Earthquake Geotechnical Engineering for Protection and Development of

Environment and Constructions – Silvestri & Moraci (Eds)

© 2019 Associazione Geotecnica Italiana, Rome, Italy, ISBN 978-0-367-14328-2

Dynamic soil-structure interaction of piles in soft soils

X.F. Vera-Grunauer, M.A. Chavez-Abril, J. Ordoñez-Rendon, O. Gonzalez-Carrion &

S. Lopez-Zhindon
Geoestudios S.A., Guayaquil, Ecuador

ABSTRACT: During a seismic event, a coupled behavior known as soil-structure inter-

action (SSI) may be highly relevant if the given structural stiffness is large in comparison to
the underlying soil rigidity. Regarding a foundation’s structural seismic design, some method-
ologies are usually employed in professional practice, which depend on the considered degree
for controlling the foundation’s non-linear (inelastic) structural behavior. SSI could be ana-
lyzed either by a fully coupled analysis or a decoupled SSI study. This research provides a
framework applied for deep foundations as a means to realize an enhanced decoupled SSI
analysis through an inertial co-efficient β, which considers the lack of simultaneity between
the inertial and kinematic inter-action effects and in estimating the structural demand of piles
with limited or no nonlinear behavior. Numerical and analytical methods were employed in
addition of a comprehensive site investigation, involving a deltaic-estuarine environment in
the city of Guayaquil, Ecuador.

1 INTRODUCTION

Soil-structure interaction (SSI) becomes a relevant issue in geotechnical earthquake-engin-

eering when a relatively stiff structural system overlying a relatively loose/soft soil is sub-
jected to a seismic event. SSI is composed of two mechanisms, an inertial and a kinematic
component. Inertial interaction describes how inertial forces generated in the structure
induce foundation displacements and rotations that would not occur if the structure had a
fixed base, resulting in additional displacements in the structure and a change of fundamen-
tal frequency or “period lengthening.” Kinematic interaction describes how waves propagat-
ing in the free field differ from the motion of the foundation(s) due to a differential
foundation–soil stiffness, ground-motion incoherence, and wave-scattering effects (Boulan-
ger et al. 2007). Inertial interaction analysis is the conventional practice when studying a
complete structural system under seismic loading, often ignoring the kinematic interaction,
which could be a detrimental decision.
Relevant SSI case histories involving pile foundations such as the presented works by Gaze-
tas, et al. (1993) and Koyamada, et al. (2006) present the highly relevant role that kinematic
interaction had at two Japanese locations. One involved the heavy damage upon the given pile
foundation for the Obha Ohashi Bridge and partial destruction of the pile foundation from
the Konan junior high school. Both locations exhibited a 20 m depth soil profile of very soft
soils with characteristic shear wave velocities lower than 100 m/s. The Japanese bridge case
history involved the modeled and recorded (gages) bending strains of Ohba Ohashi’s pile
foundation. The second case history involved modeled bending moments of the collapsed pile
foundation of Konan junior high school. Kinematic bending strains and moments modeled
from both case histories present a highly significant portion of the total effects upon the piles
both at the pile head and at great depths wherever interfaces of soil layers with sharply differ-
ing stiffness exist, which meant deformations and internal forces that were beyond the ultim-
ate limits of the affected piles.

5522
 Guayaquil is one of the most important cities in Ecuador and has a significant earthquake
hazard arising mostly from a subduction tectonic environment. Most of the city is located in a
deltaic estuarine environment of the lower Guayas River, and the soil stratigraphy consists of
deep soft sediments over soft rocks from the Cretaceous period. These deltaic-estuarine sedi-
ments were deposited in a brackish environment, and they are unusually week and highly
compressible. The involved project for the study consists of an aerial cableway which is
planned to connect several locations from the downtown of Guayaquil, crossing a river of
approximately 1500 m width to a nearby city, Duran. The air cableway project will be sup-
ported by various checkpoint stations, which are structures that are raised from the street sur-
face through massive columns, as well as having other isolated massive columns that will serve
as a necessary component for this mentioned project. The involved SSI analysis will focus on
the study regarding one massive representative column (of the ones mentioned above), as well
as its underlying foundation when a seismic event occurs.

2 MOTIVATION

The primary motivation for employing the proposed framework in this project was to validate
recent studies performed by Khosravifar & Nasr (2018), regarding decoupled simplified SSI
analysis results and how to enhance them for being closer to a more realistic and less conser-
vative approach of deep foundation designs. Assuming 100% of the inertial component and
100% of the kinematic component during a seismic event involving a pile group foundation is
considered a significantly conservative approach. Conservatism occurs due to an idealization
of simultaneity regarding internal forces affecting the piles through both inertial and kine-
matic interactions.
Some previous studies performed by authors such as Chang et al. (2006), Tokimatsu et al.
(2005), Brandenberg et al. (2005) and Boulanger et al. (2007) for elastic piles with lateral
spreading in liquefiable soils converged to a general approach of adopting 100% of kinematic
component and 50% of an inertial component (moment and shear demand). The inertial com-
ponent of 50% may be referred to an inertial coefficient β which considers the lack of simul-
taneity in estimating the structural demand of piles with limited or no nonlinear behavior.
Khosravifar & Nasr (2018) obtained results regarding a back-calculated coefficient β for
crustal and subduction zones (i.e., near field and far field seismic input motions, respectively),
from which values β values of 0,60 (near field) and 0,75 (far field) were obtained from a given
statistical trend study.
The inertial coefficient β, presented in Equation 1, shows how the internal forces of a pile
foundation are represented in terms of soil-structure interaction response (kinematic and iner-
tial components) for a simplified decoupled analysis.

Mtot ¼ Mkin þ β  Min ð1Þ

Where Mtot = total moment generated at a particular pile location, Mkin =kinematic moment
component, Min =inertial moment component, and β=inertial coefficient.
Based on the aforementioned studies, a need to obtain appropriate design loads for the
given deep foundations of the related air cableway project was needed. Employing 100% of
both inertial and kinematic loads may be too conservative, while using 100% kinematic and
50% inertial may be on the unconservative side, especially for soft soil deposits (i.e., deep clay
deposits).

3 FRAMEWORK

A comprehensive description of the developed methodology for this research project is pre-
sented in this section.

5523
 The initial and most relevant component was to perform sufficient site investigation, which
involved in-situ and laboratory tests (i.e., SPT tests, CPTu tests, conventional monotonic soil
mechanic laboratory tests), as well as geophysical survey. The geophysical tests included Sur-
face Wave Analysis (MASW+MAM) and microtremor H/V technique (Nakamura method)
for correctly assessing the existing shear wave velocity profiles, fundamental elastic site
period, as well as the approximation of the rigid base’s (half-space) depth. SPT and CPTu
tests provided the means to reasonably characterize the soil profiles regarding their shear
strength, stiffness and liquefaction potential. Additionally, typical static soil laboratory tests
such as oedometer, unconfined compression, Atterberg limits, and grain size distribution were
employed in this study. Monotonic direct simple shear (DSS) tests (Vera-Grunauer, 2014)
were employed for the advanced constitutive soil model calibration of the numerical simula-
tions from this project. Cyclic soil tests (Vera-Grunauer, 2014) were used as an additional ref-
erence for the calibration of shear stiffness G/Gmax modulus reduction curves, as well as given
material damping ratio (ξ).
The Ecuadorian construction standard, NEC (2015), provides a design response spectrum
at outcropping rock; as a result, a large group of earthquake motions that were seismologic-
ally compatible with events governing the hazard at the structure spectral ordinate from the
design response spectrum were considered. The selected and scaled earthquake motions from
25 subduction events and 25 crustal faults events were chosen. The longer duration signals
(far field) represent subduction interplate zones, whereas shorter duration signals (near field)
represent near crustal faults. Proper site response analyses (SRA) were performed with the
given seismic data, through one-dimensional nonlinear effective stress analysis (software
DEEPSOIL). The soil parameters used for the mentioned 1-D SRA were based on the
obtained soil parameters from in situ geophysical tests, as well as laboratory/field tests and
empirical data regarding the modulus reduction curves (G/Gmax and ξ). Darendeli (2001) and
Menq (2003) were the empirical basis to formulate the required modulus reduction curves for
fine-grain soils (using the data from Vera-Grunauer, 2014) and granular soils respectively.
The results from DEESOIL of the several input motions were the main basis to properly simu-
late the framework of this project, which involved a direct or fully coupled SSI analysis by
means of a finite element method (FEM). Two particular seismic input motions, a near field,
and a far-field motion were the characteristic input motions of this study, as they represented
actual seismic results of the analyzed site and thus were used as the primary reference point
for this research project. The name fpnf represents the local characteristic near-field motion of
an event with a Mw=6.7, whereas ewff represents local far-field motion of an event with a
Mw=7.8. The elastic site period interpreted from the geophysical survey yielded an approxi-
mate value of Te=1.50 s. Soft normally consolidated soils with PI=25-50%, Vs30=130 m/s and
with little to none granular soils are present in the first 30 meters of soil profile at the studied
site location. At further depths beyond 30 meters and up to 145m (half-space location), inter-
calations of thick hard clay (with unconfined compression strength of 0.5 MPa) and very
dense sand layers are present.
The chosen framework for the FEM analysis was through PLAXIS 2D, employing the
hardening soil model with small strains (HS small). This advanced constitutive soil model was
of reasonable application for this case study due to the absence of liquefiable soil, meaning no
need for more advanced models to capture such behavior (e.g., UBCSAND, Hypoplasticity
with intergranular strains, PM4SAND). The HS small model is a nonlinear elasto-plastic
framework that allows considering stress dependency-stiffness, shear and volumetric harden-
ing (mobilization of shear strength and yield surface), and more importantly for dynamic
applications, it may consider small-strain stiffness. In a similar manner to the preliminary
SRA analysis in DEEPSOIL, proper calibration of modulus reduction curves (G/Gmax and ξ)
was performed, as well as a calibration of monotonic shear strength and stiffness parameters.
Available monotonic simple shear test data for characteristic fine soils of the studied area
were used, these tests represented lightly apparent (due to the presence of Pyrite cementation)
over-consolidated clays (OCR=1,7) at vertical effective stresses of 105 kPa approximately.
The HS small model was calibrated with the given DSS results in terms of effective stress

5524
paths, maximum shear stresses, and excess pore pressure generation, yielding satisfactory
results.
An initial model was conceived in PLAXIS 2D by means of a simplified 1-D soil column to
validate the numerical results of this nonlinear effective stress analysis with the ones obtained
from DEEPSOIL, generating satisfactory results for both motions fpnf and ewff. The bound-
ary conditions of the FEM model were similar to the applied ones in DEEPSOIL, using a
rigid base as bottom boundary and free field conditions for the lateral boundaries. After
proper validation of the numerical SRA model, a 2D FEM model was generated, including
the structural system as a purely elastic and simplified single degree of freedom (SDOF)
model, including as well the pile cap and pile foundation group. The structures were modeled
as plate element, which represented a massive reinforced concrete column (100 x 200 cm) of 10
meters height. The pile cap of 120 cm thickness was also modeled as a plate element. For the
deep foundations (pile group), four massive reinforced concrete sections of 50 x 50 cm with
approximately 30 meters of length were modeled by means of embedded beam elements. The
axial load that represented the modeled structure (massive column) was of 2500 kN with a
defined structural natural period of Tcol= 1,10s. The material damping assigned to the
involved structures was of 5%, as commonly adopted for reinforced concrete structures. A
fully fixed pile-head connection was adopted at the numerical models (PLAXIS) through rigid
connections between the plate (cap) and embedded beam (piles).
From the SSI fully coupled FEM dynamic analysis (PLAXIS 2D), results involving internal
forces of the piles, as well as free field stress-strain results were obtained to apply an inertial
and kinematic response upon a simplified and decoupled SSI model of equivalent nature to
the one employed for the FEM framework. This decoupled method, which is an equivalent
static (pseudo-static) analysis (ESA), was realized by means of lateral load pile analysis
through a beam on elastic foundation (BEF) analysis through the commercial software
LPILE. The indirect SSI method was composed of appropriate P-Y curves applied to this
study, as well as kinematic and inertial input demands (obtained from FEM results). The
applied kinematic input was a soil lateral displacement profile obtained from the integral of
the shear strains over the entire soil profile at free field conditions. The used inertial response
of the ESA framework for a fixed pile head connection solely involved the axial and shear
forces obtained from the numerical analysis, whereas no bending moments obtained from the
numerical analysis were considered as input in the ESA analysis. Due to the significant differ-
ence in rigidity between the pile cap (higher stiffness) and pile elements (lower stiffness), the
applied bending moments are taken by the pile cap and are transmitted to the piles as axial
forces (compression and tension). As a result, bending moments at the pile heads are con-
sidered in the ESA framework by means of the input axial forces. It must be noted that both
the ESA and FEM dynamic analyses were based on purely elastic structural elements.
The outcome of the ESA analysis yields results of internal forces (bending moments and
shear forces), which are employed to back-calculate the results of the commented ESA ana-
lysis in such a way that the magnitude of bending moments of the ESA approximately
matches the ones of the FEM dynamic analysis. The procedure for handling such back calcu-
lation is through an inertial coefficient β, which considers the lack of simultaneity and model
limitation by means of ESA in estimating the structural demand of piles with limited nonlinear
behavior and thus, the inertial demand used as input for ESA analyses are affected by β. The
aforementioned back analysis was realized for the 50 adopted input motions, which include
the two local characteristic motions for Guayaquil. Finally, based on statistical trends of the
50 input motions and their respective analytical/numerical simulations, a coefficient β was
obtained for the involved research project.

4 RESULTS

The advanced constitutive soil model (HS small) used for FEM simulations, was initially cali-
brated with respect to a monotonic DSS test in terms of excess pore pressure, shear strength
and stiffness of a chosen characteristic soil sample from the studied site. Figure 1 illustrates

5525
Figure 1. Numerical calibration of HS small model with empirical data of BSF clay sample at a 12m
depth layer, OCR=1.7 and σ;vo =105 kPa.

Figure 2. Comparison between SRA simulations for characteristic near-field signal fpnf.

the numerical calibrations which show a reasonable match with respect to the available experi-
mental data (Vera-Grunauer, 2014) of a slightly over-consolidated clay sample named BSF.
Simulated results regarding the FEM and BEF models applied for this study are depicted in
terms of the characteristic input signal for near-field fpnf. Results from the characteristic far
field input signal ewff show a similar trend to its near field counterpart, which is why only the
results of the signal fpnf have been shown. Modelled shear strains, shear stresses and peak
ground acceleration spectra are shown in Figure 2, from which the results of DEEPSOIL,
PLAXIS 1-D simplified soil column and PLAXIS 2D fully coupled SSI model (at free field
condition) depict reasonable comparisons for a given case of a seismic input motion (near
field fpnf motion).
Although an inertial coefficient β was required for enhancing a complete SSI based on an
ESA, the results of the kinematic interaction for the simplified BEF model in comparison to
the numerical simulations in PLAXIS 2D yielded reasonable results in terms of the maximum
magnitude of internal forces. Figure 3 shows the comparison of shear forces and bending
moments for the modeled piles subjected to purely kinematic interaction through the adopted
characteristic near field motion fpnf. The calculated bending moments (Figure 3) and shear
forces through FEM (PLAXIS 2D) are represented as two peak-value functions that depict
the maximum positive and negative scalar results of internal forces over the entire dynamic
simulation. ESA (LPILE) internal forces are represented (Figure 3) as a single output function
of depth. The obtained ESA kinematic response results (Figure 3) are generated due to the
applied input prescribed lateral displacement profile (monotonically applied maximum lateral
deformation) obtained from the numerical output of the fully coupled analysis at free field
conditions.

5526
Figure 3. Comparison between BEF and FEM analysis of purely kinematic interaction for local site
characteristic near-field signal fpnf.

Figure 5. Mismatch of horizontal accelerations at pile head and top of the structure (column) locations
for characteristic near-field signal fpnf.

Figure 6. Mismatch of lateral displacements at pile head and top of the structure (column) locations for
characteristic near-field signal fpnf.

A fundamental keynote in this study is to compare the dynamic behavior of both the pile
foundation and the overlying structure in terms of peak accelerations and lateral displace-
ments over the analyzed time interval. Figures 5 and 6 depict a mismatch between both the
pile head and the structure’s upper location when referring to horizontal peak accelerations

5527
and horizontal displacements. This mismatch represents a lack of simultaneity in the occur-
rence of the dynamic maximum dynamic effects between the pile and overlying structure. If
both peaks matched, severe structural damage could compromise the structure’s performance
by the given amplified effects during a seismic event.
It must be noted that the peak accelerations observed in Figure 5 should not be related to
the time lapse in which a maximum curvature ϕ=M/EI is developed for a given case such as
the one involved in this study. In a similar way to what is shown in Figure 5, Figure 6 depicts
mismatching maximum pile head and overlying structure’s lateral displacements, meaning
that both structures (underlying and overlying) do not experience peak dynamic-induced
deformations simultaneously (i.e., structural damage is not at its maximum potential). As
curvature ϕ depends on bending moments M and bending stiffness EI, it implies that the max-
imum moments are not expected to occur necessarily at coinciding time lapses of peak acceler-
ation, but instead at time lapses of maximum curvature. During an earthquake event and an
existing pile foundation, pile maximum bending moments may not necessarily occur in one
particular direction of ground-structure motion, but instead in a different direction. For
instance, since earthquake motions do not necessarily travel unilaterally through soil or rock,
a pile foundation may experience the largest bending moment or curvature while simultan-
eously the maximum acceleration is occurring in a different direction.
The applied framework of this study oversees the aforementioned behavior regarding max-
imum bending moments since it relies on a two-dimensional basis. As such, it is not possible
to model input motions which may be acting during an earthquake event in more than one
direction, as well as the involved structural internal forces that develop due to it. The pre-
sented framework depends on a prescribed dynamic motion direction, in which maximum
structural internal forces and deformations, as well as maximum accelerations occur over the
same defined direction.
For shear waves in an unbounded homogeneous medium, the shear strains are directly pro-
portional to the particle velocity at all times. As a result, the shear strain time series of an
earthquake can be calculated as the velocity time series divided by the shear wave velocity (as
a proxy of a characteristic induced shear strain, PGV/Vs30, PGV at outcropping rock input
motion and site dependent Vs30 were considered). Figures 7 and 8 depict the overall results
involving the estimation of the inertial coefficient β by means of back analysis calculations
through the ESA adopted, as well as the input inertial and kinematic loads obtained from the
FEM analysis of every single input motion out of a total of 25 near field NF input motions

Figure 7. Trends for inertial coefficient for 25 far-field (FF) input motions and 25 near-field (NF) input
motions with respect to a characteristic induced shear strain, PGV/Vs30, and wavelength (λ=Tm*Vs30).

5528
Figure 8. Trends for inertial coefficient for 25 far-field (FF) input motions and 25 near-field (NF) input
motions with respect to a characteristic induced shear strain, and wavelength (λ=Tm*Vs30).

(crustal zones) and 25 far field FF input motions (subduction zones). Evident dispersion of
data from Figures 7 and 8 is observed, in one of the cases showing a trend of decreasing β
with the increment of the characteristic induced shear strain. In Figure 7 an additional data
dispersion is shown by means of the mean period of the input motions Tm and the site-depend-
ent Vs30 which relate to a particular wavelength (λ=Tm*Vs30) per input motion (Figure 8).
Nevertheless, a median value out of the 50 data points was calculated, which is equal to
β=0,75. It is further commented that the characteristic near field input motion fpnf yielded a
coefficient β=0,65 and for the characteristic far field input ewff an inertial coefficient β=0,70
was obtained.
A relevant aspect of this study is that the FEM and BEF analyses were performed for ideal-
ized purely elastic structural materials, from which significant limitations are expected in
terms of actually considering plastic inertial forces generating the imminent failure of the
structural system. If a limited nonlinear behavior is chosen as a seismic design approach, an
appropriate amount of ductility capacity must be considered in the design.
Figure 8 depicts a trend of β values above unity at relatively short wavelengths, which repre-
sent an underestimation of pile kinematic demands as well as an overestimation of inertial
demands, and this occurs due to the way in which lateral pile displacements are being con-
sidered. The proposed framework depends on a soil lateral displacement via integration of
shear strains that monotonically increase from the pile tip up to the pile head. A maximum
curvature ϕ or bending moment M induced over a pile foundation due to dynamic excitation
does not necessarily generate as a result of a curvature ϕ developed from a monotonically
increasing horizontal soil deformation, but instead may depend on the largest bending
deformation differential over a total given pile length. An input motion with short wavelength
would induce a more prominent deformed (due to an underestimation of kinematic demand)
pile with several curvature inversions, rendering the monotonically increasing lateral soil dis-
placement approach adopted in this framework as a limitation in accurately predicting max-
imum bending moments and pile kinematic demands.

5 CONCLUSIONS

The characteristic inertial coefficients from seismic data related to Guayaquil’s area yielded
comparable results to the median from the 50 input scaled motions used for this project, thus
evidencing reasonable results within the scope of this study. A presented framework for soil

5529
structure interaction was proposed and to a certain degree validated in comparison to earlier
works regarding the same topic of relating an inertial coefficient β within a simplified SSI
framework. When employing a seismic design methodology with limited nonlinear behavior,
care should be taken in the proper selection of reduction factors to allow such design to have
the required ductility.
Some limitations are given for the performed analyses, such as relying on a purely elastic
structural modeling, as well as the idealization of a single degree of freedom, whereas struc-
tures, in reality, do not represent such conditions. Prediction of maximum bending moments
had a limitation regarding seismic input signals of relatively high frequencies (small wave-
lengths) since the kinematic demand of the modeled piles was underestimated and thus larger
curvature ϕ values (bending moments) were generated. One more relevant limitation was the
lack of sufficient input signals from the local site, which forced the employed framework to be
completed mostly (except for the two local characteristic input motions) with scaled input
motions of close regions and similar tectonic mechanisms of failure. More reliable structural
seismic responses may be expected for upcoming numerical models that involve three-dimen-
sional frameworks, thus being able to capture structural forces significantly better. SSI ana-
lyses of pile foundations could be better analyzed through three-dimensional frameworks,
although, significant limitations regarding dynamic boundary conditions in 3D models still
prove a challenge for such frameworks to be as reliable as the current state of the art. A rec-
ommendation for future investigations may be to compare decoupled SSI analyzes with two
combinations to evaluate which yields larger results, the first one involving solely 100% of
inertial interaction, and the second one involving a 100% kinematic interaction + inertial
interaction multiplied by an adopted β coefficient.

REFERENCES

Chang, D., Boulanger, R. W., Brandenberg, S. J., & Kutter, B. L. 2006. Dynamic analyses of soil-pile-
structure interaction in laterally spreading ground during earthquake shaking. Seismic Performance
and Simulation of Pile Foundations in Liquefied and Laterally Spreading Ground (pp. 218–229).
Boulanger, R. W., Chang, D., Brandenberg, S. J., Armstrong, R. J., & Kutter, B. L. 2007. Seismic design
of pile foundations for liquefaction effects. Earthquake Geotechnical Engineering (pp. 277–302).
Springer, Dordrecht.
Brandenberg, S. J., Boulanger, R. W., & Kutter, B. L. 2005. Discussion of “Single Piles in Lateral
Spreads: Field Bending Moment Evaluation” by Ricardo Dobry, Tarek Abdoun, Thomas D.
O’Rourke, and SH Goh. Journal of Geotechnical and Geoenvironmental Engineering 131(4),529–531.
Darendeli, M. B. 2001. Development of a new family of normalized modulus reduction and material
damping curves.
Gazetas, G., Tazoh, T., Shimizu, K., & Fan, K. 1993. Seismic response of the pile foundation of Ohba-
Ohashi bridge.
Khosravifar, A., & Nasr, J. 2018. Modified design procedures for bridge pile foundations subjected to
liquefaction-induced lateral spreading. DFI Journal-The Journal of the Deep Foundations Institute
1–14.
Koyamada, K., Miyamoto, Y., & Tokimatsu, K. 2006. Field investigation and analysis study of damaged
pile foundation during the 2003 Tokachi-oki Earthquake. Seismic Performance and Simulation of Pile
Foundations in Liquefied and Laterally Spreading Ground (pp. 97–108).
Menq, F. Y. 2003. Dynamic properties of sandy and gravelly soils. The University of Texas at Austin.
Tokimatsu, K., Suzuki, H., & Sato, M. 2005. Effects of inertial and kinematic interaction on seismic
behavior of pile with embedded foundation. Soil Dynamics and Earthquake Engineering 25(7–10),
753–762.
Vera-Grunauer, X. F. 2014. Seismic Response of a Soft, High Plasticity, Diatomaceous Naturally
Cemented Clay Deposit. Ph.D. dissertation, University of California, Berkeley, California, USA.

5530