Dynamic Soil-Structure-Soil-Interaction Analysis of Structures in Dense Urban Environments
- Author(s): Jones, Katherine Carys
- Advisor(s): Bray, Jonathan D
- et al.
Urban centers are increasingly becoming the locus of enterprise, innovation, and population. This pull toward the center of cities has steadily elevated the importance of these areas. Growth has necessarily spawned new construction. Consequently, modern buildings are often constructed alongside legacy structures, new deep basements are constructed alongside existing shallow foundations, and city blocks composed of a variety of building types result. The underlying soil, foundation, and superstructure of each of these buildings can interact and combine to yield unique seismic responses.
Since the seminal work of researchers such as Luco and Contesse (1973) and Wong and Trifunac (1975), researchers have investigated the effects of soil-structure interaction (SSI). This phenomenon refers to the interaction between a single building, its foundation, and the underlying soil during a seismic event. However, as the trend toward urbanization continues, a shortcoming of this conventional SSI approach is that in reality, a structure will almost certainly be located near other structures in metropolitan areas.
In this line of research, the interaction of multiple, adjacent buildings during a seismic event, a phenomenon known as structure-soil-structure interaction (SSSI), is investigated. This topic does not yet command the level of attention given to SSI. However, SSSI has the potential to be significantly detrimental or beneficial, depending on the configuration and dynamic properties of the buildings and their foundations in dense urban environments. It is important to understand SSSI effects so that earthquake engineers can make informed decisions about the design and construction of structures in increasingly dense urban areas.
As part of a larger, multi-university National Science Foundation (NSF)-supported Network for Earthquake Engineering Simulation Research (NEESR) project, a series of centrifuge experiments were performed at the NEES-supported Center for Geotechnical Modeling (CGM) at the University of California, Davis. Each of these experiments examined aspects of SSI or SSSI through the use of nonlinear structural model buildings situated on different foundations that were supported on deep sand deposits. The centrifuge experiments created a suite of small-scale physical model "case histories" that provided "data" and insight that could be extended through calibrated numerical simulations. The results of the first three centrifuge experiments in the test series (i.e., Test-1, Test-2, and Test-3) were utilized in this dissertation.
Numerical analyses are usually only performed for high-profile projects. The effort, expertise and resources required to calibrate and to perform detailed numerical simulations is often prohibitive for typical low- to mid-rise structures. There is a need for a more accessible numerical tool that both geotechnical and structural engineers can utilize to gain insight. In this research, the FLAC finite difference program (Itasca, 2005) with a fully nonlinear effective stress soil constitutive model was used to analyze the centrifuge test-generated "case histories."
Test-1 and Test-2 examined SSI and SSSI effects of two moment-resisting frames (MRFs). Test-1 employed a solitary 3-story (prototype) MRF founded on shallow spread footings and a solitary 9-story (prototype) MRF founded on a deep basement (equivalent to 3-stories, prototype) to investigate SSI effects. In Test-2, the 3-story (prototype) and 9-story (prototype) MRFs were placed immediately adjacent to one another to examine SSSI effects. Kinematic interaction effects were primarily observed in these tests. Hence, Test-3 was designed to investigate inertial interaction effects. Three structures were included in Test-3: two MRFs founded on shallow spread footings and one elastic shear-wall structure on a mat foundation. Each of these structures was designed to maximize inertial interaction by: (1) matching the flexible base period of each structure to the soil column to induce resonance, and (2) optimizing structural properties to increase inertial interaction effects. One MRF was positioned alone at one end of the centrifuge model, a SSI condition, and the other MRF and the elastic shear-wall structure were positioned immediately adjacent to each other in the other end of the centrifuge model, a SSSI condition.
The rich data set developed through the centrifuge experiments formed the basis of the initial FLAC analyses. A critical aspect of any seismic analysis is the constitutive model used to capture the soil response to cyclic loading. Several soil models were examined during an initial seismic site response analysis. Free-field data from sensors located within the centrifuge soil column were used to quantify the vertical propagation of ground motions through the soil profile. The best model for the dense (Dr = 80%), dry sand used in the centrifuge for Test-1 through Test-3 was a Mohr-Coulomb based model with hysteretic damping, UBCHYST (Naesgaard, 2011). Pseudo-acceleration response spectra and acceleration time histories at the base and at the free-field surface from the centrifuge and the numerical model were compared. The numerical simulations successfully captured the key aspects of the observed seismic site-response for both near-fault pulse-type motions and ordinary motions at a variety of intensities.
After successfully capturing the free-field seismic site responses of Test-1 and Test-2, the dynamic responses of the structural models were examined. Each structure was modeled satisfactorily with a two-dimensional, plane-strain numerical model. Engineering design parameters (EDPs) were computed for key structural responses, including (1) transient peak roof drift, (2) residual roof drift, (3) transient peak displacement and (4) peak acceleration at the center of mass of the structure. Additionally, the acceleration time histories and pseudo-acceleration response spectra at the center of mass of the structure for each motion were examined. These metrics were used to compare the numerically estimated dynamic responses with those recorded in the centrifuge experiments. The dynamic response of the 3-story (prototype) MRF estimated with the numerical model was in close agreement with the observed experimental data for both the SSI (Test-1) and SSSI (Test-2) configurations. The more complicated 9-story (prototype) model exhibited greater sensitivity to numerical system inputs, including fixed-base fundamental period and applied structural Rayleigh damping. However, the majority of its recorded dynamic responses were well-matched by the numerical model.
The resonant condition created in Test-3 proved challenging to model numerically. The two Test-3 conditions (i.e., SSI and SSSI) were analyzed separately. Significant inertial interaction, including rocking, was observed during the centrifuge test and in the post-processing of data; pseudo-acceleration responses three to five times those recorded in Test-1 and Test-2 were recorded. While the shapes of the pseudo-acceleration response spectra, periods of amplification, and time-histories were well-captured, the numerical model estimated significantly lower amplitudes of the responses for the structures than were observed during the centrifuge test. A sensitivity study was performed to evaluate the influence of several parameters, including (1) the shear wave velocity profile, (2) interface elements, (3) fixed-base fundamental period estimate, and (4) constitutive model parameters. Some of the relative lack of amplification in the numerical simulations was due to over damping in the constitutive model. This was addressed by altering the shear modulus and material damping curves for the soil directly beneath the structures' foundation elements. However, the primary reason for the lower amplitude estimated by the numerical model appeared to be due to the difficulty of capturing the seismic responses of structures in the resonant condition. Shifting the period of any component of the soil-structure system would necessarily have a significant impact on the dynamic response by shifting the system away from resonance. Despite this challenge, the numerical simulations yielded important insights. While the amplitudes of dynamic responses were underestimated for most of the ground motions, the changes in response of the 3-story (prototype) MRF between SSI and SSSI were captured. The elastic shear wall displayed similar behavior; while the spectral shapes were matched for most motions, the amplitudes estimated by the numerical simulations were consistently below those observed in the centrifuge. Comparison of overall change from low- to high-intensity motions or trends from SSI to SSSI could be captured with the model; however, the amplitudes of the responses were generally underestimated. This set of analyses highlighted the challenge of modeling a resonant condition. Additional work is needed to explore the characteristics of the centrifuge when intense input motions are used which are in resonance with the soil in the model.
Finally, two prototypical structures were examined. The first, a 3-story MRF, was the model upon which the centrifuge 3-story (prototype) model was based (Ganuza, 2006). Both solitary (SSI) and adjacent (SSSI) configurations were considered for this prototypical 3-story MRF founded on a dense sand soil column. The dynamic responses of the MRF for the solitary (SSI) condition paralleled those observed in the centrifuge experiments. For the considered configurations of adjacent low-rise structures, SSSI effects were found to be either negligible or only slightly beneficial or detrimental for the five ground motions utilized for dynamic analysis. The other prototypical MRF, a 5-story structure, was a simplified version of a typical, medium-rise structure (Ganuza, 2006). The 5-story MRF exhibited dynamic responses consistent with previous work. Amplification was observed at (1) the first and second modes of vibration of the structure, (2) the site period, and (3) the mean period of the motion. Further research is needed to study and more fully quantify SSSI effects for a wider set of structures, adjacent configurations, and ground motions.