Professional Documents
Culture Documents
Abstract: Commonly used simplified one-dimensional nonlinear seismic site response analyses employ constitutive models based on a
Downloaded from ascelibrary.org by La Trobe University on 05/20/16. Copyright ASCE. For personal use only; all rights reserved.
variation of the hyperbolic model to represent the initial stress-strain backbone curve. Desirable features of the backbone curve include
provision of (1) an initial shear modulus at zero shear strain, (2) a limiting shear stress at large shear strains, and (3) flexible control of
the nonlinear behavior between those boundary conditions. Available hyperbolic models have combinations of two of these features.
A new general quadratic/hyperbolic (GQ/H) model is developed from the bivariate quadratic equation to provide all desired features.
Nonlinear behavior is controlled by a shear-strain-dependent curve-fitting function. The model’s unload-reload rules and coupling with
pore-water pressure generation are also presented. Several total-stress site response analyses are presented to demonstrate the performance
of the GQ/H model relative to a commonly used hyperbolic model in which the maximum shear stress cannot be defined. The analyses
show the importance of properly representing the maximum shear stress in the constitutive model because it may lead to underestimation
or overestimation of the computed site response. DOI: 10.1061/(ASCE)GT.1943-5606.0001496. © 2016 American Society of Civil
Engineers.
and as such, primarily employ constitutive models derived from the Gmax 1 þ Γ
original Kondner–Zelasko model (Kondner and Zelasko 1963),
hereafter called the KZ model. The original KZ model yields a hy- in which Γ = normalized shear strain function that varies between
perbolic relationship between shear stress and shear strain that can the reviewed models. However, plotting G=Gmax against Γ results
be defined as in a unique normalized shear modulus curve shown in Fig. 1 that is
γ
equivalent to the KZ model when Γ ¼ γ=γ r (i.e., a 1∶1 mapping of
Gmax γ τ γ G 1 τ G γ γ=γ r ). Table 1 summarizes the values of Γ for the models proposed
τ¼ Gmax ⇔ τ ¼ rγ⇔ ¼ γ ⇔ ¼ ·
1 þ τ max γ max 1 þ γ r Gmax 1 þ γ r τ max Gmax γ r by Hardin and Drnevich (1972), Matasovic (1993) (the MKZ
model), Hashash and Park (2001), and Darendeli (2001).
ð1Þ 1. Hardin and Drnevich (1972) altered the KZ model by modifying
the denominator γ=γ r term with a hyperbolic strain function, γ h
in which τ = shear stress; γ = shear strain; τ max = shear stress at (Table 1) in which aHD and bHD are curve-fitting parameters,
failure; and Gmax = maximum shear modulus. which results in the KZ model when aHD ¼ 0. γ r is the shear
Hardin and Drnevich (1972) introduced the concept of the refer- strain at which τ =τ max ¼ G=Gmax , but the respective values are
ence strain, γ r , noting that if a “line through the origin [of the shear not necessarily 50% of the maximum values as with the KZ
stress-strain curve] with slope Gmax is extended to intersect the line model. The primary benefit in the inclusion of the hyperbolic
τ ¼ τ max , the intersection defines a strain as”: strain function is that it allows for variation of the normalized
τ max shear stress and normalized shear modulus curves with respect
γr ¼ ð2Þ to γ r . The maximum shear modulus occurs at zero shear strain,
Gmax
and the maximum shear stress is approached as shear strains
This definition of reference strain was first introduced by Hardin approach infinity. However, the selection of appropriate para-
and Drnevich (1972) and will be maintained throughout the paper. meters in Table 1 can be quite challenging as aHD and bHD
simultaneously control the location and curvature of the
1 stress-strain curve. Both τ max and Gmax can be specified in
the model when bHD is positive.
0.8 2. Matasovic (1993) noted that the seismic response of soil depos-
its typically do not exceed the dominant shear strain range of
max
0.6
G/Gmax
1–3%. To improve the ability to capture small-strain behavior,
Matasovic (1993) introduced a modified version for the KZ
G/Gmax
not defined, and thus the model will typically overestimate the mine when the weighting function should be activated and what
maximum shear stress but can also underestimate this shear form it should take.
stress within the relevant range of shear strain. With s > 1 the
model may overestimate maximum shear stress within the rele-
vant range of shear strain before decreasing to zero. The MKZ General Quadratic/Hyperbolic (GQ/H) Model
model represents an improvement in controlling the stress-strain Backbone Curve
curve over the Hardin and Drenvich model. However, a maxi-
mum shear stress can only be prescribed by the user if s ¼ 1. The simplest nontrivial shear stress-strain response of a material
3. Hashash and Park (2001) suggested an alternate formulation of with finite strength is elastic-perfectly plastic behavior. The behav-
the MKZ model’s reference strain to include the effects of con- ior is linear elastic until the point of yielding or failure is reached,
fining pressure on modulus reduction. However, the model has and then the material deforms under constant shear stress. Both the
the same limitations as the MKZ model regarding controlling shear stiffness of the material (Gmax ) and the shear strength at fail-
for the maximum shear stress. ure (τ max ) must be known, as shown in Fig. 2(a), to capture such
4. Darendeli (2001) suggested an alternative formulation shown in behavior. Further, because the response consists only of two inter-
Table 1. Similar to the MKZ model, the implications of intro- secting lines, a quadratic model can be used to join the two lines
ducing the exponential s parameter are the following: (1) if into one curve. In terms of shear stress, τ , and shear strain, γ, a
s > 1, then τ → 0 as γ → ∞; (2) if s < 1, then τ → ∞ as quadratic model can be described as a second-degree equation
γ → ∞; and (3) if s ¼ 1, then τ → Gmax × γ r;SD as γ → ∞. in Cartesian coordinates
The MKZ model reduces to the Stokoe and Darendeli model
when β ¼ 1. As with the MKZ model, a maximum shear stress Aγ 2 þ Bγτ þ Cτ 2 þ Dγ þ Eτ þ F ¼ 0 ð4Þ
can only be specified if s ¼ 1.
These models tend to give reasonably accurate results within in which A, B, C, D, E, and F are constants.
the range of shear strains they consider to be of practical interest.
However, several of these models sacrifice the capability of correctly 1.2
modeling the soil behavior at large shear strains and hence the im-
1
plied shear strength of soil (τ at γ ¼ 10%). Others sacrifice the
capability of accurately modeling nonlinear soil behavior at small 0.8
strains while representing the soil shear strength. This is a significant
max
0.6
shortcoming when using these models in site response analyses G = Gmax
max
max
where large strains are anticipated. Other models have been devel- 0.4
1
oped which address some of these issues by either a bounding
surface plasticity approach [e.g., Borja and Amies (1994), 0.2
Boulanger et al. (2011)] or a hybridized combination of models 0
and transition functions as described in the following. 0 0.5 1 1.5 2
(a)
Yee et al. (2013) proposed a procedure that captures small-strain r
of each model are set so that the slopes of the two shear stress-strain 0.6 G/Gmax
curves are equivalent at the transitional shear strain. This allows for 0.4
a smooth, continuous transition between the two curves. However,
such a procedure is difficult to implement in site response analysis 0.2
software because it requires two backbone curves to be defined 0
and monitored throughout the analysis. Therefore, each backbone 0.0001 0.001 0.01 0.1 1 10 100 1000 10000
curve also requires its own unload-reload curve and the transition (b) r
between these curves must be tracked. The Yee et al. (2013) pro-
Fig. 2. Normalized: (a) shear stress; (b) shear modulus boundary
cedure is similar to a previous model proposed by Hayashi et al.
curves with respect to normalized shear strain for the GQ/H model
(1994) in which two weighted hyperbolic functions were utilized to
and to reach but not exceed a maximum positive shear stress (hori-
zontal asymptote) τ max , while maintaining the limiting condition of
elastic-perfectly plastic behavior, then the model can be reduced to in which θ1 , θ2 , θ3 , θ4 , and θ5 are curve-fitting constants chosen to
provide the best fit to the normalized shear modulus versus shear
γ strain curves over a defined strain range. With this formulation,
θτ 2 − 1 þ τ þ Gmax γ ¼ 0 ð7Þ ∂θ=∂γ → 0 as γ → ∞. The effects and purpose of the curve-fitting
γr
constants are shown in Fig. 3 and summarized in the following:
in which θ = primary curve-fitting function of this model. It is 1. θ1 = Lower-bound of θτ (i.e., θτ ¼ θ1 at γ=γ r ¼ 0);
shown in Appendix S1 that θ ≤ 1=τ max to ensure that τ ≤ τ max 2. θ2 = Range of θτ (i.e., θτ ¼ θ1 þ θ2 as γ=γ r → ∞);
for all γ, and also that θ can be presented as a function of γ 3. θ3 = Reference abscissa of θτ {i.e., when γ=γ r ¼ θ3 , θτ ¼ θ1 þ
under the condition that θ approaches a constant value as shear θ2 ½θ4 =ð1 þ θ4 Þ};
strains approach infinity (i.e., ∂θ=∂γ → 0 as γ → ∞). These 4. θ4 = Reference ordinate of θτ {i.e., when γ=γ r ¼ θ3 , θτ ¼ θ1 þ
constraints must be satisfied to obtain the large-strain maximum θ2 ½θ4 =ð1 þ θ4 Þ}; and
shear stress, τ max . 5. θ5 = Controls the slope of θτ at γ=γ r ¼ θ3 .
This equation produces a hyperbola for any finite value of θ. It should be noted that θ3 and θ4 are mathematically redundant
When θ ¼ 0, the asymptotes of the hyperbola are orthogonal (i.e., similar curves can be obtained for two specific combinations
and result in a rectangular hyperbola. When θ ¼ 1=τ max , the of θ3 and θ4 ). This is intentional to aid in manual curve-fitting pro-
asymptotes are straight lines which intersect at a point and thus cedures. For example, θ3 and θ4 can be set to lock in a data point.
result in a degenerate, two intersecting line, hyperbola (i.e., elastic- The parameter θ5 should then be used to adjust the slope and cur-
perfectly plastic behavior). For all other cases where θ ≤ 1=τ max is vature of the model about the “locked-in” data point. When using
finite, the asymptotes of the hyperbola are nonorthogonal and θ curve-fitting algorithms, it is recommended that θ3 or θ4 be set
governs the degree of nonorthogonality. equal to 1.
Eq. (7) can be presented in a normalized form as Whereas the GQ/H model will always be asymptotic to τ max at
infinite shear strain, values of θ1 , θ2 , θ3 , θ4 , and θ5 may be selected
τ 2 γ τ γ during curve fitting such that the implied shear strength at γ ¼ 10%
θτ − 1þ þ ¼0 ð8Þ
τ max γ r τ max γr is less than 50% of the target shear strength, τ max . To ensure agree-
ment between the implied shear strength and the target shear
in which θτ ¼ θ × τ max . Note that θτ is simply θ multiplied by the strength, it is recommended that users of the GQ/H model only
maximum shear stress, τ max . Because the constraint exists whereby allow curve fits where the implied shear strength reaches a desired
θ ≤ 1=τ max , the corresponding constraint on θτ is that θτ ≤ 1. fraction of the maximum shear strength at γ ¼ 10%. When using
The general solution of Eq. (8) yields the backbone curve. For correlations such as Darendeli (2001), it is recommended that curve
the case where θτ ≠ 0, the solution is fitting be performed for small-strain data up to the range of appli-
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi cability for the particular correlation while imposing a condition
τ 1 þ ðγ=γ r Þ − ½1 þ ðγ=γ r Þ2 − 4θτ ðγ=γ r Þ that shear stresses achieve a minimum value at γ ¼ 10%. In this
¼ ð9Þ
τ max 2θτ paper, curve-fitting was performed for a shear strain range up to
0.1% under the condition that the shear stresses reach 95% of the
which, due to the radical term, can alternatively be presented as maximum at γ ¼ 10%.
3
4
4
1
4
Degradation Indices
4
1
pressure during dynamic loading causes degradation of the shear
-3 strength and shear stiffness of the soil. Matasovic (1993) represents
-3.5 Reference the degradation of these within the MKZ model by inclusion of two
(a) -4 degradation indices
0 pffiffiffiffiffiffiffiffiffiffiffiffiffi
-0.5 δ G ¼ 1 − u ð14Þ
-1
-1.5
-2 3 = 0.1 and
-2.5 =1
Downloaded from ascelibrary.org by La Trobe University on 05/20/16. Copyright ASCE. For personal use only; all rights reserved.
3
-3 3 = 10 δ τ ¼ 1 − ðu Þν ð15Þ
-3.5
(b) -4
0 in which δ G = shear modulus degradation function; δ τ = shear stress
-0.5 degradation function; u = excess pore water pressure normalized
-1
-1.5 by initial effective overburden stress; and ν = curve-fitting param-
=3
-2 4
eter to better model the degradation of shear strength with excess
-2.5 =1
-3
4
pore pressure generation. When combined with an excess pore-
4 = 0.333
-3.5 water pressure generation model [e.g., Matasovic (1993) and Green
(c) -4 et al. (2000)], the inclusion of these degradation indices in the GQ/
0
-0.5 H model allows for changes in soil stiffness due to pore-water pres-
-1 sure generation analogous to the MKZ model. Other formulations
-1.5
-2 5 =2 for these degradation parameters are available and may be used in-
-2.5 5 =1 terchangeably in the following equations. These parameters are
-3 = 0.5 implemented in the GQ/H model with the following formulation:
5
-3.5
-4 2
0.0001 0.001 0.01 0.1 1 10 100 1000 10000 τ δ γ τ δ γ
(d) δ τ · θτ − 1þ G · þ G· ¼0
r
δ τ · τ max δτ γ r δ τ · τ max δτ γr
Fig. 3. Parametric evaluation of nonlinear function θτ ð16Þ
shows the normalized shear stress curve plotted versus the normal-
1 þ δδτGγγr − 1 þ δδτGγγr − 4δτ θτ δδτGγγr
ized shear strain in arithmetic scale for the condition of θτ ¼ 1. This τ
¼ ð17Þ
case produces a degenerate hyperbola which represents an elastic- δ τ τ max 2δ τ θτ
perfectly plastic material response. As shear strains approach the
reference shear strain, the shear stress approaches the maximum and
shear stress at a linear slope of G=Gmax ¼ 1. The maximum shear 2
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
n
o2
3
stress is reached at the reference shear strain and then remains con- δG γ
6 1þ − 1 þ δδτGγγr − 4δτ θτ δδτGγγr 7
stant as shear strains approach infinity. Fig. 2(b) shows the semilog G 1 6
δτ γr
7
¼
plot of the corresponding normalized shear modulus reduction δ G Gmax δ G γ=δ τ γ r 4 2δ θ τ τ
5
curve plotted against the normalized shear strain for the condition
of θτ ¼ 1. It can be seen that the normalized shear modulus is equal
ð18Þ
to unity for shear strains less than the reference shear strain, and
transitions into the reciprocal of the normalized shear strain for
shear strains larger than the reference shear strain.
The normalized shear stress curve shown in Fig. 2(a) and the Hysteretic Behavior in GQ/H Model
normalized shear modulus reduction curve shown in Fig. 2(b) re- In the widely used nonlinear time-domain site response analysis
present the bounding (i.e., elastic-perfectly plastic) behavior of the codes (e.g., DESRA, DMOD, and DEEPSOIL), the extended un-
GQ/H model. All behavior must fall on or below these boundary load–reload Masing rules (Masing 1926) are used to model hyster-
curves. These boundaries cannot be exceeded without compromis- etic behavior. The extended Masing rules are commonly stated as:
ing the model (e.g., selecting θτ > 1). In other words, the simplest 1. For initial loading, the stress-strain curve follows the backbone
behavior that can be obtained with the GQ/H model is an elastic- curve:
perfectly plastic response. Furthermore, the bounding curves offer
an opportunity to validate the modeled behavior. τ ¼ Fbb ðγÞ ð19Þ
The Yee et al. (2013) model composed of two simplified
hyperbolic functions can be represented by a single quadratic func- in which τ is the shear stress, and Fbb ðγÞ is the backbone curve
tion with the corresponding θ functions shown in Appendix S1. function.
Additional parameterization possibilities of the GQ/H model are 2. If a stress reversal occurs at a point (γ rev , τ rev ), the stress-strain
provided in Appendix S2. curve follows a path given by
The unloading-reloading curve can be obtained by applying the applied to the MRDF unloading-reloading equation as
second Masing rule to the backbone equation τ δ γ − γ rev
τ ¼ Fðγ max Þ − max 1 þ G
τ γ − γ rev θτ δτ 2γ r
τ ¼ − max 1 þ sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2
θτ 2γ r δ γ − γ rev δ γ − γ rev
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
− 1þ G − 4δ τ θτ G
γ − γ rev 2 γ − γ rev δτ 2γ r δτ 2γ r
− 1þ − 4θτ þ τ rev ð22Þ
2γ r 2γ r
− Gγmax δ G ðγ − γ rev Þ þ Gγmax δ G ðγ − γ rev Þ þ τ rev ð27Þ
in which γ = given shear strain; γ r = reference shear strain; γ rev =
reversal shear strain; and τ rev = reversal shear stress.
However, the hysteretic damping calculated using the unloading- Calibration of GQ/H Model Parameters
reloading stress-strain loops based on the Masing rules results in
overestimation of damping at large shear strains. Phillips and Generic modulus reduction and damping curves by the Electric
Hashash (2009) proposed an expression that modifies the Masing Power Research Institute (EPRI 1993), Vucetic and Dobry (1991),
unloading-reloading rules and provides better agreement with the and Darendeli (2001) are commonly used in site response analysis
damping curves at large shear strains, commonly called the modulus to represent dynamic soil behavior when no site-specific curves are
reduction and damping factor (MRDF) approach. The Phillips and available. The authors used the curves proposed by Darendeli
Hashash (2009) MRDF reduction factor is given by (2001), which are considered to be an improvement over curves
proposed in prior studies, to illustrate the application of the GQ/H
Gγ max p3
Fðγ max Þ ¼ p1 − p2 1 − ð23Þ model. The Darendeli (2001) modulus reduction curves require
Gmax 0
the initial effective vertical stress (σv0 ), the plasticity index (PI),
in which p1 , p2 , and p3 are nondimensional parameters selected to the overconsolidation ratio (OCR), and the initial coefficient of
obtain the best fit with the target damping curve. Eq. (23) shows that effective lateral earth pressure (K 0 ) as input.
the reduction factor is dependent on the normalized modulus reduc- Fig. 4 shows the GQ/H model fitting procedure for a modulus
tion curve. Thus, the user should first determine the constitutive reduction curve for sands which will be used in site response
model parameters which provide the best fit to the backbone curve. analyses for the Meloland overcrossing station in California (de-
Then, the reduction factor parameters p1 , p2 , and p3 can be selected scribed in the next section). The reference curve was obtained from
0
to provide the best fit of the damping curve without modification of Darendeli (2001) for an σv0 of 521 kPa, OCR of 1, K 0 of 0.65, PI of
the fit to the backbone curve. Other formulations for the reduction 0, and shear strength (τ max ) of 300 kPa [Fig. 4(a)]. The target shear
factor, Fðγ max Þ, such as Darendeli (2001) are available and may be stress (τ max ) was estimated by the Mohr–Coulomb failure criterion
used interchangeably in the following equations. with assumed frictional angle of 30° and zero cohesion. The initial
The MRDF approach is implemented in the 1D nonlinear site maximum shear modulus was estimated using a shear wave veloc-
response analysis program DEEPSOIL by including the reduction ity (V s ) of 260 m=s and a unit weight of 18 kN=m3 . The θτ can be
factor, Fðγ max Þ. The modified unloading-reloading equation can be computed using Eq. (12), which is fitted by the hyperbolic relation-
expressed as ship Eq. (13) [Fig. 4(b)] using a curve-fitting algorithm. When
using the reference curves proposed by Darendeli (2001), it is rec-
τ γ − γ rev ommended to fit θτ for shear strains between 0.001 and 0.1–0.3%
τ ¼ Fðγ max Þ − max 1 þ
θτ 2γ r because the proposed reference curves were based on a database
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2 that becomes sparse for shear strains beyond 0.1%, and the fitting
γ − γ rev γ − γ rev
− 1þ − 4θτ − Gγ max ðγ − γ rev Þ of θτ for shear strains less than 0.001% has limited influence on
2γ r 2γ r G=Gmax . The recommended absolute maximum bound of θτ fitting
þ Gγ max ðγ − γ rev Þ þ τ rev ð24Þ is 0.3% for the Darendeli (2001) reference curves, and because of
the limited influence of the fit of G=Gmax to small strains, there
in which γ = given shear strain; γ r = reference shear strain; γ rev = is no recommended absolute minimum bound for θτ fitting. The
reversal shear strain; τ rev = reversal shear stress; and Gγmax = shear shear stress is obtained by Eq. (10). The MKZ and GQ/H models
modulus experienced at γ max on the backbone curve. both match well with the reference G=Gmax curve [Fig. 4(a)]. The
The stress and modulus degradation indices can be applied to MKZ model does not have the capability to match the target shear
the backbone and the unloading-reloading equations to account stress at large shear strains. However, the GQ/H model yields shear
for generated excess pore water pressure as stress that approaches the target shear strength at large shear strains.
0.8 0
-20
0.6
G/Gmax
-40
0.4
-60
0.2 -80
0 -100
0.0001 0.001 0.01 0.1 1 10 10-4 10-3 10-2 10-1 100 101 102
(a) Shear strain, (%) (b) Normalized shear strain,
0.8
Reference
Downloaded from ascelibrary.org by La Trobe University on 05/20/16. Copyright ASCE. For personal use only; all rights reserved.
= 0.576
0.6 GQ/H
MKZ
0.4
0.2
0
0 2 4 6 8 10
(c) Shear strain, (%)
Fig. 4. GQ/H model fitting procedure for sands with a vertical effective stress of 521 kPa, OCR of 1, and shear strength of 300 kPa [reference curve
from Darendeli (2001)]: (a) normalized shear modulus reduction; (b) θτ ; (c) normalized shear stress (τ =σv0 0 )
In this case, the GQ/H model achieves 95% of the target shear [Fig. 5(c)]. Fig. 6 presents this data in normalized stress-strain
strength at γ ¼ 10%. space. Fig. 7 shows the MRDF hysteresis behavior of the GQ/H
Fig. 5 shows the effect of target shear strength on G=Gmax , θτ , model. The Masing rules yield damping ratios greater than those
and the shear stress generated by the GQ/H model for sands with by Darendeli (2001) which are based on dynamic laboratory
the same conditions as that for Fig. 4. The τ max slightly increases tests. When the MRDF procedure is applied to the GQ/H model,
G=Gmax at large strains (γ > 0.1%) [Fig. 5(a)]. The GQ/H model damping ratios are shown to match well with the reference curve
yields shear stresses that approach varying target shear strengths [Fig. 7(a)]. Fig. 7(b) shows the reduction factors for hysteretic
1 5
0.8
0
0.6
G/Gmax
0.4
-5
0.2
Increasing max
0 -10
0.0001 0.001 0.01 0.1 1 10 0 2 4 6 8 10
(a) Shear strain, (%) (b) Normalized shear strain,
0.6
Reference
0.4
0.2
0
0 2 4 6 8 10
(c) Shear strain, (%)
Fig. 5. Effect of target shear strength on (a) G=Gmax ; (b) θτ ; (c) shear stress for sands with a vertical effective stress of 521 kPa, and OCR of 1
[reference curve from Darendeli (2001)]
Silty clay 20
0.4 Sand
Boundary
40
Depth (m)
Clay
(a) 0
1.2 Silty sand
1 60
0.8
0.6
Silt 80
0.4
0.2
Downloaded from ascelibrary.org by La Trobe University on 05/20/16. Copyright ASCE. For personal use only; all rights reserved.
0 100
0.001 0.01 0.1 1 10 100 1000 0 500 0 1000 0 30 60
(b) VS (m/s) Implied shear Implied friction
strength (kPa) angle ( )
Fig. 6. Normalized: (a) shear modulus; (b) shear stress curves for vary-
ing τ max ; the boundary curves (θτ ¼ 1) are presented (a) (b) (c) (d)
MKZ GQ/H
damping and the corresponding MRDF fitting. The authors recom- Fig. 8. Profiles of (a) stratigraphy; (b) shear wave velocity (V S ) of sta-
mend that fitting of damping ratio be conducted for all available tion Meloland overcrossing (MEL) in California; (c) implied shear
levels of shear strain. The MRDF procedure generates reduced hys- strength; (d) implied friction angle for both MKZ and GQ/H models
teretic shear stress-strain relationships for unloading and reloading
conditions as shown in Fig. 7(c).
motions are described later. An additional site response analysis
Application of the GQ/H Model to Seismic Site was conducted for comparison to centrifuge results presented in
Response Analysis Hashash et al. (2015). The performance of the GQ/H model is
compared with the performance of the MKZ model in accordance
The performance of the GQ/H model in seismic site response with normalized shear modulus reduction curves proposed by
analysis is demonstrated at two sites whose profiles correspond Darendeli’s (2001) correlations, which utilize the MKZ model as-
to Meloland overcrossing (MEL) in El Centro, California and suming fixed model parameters. In this section, implied strength
Apeel #2 (A02) in Redwood City, California. The input ground parameters (i.e., shear strengths and friction angles) from the MKZ
40 1.2
hysteretic damping F
Reduction factor for
30
0.8
Damping ratio,
20
0.4
10
0 0
0.0001 0.001 0.01 0.1 1 10 0.0001 0.001 0.01 0.1 1 10
(a) Shear strain, (%) (b) Shear strain, (%)
300
Reference
200
Shear stress, (kPa)
GQ/H + MRDF
100
GQ/H + Masing
0
-100
-200
-300
-4 -2 0 2 4
(c) Shear strain, (%)
Fig. 7. MRDF procedure for GQ/H model for sands 521 kPa, OCR of 1, and shear strength of 300 kPa [reference curve from Darendeli (2001)]:
(a) damping ratio; (b) MRDF reduction factor, FðγÞ; (c) shear stress-strain relationship for γ rev ¼ 0.3, 1.5, and 4%
Figs. 8(a and b) show profiles of stratigraphy and shear wave implied shear strengths for the silt and clay layers after correction
velocity, respectively, for station MEL (Baturay and Stewart with the GQ/H model fit are approximately 95% of the estimated
2003). The subsurface soils of this station consist of silts and clays target shear strength (0.22σv00 ) and are less than the implied shear
with layers of sands at three different depths. The natural period of strengths of the MKZ model. However, use of the MKZ model fit of
this site is approximately 1.6 s. The authors calculated modulus the modulus reduction curves results in significant underestimation
reduction and damping curves by Darendeli (2001) for all soil of the friction angles for the sand layers. The GQ/H model large-
layers. Fig. 8(c) shows implied shear strength (τ at γ ¼ 10%) strain strength is asymptotic to the target friction angle because
by the MKZ model with depth, which is defined as the maximum τ max is a direct model input [Fig. 8(d)]. Fig. 9 shows normalized
value of ρV 2S ðG=Gmax Þγ. It is noted that the implied friction angles shear modulus, damping ratio, and shear stress curves at depths
estimated by tan−1 ðτ max =σv0 0
Þ for the sand layers for the MKZ corresponding to the mid-depths of three selected sand layers.
Downloaded from ascelibrary.org by La Trobe University on 05/20/16. Copyright ASCE. For personal use only; all rights reserved.
15 m
1 25 500
0.6 15 300
0.4 10 200
0.2 5 100
0 0 0
33 m
1 25 500
0.6 15 300
0.4 10 200
0.2 5 100
0 0 0
55 m
1 25 500
Shear stress (kPa)
0.8 20 400
Damping (%)
G/Gmax
0.6 15 300
0.4 10 200
0.2 5 100
0 -4 0 0
10 10-3 10-2 10-1 100 101 10-4 10-3 10-2 10-1 100 101 0 2 4 6 8 10
Shear strain (%) Shear strain (%) Shear strain (%)
(a) (b) (c)
Fig. 9. (a) Normalized shear modulus; (b) damping ratio; (c) shear stress curves for station Meloland overcrossing (MEL) at selected depths (15, 33,
and 55 m) for both MKZ and GQ/H models
Table 2. Summary of Earthquakes and Stations for the Selected Ground Motions (Ancheta et al. 2014)
Earthquake name M Station name Station V S30 (m=s) Epicentral distance (km) PGA (g)
1971 San Fernando 6.61 Pacoima Dam (upper left) 2,016 11.87 1.22
1978 Tabas, Iran 7.35 Tabas 767 55.24 0.81
1989 Loma Prieta 6.93 Gilroy Array #1 1,428 28.64 0.43
1989 Loma Prieta 6.93 UCSC Lick Observatory 714 16.34 0.46
1989 Loma Prieta 6.93 Los Gatos–Lexington Dam 1,070 20.35 0.44
1990 Manjil, Iran 7.37 Abbar 724 40.43 0.52
1992 Landers 7.28 Lucerne 1,369 44.02 0.73
1994 Northridge 6.69 Pacoima Dam (downstr) 2,016 20.36 0.42
1994 Northridge 6.69 Pacoima Dam (upper left) 2,016 20.36 1.39
1999 Chi-Chi, Taiwan 7.62 TCU045 705 77.50 0.49
1999 Duzce, Turkey 7.14 IRIGM 496 760 24.31 1.00
2007 Chuetsu-oki, Japan 6.80 Kashiwazaki Nishiyamacho Ikeura 655 10.47 0.86
0 2
1.5
20
1
40
Depth (m)
0.5
0
60 0.01 0.1 1 10
(c) Period (s)
80 Input
Surface (EL-MKZ)
100 Surface (NL-MKZ)
0 0.5 1 0 5
Surface (EL-GQ/H)
PGA (g) Max. shear
(a) (b) strain (%) Surface (NL-GQ/H)
Fig. 10. Results of site response analysis for station Meloland overcrossing (MEL) subject to the ground motion measured at Pacoima Dam during the
1971 San Fernando earthquake: (a) PGA profiles; (b) maximum shear strain profiles; (c) response spectra on the ground surface
0 1.5
1 EL
ln(SaMKZ / SaGQ/H)
0.5
20
0
-0.5
40 -1
Depth (m)
-1.5
(c)
1.5
60
1 NL
ln(SaMKZ / SaGQ/H)
0.5
80 0
-0.5
EL NL -1
100 -1.5
-20 0 20 -20 0 20 0.01 0.1 1 10
Period (s)
(a) (b) (d)
GQ=H
Fig. 11. Differences between maximum shear strains estimated by MKZ (γ MKZ max ) and GQ/H (γ max ) models for both (a) equivalent-linear (EL); and
(b) nonlinear (NL) analyses, and residuals of spectral accelerations on the ground surface; (c) lnðSaMKZ =SaGQ=H Þ for EL; (d) NL analyses for station
Meloland overcrossing (MEL)
0 1
ln(SaMKZ / SaGQ/H)
EL
0.5
20
0
-0.5
40
Depth (m)
-1
(c)
1
60
ln(SaMKZ / SaGQ/H)
NL
0.5
0
80
-0.5
EL NL
100 -1
-20 0 20 -20 0 20 0.01 0.1 1 10
Period (s)
(a) (b) (d)
GQ=H
Fig. 13. Differences between maximum shear strains estimated by MKZ (γ MKZ max ) and GQ/H (γ max ) models for both (a) equivalent-linear (EL);
and (b) nonlinear (NL) analyses, and residuals of spectral accelerations on the ground surface; (c) lnðSaMKZ =SaGQ=H Þ for EL; (d) NL analyses
for station A02
Fig. 14(c) shows the implied friction angle for the MKZ model
Depth (m)
Sand without strength correction and the GQ/H model. The use of the
MKZ model fit of the modulus reduction curves results in signifi-
15 cant underestimation of the friction angle, whereas the implied fric-
tion angle of the GQ/H model reaches approximately 95% of the
estimated target friction angle at γ ¼ 10%. Figs. 15(a and b) show
the PGA and maximum shear strain profiles, respectively, of site
20 response analyses conducted with the MKZ and GQ/H models us-
ing the ChiTCU input motion. The softer response of the nonlinear
MKZ model generally underestimates the PGA above a depth of
10 m relative to the centrifuge measurements. The nonlinear MKZ
25 model underestimates the surface spectral response at periods up to
0 500 0 20 40
VS (m/s) Implied friction
approximately 1 s, as shown in Fig. 15(c). The nonlinear GQ/H
angle ( ) model, however, exhibits a much-stiffer response that is in good
agreement with the centrifuge measurements of PGA and spectral
(a) (b) (c)
response. Neither of the equivalent-linear models correctly esti-
MKZ GQ/H mated the response; however, the results of the equivalent-linear
GQ/H model generally exhibited behavior closer to the measured
response.
Fig. 14. Profiles of (a) stratigraphy; (b) shear wave velocity (V S );
Three total stress site response analyses were performed to com-
(c) implied friction angle for both MKZ and GQ/H models of the
pare the results of two types of nonlinear models: one which allows
Hashash et al. (2015) centrifuge model
maximum shear strength to be specified (GQ/H model); and the
0 4
Spectral acceleration (g)
3
5
2
10
Depth (m)
0
15 0.01 0.1 1 10
(c) Period (s)
20
Input
Surface (EL-MKZ)
25 Surface (NL-MKZ)
0 0.5 1 0 2 4 6 Surface (EL-GQ/H)
PGA (g) Max. shear
(a) (b) strain (%) Surface (NL-GQ/H)
Surface (Centrifuge)
Fig. 15. Results of site response analysis for Hashash et al. (2015) centrifuge model subject to the ChiTCU input motion: (a) PGA profiles; (b) max-
imum shear strain profiles; (c) response spectra on the ground surface
τ 1 = transitional shear stress of the Yee et al. (2013) Advances in Geotechnical Earthquake Engineering and Soil Dynamics,
model; Missouri Univ. of Science and Technology, Rolla, MO.
τ max = maximum shear stress; Hayashi, H., Honda, M., Yamada, T., and Tatsuoka, F. (1994). “Modeling
τ mo = reference shear stress of the MKZ model; and of nonlinear stress strain relations of sands for dynamic response analy-
τ rev = shear stress at reversal. sis.” Proc., 10th Earthquake Engineering World Conf., Balkema,
Rotterdam, Netherlands.
Kondner, R. L., and Zelasko, J. S. (1963). “Hyperbolic stress-strain formu-
Supplemental Data lation of sands.” 2nd pan American Conf. on Soil Mechanics and Foun-
dation Engineering, Associação Brasileira de Mecânica dos Solos, São
Appendixes S1 and S2 are available online in the ASCE Library Paulo, Brazil.
(www.ascelibrary.org). Ladd, C. C. (1991). “Stability evaluation during staged construction.” J.
Geotech. Eng., 10.1061/(ASCE)0733-9410(1991)117:4(540), 540–615.
Masing, G. (1926). “Eigenspannungen und Verfestigung beim Mess-
References ing.” 2nd Int. Congress on Applied Mechanics, Orell FüssliZurich,
Switzerland.
Ancheta, T. D., et al. (2014). “NGA-west2 database.” Earthquake Spectra, Matasovic, N. (1993). “Seismic response of composite horizontally-layered
30(3), 989–1005. soil deposits.” Ph.D. thesis, Univ. of California, Los Angeles.
Baturay, M. B., and Stewart, J. P. (2003). “Uncertainty and bias in ground- Phillips, C., and Hashash, Y. M. A. (2009). “Damping formulation for non-
motion estimates from ground response analyses.” Bull. Seismol. Soc. linear 1D site response analyses.” Soil Dyn. Earthquake Eng., 29(7),
Am., 93(5), 2025–2042. 1143–1158.
Borja, R., and Amies, A. (1994). “Multiaxial cyclic plasticity model Ramberg, W., and Osgood, W. (1943). “Description of stress-strain curves
for clays.” J. Geotech. Eng., 10.1061/(ASCE)0733-9410(1994)120: by three parameters.” Rep. No. NACA-TN-902, National Advisory
6(1051), 1051–1070. Committee for Aeronautics, Washington, DC.
Boulanger, R., Kamai, R., and Ziotopoulou, K. (2011). “Numerical mod- Stewart, J. P., and Kwok, A. O. (2008). “Nonlinear seismic ground response
eling of liquefaction effects.” 4th IASPEI/IAEE Int. Symp., Univ. of analysis: Code usage protocols and verification against vertical array
California, Santa Barbara.
data.” Geotechnical Engineering and Soil Dynamics IV, D. Zeng, M. T.
Chiu, P., Pradel, D. E., Kwok, A. O.-L., and Stewart, J. P. (2008). “Seismic
Manzari, and D. R. Hiltunen, eds., ASCE Geotechnical Special
response analyses for the silicon valley rapid transit project.” Geo-
Publication, Sacramento, CA.
technical Earthquake Engineering and Soil Dynamics IV, ASCE,
Sacramento, CA. Vucetic, M., and Dobry, R. (1991). “Effect of soil plasticity on cyclic re-
Darendeli, M. B. (2001). “Development of a new family of normalized sponse.” J. Geotech. Eng., 10.1061/(ASCE)0733-9410(1991)117:
modulus reduction and material damping curves.” Ph.D. thesis, Univ. 1(89), 89–107.
of Texas, Austin, TX. Yee, E., Stewart, J., and Tokimatsu, K. (2013). “Elastic and large-strain
EPRI (Electric Power Research Institute). (1993). “Guidelines for deter- nonlinear seismic site response from analysis of vertical array record-
mining design basis ground motions.” Rep. No. TR-102293, Palo Alto, ings.” J. Geotech. Geoenviron. Eng., 10.1061/(ASCE)GT.1943-5606
CA. .0000900, 1789–1801.
Gingery, J., and Elgamal, A. (2013). “Shear stress-strain curves based on Zhang, J., Andrus, R. D., and Juang, C. H. (2005). “Normalized shear
the G/Gmax logic: A procedure for strength compatibility.” IACGE modulus and material damping ratio relationships.” J. Geotech. Geoen-
2013, ASCE, Reston, VA, 721–729. viron. Eng., 10.1061/(ASCE)1090-0241(2005)131:4(453), 453–464.