In their quest to design buildings and other structures that can withstand the most severe earthquakes without loss of life or damage, researchers at Lehigh have developed a novel testing technique that uses high-performance computing (HPC).
On a three-story building, the researchers have installed diagonal braces fitted with dampers made of elastomeric, viscous, magnetorheological (MR) or other “smart” materials that are currently used in shock absorbers and suspension systems.
The researchers combine physical experiments with numerical models in a “hybrid simulation” that evaluates the performance of the braces, other structural elements and the overall building itself under earthquake conditions.
The project is being carried out at Lehigh’s ATLSS Center
(Advanced Technology for Large Structural Systems) under the direction of James Ricles, the Bruce G. Johnson Professor of structural engineering, and Richard Sause, ATLSS director and Joseph T. Stuart Professor of structural engineering.
The project is funded by the National Science Foundation through the NEES
(George E. Brown Jr. Network for Earthquake Engineering Simulation) program. Established and supported by NSF, the 14 member universities in NEES collaborate on earthquake-engineering studies.
Hybrid simulation, says Ricles, is a useful tool when some parts of a structural system are understood well enough to be mathematically modeled while other parts are not and must be tested physically in the lab. In the current project, engineers are not as familiar with the behavior of the new dampers as they are with other structural elements and materials.
A greater degree of freedom
A unique feature of the current project, says Ricles, is a multitasking grid of parallel processors and an experimental coordinator, which orchestrates the test. The coordinator simultaneously imposes the effects of earthquake forces, or loading, on the braces and dampers in the lab and on a computer model of the building and its remaining elements. The laboratory and computer models are coupled through their common degrees of freedom, or equations. The coordinator and both portions of the test are linked by a local area network that transmits information in nanoseconds over a fiber-optic cable.
“In a hybrid simulation experiment,” says Ricles, “data needs to be shared and exchanged between workstations almost instantly, so quickly it’s almost like shared memory.”
The test is run in “real time,” lasting in duration as long as a typical earthquake of about 30 seconds. Previous tests were slowed down to allow for transmission of information across a slower network and for interruptions. Real-time tests yield more accurate data about the performance of the dampers, which are made of “rate-dependent” materials that respond both to the cumulative effect of stresses and to the speed at which they are imposed.
During the test, the experimental coordinator commands hydraulic actuators to impose deformations on the dampers and braces approximately 1,000 times a second and from multiple directions. These “command displacements” simulate the multidirectional loading of earthquakes as well as the ground motion, or acceleration, caused by an earthquake.
Eliminating "down time" with dampers
One objective of the test is to determine how well the dampers improve the seismic performance of a building structural system. If successful, says Ricles, the dampers will enable beam-to-column connections and other structural elements to move and then return to their original position following an earthquake. This would allow a structure to survive an earthquake without damage or loss of life while avoiding costly “down time” during which it is being repaired and cannot be used or occupied.
In the hybrid simulation, the experimental coordinator uses a numerical integration algorithm to calculate the command displacements and the responses that these displacements elicit from the structural elements, dampers and braces.
“The experimental coordinator issues command displacements for each substructure – analytical and experimental – for discrete points in time, or time steps,” says Ricles. “It calculates the response to each command using an explicit numerical integration algorithm to integrate equations of motion. It then uses this information to calculate the command displacement for the next time step. Each response affects the next command.
“The coordinator must solve the equations of motion and prepare command displacements every one-thousandth of a second in a recursive manner.”
The goal of the project, says Ricles, is to develop a high-fidelity numerical model for the hybrid simulation. The model will contain thousands to millions of degrees of freedom (equations) and will enable researchers to obtain accurate response predictions for a complex structural system.
“The number of degrees of freedom is equivalent to the number of equations that have to be solved. This number, in our simulations, is limited by the available computational power of the grid that we use in the simulation.”
A crucial role for parallel processors
The accuracy, synchronization with real time, and completeness of each calculation are critical, says Ricles.
“If the actuators do not move correctly, the experimental coordinator will not have the information it needs to issue the next command displacement correctly. The same is true with the computer model.
“To guard against errors, and to enable faster computation, we use parallel processors to ensure that the calculations are done quickly and correctly. HPC allows us to do this and thus raise the level of performance of large structural systems by being able to perform large-scale simulations.”
The researchers are seeking to validate hybrid simulation as an effective method of evaluating various types of structural design and their ability to mitigate or prevent damage from earthquakes and from loading imposed by wind and explosions.
“This is the first of many tests,” says Ricles. “Successive tests will be performed for each type of earthquake ground motion and for a variety of structural designs. They will tell us quickly and efficiently the degree of damage and its probability under a variety of possible scenarios.
“Hybrid simulation will obtain very accurate information that will help us refine performance-based design procedures for next-generation building systems.”
High-performance computing, says Ricles, will enable researchers to “scale up” future hybrid simulation tests to evaluate more complicated systems with more variables and more equations to solve.
“We are expanding the processing capacity of our grid. That will enable us to expand the scope of future hybrid simulations.”
In one test, the researchers evaluated a two-story building with four bays and with damper-fitted braces on both levels.
“The test helped us verify that we could reduce the weight of the building’s structural elements by 30 percent and obtain a better performance under earthquake conditions,” says Ricles. “So in the end, the dampers paid for themselves.
“This is a very multidisciplinary project, combining HPC, advanced devices, advanced materials, structural engineering, numerical methods, hardware, software, databases and servohydraulic systems.”
The hybrid simulation project is being conducted in Lehigh’s Real-time Multidirectional (RTMD) Earthquake Facility
, which is part of the NEES network.
Other contributors to the project include ATLSS research scientists Chen Cheng and Theodore Karavasilis and Thomas Marullo, who is RTMD information technology systems manager.
The operations manager for the RTMD Earthquake Facility is Gary Novak.
Story by Kurt Pfitzer
Posted on Tuesday, September 15, 2009