Seismic performance evaluation of sheet-pile wall with liquefiable backfill

Abstract

Sheet-pile walls are increasingly used as retaining structures at ports and harbors. Many port and harbor facilities located in seismically active zones are susceptible to earthquake-induced damages due to excess pore water pressure (Δu) buildup and liquefaction triggering in the surrounding soil. As damage is caused by excessive deformations and not catastrophic collapse, performance-based design methods are well suited over the conventional force-balance approach. A key step in performance-based design is the analysis method used to obtain wall displacements. The displacement-based sliding block method is attractive for design, however, there are several inherent limitations that makes it unsuitable for predicting the displacements of a sheet-pile with saturated, liquefiable backfill. A comprehensive analysis of soil-structure system with liquefiable soils is possible only by conducting a non-linear effective stress analysis (ESA) using a constitutive model for liquefaction simulation, such as PM4Sand. However, the results are sensitive to the uncertainties in the input parameters, calibration of the constitutive model, and the numerical simulation procedure. So, ESA needs to be properly validated using centrifuge tests or well-documented case-histories to increase its reliability in performance-based design. This study evaluated the seismic performance of a sheet-pile wall with liquefiable backfill using dynamic centrifuge tests and numerical simulations using PM4Sand model in FLAC2D. Seven centrifuge tests with saturated backfill and one with a dry backfill were performed as part of the Liquefaction Experiments and Analysis Project (LEAP) to generate a validation database and investigate the effect of different parameters on the wall movement. The displacement of the sheet-pile with a dry backfill decreased during multiple shaking events due to the mobilization of passive pressure in the embedded portion of the wall. The sheet-pile displacement with saturated backfill was ten times larger than with dry backfill for equivalent test conditions due to significant softening of the soil associated with the generation of Δu and liquefaction triggering. Also, backfill relative density (DR), peak ground acceleration of the base motion, and pre-seismic wall displacement played an influential role in co-seismic wall displacements. The numerical predictions agreed with the centrifuge results to a varying degree depending on the calibration method, the type of predicted response, and the test conditions. One of the major limitations of the PM4sand model is that it could not be calibrated for the cyclic strength (CRR15), the slope of the liquefaction strength curve (CRR-N), and the rate of post-liquefaction shear strain (γ) accumulation simultaneously. The simulations calibrated for the post-liquefaction γ response and CRR15 (C3 calibration) agreed with the centrifuge results in terms of wall displacement in most of the tests, but underpredicted the backfill settlements and Δu in the backfill. Likewise, the simulations calibrated for the slope of CRR-N curve and CRR15 (C2 calibration) overpredicted the wall displacement in all the tests, but gave closer predictions of backfill settlement and Δu response. The default C1 calibration for CRR15 gave intermediate results between the C2 and C3 calibrations. Sensitivity analyses showed that the predictions were most sensitive to the uncertainties in the DR and the CRR15 of the soil. Finally, numerical simulation using the C2 calibration is recommended for design practice because it gave a conservative estimate of wall displacements in all the simulated tests.