Mechanisms of earthquake-induced deformation in slopes and embankments

Earthquake-induced deformation in slopes and embankments results from several mechanisms acting alone or in combination: (1) sliding displacement along a localized failure surface [i.e., "sliding block" displacement], (2) deformation resulting from the densification of unsaturated soil termed "seismic compression" and (3) deformation resulting from accumulation of plastic strains in highly stressed regions of a slope. In contrast to sliding, which involves highly localized deformations, the latter two mechanisms are collectively referred to as "distributed deformations" and can by themselves result in damaging displacements to facilities and infrastructure. This research uses physical and numerical modelling in parallel to investigate the mechanisms governing earthquake induced deformations in slopes. The physical modelling portion of this research utilized both geotechnical centrifuge and 1-g shaking table experimental methods to examine the contribution of distributed deformation to overall earthquake-induced displacement in slopes. The numerical analyses, which were calibrated to physical model experimental data, provided additional insight into the seismic performance of the models. This research was precipitated, in part, by observations from past earthquakes suggesting that mechanisms other then sliding may be responsible for seismically induced displacement in unsaturated granular slopes (e.g. embankments). In the absence of fully documented case histories to identify and document these mechanisms, centrifuge model experiments were performed to investigate the mechanism(s) that govern the seismic performance of these types of slopes. The experimental results suggests that, for the range of test conditions considered in this study (slope geometry; amplitude, frequency and duration of input motions), seismic compression was the principle mechanism of deformation in the unsaturated granular soil; deformation resulting from the shearing, as indicated by the development of localized surfaces, was not observed. The centrifuge model response was numerically simulated using the FINN soil constitutive as implemented in the computer code FLAC. The numerical simulations, which allowed for the one-step, fully coupled analysis of seismic compression, matched well with both the dynamic and deformation response of the model slopes. Earthquake reconnaissance reports also indicate that distributed deformations can occur in slopes comprised of cohesive soils. To investigate this phenomena, a second phase of study was undertaken where shaking table tests were performed on small scale cohesive slopes comprised of "model clay" (a saturated mixture of kaolinite and bentonite). Parallel numerical simulations of the shaking table test were performed using a strain-softening soil constitutive model as implemented in FLAC. The numerical simulations and shaking table experimental results were in very good agreement. The physical model tests provide a unique dataset that allowed topographic amplification to be considered empirically. Analyses of the experimental results indicate that amplification at the slopes crest occurs almost entirely as a result of site effects; the contribution from topographic amplification was negligible for most of the cases. Overall amplification was consistently higher at the steeper crest of the two slopes for all the tests (both centrifuge and shaking table), most likely as a result of topography.