Professional Documents
Culture Documents
EARTH STRUCTURES
By Ellen M. Rathje1 and Jonathan D. Bray2
ABSTRACT: Earthquake-induced sliding displacements of earth structures are generally evaluated using sim-
plified sliding block analyses that do not accurately model the seismic response of the sliding mass nor the
seismic forces along the slide plane. The decoupled approximation introduced to capture each of these effects
separately is generally believed to be conservative. However, recent studies using linear viscoelastic sliding mass
models have revealed instances where the decoupled approximation is unconservative. In this paper, a coupled
analytical model that captures simultaneously the fully nonlinear response of the sliding mass (necessary for
intense motions) and the nonlinear stick-slip sliding response along the slide plane is presented. The proposed
sliding model is validated against shaking table experiments of deformable soil columns sliding down an inclined
plane. The effect of sliding on the response of earth structures is evaluated, and comparisons are made between
sliding displacements calculated using coupled and decoupled analytical procedures with linear and nonlinear
material properties. Nonlinearity resulting from stick-slip episodes is often the dominant source of nonlinearity
in this problem. The decoupled approximation was unconservative primarily for intense ground motions for
systems with low values of ky, larger values of ky /kmax, and high period ratios (Ts /Tm). Results indicate that a
decoupled analysis is adequate for earth structures that are not expected to experience intense, near-fault motions.
However, for projects undergoing intense, near-fault ground motions, a fully nonlinear, coupled stick-slip analysis
is recommended.
冘
is defined as follows: nonlinear soil properties and coupled stick-slip sliding).
C 2i (1/fi ) The analytical models displayed in Fig. 1 all represent the
冘
Tm =
i
(1) sliding mass in one dimension. Studies have revealed that 1D
C i2 analysis provides a reasonably conservative estimate of the
i
冉 冊
= s⬘ (10)
s̈ and substituting the result into (4) to create the following Gmax␥
expression: 1⫹
mo
M*ü ⫹ Cu̇ ⫹ Ku = M ⭈ 1 ⭈ g (6) where Gmax = small strain shear modulus; ␥ = shear strain;
where M* is defined as follows: = shear stress; mo = shear stress capacity of the soil at a spec-
ified reference shear strain ␥mo; and  and s⬘ = curve fitting
1 parameters. These parameters are chosen to provide the best
M* = M ⫺ ⭈ M11T M (7) fit with the nonlinear strain-dependent shear modulus reduc-
MT
tion and damping curves of the soil. The MKZ model is equiv-
It is important to note that there are two specific cases where alent to the hyberbolic model when  = s⬘ = 1.0. The addi-
the mass matrix M* as defined in (7) may be singular. These tional MKZ parameters,  and s⬘, allow for more control over
cases are systems with 1- or 2-DOFs, where MT is exactly the shape of the stress-strain curve and, hence, provide a more
equal to the sum of the lumped masses contained in M. When accurate representation of the variation of shear modulus and
studying discrete sliding systems, previous researchers have damping ratio with shear strain.
circumvented this problem by assigning some mass directly to D-MOD solves the equations of motion in the time domain
the sliding interface [e.g., Mostaghel and Tanbakuchi (1993) using the Newmark average acceleration integration proce-
and Kramer and Smith (1997). In the computer code devel- dure, updating the soil stiffness at each time step. D-MOD and
oped for this study, only half of the mass of the base element other DESRA-based nonlinear codes have been shown to cal-
(i.e., directly above the sliding interface) is assigned to the culate seismic responses similar to the established equivalent-
lumped mass for that element, and the other half is assigned linear program SHAKE at low acceleration levels where non-
to the sliding interface. This adjustment assures that MT is not linearity is less pronounced [e.g., Finn (1988), Matasovic
1004 / JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING / NOVEMBER 2000
(1993), and Bray and Rathje (1998)]. However, D-MOD cal-
culates a lower high frequency response and higher low fre-
quency response at high acceleration levels in agreement with
recorded strong motion data at rock and soft soil sites [e.g.,
Borcherdt (1994)]. Back-analysis of several well-documented
case histories, such as Treasure Island during the Loma Prieta
earthquake and the Wildlife Liquefaction site during multiple
earthquakes, have validated the use of D-MOD for ground
response analysis [e.g., Matasovic (1993) and Matasovic and
Vucetic (1995)].
interface, but is a function of the entire nonuniform accelera- ysis peak at a smaller period ratio than those from linear cou-
tion distribution within the soil column. Eqs. (3) and (5) de- pled analysis. The shift in the peak is largely due to nonlinear
scribe this condition. Consequently, accelerations greater than analysis capturing the reduction in stiffness experienced during
the yield acceleration are observed at the base of the deform- shaking. Interestingly, although the nonlinear and linear peaks
able sliding soil column during sliding, while the force at the in Fig. 4 occur at different period ratios, their magnitudes are
sliding interface remains limited to MT g. Data from the shak- reasonably similar. The linear analysis peak is slightly smaller
ing table experiments confirm this observation (Wartman than the nonlinear peak because a constant value of 15%
1999; J. Wartman et al., unpublished paper, 2000). Accord- damping was used in the linear analyses, which is higher than
ingly, numerical procedures that attempt to limit the acceler- the typical level of material damping developed in the nonlin-
ation transmitted to the base of a sliding deformable soil col- ear analysis at low period ratios. Additionally, the linear anal-
umn do not accurately model the overall dynamic response of ysis does not peak at Ts /Tm equal to 1.0 due to nonlinearity
a sliding system. introduced by sliding. The similarity in the linear and nonlin-
The favorable comparison between results of the coupled ear results at period ratios <1.0 indicates that nonlinearity in-
stick-slip numerical model presented and the deformable slid-
ing block experiments provides justification for the use of this
formulation. Previous research [e.g., Finn (1988), Matasovic
and Vucetic (1995), and Bray and Rathje (1998)] has already
provided evidence that a fully nonlinear dynamic analysis can
capture the nonlinear dynamic material response of earth and
waste structures under strong earthquake shaking. Therefore,
the proposed nonlinear, coupled stick-slip analytical procedure
is judged to be sufficiently reliable to estimate the seismically
induced permanent displacement of 1D systems that exhibit
significantly nonlinear material responses. This formulation
provides the opportunity to evaluate the relative importance of
material nonlinearity and nonlinearity introduced by stick-slip
episodes. Additionally, the decoupled approximation for non-
linear systems under intense ground shaking can be investi-
gated.
FIG. 8. Effect of Shear-Wave Velocity on Displacement Differ- FIG. 9. Coupled and Decoupled System Response for: (a and
ence for Superstition Mountain Input Motion and k y ⴝ 0.05 b) Ts /Tm ⴝ 0.5; (c and d) Ts /Tm ⴝ 2.0
FIG. 11. (a) Displacement Ratio versus Decoupled Displacement; (b) Displacement Difference versus Decoupled Displacement