Liquefaction-induced settlement of shallow-founded buildings continues to produce significant damage during earthquakes. The state-of-the-practice for estimating liquefaction-induced settlements relies on simplified procedures that do not capture the important shear-induced mechanisms that often control structural settlements. Consequently, building settlement is often underestimated. Performance-based design requires an improved assessment of liquefaction-induced building settlement. Nonlinear dynamic soil-structure-interaction (SSI) effective stress analyses can capture shear-induced liquefaction building settlement mechanisms. However, they are not commonly used in engineering practice due to their lack of validation. Well-documented field case histories of building performance at sites with liquefiable soil provide the opportunity to validate available analytical tools. In this study, five significant buildings with shallow foundations affected by 2010-2011 Canterbury earthquake sequence are back-analyzed to evaluate the capabilities of dynamic SSI effective stress analysis and to gain insights into the mechanisms controlling liquefaction-induced building settlement.

Before the back-analyses of field case histories are performed, 36 model case histories of structural performance from a series of geotechnical centrifuge experiments are analyzed. The centrifuge experiments provide a wealth of quantitative time-varying parameters (e.g., pore water pressure, acceleration, and displacements) for detailed examination of the capabilities of the employed analytical model and procedures. The free-field responses measured in the centrifuge experiments are captured well in the numerical analyses, especially in terms of acceleration-time histories and pore water pressure generation during strong shaking. The analyses also captured liquefaction-induced building settlement in the centrifuge experiments reasonably well, although there was a tendency for it to overestimate the amount of measured building settlement. The tendency for and amount of overestimation were greater for cases in which the ground motions induced relatively small settlements (< 200 mm).

Although the field case histories contain significantly more uncertainty in terms of the earthquake ground motions, soil properties, and structural response than the centrifuge experiments, they provide important insights not captured commonly in the centrifuge experiments (e.g., effects of sediment ejecta, variable ground conditions, and naturally deposited soil). Importantly, advanced analytical methods will not be employed in engineering practice until they can be shown to capture key aspects of building performance during earthquakes in the field. Thus, the primary objective of this research effort is to perform back-analyses of well documented case histories of liquefaction-induced building settlement in the Central Business District (CBD) of Christchurch, New Zealand. The Christchurch case histories include vast amounts of detailed information about the earthquake ground motions, site characterization, structural configurations, and observed seismic performance. Back-analyses were performed for three events of the Canterbury sequence of earthquakes: (1) the 4-SEP-2010 Mw 7.1 Darfield earthquake that produced peak ground accelerations (PGA) in the CBD of 0.16-0.28 g, (2) the 22-FEB-2011 Mw 6.2 Christchurch earthquake that produced PGAs of 0.35-0.55 g in the CBD, and (3) the 13-JUN-2011 Mw 6.0 earthquake that produced PGAs of 0.18-0.30 g in the CBD. In addition to having different intensities of strong shaking, the earthquakes also produced ground motions with different frequency contents and significant durations. The careful documentation of the effects of a sequence of three major earthquakes on the ground and structures in a modern city is unprecedented. Hence, these field case histories represent a unique opportunity to evaluate the capabilities of advanced numerical simulations of liquefaction effects on buildings.

The field case histories analyzed in this study consist of multi-story buildings with shallow foundations over soil deposits which include soil layers prone to liquefaction. Site-specific cone penetration tests (CPT) and laboratory test data, especially for loose-to-medium dense soil units that control the seismic response of the ground and building, are essential in refining the calibration of the PM4Sand model. Understanding site geology is also critically important when developing the FLAC model. The CPT investigations confirmed that shallowly buried streams were beneath parts of some of the buildings. Thus, the buried stream channels had to be included in the heterogeneous soil profiles modelled in the back-analyses. During the model calibration process, the free-field ground response was shown to compare well with field observations and the results of established simplified procedures in terms of pore water pressure ratios, shear strains, and factors of safety against liquefaction. The 5%-damped acceleration response spectra for the motions calculated at the ground surface also compared favorably with the response spectra of the nearby recorded free-field motions.

The CTUC building was a reinforced-concrete, six-story structure founded on footings connected with tie-beams. The site conditions include a buried stream that crosses underneath a corner of the building where most of the damaged was observed. Analyses show that the building underwent a bearing capacity-type of failure during the Christchurch earthquake, which led to significant differential settlement whose magnitude was consistent with field observations. The FTG-7 building was a moment resisting steel-frame structure founded on strip footings in one direction that were tied together with grade beams in the other direction. The soil deposit has fairly uniform, thick liquefiable layers. After the earthquakes, differential settlement, tilting, and structural damage were observed. The analyses indicated that SSI-induced ratcheting is the primary mode of deformation, which is observed by the rocking of the building’s perimeter columns moving vertically in opposite directions during the same cycle of loading. The PWC building and CTH auditorium, which are located close to the Avon river, are also analyzed. Having a free-face near the structures added lateral and vertical movements associated with lateral spreading. The performance of the PWC building is influenced by several factors including the shape of the basement, a medium dense sandy soil layer located close to the base of the foundation, lateral movements towards the river, etc. The performance of the CTH building was affected by shear-induced settlements that produced differential settlement of adjacent columns, as well as soil-ejecta-induced and volumetric-induced settlements, and vertical movements resulting from lateral spreading. For these two buildings, a single controlling mechanism is not clearly identified; it is likely that the observed building movements resulted from a combination of ground deformation mechanisms. In the last case, the difference in weight and bearing pressures of each side of the C building in the west and east direction and the unintended consequence of soil improvement due to installing tie-downs to resist static buoyant water pressures under the western part of the facility that did not have a structure atop of the basement caused differential settlement that induced structural cracking of some elements. The nonlinear dynamic SSI effective stress analyses were able to capture the tendencies of the basement mat to uplift on its western end and to settle on its eastern end.

Good agreement between the calculated and measured building settlements was obtained for these buildings for the Christchurch earthquake, which shook them most intensely. The analyses overestimated building settlements for the lower intensity Darfield and 13-JUN-11 earthquakes. The overestimation of building settlements for the Darfield earthquake was relatively minor. The overestimation of building settlements for the 13-JUN-11 event was more significant, and it was judged to occur because the analyses overestimated the free-field response recorded at nearby strong motion stations for this event.

The back-analyses of field case histories provide valuable insights into the mechanisms causing liquefaction-induced building settlements. The satisfactory comparison of the calculated and measured responses provides confidence in the use of nonlinear dynamic SSI effective stress analyses as a decision-making tool in performance-based design. One of the shortcomings of these continuum-based analyses of liquefaction-related phenomena is their inability to capture the effects of soil ejecta.