Seismic Performance and Simulation of Pile Foundations in Liquefied and Laterally Spreading Ground

The mechanism of subgrade reaction development in liquefied and laterally spreading ground is discussed through large shaking table tests. With increasing relative displacement, the extension stress state develops on one side of a pile with significant pore water pressure reduction, while the compression stress state develops on the other side with insignificant pore water pressure reduction. The subgrade reaction is induced by the difference in stress sates in both sides, i.e., the pile is pulled by the soil on the extension side. In liquefied level ground, a pile is pulled by the soil on right and left sides alternately because the extension and compression stress states alternately develop on both sides. In laterally spreading ground, a pile is pulled only by the downstream soil because the extension stress state develops on the downstream side of the pile only when the ground moves downstream. The subgrade reaction in the laterally spreading ground consists of cyclic as well as permanent components, of which contribution depends on pile stiffness. The cyclic one with pore water pressure reduction, induced by the cyclic ground deformation, becomes large in stiff piles, while the permanent one, induced by the permanent ground deformation, becomes large in flexible piles.

To aid in the design of the Ravenel Bridge in Charleston, South Carolina, lateral 1load tests were performed on a 2.59 m diameter drilled shaft foundation before and after blasting to induce liquefaction. Excess pore pressure ratios between 75 and 100% were induced to a depth of 13 m using explosive charges. Following blasting, the stiffness of the load-deflection curve decreased to about 15% of its pre-blast value and the reduced lateral resistance led to a 100% increase in the maximum bending moment. Back-calculated p-y curves generally had concave upward curve shapes similar to those observed in previous blast-liquefaction tests at Treasure Island. The increased resistance for the 2.59 m diameter shaft could be accounted for using a constant pile diameter correction factor of nine relative to a 0.3 m diameter pile.

Piles passing through laterally spreading slopes can be subjected to considerable loads by the soil flowing past them. Many case histories have been documented of piles which suffered failure as a result of horizontal loads exerted by the flowing soil. This paper details the results of a series of dynamic centrifuge tests carried out at Cambridge University Engineering Department, to investigate the transfer of load from the spreading soil to the piles passing through it, with particular emphasis on the effective stress state of soil elements immediately upslope and downslope of the pile. This soil stress state can be calculated by virtue of instrumentation measuring both horizontal total stress and pore pressures at locations close to the upslope and downslope faces of the piles. By comparison of results obtained for both rigid and flexible piles, conclusions will be drawn as to the effects of pile flexibility on modifying the behavior of the soil-pile system.

Two dimensional model tests are performed on a horizontal cross section of a soil-pile system in a pile foundation. The objective of the model tests is to evaluate local soil displacement field in the vicinity of the piles associated with a global displacement of soil around the pile foundation. Two dimensional effective stress analyses in horizontal plane are also performed to generalize the findings from the model tests. An effective stress model based on multiple shear mechanism is used through a computer code FLIP. Primary findings from this study are as follows: (1) In dry condition, displacement vectors are directed away from pile front, and displacement at pile side rapidly decreases with an increasing distance from soil-pile interface. In undrained condition, displacement field shows vortexes at pile side associated with push-out/pull-in pattern of displacements in front of and behind the pile. (2) Distribution of local soil displacement between piles deployed perpendicular to direction of global displacement of soil shows high strain concentration (i.e. discontinuity in displacement) at soil-pile interface.

Liquefaction-induced lateral spreading continues to be a major cause of damage to deep foundations. Currently there is a huge uncertainty associated with the maximum lateral pressures and forces applied by the liquefied soil to deep foundations. Furthermore, recent centrifuge and 1g shaking table tests of pile foundations indicate that the permeability of the liquefied sand is an extremely important and poorly understood factor. This article presents experimental results and analysis of one of the centrifuge tests that were conducted at the 150 g-ton RPI centrifuge to investigate the effect of soil permeability in the response of single piles and pile groups to lateral spreading.

Three design specifications (highway bridges, railway facilities, and building foundations) currently used for pile foundations in liquefiable and laterally spreading ground in Japan are described and compared in this paper. Although all the specifications take into account the effects of ground deformation due to liquefaction or lateral spreading on the performance of the pile foundation, the coefficient of subgrade reaction used in the three specifications varies considerably. The effect of the reduction of the coefficient of subgrade reaction due to liquefaction on the response of the pile is investigated for two types of piles. Finally, a design chart of the pile foundation in liquefiable and laterally spreading ground studied in the research committee of the Japanese Geotechnical Society is introduced.

This paper presents a simplified procedure for computing liquefaction-induced lateral spreading forces on piles or abutment walls supported on piles. This approach is in essence to develop two curves of load versus displacement behavior to consider the soil-pile-structure interaction in a "decoupling" way. A curve of lateral spreading force versus displacement is first established by slope stability analysis for the soil mass behind the wall based on the Newmark chart that relates the standardized maximum displacement of the soil mass to the limiting acceleration coefficient. A second curve of lateral load versus di splacement for the abutment wall supported on piles is then obtained by push analysis of the wall structure under the lateral spreading force. The intersection between the two load-displacement curves is considered as the lateral spreading force on the abutment wall. Results of analyses performed for one bridge are presented in this paper to illustrate the proposed method. It appears that the simplified method of accounting for the pile-wall-soil interaction can satisfactorily predict the lateral spreading forces on abutment walls supported by piles without the uncertainty in determining the seismic coefficient related to the use of the Mononobe-Okabe method.

This paper examines earth pressure acting on an embedded footing and its effects on pile forces, based on both liquefaction and non-liquefaction tests using a large-scale laminar shear box. The following conclusions are drawn: (1) The total earth pressure defined by the difference in earth pressure between passive and active sides in the non-liquefaction tests varies significantly depending on its phase relative to the soil inertia around the embedded footing as well as on the relative displacement between soil and footing; (2) The total earth pressure in the liquefaction test, by contrast, depends mainly on the relative displacement because the soil inertia gets small in liquefied soil; (3) The total earth pressure in the non-liquefaction tests tends to be out of phase by 180 degrees with the superstructure inertia, reducing the shear force and bending moment at the pile head; and (4) The total earth pressure in the liquefaction tests tends to be in phase with the superstructure inertia, making the bending moment at the pile head large. A method for estimating the total earth pressure considering its phase relative to the superstructure inertia as well as the effects of soil inertia has been proposed. The proposed method gives a reasonable explanation of the difference in earth pressure between different tests.

Konan junior high school in Hokkaido, the northernmost main island of Japan, suffered severe damage to its superstructure and pile foundations during the Tokachi-oki Earthquake (M=8.0), on September 26, 2003. The objective of this study is to verify the main factors causing this damage by field investigations and earthquake response analyses. Detailed field surveys were carried out on the superstructure and pile foundations immediately after the quake. Analyses of the structure of the pile foundation were conducted using a numerical model taking into account the effect of soil nonlinearity. Strong motions observed near the school were employed as the input motion to the analysis model. It is found that compression failures at the pile heads were caused not only by the inertial force from its superstructure but also by large displacement of a very soft clay, leading to differential settlement as well as shear cracks in walls of the superstructure.

This paper presents the assessment of push-over analyses of single pile and pile groups in laterally spreading ground using p-y approach. The results from full-scale blast induced lateral spreading experiments in Japan were used to evaluate the numerical model. For the single pile, the pile responses were determined by imposing the measured free field displacements to the Winkler spring model. For pile groups, the piles in the group were modeled as an equivalent single pile, together with reduced stiffness p-y springs to account for the effect of pile group. A rotational spring was also introduced into the model to represent the rotational pile group stiffness. The results from the analyses for both single pile and pile groups were in good agreement with the measured responses from the experiments, thus suggesting that this method may be used in common design practice for piles in laterally spreading ground problems.

Three case histories of failure of piles caused by ground lateral spreading are back studied using the beam on Winkler foundation method including nonlinear soil springs and a nonlinear moment-curvature relation for the pile. Soil loads on the pile are modeled based on the modified Bouc-Wen model. The lateral spreading effect is modeled by laterally displacing the soil spring support by an amount equal to the free field permanent deformation. Design procedures suggested by Tokimatsu et al. and by JRA were also used for case histories evaluation and compared to available observation results. Buckling analysis was performed to determine critical buckling load. Whether the cause of failure for the evaluated cases was due to either bending or buckling is also discussed in the paper.

This paper describes the results of an investigation of the dynamic stiffness of pile foundations in liquefiable soils. Such soils frequently exist near bodies of water where bridges are required. Because potentially liquefiable soils are generally weak and compressible even under static conditions, bridges founded on them are usually supported on pile foundations. During earthquake shaking, the excess porewater pressure that builds up in liquefiable soils influences both the seismic response of the soil deposit itself, and also the local interaction with any foundations that extend through the liquefiable soil. These phenomena can strongly influence the stiffness of pile foundations in liquefiable soils during earthquakes, and need to be accounted for if accurate evaluations of pile stiffness are to be obtained. In this work, tools and procedures were developed for evaluation of the stiffness of pile foundations in liquefiable soils during earthquakes.

The load-displacement response of liquefied soils surrounding deep foundations has been investigated previously by conducting lateral load tests on a full-scale 3 x 3 steel pipe pile group embedded in sands liquefied by controlled blasting. Based on this testing, p-y curves for fully liquefied sand were developed which included stiffness degradation due to multiple large-displacement load cycles. Additional study of the soil resistance-displacement time histories used to develop these p-y curves indicates that p-y curves for liquefied sand exhibit a generally linear response during an initial, relatively large-displacement cycle of loading. As the number of loading cycles increases, the stiffness of the soil response decreases, and p-y curves become increasingly shaped concave-up. The shapes of the p-y curves during initial and subsequent loading cycles are markedly different than that of a p-y curve for soft clay in the presence of water. One reason for this difference is that the soft clay model cannot reproduce the effects of contractive/dilative phase transformations. The lateral subgrade modulus for liquefied soil decreases rapidly with the first few loading cycles, losing between 50 to 70% of its initial value by the third cycle for a wide range of deflection levels.

Static seismic analysis procedures are evaluated against results from eight dynamic model tests performed on a 9-m-radius centrifuge to study the behavior of pile groups in liquefiable and laterally spreading ground. Either 0.73-m-diameter piles or 1.17-m-diameter piles were fixed into a large embedded pile cap in each test. The soil profile consisted of a gently sloping nonliquefied crust over liquefiable loose sand over dense sand. Each model was tested with a series of realistic earthquake motions with peak base accelerations ranging from 0.13 g to 1.00 g. Representative data that characterize important aspects of soil-pile interaction mechanics in liquefied and laterally spreading ground are presented. The observed results of static seismic analyses using Beam on Nonlinear Winkler Foundation (BNWF) models are compared with the centrifuge test data. The analyses are shown to reasonably predict peak bending moments and pile cap displacements when input parameters correspond well with test measurements, but significant errors arise when input parameters do not match the test observations even though they do correspond with commonly-used design assumptions.

The utilization of performance-based seismic design concepts at major ports often necessitates a reliance on numerical models for simulating the dynamic soil-foundation-structure interaction of piers and wharves. Construction and siting of pile supported wharves, which comprise the most common type of waterfront structure in cargo handling portions of major U.S. ports, includes pile embedment in submarine slopes routinely comprised of rock-fill underlain by weak marine soils. Field observations at ports around the world lead to the general conclusion that incipient, permanent slope deformations can be expected at ground motion levels as low as 0.15g, well below design levels in many regions. In light of the resources required to prevent slope deformation small permanent slope deformations are considered acceptable for most waterfront projects. The allowance in seismic performance criteria of limited, yet permanent, deformations of the slope and piles makes the design of pile foundations for port waterfront structures rather unique. Seismic analysis requires that both the kinematic and inertial aspects of the seismic loading of pile foundations are evaluated in a coupled manner. This paper addresses the seismic analysis of pile supported wharves; however, aspects of this work can be readily applied to other coastal infrastructure. In order to identify the strengths and limitations of practice-oriented methods for evaluating the seismic performance of pile supported wharves centrifuge modeling was employed to supplement field case history data for validation of two deformation-based analysis methods; a numerical 2D nonlinear, effective stress model, and a simple, straightforward procedure based on the rigid, sliding block concept. Modeling results demonstrate that the practice-oriented, sliding block model can yield suitably accurate deformation estimates, provided that the analysis method accounts for excess pore pressure generation and the stabilizing effect of the piles in the slope. The numerical procedure, which utilized a simplified stress-based pore pressure generation constitutive model, resulted in reliable deformation patterns and acceleration estimates, and pore pressures were somewhat well reproduced; however the maximum dynamic and residual moments in the piles were not well replicated. Shortcomings of the numerical geomechanical modeling are addressed. The limitations notwithstanding, the numerical procedure is advantageous compared to the sliding block method, as the pattern of soil deformation and pile curvature are more accurately simulated, a more complete picture of soil-foundation-structure interaction is provided, and the coupled performance of the waterfront slope and wharf is more clearly demonstrated.

An overview of nonlinear approach currently used by one group of practitioners in designing foundations for port facilities involving piles through potentially liquefiable soils is presented. The computer program FLAC is used as the overall model in this nonlinear approach, but other programs can be used as well. The approach consists of extensive parametric analyses using relatively simple models anchored in the knowledge of fundamentals and experience evaluating case histories. Given a postulated input ground motion, some parameters in the models are specified based on one-dimensional analysis using SHAKE and FLAC. Then, parametric gravity turn on analysis and parametric dynamic analysis are performed on a two-dimensional section without piles using pre-liquefied soil strength. Next, the section with piles is analyzed first using pre-liquefied soil shear strength. This process and the results from these analyses are used to develop required geotechnical design parameter values. Finally, the designed section is analyzed modeling an excess pore water pressure generation (liquefaction process) and design parameter values are adjusted as necessary.

The purpose of this study is to clarify the loading rate dependency of subgrade reaction of a single pile in saturated or liquefied soil. The parametric studies were performed by incorporating the soil-pore water coupled formulation and a newly proposed constitutive model for sand. The model can explicitly treat the degree of liquefaction by changing the lower limit of the mean effective stress. The liquefied soil at a certain depth around a pile is modeled with finite elements under plane stress condition. The subgrade reaction of a pile is calculated under various loading frequencies, permeability and degrees of liquefaction. The results of the parametric studies show that the subgrade reaction varies with loading rates, permeability and degrees of liquefaction, and the dimensionless subgrade reaction normalized by that without seepage has a unique relation with dimensionless time factor which consists of pile diameter, loading rate, permeability coefficient and coefficient of volume compressibility.

Nonlinear dynamic analyses using the finite element (FE) method are compared to the recorded responses of pile-supported structures in soil profiles that develop liquefaction and lateral spreading during earthquake shaking in centrifuge model tests. The centrifuge tests included a simple superstructure supported on a group of six piles. The soil profile consisted of a gently sloping nonliquefiable clay crust over liquefiable loose sand over dense sand. The FE models consisted of a two-dimensional model of the soil profile with a beam-column model of the structure and pile foundation that was attached to the soil profile by soil springs. Centrifuge recordings were used to calibrate the nonlinear FE models. Parametric analyses were used to evaluate the sensitivity of the FE results to variations in the modeling parameters and FE representation. With appropriate calibrations, the FE models were able to reasonably approximate the essential features of soil and structural response.

The paper presents application of beam on non-linear Winkler foundation method to piles in laterally spreading liquefied soils. In many cases, parameters used in the method are determined to fit particular case studies and are hardly applicable to the other separate cases, due to difficulties in the modelling of complicated phenomena with the simplified method. In this study, variation of p-y curve parameters for piles in laterally spreading liquefied soils is systematically examined by numerically analysing physical model tests undertaken by independent researchers.

In the simplified analysis of piles subjected to lateral spreading, a single pile model is commonly used, and therefore, one of the key issues is how to apply this model to the analysis of pile groups, particularly in the case when individual piles in the group are subjected to different lateral ground displacements. This paper presents a simple concept for analysis of pile groups by using a single pile model. The evaluation of the pile group response by means of the single pile analysis is carried out in two steps, each consisting of several iterative calculations. The objective of the first series of analyses is to estimate the lateral displacement of the pile group at the pile-head level while in the second series of analyses the response of each individual pile is evaluated by considering its particular position within the group.

Centrifuge testing has been undertaken to investigate instability failure of pile groups during seismic liquefaction, with specific reference to the `top-down' propagation of liquefaction during the earthquake and to account for initial imperfections in pile geometry. The results of these tests were used to validate numerical models within the finite element program ABAQUS, based on the popular p-y analysis method. Pseudostatic classical and post-buckling analyses were conducted to examine the collapse behaviour of the pile groups and were found to give reasonable predictions of collapse load and conservative predictions of the associated deflection conditions. This numerical model was compared to currently published methods which were found to over-predict collapse loads. The resulting insights into the collapse of axially loaded pile groups revealed that the failure load is strongly dependent on both the depth of liquefaction propagation and initial imperfections, which reduce the collapse load.

This paper describes horizontal and vertical 2D shaker that can make horizontal and vertical motions of the shaking table independently in the TITech Mark III Centrifuge. In the 2D shaker, two inclined servo-hydraulic actuators with horizontal angle of 30 degrees are symmetrically mounted underneath the shaking table. Vertical static load of the moving mass is supported with four air springs so that the force from the actuators underneath the shaking table can act mainly for the dynamic motions of the mass. Results on a centrifuge tests using the shaker are also introduced. In the test effectiveness of active drainage as a countermeasure against liquefaction for a piled foundation was attempted to study.

This paper discusses the possibility and effectiveness of soil desaturation as a liquefaction countermeasure technique. Soil type and soil characteristics that can be effectively desaturated during air injection, and longevity of air bubbles in the soil are described first. Inspection of samples taken by ground freezing method from SCP improved ground revealed that degree of saturation (S_r) shortly after SCP was lower than 91 % in the sand layers and a good correlation was found between S_r and 5% diameter of the soil. Degree of saturation of soils after several years ago was noticeably, but not significantly, higher as compared with that shortly after ground improvement, indicating longevity of air bubbles injected in the improved soil. Next, a series of shaking table tests of desaturated ground with and without rigid foundation resting on it is discussed. In the tests, vacuum pressure was introduced to the models to avoid distorted scaling factor.

Results of a series of centrifuge experiments to study the dynamic response of pile foundations under lateral spreading were compared to the results of numerical analysis. Experiments were carried out under the centrifugal acceleration of 40 G. Piles in the model foundation were lined up 3 by 3 pattern with a spacing of three pile diameter, and both pile head and bottom were rotation fixed. Piles were placed in the inclined ground of saturated sands and applied lateral loads due to the ground deformation. The effective stress finite element analysis was conducted to simulate these experiments. Computed time histories of pile head acceleration and displacement were consistent with those obtained from experiments. However, much smaller surface ground deformation in simulation might cause small amplitude of bending moments. Some calibration in the numerical modeling may be required to have more consistent results on bending moments.

Earthquake-induced deformations of a bridge abutment underlain by liquefied soil may be reduced by the restraining forces that come from the pile foundation and bridge superstructure. The reduced abutment displacements may, in turn, reduce the loads or displacement demands that are imposed on the piles. Design methods that account for this "pile pinning" effect have been used in practice and included in the recent NCHRP 472 guidelines, but have not previously been evaluated against physical data. An initial evaluation of these pile pinning analysis methods is presented via comparison against the results of a recent dynamic centrifuge model test. Four aspects of these analysis methods are identified as requiring modification or adjustments to avoid an unconservative estimate of pile foundation performance. These and other design issues are discussed.