UC Berkeley

Observations from recent earthquakes show that all types of retaining structures with non-liquefiable backfills perform very well and there is limited evidence of damage or failures related to seismic earth pressures. Even retaining structures designed only for static loading have performed well during strong ground motions suggesting that special seismic design provisions may not be required in some cases. The objective of this study was to characterize the seismic interaction of backfill-wall systems using experimental and numerical models, with emphasis on cohesive soils, and to review the basic assumptions of current design methods.

In the experimental phase of this research, two sets of centrifuge models were conducted at the Center for Geotechnical modeling in UC Davis. The first experiment consisted of a basement wall and a freestanding cantilever wall with level backfill, while the second one consists of a cantilever wall with sloping backfill. The soil used in the experiments was a compacted low plasticity clay. Numerical simulations were performed using FLAC2-D code, featuring non-linear constitutive relationships for the soil and interface elements. The non-linear hysteretic constitutive UBCHYST was used to model the level ground experiment and Mohr-Coulomb with hysteretic damping was used to model the sloping backfill experiment. The simulations captured the most important aspects of the seismic responses, including the ground motion propagation and the dynamic soil-structure interaction. Special attention was given to the treatment of boundary conditions and the selection of the model parameters.

The results from the experimental and numerical analysis provide information to guide the designers in selecting seismic design loads on retaining structures with cohesive backfills. The experimental results show that the static and seismic earth pressures increase linearly with depth and that the resultant acts at 0.35H-0.4H, as opposed to 0.5-0.6H assumed in current engineering practice. In addition, the observed seismic loads are a function of the ground motion intensity, the wall type and backfill geometry. In general, the total seismic load can be expressed using Seed and Whitman's (1970) notation as: Pae=Pa+dPae, where Pa is the static load and dPae is the dynamic load increment. While the static load is a function of the backfill strength, previous stress history and compaction method, the dynamic load increment is a function of the free field PGA, the wall displacements, and is relatively independent of cohesion. In level ground, the dynamic load coefficient can be expressed as dKae=1/2gH2(0.68PGAff/g) for basement walls and dKae=1/2gH2(0.42PGAff/g) for cantilever walls; these results are consistent with similar experiments performed in cohesionless soils (Mikola & Sitar, 2013. In the sloping ground experiment the seismic coefficient came out to dKae=1/2gH2(0.7PGAff/g), which is consistent with Okabe's (1926) Coulomb wedge analysis of the problem. However, that slope was stable under gravity loads even without the presence of the retaining wall (FS=1.4). Measured slope displacements were very small and in reasonable good agreement with the predictions made with the Bray and Travasarou (2007) semi-empirical method.

The experimental data was not sufficient to determine accurately the point of action of the seismic loads. However, the numerical simulations and Okabe's (1926) limit state theory suggest that the resultant acts between 0.37H-0.40H for typical values of cohesion. While the resultant acts at a point higher than 0.33H with increasing cohesion, the total seismic moment is reduced due to the significant reduction in the total load Pae, particularly for large ground accelerations. The results also show that typical retaining walls designed with a static factor of safety of 1.5 have enough strength capacity to resist ground accelerations up to 0.4g. This observation is consistent with the field performance of retaining walls as documented by Clough and Fragaszy (1977) and the experimental results by al Atik and Sitar (2010) and Geraili and Sitar (2013).

The evaluation of earth pressures at the wall-backfill interface continues to be a technical challenge. Identified sources of error in the present study include the behavior of pressure sensors, the geometric and mass asymmetry of the model and the dynamic interaction between the model and the container. While these centrifuge experiments reproduced the basic response of prototype models, ultimately, instrumented full-scale structures are most essential to fully characterize the response of tall walls and deep basements with varieties of backfill.