈥檚 biggest construction companies. The more engineers look for cracked connections, he notes, the more they find. While no coordinated survey has yet been instigated in Kobe, where more than 5500 people died, the bill for repairing the damage is estimated at $100 billion and rising.
Although the quakes at Northridge and Kobe turned many homes, offices and elevated highways into piles of rubble, these failures, like those more recently on Sakhalin Island in Russia and at Egion in Greece, can be blamed on outdated design or sloppy construction. What is worrying engineers around the Pacific Rim is that the welded connections of their steel frames, which represent the state of the art, are cracking regardless of quality. Furthermore, the structures are holding together most of the commercial and office buildings completed over the past 15 years.
鈥淎 class of buildings we had a lot of faith in has been proven pretty dangerous,鈥� says Stephen Mahin, professor of structural engineering at the Earthquake Engineering Research Center of the University of California at Berkeley. In Japan, there is similar disillusionment. 鈥淚鈥檓 very worried about this,鈥� says Hanji Hattori, a consulting engineer and head of a subcommittee studying earthquake damage to structures for the Japan Institute of Architects. 鈥淚 think most of the profession is worried about it.鈥�
Moreover, there is no consensus on how to fix the cracked connections or design better ones. The US Federal Emergency Management Agency, based in Washington and reporting directly to the White House, is trying to work out how best to spend up to $10 million on a programme of research into the problems, but the results will not be ready for at least three years. 鈥淚t鈥檚 going to take years of work to really understand what went wrong and what needs to be done,鈥� says Michael Engelhardt, professor of civil engineering at the University of Texas, Austin. In the meantime, about $1 billion of planned government construction in Los Angeles is on hold.
This is not the first time engineers have been surprised by what earthquakes do to their structures. 鈥淭he history of the seismic [design] codes is the history of disasters,鈥� says Mary Comerio, professor of architecture at the University of California at Berkeley and an expert on the evolution of the seismic provisions of building codes. Despite giant shaking tables and other testing rigs, real buildings are only put through their paces in real earthquakes. Until then, she says, 鈥渋t鈥檚 all theoretical鈥�.
Seismic provisions were first added to Japan鈥檚 building code in 1924, the year after the Great Kanto Earthquake destroyed much of Tokyo and left 140 000 people dead. The US followed suit a decade later when California became the first state to broaden the scope of its building regulations after an earthquake in Long Beach toppled hundreds of buildings.
In those early provisions, engineers estimated that earthquakes could generate ground accelerations as high as 0.3g. But a building鈥檚 inertia and the acceleration鈥檚 short duration, they reasoned, would reduce the effect to 0.1g. They proposed that designers use this figure, multiplied by the building鈥檚 mass, to estimate the horizontal force that is distributed uniformly down the side of the building during an earthquake.
It all sounded too simple and those pioneers knew it, says George Housner, emeritus professor of civil engineering at the California Institute of Technology. But they had so little research to go on, he adds, that they had no choice. For a start, the engineers were treating earthquake loads as unvarying static forces rather than as dynamic forces of changing magnitude. The provisions also failed to take into account the resonance of buildings, which depends on their size and shape, and on whether they are made from stiff concrete or flexible steel.
Research in the US led to Los Angeles adopting new building codes in 1944. While these revisions did not change the figure for the ground acceleration used in design calculations, the horizontal force estimated to be acting on the building was modified to take into account the building鈥檚 shape and the materials it was made of. By 1950, Japan had made similar changes, but it also doubled the recommended design value of ground acceleration to 0.2g.
Learning process
Engineers were shocked. In 1968, an earthquake of magnitude 7.9 hit the northern end of Japan鈥檚 Honshu island, killing 52 people and destroying more than 3500 buildings throughout the rural area. Three years later, a quake of magnitude 6.7 struck San Fernando in California. Sixty-five people died and there was extensive damage to structures, including those built to the latest standards.
The devastation shocked engineers for two reasons. In California, instruments measured accelerations as high as 1.25g 鈥� not only far higher than the figures in the building codes in the US and Japan, but well above the level that many engineers thought possible. The earthquakes also toppled new structures designed to withstand them. On Honshu, columns in new schools fractured, while in California, a highway flyover collapsed before it had opened to traffic, and a six-storey hospital was wrecked just 30 days after admitting its first patients. All the structures were made of concrete, which was cheaper than steel at the time.
Engineers were again forced to confront the recurring dilemma of balancing costs and safety. Traditionally, beams and columns are designed to behave elastically so that they bend and stretch when loaded but recover their original shapes when the loads are removed. Such a design can withstand the effects of the moderate earthquakes that rattle windows and topple bookshelves once or twice during the working life of every building in seismically active regions. But designing buildings to behave elastically even during the biggest earthquakes that strike a region only once every century or two is not economically feasible, they concluded. The beams and columns would need to be huge, and the resulting buildings would be very expensive.
So engineers adopted an approach, known as ductile design, that makes the most of a failure mode. They made a virtue of the fact that beams and columns may bend and stretch permanently when subjected to loads beyond their design limits. This so-called plastic deformation would absorb the dynamic energy generated by an unusually powerful earthquake and dampen the vibrations of the structure. The larger the earthquake, the greater the deformation. Although the building would be damaged beyond use, lives would be saved if it resisted collapse.
While concrete is not inherently ductile, reseachers discovered in the 1960s that they could make it behave in a ductile way by substantially increasing the volume of steel reinforcement buried within it at critical locations.
Construction specifications drawn up by the California Department of Transportation dramatically illustrate the effect of the change. Before 1971, when engineers seriously began to consider using ductile designs, the department had demanded that steel bars of 0.5-inch (1.27-centimetre) diameter be spaced every 6 inches throughout the height of any concrete column supporting an elevated highway. Now, after several successive increments in requirements, highway columns must have steel bars of 0.75-inch diameter in a continuous spiral with a 3-inch pitch, which more than quadruples the volume of steel necessary.
Japanese authorities similarly boosted requirements in the early 1970s, though they tended to design structural elements to be stronger as well as more ductile. In Kobe, however, the Hanshin Expressway had already been built and, as infamously captured on camera, a 500-metre stretch of the highway simply flopped over on its side when this year鈥檚 quake struck.
While the San Fernando earthquake provided many examples of how not to do things, it also yielded clues about how to improve structural design. In the hope of capturing an earthquake in the act, researchers from Caltech had equipped a 10-storey steel building at the Jet Propulsion Laboratory in Pasadena with dozens of instruments to record the structure鈥檚 motion during a quake. These recordings helped engineers to fine-tune their computer models, which could then be used to test designs. But there are limitations. The models still cannot predict a building鈥檚 motion after plastic deformation begins. They also assume that connections are seamless, instead of the weak links that experience has shown them to be.
Steel鈥檚 ability to deform without fracturing, the property on which ductile design depends, persuaded many engineers in the 1970s that structural frames built from steel would resist earthquakes naturally. They assumed the welded connections would be as strong and deformable as the steel girders they joined together. But they were wrong, engineers now say, because the assumption was based on tests of girders that were smaller than those used in practice. 鈥淣orthridge was the full-scale test of what we were doing,鈥� says Engelhardt, 鈥渁nd we didn鈥檛 really pass that test.鈥�
After Northridge, Mahin tested 24 full-size connections at Berkeley. All cracked before the expected plastic deformation set in, some at loads just one-sixth of the maximum load assumed in the design. For reasons that are still not clear, connections that work for smaller beams and columns fail when used for larger members.
Nevertheless, Mahin remains confident that a new connection will be developed. One approach involves trying to devise some form of structural insulation for the welds to ensure that the stresses in them are kept below those in the nearby steel girders. In this way, plastic deformation should occur in the girder before the welds are stressed to breaking point. But any new connection is likely to be more complicated, and costly.
Connections aside, however, the new buildings performed well. Immediately after the Kobe quake, Obayashi Corporation, one of Japan鈥檚 biggest construction companies, checked 223 major buildings it had designed and built in the area from before 1971, when regulations stipulated more steel in concrete to increase the material鈥檚 ductility, to after 1981, when the building codes in Japan stipulated a ductile design. Although 36 per cent of those built before 1971 were too damaged to enter, just 11 per cent of those built between 1971 and 1980 were unsafe and only 6 per cent of those built after 1981 were too dangerous. While Obayashi has not yet systematically logged the condition of the welds in both visually damaged and undamaged structures, cracked welds alone rarely demand immediate evacuation. 鈥淚 haven鈥檛 seen anything [in the damage in Kobe] that would change our approach to ductile design,鈥� says Watabe of Shimizu. The main problem at Kobe, as with most cities, was that only a fraction of a city鈥檚 building stock meets the current building code.
Based on instrument recordings of ground accelerations of 1.8g at Northridge, seismologists are trying to convince engineers that the shaking generated close to a rupturing fault, within 10 kilometres say, can be much greater than previously thought. Engineers are not totally convinced, however. 鈥淢aximum ground acceleration doesn鈥檛 mean much to us,鈥� says Shunsuke Otani, professor of architectural engineering at the University of Tokyo. Peak accelerations can be so fleeting that they have little effect on structures, say engineers. Other components of the ground motion, such as the amplitude and frequency of a shock wave, need to be considered, they insist, and many engineers are already trying to incorporate all of these parameters into their computer models.
For Charles Kircher, a member of the Seismology Committee of the Structural Engineers Association of California, the level of ground motion predicted in the current code is probably accurate 鈥� but only for locations between 10 and 15 kilometres away from the ruptured fault. The code probably overestimates ground motions for locations farther away but underestimates it for sites that are closer, he says. He expects the committee to recommend that the maximum design load for structures sited next to active faults be doubled.
Japanese engineers say that such a proposal is impractical. 鈥淎ll of Japan is too close to faults,鈥� says Masayoshi Nakashima, associate professor at Kyoto University鈥檚 Disaster Prevention Research Institute. The plastic deformation dictated by ductile design, he adds, can accommodate any force higher than expected.
Awareness campaign
In the US, where the cost of the Northridge disaster has disrupted the state鈥檚 economy, engineers are considering tightening the code so as to mitigate damage as well as save lives. If this proposal gathers momentum, engineers could start by bringing their clients into the discussions of the risks inherent in balancing costs and earthquake resistance. Many owners of buildings at Northridge thought their buildings should have come through with less or no damage 鈥� many were unaware that engineers expect even moderate earthquakes to cause significant nonstructural damage.
There will always be limits, however. Uninterrupted operation of a building is something no engineer would guarantee for anything but relatively mild earthquakes, says Christopher Arnold, an architect based in California. Despite all the progress, a phrase never uttered by engineers is 鈥渆arthquake-proof鈥� (see Diagram).
And they all come tumbling down
MANY structures do not receive the sophisticated analysis given to prestigious high-rise towers because of the cost involved. Single-storey family houses, for example, are built following simple guidelines that barely improve on traditional construction methods. In Japan, traditional homes with heavy clay tile roofs supported on unbraced wooden posts present little resistance to earthquakes. In Kobe they collapsed, and accounted for about 90 per cent of the casualties. Californians are luckier. The plywood sheets used to build walls in typical North American homes provide resilient lateral resistance.
Engineers are involved in the design of smaller commercial and institutional buildings, but they tend to rely on all-purpose equations rather than specific computer models on which larger structures can depend. The equations assume that structures have simple, rectangular plans, and they are not easily applied to those with unusual or irregular shapes, as earthquakes can illustrate.
The most common building irregularity is the 鈥渟oft storey鈥�. This occurs typically on the ground floor when a large shop window or garage entrance interrupts the solid wall or bracing used to stiffen the upper floors, making the ground floor more flexible, or softer, than the rest of the building. The equations assume that the sway of a building will be distributed uniformly throughout its height. In fact, the most flexible level attracts the motion. As a result, ground-level carports collapsed in Northridge and small shops crumbled in Kobe.
While designers have come to recognise this problem, the influence of an atypical building configuration can be more subtle, as the partial failure of a six-storey office block on a prominent corner in Kobe demonstrated. Although the building had a simple, rectangular plan, the two sides facing the streets had concrete columns to allow for windows, while the other two sides had solid concrete walls. Because walls are much stiffer than columns, the centre of the building鈥檚 rigidity was near the corner where the two solid walls met, but the centre of the building鈥檚 mass was much closer to the geometrical centre of the building. The offset between those two centres enabled the earthquake to generate torsional forces that fractured the columns all the way up the building, causing the corner nearest the street intersection to disintegrate and drape from the corners near the solid wall, which were left almost intact.
Putting a damper on things
INSTEAD of designing structures to absorb the effects of earthquakes, engineers can isolate them from their foundations and thus insulate them from the shocks.
The most common form of base isolation, as the practice is known, places multilayered sandwiches of rubber bonded to steel plate between a structure and its foundations. The steel plates make the rubber pads stiff enough to support the weight of the structure, but leave them flexible enough to cut down the ground motion that would otherwise pass into the structure.
Developed from the mid-1970s onwards, rubber base isolation systems use a variety of different mechanisms to damp their motion. They incorporate hydraulic pistons, springs or a compressible core plugged with lead. Alternatively, the isolator itself may be made from a rubber that deforms plastically to damp the motion before regaining its original shape.
The largest building that relies on base isolators was completed in Kobe last October. 鈥淚 never imagined it would get this test in just three months,鈥� says Shoichi Yamaguchi, president of Tokyo-Kenchiku Structural Engineers, its designer. The six storeys of the West Japan Building of the Ministry of Posts and Telecommunications sit on 120 isolators that range from 80 to 120 centimetres in diameter and from 15 to 23 centimetres in thickness. Yamaguchi was pleased with the building鈥檚 response to the earthquake, the epicentre of which was 30 kilometres away. The isolators cut peak ground accelerations of about 0.4g on the foundations to 0.12g on the first floor, he says. This left the building unscathed, while partitions and ceilings in an adjacent building without any isolation system were damaged.
Base isolation is also a useful way of bringing existing buildings up to current seismic standards. Engineers have done this at historic sites, such as the city halls in Salt Lake City and San Francisco, and Parliament House in Wellington, New Zealand. But it is an expensive business. Even for new buildings, base isolation can add between 5 and 15 per cent to the overall cost. Furthermore, isolators cannot be used for buildings above 15 storeys because they make the structures too flexible.
Other base-isolation schemes are used less frequently. With the friction-pendulum system, the building rests on low-friction Teflon bearings that slide back and forth within concave saucer-like supports. The pile-in-sleeve system supports the building on piles that run through open tubes, which isolate the structure from surface ground movements.
A number of Japanese companies have also developed damping systems that rely on a heavy pendulum on the roof to counter a building鈥檚 motion. The most ambitious of them uses pistons, controlled by a computer that picks up signals from instruments recording the motion, to drive the pendulum one way while the structure swings the other. But these tuned, mass dampers are more effective in countering wind-induced vibrations, and their use against earthquakes remains experimental.