Professional Documents
Culture Documents
Abstract: During earthquake ground shaking earth pressures on retaining structures can cyclically increase and decrease as a result of inertial
forces applied to the walls and kinematic interactions between the stiff wall elements and surrounding soil. The application, based on limit
equilibrium analysis, of a pseudostatic inertial force to a soil wedge behind the wall [the mechanism behind the widely-used Mononobe–
Downloaded from ascelibrary.org by Xavier Vera on 07/02/15. Copyright ASCE. For personal use only; all rights reserved.
Okabe (M–O) method] is a poor analogy for either inertial or kinematic wall–soil interaction. This paper demonstrates that the kinematic
component of interaction varies strongly with the ratio of wavelength to wall height (λ=H), asymptotically approaching zero for large λ=H,
and oscillating between the peak value and zero for λ=H < 2.3. Base compliance, represented in the form of translational and rotational
stiffness, reduces seismic earth pressure by permitting the walls to conform more closely to the free-field soil displacement profile. This
framework can explain both relatively low seismic pressures relative to M–O predictions observed in recent experiments with λ=H > ∼10,
and relatively high seismic earth pressures relative to M–O from numerical analyses in the literature with λ=H ¼ 4. DOI: 10.1061/(ASCE)
GT.1943-5606.0001312. © 2015 American Society of Civil Engineers.
Author keywords: Wall; Seismic earth pressure; Wave; Analysis; Dynamic testing.
inclined or incoherent waves (e.g., Veletsos and Prasad 1989). The is a simplifying assumption commonly used in soil–structure inter-
kinematic component of seismic earth pressures accounts for the action problems because it permits development of tractable solu-
interaction between the free-field motion ug ðzÞ and the structural tions. The values assigned to the stiffness terms should account for
wall elements, apart from their inertia and any external inertial coupling between various foundation vibrations modes, as de-
loads imposed upon the system. scribed in a subsequent paragraph. Although these assumptions
As shown in Fig. 1(b), the inertial interaction problem involves may appear to be limiting, the method can be readily extended
computation of the response of a structure and its foundation to to a wide range of practical conditions (including nonrigid founda-
the kinematic ground motions. Inertial forces from the structure tions as well as nonlinear and nonuniform soil) in a manner typical
cause additional relative displacements between the foundation of soil–structure interaction applications (NIST 2012) as illustrated
and the free-field, and additional increments of seismic earth pres- subsequently.
sure. The springs and dashpots in Fig. 1(b) represent the impedance The model derivation is described in two stages, as follows:
of the foundation from translation and rocking vibration modes (1) wall pressures and their resultant demands (forces and mo-
(e.g., Pais and Kausel 1988; Gazetas 1991). ments) are derived from the product of differential wall/free-field
In light of the previously mentioned soil–structure interaction displacement and wall-soil stiffness, and (2) equations for the stiff-
framework, the soil wedge concept currently used to evaluate seis- ness terms are developed (which is essential for analysis of force/
mic earth pressures will seldom have relevance to the physical moment demands and differential wall/free-field motions). Funda-
mechanisms producing those pressures. Even in cases where a state mental characteristics of wall–soil interaction derived from these
of active earth pressure (and its associated soil wedge) exists prior analyses are then described and illustrated using example solutions,
to seismic shaking, increments of earth pressure from earthquake which demonstrate that the wall–soil interaction response depends
ground shaking will arise from relative displacements between the strongly on the ratio of wavelength to wall height.
wall and free-field soil associated with kinematic and inertial inter-
action, which is not well-represented by a seismic coefficient acting
on an active wedge. Inertial interaction can mobilize large relative Wall–Soil Interaction Forces and Displacements
displacements when, for example, a massive structure is connected
to the wall elements and base shear mobilizes reaction stresses at A rigid U-shaped structure with vertical walls embedded in a soil
the soil/wall interface. Such effects can be evaluated as part of seis- profile experiencing vertically propagating harmonic free-field
mic structural response analysis if soil springs are included in the shear waves is shown in Fig. 2. The free-field ground motion is
structural model. Free-standing walls or basement walls not struc-
turally connected to lateral force resisting elements in structures
would have seismic earth pressures dominated by kinematic inter-
action, which is the topic addressed in the remainder of this paper.
Model Derivation
ing that the free-field ground motion is input to the free-ends of the
of wall area (superscript i denotes stiffness intensity measured in
spring elements. Substituting Eq. (4) into Eqs. (1) and (2), and re-
units of F=L3 ; details in the subsequent text); k ¼ 2π=λ = wave
quiring horizontal force and moment equilibrium between the wall
number; and λ = wavelength of the shear wave propagating
resultants and base reactions provides
vertically through the soil. The moment applied by the horizontal
Z H
soil–wall interaction stresses relative to the foundation slab base
elevation is PE ¼ kiy ½ug0 cos kz − uFIM − θFIM ðH − zÞdz
0
Z H Ky
ME ¼ kiy ðH − zÞ½ug0 cos kz − uw ðzÞdz ð2Þ ¼ ðu − ug0 cos kHÞ ð5aÞ
0
2 FIM
Z H
Eqs. (1) and (2) can be combined to calculate the location of ME ¼ kiy ½ug0 cos kz − uFIM − θFIM ðH − zÞðH − zÞdz
resultant PE , measured as distance h upwards from the base of 0
the wall as K xx θFIM
¼ þ kiz HB2 θFIM ð5bÞ
2
h M
¼ E ð3Þ Stiffness terms K y and K xx are multiplied by one-half to account
H PE H
for two vertical walls being attached to a single rigid base. The kiz
The depth-dependent wall displacement uw ðzÞ for a rigid wall term on the right side of Eq. (5b) represents the moment induced by
and foundation system is vertical tractions acting on the walls. By evaluating the integrals
and rearranging terms, the subsequent solution is obtained for
uw ðzÞ ¼ uFIM þ θFIM ðH − zÞ ð4Þ foundation displacements
uFIM ½6H2 ðkiy Þ2 þ 3k2 K y ðK xx þ 2kiz HB2 Þ þ 2k2 H3 K y kiy cos kH þ ½4kH 3 ðkiy Þ2 þ 6kðK xx þ 2kiz HB2 Þkiy sin kH − 6H2 ðkiy Þ2
Hu ¼ ¼
ug0 k2 ½H4 ðkiy Þ2 þ 2H3 K y kiy þ 6Hkiy þ 3K y ðK xx þ 2kiz HB2 Þ
ð6aÞ
θFIM B ½6k2 H2 K y kiy þ 24Hðkiy Þ2 þ 12K y kiy sin2 ðkH=2Þ − 6kH 2 ðkiy Þ2 sin kH − 3H2 k2 K y kiy
Hθ ¼ ¼B ð6bÞ
ug0 k2 ½H4 ðkiy Þ2 þ 2H3 K y kiy þ 6Hkiy þ 3K y ðK xx þ 2kiz HB2 Þ
These foundation displacements can then be inserted into and the base slab. Such partitioning is required to obtain the distri-
Eqs. (5a) and (5b) to obtain PE and M E for a compliant base con- bution of earth pressure acting on the vertical walls, which is our
dition. Varun et al. (2009) developed a solution for the kinematic objective. To overcome this problem, available solutions are first
displacements of axisymmetric caissons that is analogous to the used to define stiffness terms for individual foundation components
wall displacements derived here. under the assumption of no interaction between vibration modes
(i.e., the components are independent). Next, modification factors
χy and χxx are introduced to account for interaction between the
Stiffness of Wall–Soil System translation and rotation terms, respectively, such that the resulting
Having formulated the solution for PE and ME , the stiffness terms, global foundation stiffness matches published equations for em-
kiy , kiz , K y , and K xx are now evaluated. Classical inertial soil- bedded foundations. For simplicity, the base and wall stiffnesses
structure interaction (SSI) literature (e.g., summarized by Gazetas are both modified by the same χy and χxx terms.
1983; Mylonakis et al. 2006; NIST 2012) provides equations for
the overall stiffness of embedded foundations representing the inter- Horizontal Wall–Soil Stiffness Intensity kiy
action of the soil with the entire foundation system, but the global Kloukinas et al. (2012) developed a simple analytical expression
stiffness is not partitioned into contributions from the vertical walls for kiy for kinematic interaction between rigid vertical walls and
complicated conditions including soil layering and propagation ated substantial site response due both to the infinite impedance
of surface waves, shear strain has been found to range from 0.2 contrast (from the rigid base) and significant energy in the input
times to 1.7 times PGV=V s , with 1.0 being a commonly used value motions at the fundamental frequency of the backfill (where
for horizontal-component ground motions (Trifunac et al. 1996; λ=H ¼ 4).
Brandenberg et al. 2009), which provides an estimate of peak shear Five of the free-field surface motions were obtained from
strain consistent with the assumed shape of the soil displacement Ostadan (personal communication, 2013) and used to compute
profile. ug0 ðtÞ by double-integrating the surface accelerations in time.
Peak shear strain computed using these procedures can then be Those free-field motions were then applied using the proposed
converted to a representative uniform strain by multiplying the peak FD and SF solutions. Since the base of the wall was rigidly con-
shear strain by ðM − 1Þ=10, where M is moment magnitude (Idriss nected to the ground, only the stiffness term kiy is needed in the
and Sun 1991). The equivalent uniform shear strain would then be solution, and the frequency-dependent value was computed using
used to compute a value of G=Gmax from a selected modulus re- Eq. (7) with χy ¼ 1. Fig. 6(a) compares maximum earth pressures
duction curve, from which reduced values of G and V s can be ob-
over the wall height from the FD solution relative to those obtained
tained for use in the analysis. This equivalent-linear procedure
by Ostadan (2005) for two of the ground motions [three are omitted
neglects local strains imposed by the wall, and is reasonable for
for clarity in Fig. 6(a)]. Table 1 presents the resultants of these
cases involving free-field ground strains smaller than about 1%.
distributions. The resultant forces are in good agreement, with
However, the procedure may become erroneous at larger strains
errors ranging from −10 to þ12%.
corresponding to ground failure. Free-standing retaining walls that
In the SF solution, the surface displacement is computed as
rotate or translate significantly may mobilize such large shear
ug0 ¼ PGV · T=2π, where PGV is taken from ground-surface mo-
strains, but this will rarely be the case for stiff building basement
tions, and period T is taken as 4H=V s due to the strong impedance
walls.
contrast at the base of the soil layer. The agreement with the
The solution presented in this paper assumes perfect contact be-
Ostadan (2005) solution is reasonable, but not as good as the FD
tween the soil and the vertical walls. In reality, a gap might form in
solution, with errors ranging from −12 to þ57%. The Mononobe–
cohesive soils at this interface if PE is negative (i.e., the wall is
Okabe earth pressure resultant presented by Ostadan (2005),
moving away from the soil) and its absolute value is larger than
the initial earth pressure on the wall. Gapping may theoretically 160 kN=m for all of ground motions, underpredicts the earth pres-
cause pounding and additional stresses on the wall beyond those sures in every case.
considered in this paper. However, it is likely that peak earth pres- The conditions considered by Ostadan (2005) are nearly optimal
sures will occur when PE is positive (i.e., when the free-field soil for generating large kinematic wall pressures (i.e., λ=H ¼ 4, asso-
moves towards the wall), which is considered in the present analy- ciated with first mode response of the backfill, lies near the peaks
sis. The efficacy of the proposed procedure is demonstrated in the of the curves in Fig. 4). Such conditions cause the mobilized earth
subsequent section and will be tested further over time as additional pressures to exceed those from the M–O theory. The Ostadan
experimental data become available. (2005) results are broadly consistent with previous findings by
Arias et al. (1981) and Veletsos and Younan (1994) obtained by
analytical closed-form solutions for similar configurations.
Comparison to Experimental-Based and
Simulation-Based Results in Literature Al Atik and Sitar (2009, 2010) Experimental Results
In this section, two prior studies that reached strongly divergent Al Atik and Sitar (2009, 2010) performed centrifuge experiments
conclusions about the levels of seismic earth pressures acting on on relatively rigid and flexible U-shaped walls with prototype di-
retaining walls are interpreted using the proposed methodology. mensions of H ¼ 6.5 m and B ¼ 5.3 m embedded in a profile of
In the first study, Ostadan (2005) performed elastic wave propaga- medium dense sand with thickness D ¼ 19 m, and γ ¼ 17 kN=m3 .
tion analysis using a numerical finite-element code (SASSI; Lysmer The average small-strain shear wave velocities given by Al Atik
et al. 1999) to investigate the kinematic interaction between free- and Sitar (2009, 2010) were V s ¼ 170 m=s behind the walls
field site response and a massless embedded structure connected to and V s ¼ 260 m=s for the depth interval from the base of the wall
a rigid base and fixed against rotation. Ostadan (2005) concluded to the essentially rigid base of the container. The FD and SF so-
that M–O earth pressure theory significantly underpredicts the mo- lutions are compared with results of experiments performed
bilized earth pressures by factors ranging from 2 to 4 depending on using motions denoted Loma Prieta SC1, Loma Prieta SC2, and
ground motion characteristics. In the second study, Al Atik and Kobe PI2.
Fig. 6. Maximum seismic earth pressure increments computed by the proposed kinematic methodology compared with (a) numerical solutions by
Ostadan (2005); (b) experimental results by Al Atik and Sitar (2009, 2010)
For these experiments, ug0 was obtained by digitizing and the predicted distributions being approximately zero at the base
double-integrating in time the plots of free-field surface accelera- of the wall and having their maximum at the ground surface. This
tion presented by Al Atik and Sitar (2009). These motions induced mismatch may result in part from the assumption of depth-invariant
nonlinear response in the sand, and measured shear strains and kiy , whereas the shear modulus of sand in the centrifuge models
the interpreted modulus reduction (G=Gmax ) curve by Al Atik and increases with depth. A more robust solution would utilize kiy val-
Sitar (2009, 2010) were used to estimate representative values of ues that increase with depth in accordance with the variation in soil
G=Gmax ¼ 0.28, 0.25, and 0.10 for the SC1, SC2, and PI2 ground shear modulus, combined with a site response study that captures
motions, respectively. Comparisons between computed (FD solu- the influence of these variations on the free-field displacement
tion) and measured maximum earth pressures for the three digitized profile. The authors lacked the required data to perform such an
ground motions are shown in Fig. 6 for SC2 and PI2 (SC1 omitted analysis. The modulus reduction was an important part of this
for clarity). Resultant forces for all three motions are shown in analysis; if taken as unity (linear soil) earth pressures are signifi-
Table 2. Resultant force errors range from −7 to þ23% for the cantly overpredicted.
FD solution and from þ6 to þ23% for the SF solution. Although Mononobe–Okabe earth pressures presented by Al Atik and
the earth pressure resultants are predicted quite well, the shape of Sitar (2009, 2010) were computed using the ground surface PGA
the pressure distributions differ significantly, with the reported dis- and 0.65PGA. For consistency with the Ostadan (2005) compari-
tributions from measurements increasing linearly with depth and sons, the PGA-based M–O estimates are presented here. As shown
of kinematic and inertia loading, and peak moments were out-of- Acknowledgments
phase with peak kinematic earth pressures. The evaluation of these
inertial effects is a straightforward extension of the proposed meth- The authors thank Farhang Ostadan for sharing the ground motion
odology, but is not considered in this paper for brevity and because data utilized in a 2005 paper, and the two anonymous reviewers
the authors lacked the required data. whose comments helped improve the paper.