Earthquake Protection for Data Systems
- Published on Sunday, 28 October 2007 23:44
Understanding Seismic Motions
Earthquake exposure varies across the United States. Seismic zoning maps are the most familiar way of demonstrating this variance. Building codes are based on the information from these maps, and for relatively simple structural design and equipment requirements, use of these design specifications is recommended.
Variables which should be considered when estimating the amount of damage an EDP facility are the probable location of future earthquakes, their expected frequency of occurrence, soil profile and building height. The selection of a design method and criteria will be, to a large extent, controlled by the expected performance criteria for the building and its contents. The choice of design approach to use should be based on the importance and complexity of the proposed facility.
For major industrial facilities, such as dams and power plants, and for public facilities, such as hospitals, which should be operational after an earthquake, detailed hazard studies should be completed prior to design. When a major data processing center is considered to be of equivalent importance, detailed seismic hazard studies may be necessary.
Building Seismic Design Considerations
The ground motions created by earthquakes produce friction and displacement in structures and the contents within them. It is important when designing for earthquake resistance that all components of a building be provided with adequate strength, be tied together, and provisions be made for differential motions and displacements.
Historically, the primary purpose of U.S. seismic codes has been to protect life and safety of building occupants and those immediately outside the building. Therefore, buildings built to Uniform Building Code standards or its equivalent would withstand minor to moderate earthquakes with relatively little to no damage and major earthquakes without major structural collapse. In recent years, however, other considerations have become important, such as keeping exits open and essential facilities, such as hospitals and communication centers, functional during and after earthquakes. As costs for repairs have risen, non-structural elements have also received attention.
Performance of buildings in earthquakes is influenced by a number of factors. These include the structural system used for the building, the type of construction material, flexibility of the building frame, whether or not many brittle-type components, such as glass windows or ceramic tile finishes, are used, age, and location.
A frame structure is usually the most flexible, and it dissipates energy by deforming in bending or flexure. However, the non-structural components and systems must be designed to accommodate the expected deformations. A shear wall structure, on the other hand, is quite rigid and deflects less than a comparable frame structure. Less deformation in shear wall structures can mean savings on connections of exterior walls, windows, and interior partitions.
Structural framing systems can be stiffened by using cross bracing or eccentric bracing, or shear walls to reduce deformations induced by seismic motion. However, braced steel framing has much less ductility than a structural steel frame. Well designed structural steel framed buildings have generally performed well during earthquakes, providing the non-structural components have been properly connected to the structural frame. Concrete buildings when properly designed have also functioned well during earthquakes. However, there have been catastrophic collapses when the design has been deficient.
Masonry structures made either of clay brick or concrete block, and which are adequately reinforced and tied together, have suffered minor to moderate damage during earthquakes. Unreinforced masonry construction has a poor record of earthquake resistance. Wooden structures have an inherent energy absorbing capacity, especially for relatively small structures (up to three stories in height).
The basic layout of a building also has a great deal to do with its seismic resistance. Forces induced in non-symmetrical buildings are considerably greater than those in symmetrical structures and hence damage is generally more extensive. Also, earthquake motions generally amplify with building height. As a result, equipment and building components can experience stronger motions the higher their location in the facility.
There are two basic methods for the seismic analysis of buildings: equivalent lateral static force and dynamic analysis.
A computer is needed to determine the detailed dynamic behavior of structures under the influence of seismic motions. Engineers have sought to simplify and reduce the dynamic solution to a static force approach. The static design forces used are a function of the seismic risk zone, building framing type, size and height of building, building weight, site soil conditions, and occupancy.
Dynamic analysis is an alternative that considers a structure’s dynamic characteristics when it is subjected to assumed earthquake ground motions. Dynamic analysis is an extremely powerful tool to estimate the response of a complex structure. Its use is generally not justified for simpler, smaller buildings, or for symmetrical buildings, except for those that are multistory (four or more stories).
In the 1971 San Fernando, California earthquake there was significant damage to utility systems. The electrical power system had several electrical substations with extensive equipment damage and several hundred pole transformers dislodged. Power distribution was disrupted to nearly 350,000 customers from a few minutes to several hours.
A telephone system serving 10,000 subscribers was down for about six weeks because of damage to central switching equipment. The water supply system sustained 22 breaks in large trunk lines, and nearly 400 breaks occurred in distribution mains. There was heavy damage to the sewer system.
The natural gas system had over 800 breaks in lines serving 17,000 customers. Several gas line breaks erupted in flame. Vehicle access to the areas was impeded by damage to numerous highway overpasses and roadways.
The lesson of the San Fernando quake is that it is prudent to carefully evaluate the potential seismic performance of required support utilities, both off-site and on-site.
Other earthquakes have demonstrated both the dependence on and fragility of these systems. Such information enables management to make better decisions on what back-up measures are necessary for the EDP facility.
The utility services for an office building are usually supplied by a utility company or government agency.
The EDP owner or his or her architect-engineer should meet with the utilities providers to discuss their systems. Information should be requested about the reliability of the supply transmission system, redundancy and alternate routing, local distribution system, time required for system repair or restoration of service, and the priority that the EDP center would have in this process.
If the reliability of the utility systems cannot be reasonably ascertained, consideration should be given to adding back-up systems for the EDP facility.
The power system includes storage batteries for uninterrupted power supply (UPS), the main power supply, and a back-up generator system for critical support systems. All systems should be designed with adequate seismic resistance.
The need for a back-up water supply system is dependent on the type of cooling system: cooling tower or chiller.
A supply of potable water for personnel should be provided so the EDP center’s staff can continue to function. Consideration should be given to installing gas shut-off valves if the facility is located in a high seismic zone.
A building and its contents are often described in three major groups: structural, non-structural, and furniture and supplies. Nonstructural elements include the mechanical, electrical, plumbing, and lighting systems; partitions and ceilings; and most permanently placed equipment related to the building use.
Damage to nonstructural elements of buildings is caused in two primary ways: (1) damage related to differential distortion of the structure, and (2) damage related to shaking of elements.
Distortion damage occurs to any element that is forced to undergo the same deformations as the structure, and is not able to do so without damage. This includes filled walls or other stiff walls, stairways, and vertical service distribution elements, like elevators.
Shaking damage or disruption primarily happens to elements that may respond to the vibratory motion of the structure by sliding, overturning, or swinging.
Another more subtle effect related to the building motion is internal damage to equipment, regardless of whether the element slides, overturns or swings.
Seismic protection standards have only recently taken nonstructural elements into consideration. The current accepted standard is that nonstructural elements must be mounted so that the element as a unit will meet seismic force requirements. Interior contents of the elements are not protected by regulation, except to the extent that they should not fall out or otherwise immediately increase personnel hazard. This rationale is based on the theory that most equipment will experience higher peak accelerations during transportation than will occur in an earthquake.
Both the design forces and anchoring or restraining techniques must be considered. The following 13 types are included in the guidelines. Three levels of protection are discussed for each: minimum, intermediate and maximum. It is then up to the users of this manual to select the level of protection they wish to achieve.
The appropriate level of protection for each element and each installation must be determined individually, considering its importance and the likely damage and time to repair or replace it.
The 13 elements covered include: (1) fixed, floor mounted elements; (2) fixed, suspended elements; (3) fixed elements on access floors; (4) isolated (vibration) floor mounted; (5) Isolated (vibration) hung elements; (6) Storage racks or shelves; (7) ceilings; (8) light fixtures; (9) partitions; (10) piping systems; (11) air distribution systems; (12) electrical distribution systems; and (13) elevators.
Raised computer floors are a critical element in the data processing center. They are the basic support for the equipment and a shield for subfloor utilities needed to operate the equipment. The collapse of these floors in a major earthquake means extensive damage to equipment and utilities. Recovery could take months.
While data on floor performance under extreme conditions is lacking, cyclic loading tests have been conducted on raised floor systems and various components. These tests have shown that the typical raised floor system may behave in a brittle fashion, exhibiting little reserve capacity beyond initial yielding. Tests have shown that with corrective design, the behavior of these floors may be modified into an energy absorbing form capable of resisting seismic loads.
Earthquake motions induce lateral and vertical loads on raised floors. The lateral loads are transferred to the building floor system by a supporting floor structure of stringers, pedestals and braces.
The four most commonly used floor systems are cantilever pedestals, braced pedestals, braced panels, and pedestal-stringer frames. Each type of system, if properly designed, can be seismically resistant.
Existing raised computer floors should exhibit the same performance characteristics as new floors, if the level of acceptable interruption is to be met. The three steps for evaluating raised floors include a preliminary survey, an analytical evaluation, and a test. The preliminary survey can be a means of setting priorities for the various elements requiring later detailed evaluation. It can help management identify those areas or elements which, because of the importance of their function or their vulnerability to seismic vibrations or motions, are most in need of attention.
Floor systems may be analyzed for earthquake performance by either a static load or dynamic analysis. The process should include a visual inspection by an engineer to verify that the installed and maintained floor system is similar to the system that has been analyzed.
The most direct and complete method of evaluating existing floor systems is to perform physical tests. If the floor system is not capable of resisting the loads and satisfying the deflection requirements, then a program of strengthening and or stiffening should be implemented.
Tests performed to simulated earthquake motions have shown that computer equipment, when permitted to roll and isolate itself from strong, intense shaking survives in a functioning mode more readily than equipment that is rigidly constrained or anchored to the structure.
It may be too costly to anchor computer equipment that is frequently being reconfigured to meet processing change in system requirements. A high degree of earthquake protection may be provided by relatively simple cost effective measures, such as guarding floor penetrations, providing adequate space between the equipment and furniture or building elements, tying equipment together in clusters to increase its stability, and bracing tall top-heavy equipment to prevent it from tipping.
Vibratory testing of peripheral computer equipment is normally performed by the computer manufacturers as part of their design process.
This testing is intended to assure functioning of the equipment under normal operating conditions; yet, few of the peripherals, and even fewer of the computer mainframes are tested under simulated earthquake conditions.
The methods for analyzing and testing new equipment vary significantly from component to component.
Typical electronic components do not always obey the same laws of structural mechanics used in the design of computer cabinets and other parts of the system.
The five methods of analysis and testing include the following: (1) static building code analysis with a lateral force coefficient; (2) rigid body dynamic analysis (simplified method); (3) elastic and nonlinear dynamic analysis; (4) dynamic test of components; and (5) dynamic test of assembled unit.
Most magnetic storage media used in electronic data processing systems are susceptible to damage from earthquakes.
Many data libraries have offsite back-up locations to prevent loss of stored data at the main site. This is generally an adequate safety precaution for fire, theft, or sabotage. However, in the event of an earthquake, depending on its location, the offsite storage facility may be equally vulnerable to earthquake motion. Tape and disk storage vaults deserve special design consideration for earthquakes.
The most common form of seismic protection for the tape storage racks is an overhead brace that extends from the top of one rack to the tops of adjacent racks. Often this brace is tied to a wall of the storage vault. The racks also may be anchored to the floor. However, no special precautions are provided to prevent the tapes that are hung or on shelves from shaking free during earthquakes.
One of the safer forms of tape storage is the enclosed shelf unit with doors that latch shut. If this unit is adequately anchored to a strong structural wall, or properly tied to a floor system, the tapes stored in it have a reasonable chance of surviving damage.
Most tape storage racks and disk storage shelf units have been designed to resist vertical loading, but they often lack adequate lateral support. The industry recognizes that racks may tip if bumped. Bolt holes for hold down anchors and overhead bracing to minimize accidental tipping of the racks are usually provided.
Shake table tests may be used to evaluate the seismic resistance of tape storage racks and disk storage shelf units.
Such tests provide a fairly realistic assessment of the expected earthquake performance of tape storage systems. However, past earthquake experience has shown that many of the elements may fail prematurely, producing unusual and almost unpredictable behavior. Testing of this equipment is one of the best ways of verifying whether the rack support system is adequate for protecting the tape library.
Also, a surprising number of failures in data library storage systems have been attributed to connection failures between the overhead bracing and the surrounding building walls. Thus, it is critical that the building elements that support or brace the rack system be reviewed by a qualified structural engineer.
Relatively frequent lower magnitude earthquakes can disrupt or damage data processing centers. Many regions of the U.S. are susceptible to ground motions strong enough to result in such effects.
These guidelines for protecting EDP systems are the first comprehensive attempt to help users minimize their losses. Although they represent the current state of knowledge, performance data from actual earthquakes and more research are needed.
Robert A. Olson is President of VSP Associates, Inc., located in Sacramento, California.
This article adapted from Vol. 5 #2.