Sie sind auf Seite 1von 16

ARTICLE IN PRESS

Soil Dynamics and Earthquake Engineering 29 (2009) 1143–1158

Contents lists available at ScienceDirect

Soil Dynamics and Earthquake Engineering


journal homepage: www.elsevier.com/locate/soildyn

Damping formulation for nonlinear 1D site response analyses


Camilo Phillips , Youssef M.A. Hashash 1
Department of Civil and Environmental Engineering, University of Illinois at Urbana-Champaign, 205 N. Mathews Avenue, Urbana, IL 61801, USA

a r t i c l e in fo abstract

Article history: Measurements and observations of ground shaking during large earthquakes have demonstrated the
Received 29 November 2008 predominant role of site effects in the response of infrastructure during a seismic event. Despite
Accepted 29 January 2009 significant efforts to model the hysteretic response and nonlinearity of soils due to medium and large
ground motions, the most widely accepted nonlinear site response methods are not able to represent
Keywords: simultaneously the changes of stiffness and energy dissipation (damping) observed in both laboratory
Small strain damping tests and during earthquake events. This paper presents two new soil damping formulations
Hysteretic damping implemented in nonlinear one-dimensional site response analysis for small and large strains. The first
Masing rules formulation introduces an approach to construct a frequency-independent viscous damping matrix
Nonlinear site response analysis
which reduces the over-damping at high frequencies, and therefore, the filtering at those frequencies.
The second formulation introduces a reduction factor that modifies the extended Masing loading/
unloading strain–stress relationship to match measured modulus reduction and damping curves
simultaneously over a wide range of shear strains. A set of examples are introduced to illustrate the
effect of using the two proposed formulations, separately and simultaneously, in nonlinear site response
analyses.
Published by Elsevier Ltd.

1. Introduction layers) in which an equivalent soil stiffness and damping for each
layer cannot represent the behavior of the soil column over the
Numerous seismic events, such as the 1985 Michoacan earth- entire duration of a seismic event. In these cases, a nonlinear
quake, the 1989 Loma Prieta earthquake, the 1994 Northridge time-domain solution is used to represent the variation of the
earthquake, the 1995 Kobe earthquake and the 1999 Chi-Chi shear modulus (G) and the damping ratio (x) during shaking.
earthquake have demonstrated the relevance of local geologic and In nonlinear analysis, the following dynamic equation of
geomorphologic conditions on the seismic ground response. The motion is solved [2]:
changes in the intensity and the frequency content of the motion € þ ½Cfug
½Mfug _ þ ½Kfug ¼ ½MfIgu€ g (1)
due to the propagation of the seismic waves in soil deposits and
the presence of topographic features, commonly referred to as site where [M] is the mass matrix, [C] is the viscous damping matrix,
effects, have a direct impact on the response of the structures [K] is the stiffness matrix, {ü} is the vector of nodal relative
during an earthquake event. One-dimensional site response acceleration, {u̇} is the vector of nodal relative velocities and {u} is
analysis methods are widely used to quantify the effect of soil the vector of nodal relative displacements. {üg} is the acceleration
deposits on propagated ground motion. These methods can be at the base of the soil column and {I} is the unit vector. [M], [C]
divided in two main categories: (1) frequency domain analyses and [K] matrices are assembled using the incremental response of
(including the equivalent linear method, e.g SHAKE [1]) and (2) the soil layers. The soil response is obtained from a constitutive
time-domain analyses (including nonlinear analyses). model that describes the cyclic behavior of soil. The dynamic
Frequency domain methods are the most widely used to equilibrium equation, Eq. (1), is solved numerically at each time
estimate site effects due to their simplicity, flexibility and low step using the Newmark b method [3].
computational requirements. However, there are cases (i.e. high The geologic column is discretized into individual layers using
seismic intensities at rock base and/or high strain levels in the soil a multi-degree-of-freedom lumped parameter model shown in
Fig. 1, alternatively it can be represented using finite elements [2].
Each individual layer i is represented by a corresponding mass,
 Corresponding author. Tel.: +1 217 333 6943; fax: +1 217 333 9464. nonlinear spring, and a dashpot for viscous damping. Lumping
E-mail addresses: cphilli3@illinois.edu (C. Phillips), hashash@illinois.edu
half the mass of each of two consecutive layers at their common
(Y.M.A. Hashash). boundary forms the mass matrix. The stiffness matrix is updated
1
Tel.: +1 217 333 6986. at each time increment to incorporate nonlinearity of the soil.

0267-7261/$ - see front matter Published by Elsevier Ltd.


doi:10.1016/j.soildyn.2009.01.004
ARTICLE IN PRESS

1144 C. Phillips, Y.M.A. Hashash / Soil Dynamics and Earthquake Engineering 29 (2009) 1143–1158

Nomenclature ties (plasticity, effective stress, number of cycles,


mean frequency)
List of Symbols
b index used to calculate the damping matrix
c1 constant (equal to 0.1) to calculate the Darendeli’s
[C] viscous damping matrix reduction factor.
[K] stiffness matrix c scalar value used in the calculation of the confining
[M] mass matrix pressure dependent reference strain
{I} unit vector fn natural frequency of the soil column; n ¼ 1 corre-
{u} vector of nodal relative displacements sponds to the fundamental frequency
{u̇} vector of nodal relative velocities pi parameters used to calculate the reduction factor
{ü} vector of nodal relative acceleration s hyperbolic model parameter
{üg} acceleration at the base of the soil column F real modal matrix of the system
A area enclosed by the hysteretic loop used to deter- b hyperbolic model parameter
mine the hysteretic damping g shear strain
B triangular area used to determine the hysteretic gm maximum shear strain
damping gr reference shear strain hyperbolic model
DF(gm) Darendeli’s reduction factor. grev reversal shear strain
F(gm) reduction factor hysteretic damping s0 effective vertical stress
G0 initial tangent modulus of the loading curve sref reference stress
Ggm secant shear modulus for the maximum cyclic strain t shear stress
H total height of the soil column trev reversal shear stress
V̄s average shear wave velocity of the soil column xhysteretic hysteretic damping ratio
ab scalar value used in the calculation of the damping xMasing hysteretic damping ratio calculated using Masing
matrix rules
a scalar value used in the calculation of the confining xn damping ratio for a given natural frequency of the soil
pressure dependent reference strain column
b1 variable to calculate Darendeli’s reduction factor o diagonal matrix of fundamental frequencies of the
which depends on the soil and input motion proper- system

In the more widely used nonlinear time-domain site response where t is the shear stress Fbb(g) is the backbone curve
analysis codes (e.g. DESRA [4], DMOD [5] and DEEPSOIL [6]) a function.
hyperbolic model is used to represent the backbone response of (2) If a stress reversal occurs at a point (grev, trev), the stress–strain
the soil and the extended unload–reload Masing rules [7] to curve follows a path given by
model hysteretic behavior. The four extended Masing rules are
commonly stated as: t  trev g  g 
rev
¼ F bb (3)
2 2
(1) For intial loading, the stress–strain curve follows the back-
(3) If the unloading or reloading curve intersects the backbone
bone curve
curve, it follows the backbone curve until the next stress
t ¼ F bb ðgÞ (2) reversal.

m1 2
Layer
h1 k 1,c 1
1 G1, ρ1
(m1 2 + m2 2)
h2 k2 ,c2
2 G2,ρ 2
(m2 2 + m3 2)
h3 k 3,c 3
3 G3,ρ 3
Layer Properties

G: shear
sh modulus
ρ: density
hi VS: shear wave velocity
i Gi , ρ i mi = ρ i ⋅ hi h: thickness
Equivalent Lumped Mass System
hn CE = ρ E ⋅ VSE k n,c n
n Gn , ρn mn 2
k: stiffness
GE , ρE c: viscous damping

Fig. 1. Multi-degree-of-freedom lumped parameter model representation of horizontally layered soil deposit shaken at the base by a vertically propagating horizontal shear wave.
ARTICLE IN PRESS

C. Phillips, Y.M.A. Hashash / Soil Dynamics and Earthquake Engineering 29 (2009) 1143–1158 1145

(4) If an unloading or reloading curve crosses an unloading or illustrated in a series of one-dimensional wave propagation
reloading curve from the previous cycle, the stress–strain analyses.
curve follows that of the previous cycle.

The hysteretic damping calculated using the unload–reload 2. Small strain damping
stress–strain loops has the following two shortcomings:
2.1. Current representation via Rayleigh damping
(a) at very small strains the response is nearly linear (hysteretic
damping is nearly zero) and the use of velocity proportional Most of the available time-domain wave propagation codes
viscous damping is needed. Small strain damping controls include small strain damping by implementing the original
the behavior of the soil mass under dynamic loads that expression proposed by Rayleigh and Lindsay [18], in which, the
induce small deformations (i.e. weak amplitudes of seismic damping matrix results from the addition of two matrices; one
motions). Frequency-dependent Rayleigh damping [8] is proportional to the mass matrix and the other one proportional to
commonly employed which can result in over or under the stiffness matrix.
damping [8,9]
½C ¼ a0 ½M þ a1 ½K (4)
(b) the hysteretic damping can result in overestimation of
damping at large strains. Hysteretic damping controls the where [M] is the mass matrix, [K] is the stiffness matrix and a0
energy dissipation in problems which involve large strains. and a1 are scalar values selected to obtain given damping value for
two control frequencies.
Gerolymos and Gazetas [10] developed a phenomenological Small strain damping calculated using Rayleigh and Lindsay
constitutive model for the nonlinear 1-D ground response analysis [18] solution is frequency dependent (x changes depending on the
of layered sites. The proposed model, a special form of Bouc–Wen frequency of the input motion), a result that contradicts most of
viscoplastic model, is able to reproduce nonlinear hysteretic the available experimental data which show that material
behavior for different types of soils and has the ability to generate damping in soils is frequency independent at very small strain
simultaneously realistic modulus and damping curves. The model levels within the seismic frequency band of 0.001–10 Hz [19].
requires information on anisotropic behavior of the soil and shape Park and Hashash [8] and Kwok et al. [9] explained the
of the unload–reload loop. This information is not available for frequency dependency of the viscous damping matrix [C]. Hudson
most soils, thus it is difficult to quantify required model constants. et al. [20] and Park and Hashash [8] described the application of
Pyke [11] proposed an alternative hypothesis to the second the full Rayleigh formulation in site response analysis. For soil
Masing rule. The new hypothesis states that the scale of the profiles with constant damping ratio, scalar values of a0 and a1 can
stress–strain relationship for initial loading is a function of the be computed using two significant natural modes i and j
stress level on reversal for unloading and reloading and not only 2 3
1
the use of a factor of two. " # f i
( )
xi 1 6
6fi
7 a0
7
The Duncan and Chang [12] hyperbolic model using Cundall– ¼ 61 7 (5)
xj 4p 4 a1
Pyke hypothesis (HDCP model) has been implemented in the fj 5
fj
software TESS. Using Cundall–Pyke instead of the Masing rules
does not always generate a better match with laboratory dynamic where xi and xj are the damping ratios for the frequencies fi and fj
curves. Therefore, Pyke [13] proposes that the hysteretic damping of the system, respectively. For site response analysis the natural
calculated in the soil model be divided by a factor of two to frequency of the selected mode is commonly calculated as [2]
achieve a match to the laboratory measurements. To provide a
good fit to both modulus reduction and damping curves based on V̄ s
f n ¼ ð2n  1Þ (6)
laboratory tests the HDCP model implemented a shear modulus 4H
degradation scheme in which the modulus at a reversal point is where n is the mode number and fn is the natural frequency of the
not equal to G0 but is a function of strain and number of cycles corresponding mode. It is common practice to choose frequencies
[13]. The main shortcomings of using the HDCP model with shear that correspond to the first mode of the soil column and a higher
modulus degradation matching both modulus reduction and mode that corresponds to the predominant frequency of the input
damping curves are: (1) the shear modulus degradation seems motion. Kwok et al. [9] recommended a value equal to five times
excessive and therefore not always representative of soil behavior the natural frequency, Park and Hashash [8] also give a series of
and (2) the resulting damping curve in most of the cases is not a recommendations to determine these two frequencies. Equal
smooth function. values of modal damping ratios are specified for the two modes.
Muravskii [14] presented a methodology to construct loading Wilson [21] proposed to use only the stiffness proportional
and reloading curves based on a general function that becomes an damping term to solve dynamic problems involving complex
alternative to scaling the backbone by a factor of two; as is stated structural systems, in which a large number of high frequencies
in the Masing rules. Three different functions [15–17] are used to (short periods) are present. In this type of problems, periods
construct the unloading and reloading curves. These unloading smaller than the time step have a tendency to oscillate
and reloading curves allow matching the theoretical results to indefinitely after they are excited. Although the stiffness propor-
experimental data, in particular the damping properties of the tional damping with reference frequency equal to the sampling
model. In a later section of this paper a series of examples are rate frequency provides numerical stability, its behavior resem-
presented to highlight the most important features of using bles a high-pass filter, which results in a highly frequency-
Muravskii [14] method. dependent viscous damping. Common values of sampling rate
This paper describes two new formulations to model frequency (i.e. 50, 100 or 200 Hz) are higher than the upper limit
small strain and hysteretic damping. The small strain damping of the frequency content range of almost all the seismic motions
formulation relieves the problems associated with the use of and the soil deposit natural frequencies, therefore one-dimen-
Rayleigh damping. The hysteretic damping formulation sional wave propagation problems will not exhibit the aforemen-
modifies the extended Masing rules and results in better match tioned numerical instability. It should be noted that viscous
of damping curves. The effects of the proposed formulations are damping in the solution presented in this paper is included to
ARTICLE IN PRESS

1146 C. Phillips, Y.M.A. Hashash / Soil Dynamics and Earthquake Engineering 29 (2009) 1143–1158

represent soil behavior, consequently the objective solution to obtain Eq. (9) from Eq. (7) is presented in Appendix A of
should be frequency independent rather than highly frequency this paper.
dependent as the solution proposed by Wilson [21].
X
N 1
Eq. (4) can be extended so that more than two frequencies/ ½C ¼ ½M ab ð½M1 ½Kb Þ
modes can be specified, and is referred to as extended b¼0
Rayleigh formulation [i.e. Park and Hashash [8] implemented X
N 1
an extended Rayleigh scheme using four modes in the ¼ ½M ab Fo2b F1 ¼ ½Ma1=2 FoF1 (9)
software code DEEPSOIL [6]]. Using the orthogonality conditions b¼0

of the mass and stiffness matrices, the damping matrix can


consist of any combination of mass and stiffness matrices [22], 1 NX 1
xn ¼ a ð2pf n Þ2b
as follows: 4pf n b¼0 b

X
N 1 1 1
¼ ½a1=2 ð2pf n Þ ¼ a1=2
½C ¼ ½M ab ð½M1 ½KÞb (7) 4pf n 2
b¼0 ! a1=2 ¼ 2xn (10)
where N is the number of frequencies/modes incorporated. The
where (o) is the natural frequencies diagonal matrix and (F) is
coefficient ab is a scalar value assuming a constant damping ratio
the real modal matrix of the system.
throughout the profile and is defined as follows:
Eq. (10) shows that for b ¼ 12 the viscous damping of the system
is not dependant on the frequency. The numerical cost of this
1 N1X
xn ¼ a ð2pf n Þ2b (8) solution is in calculating the natural frequencies and the real
4pf n b¼0 b modal matrix of the system, which requires the calculation of the
eigenvalues and eigenvectors of the matrix [M]1[K].
Eq. (8) implies that the damping matrix can be extended to
include any number of frequencies/modes. The resultant matrix 2.3. Implementation of small strain frequency-independent damping
from Eq. (7) is numerically ill-conditioned since coefficients fn1, in nonlinear site response analyses
fn1, fn3, fn5, y, fn2n1 differ by orders of magnitude. Having more
than four frequencies/modes can result in a singular matrix All the one-dimensional wave propagation problems presented
depending on fn and ab cannot be calculated. An increase in the in this paper have been solved using the site response program
frequencies/modes used in the calculation of the damping matrix DEEPSOIL [6]. To implement the new frequency-independent
generates also an increase in the number of diagonal bands of the model, a QL/QR algorithm [24] with implicit shifts has been
viscous damping matrix, and therefore, a significant time increase included in DEEPSOIL. The QL/QR algorithm swiftly reduces the
for the solution of the wave propagation problem. Incorporating off-diagonal terms of the [M]1[K] matrix until they are negligible.
an odd number of modes is also problematic since it will result in The algorithm repeatedly applies a similarity transformation to
negative damping at certain frequencies [22]. Fig. 2 presents a the result of the previous transformation until the matrix
comparison of the effective damping obtained using one mode, converges to a diagonal form. Once [M]1[K] matrix is in a
two modes and four modes solutions. diagonal form its eigenvalues could be easily calculated, then
using the definition of eigenvector and the previously calculated
2.2. New formulation to construct the viscous damping matrix eigenvalues each one of the eigenvectors is calculated and the
construction of the F matrix is completed.
Liu and Gorman [23] extended Rayleigh damping in terms of Appendix B presents a flow chart of the method implemented
series having terms depending on two different indices (double in DEEPSOIL to obtain the damping matrix using the frequency-
series), the proposed extensions provide solutions for both independent scheme. One concern with the frequency-indepen-
negative and rational indexed series. Using the rational indexed dent viscous damping is the effect of the natural frequencies
extension and an index b equal to 12 in Eq. (7), Eqs. (7) and (8) induced by the shear modulus reduction of the soil profile as a
reduce to Eqs. (9) and (10), respectively. The complete procedure function of the strain level. A set of analyses were developed in

5
ξ Model /ξTarget

0
0.01 0.1 1 10 100
Frequency [Hz]
One Mode Two Mode Four Modes Target

Fig. 2. Effective damping for one, two- and four- (extended) mode Rayleigh formulation.
ARTICLE IN PRESS

C. Phillips, Y.M.A. Hashash / Soil Dynamics and Earthquake Engineering 29 (2009) 1143–1158 1147

which the damping matrix was calculated for each time step. where, a and c are scalars which depend on the soil type and must
These analyses showed small incidence of the change in the be determined based on laboratory test results, s0 is the effective
natural frequencies in the deposit’s response. vertical stress and sref is a reference stress.
This result agrees with the results obtained by Park and It can be observed that Eq. (12) is the result of applying the
Hashash [8]. Updating the damping matrix is computationally second Masing rule to the backbone equation of the hyperbolic
expensive (more than 10 times in the example presented later) model [Eq. (11)].
and appears to be unnecessary. Hysteretic damping (xhysteretic) is proportional to the energy
lost in each cycle of vibration and therefore is also proportional to
the area of the hysteretic loop generated from Eq. (12). Eq. (14)
3. Hysteretic damping shows the expression used to calculate xhysteretic, in which xhysteretic
is proportional to the ratio of the area enclosed by the hysteretic
3.1. Extended Masing rule hysteretic damping in the loop and the triangular area corresponding to the work developed
Hyperbolic model by an equivalent linear material (See Fig. 3).

A
The hyperbolic model is defined by using two sets of xhysteretic ¼ (14)
4p B
equations; Eq. (11) defines the stress–strain relationship (back-
where A and B correspond to area enclosed by the hysteretic loop
bone curve) for loading, Eq. (12) defines the stress–strain
and triangular area presented in Fig. 3, respectively.
relationship for unloading–reloading conditions.
Even though the nonlinear model implemented in site response
g G0 analysis should match both G & x variations with cyclic shear
t¼ (11)
1 þ b ðg=gr Þs strain, most of the available stress–strain relationships which use
the extended Masing unloading–reloading rules [2,7] provide
2 G0 ððg  grev Þ=2Þ greater xhysteretic for medium to large strains compared to the
t¼ þ trev (12) damping values obtained in dynamic tests when the calibration of
1 þ b ððg  grev Þ=2 gr Þs
the constitutive model parameter is based only in the modulus
whereby, g: given shear strain, gr: reference shear strain, b: reduction curve (MR) ([9,26,27]) as illustrated in Fig. 4.
dimensionless factor, G0: maximum shear modulus and s: This difference between xhysteretic measured in the laboratory
dimensionless exponent. [25], in order to represent the depen- tests and the values obtained using the stress–strain relationships
dence of modulus on confining pressure, propose the following based on Masing rules leads to an underestimation of shear
expression for gr: strains and/or surface intensities at the ground surface. A solution
gr ¼ aðs0 =sref Þc (13) for this mismatch could be achieved if the parameters of the soil
constitutive model are calibrated using both curves (modulus
reduction and damping curves) at the same time; however, for
G0
G simple nonlinear models such as the hyperbolic model, this
1
1 approach introduces differences in stiffness of the modeled
behavior compared with the target curves for low to medium
Shear Stress - τ-

strains.
Based on the modulus reduction and damping curves obtained
in laboratory tests it is possible to determine the parameters of
the modified hyperbolic model that provide the best fit to one or
the two curves:

A
MR Modulus reduction curve is very well matched to the
B
target curves but the damping curve is not matched due
Shear Strain - γ- to an inherent limitation in Masing load–unload criteria.
MRD Balance the mismatch between the modulus reduction
Fig. 3. Stress–strain relationship. and damping curves.

60 80
Test Data
60
50 Calculated Value
Shear Stress - τ - [kPa]

40
Damping - ξ- [%]

40
20

30 0
Δξ
-20
20
-40
10
-60

0 -80
0.0001 0.001 0.01 0.1 1 10 -0.50 -0.25 0.00 0.25 0.50
Shear Strain - γ - [%] Shear Strain - γ - [%]

Fig. 4. Overestimation of hysteretic damping using Masing rules. (a) Damping curve. (b) Hysteretic loop.
ARTICLE IN PRESS

1148 C. Phillips, Y.M.A. Hashash / Soil Dynamics and Earthquake Engineering 29 (2009) 1143–1158

1.0 30

0.8 25

Damping [%]
20
G/Go 0.6
Target 15
0.4 MR
10
MRD
0.2 5
MD

0.0 0
0.0001 0.001 0.01 0.1 1 10 0.0001 0.001 0.01 0.1 1 10
Shear Strain - γ - [%] Shear Strain - γ - [%]

Fig. 5. Hyperbolic model fitting procedure for sands (reference curve for clays from Yamada et al. [28]): (a) modulus reduction; (b) damping curve.

MD Damping curve is very well matched to the target curves 1.00


but the modulus reduction curve is not matched due to
an inherent limitation in Masing load–unload criteria. 2
R =0.95

Reduction Factor for hysteretic Damping


0.90
Fig. 5 presents a comparison of the different procedures to fit
the modulus reduction and damping curves. F( γ ) = ξ Laboratory / ξ Masing
A solution has been proposed by Darendeli [26] to provide a 0.80
2
better match for both curves (modulus reduction and damping) R =0.99
simultaneously. Using nearly 200 dynamic test results, Darendeli
[26] developed an empirically based modified hyperbolic model to 0.70 2
R =0.99
predict the linear and nonlinear dynamic responses of different
soil types. The developed model is implemented as a reduction
2
factor [Eqs. (15) and (16)] which effectively alters the Masing rules 0.60
R =0.96

xhysteretic ¼ DFðgm Þ xMasing (15)

  0.50
Gg m c 1
DFðgm Þ ¼ b1 (16)
G0

where DF(gm) is Darendeli’s reduction factor, xMasing is the 0.40


hysteretic damping calculated using the Masing rules based on 0.01 0.1 1 10
the modulus reduction curve, Ggm is the secant modulus Shear Strain - γ - [%]
corresponding to the maximum shear strain level gm, b1 is a
Vucetic & Dobry 1991- Zhang et al. 2005-
variable which depends on the soil and input motion properties Quaternary Soil PI = 0 - Fit
PI = 0 - Fit
(plasticity, effective stress, number of cycles, mean frequency) and
Vucetic& Dobry 1991 - Zhang et al. 2005-
c1 is a constant with a value of 0.1.
PI = 0 - Target QuaternarySoil PI = 0 - Target
Darendeli’s reduction factor [Eq. (16)] reduces the hysteretic
damping by 40% for small strains (go105) and by 70% for large Seed & Idriss 1970- Sand Zekkos et al. 2006- 8-25%
Mean - Fit < 20mm - Fit
strains (g4102), obtaining damping values close to the ones
measured in laboratory tests. However, using this reduction factor Seed & Idriss 1970- Sand Zekkos et al. 2006 -108-25%
results in a decrease of the damping for strains greater than Mean - Target < 20mm- Target
102 which is inconsistent with the curves obtained from Fig. 6. Reduction factor as a function of the shear strain.
laboratory tests.

The proposed functional form for F(gm) was obtained by a trial


3.2. New hysteretic damping formulation
and error procedure in which different types of expressions were
tested to fit 50 modulus reduction and damping curves (Appendix
A new expression for a damping reduction factor F(gm) is C). Fig. 6 illustrates the performance of the proposed functional
introduced, this expression modifies the Masing unloading– form using four of the fifty dynamic curves. Fig. 6 is constructed
reloading rules and provides a better agreement with the damping using a four-step process: (1) calculate the hysteretic damping
curves for larger shear strains, but preserves the simplicity of the (xMasing) by using the modulus reduction curves and the extended
solution proposed by Darendeli [26]. Eq. (17) presents the selected Masing rules, (2) determine the target damping reduction ratio at
functional form for the damping reduction factor a given strain as the ratio between the hysteretic damping
  measured in laboratory tests xLaboratory and xMasing, (3) find
Gg p3
Fðgm Þ ¼ p1  p2 1  m (17) variables p1, p2 and p3 that generate the best fit to the xLaboratory/
G0
xMasing and (4) plot the xMasing/xLaboratory ratio vs. strain. Fig. 6 also
in which, p1, p2 and p3 are non-dimensional parameters selected to includes the corresponding goodness of fit R2 of Eq. (17) using the
obtain the best possible fit with the target damping curve. p1, p2 and p3 coefficients which are consistently high. In all the
ARTICLE IN PRESS

C. Phillips, Y.M.A. Hashash / Soil Dynamics and Earthquake Engineering 29 (2009) 1143–1158 1149

4 curves studied, the proposed functional form is able to Fig. 8 compares the damping curves obtained using the new
reproduce closely the relationship between xMasing/xLaboratory and proposed reduction factor (MRDF procedure) with MRD and MR
strain, which results in a very good fit for the corresponding fitting procedures using as target the curves proposed by Yamada
damping curve. et al. [28] for clays. The new factor provides a good fit for both
Appendix C presents the results of the aforementioned modulus reduction and damping curves, even for strains greater
procedure applied to the complete set of 50 selected curves. The than 1%.
R2 coefficients obtained for both modulus reduction [using the As discussed earlier, Muravskii [14] proposed another procedure
hyperbolic backbone curve expression shown in Eq. (11)] and using three different functions [15–17] to modify the extended
damping curve [using the proposed damping reduction factor Masing’s rules and regulate the dependence of the damping ratio on
shown in Eq. (17)]. The agreement is very good to excellent for all the strain amplitude in the process of cyclic deformation of a
50 sets of curves. material. Two sets of examples compare the results of the three
The modulus reduction and damping curves fitting procedures different functions presented by Muravskii [14]. Fig. 9 presents the
using the reduction factor (MRDF) consist of the following three comparison using as target the curves proposed by Yamada et al.
steps and is implemented in DEEPSOIL: [28] for clays, Fig. 10 on the other hand presents the comparison
using as a target the curves for solid waste 8–25%o20 mm [29]. The
methodology proposed by Muravskii [14] (specifically where
1. Determine the best backbone curve parameters of the modified Muravskii [17] function is used to construct the loading and
hyperbolic model to fit the modulus reduction curve. reloading curves) generates a better fit for modulus reduction and
2. Calculate the corresponding damping curve using the backbone damping curves than the use of the MR procedure. MRDF fitting
curve (determined in the previous step) and Masing rules. procedure matches all the points of the damping curve and not only
3. Estimate the reduction factor parameters (p1, p2 and p3) that the maximum damping value as the Muravskii [14] procedure, The
provide the best fit for the damping curve. use of all the points in the curve results in a better fit for both
dynamics curves (as observed in Fig. 10).
Fig. 7 compares the damping curves obtained using the new
proposed reduction factor (MRDF procedure) with MRD and MR
fitting procedures whereby Darendeli’s curves for sand are used as 3.3. Implementation of the proposed model in nonlinear site
target curves. The MRDF approach provides an excellent match for response analysis
the target curves. The MRDF approach does not result in a
reduction in damping which is a limitation of the Darendeli The proposed model is implemented in the 1-D nonlinear site
reduction factor. response analysis (DEEPSOIL) by including the reduction factor to

1.0 30
Target
0.8 25
MR
Da mping [%]

MRD 20
0.6 MRDF
G/Go

15
0.4
10
0.2 5

0.0 0
0.0001 0.001 0.01 0.1 1 10 0.0001 0.001 0.01 0.1 1 10
Shearstrain - γ - [%] Shear Strain - γ - [%]

Fig. 7. Evaluation of proposed damping reduction factor: (a) modulus reduction; (b) using damping curve using Darendeli’s [26] sand curves as target.

1.0 30

0.8 25
Damping [%]

20
0.6
G/Go

15
0.4 Target
MR 10
0.2 MRD 5
MRDF
0.0 0
0.0001 0.001 0.01 0.1 1 10 0.0001 0.001 0.01 0.1 1 10
Shear Strain - γ - [%] Shear Strain - γ - [%]

Fig. 8. Evaluation of proposed damping reduction factor: (a) modulus reduction; (b) damping curve using clays curves proposed by Yamada et al. [28].
ARTICLE IN PRESS

1150 C. Phillips, Y.M.A. Hashash / Soil Dynamics and Earthquake Engineering 29 (2009) 1143–1158

1.0 25

0.8 20

Damping [%]
G/Go 0.6 15
Target
0.4 MRDF 10
MR
Puzrin
0.2 Davidenkov 5
Muravskii
0.0 0
0.0001 0.001 0.01 0.1 1 10 0.0001 0.001 0.01 0.1 1 10
Shear Strain - γ - [%] Shear Strain - γ - [%]

Fig. 9. Evaluation of model fitting procedures for clays Yamada et al. [28]: (a) modulus reduction; (b) damping curve.

1.0 20

0.8 Damping [%] 15


0.6
G/Go

Target 10
MRDF
0.4
MR
Puzrin 5
0.2 Davidenkov
Muravskii
0.0 0
0.0001 0.001 0.01 0.1 1 10 0.0001 0.001 0.01 0.1 1 10
Shear Strain - γ - [%] Shear Strain - γ - [%]

Fig. 10. Evaluation of model fitting procedures for solid waste 8–25%o20 mm [29]: (a) modulus reduction; (b) damping curve.

modify the unloading–reloading equations. The stress–strain not follow the second Masing rule, it follows the other three rules
model implemented in DEEPSOIL [6] is an extension of the model stated in the introduction of this paper.
developed by Duncan and Chang [12] that has been used Seismic loading involves non-symmetrical unloading–reload-
extensively in soil–structure interaction analyses of geotechnical ing cycles and there is a need to properly define of gm. gm can be
structures (referred to in this study as extended modified defined either as the maximum shear strain or the maximum
hyperbolic model). Eqs. (18) and (19) are used to represent the reversal strain. A series of one-dimensional wave propagation
loading and the unloading or reloading conditions, respectively, to analyses were developed to select which option provides a more
calculate the shear stress corresponding to a given shear strain. adequate response. Fig. 11 presents a sketch of the stress–strain
behavior computed for the case of non-symmetrical loading using
G0 g the aforementioned gm definitions. The results show that when gm
t¼ (18)
1 þ b ðg=gr Þs is defined as the maximum reversal shear strain, high levels of
permanent plastic strain are observed; these high values of
 
G0 ððg  grev Þ=2Þ G0 ðg  grev Þ permanent plastic strain are inconsistent with field observations,
t ¼ Fðgm Þ 2  therefore the option of defining gm as the maximum shear strain
1 þ b ððg  grev Þ=2 gr Þs 1 þ b ðgm  gr Þs
G0 ðg  grev Þ at any time was selected.
þ þ trev (19)
1 þ b ðgm  gr Þs

whereby, g: given shear strain, gr: reference shear strain, b: 4. 1D site response analysis using the proposed damping models
dimensionless factor, s: dimensionless exponent, grev: reversal
shear strain, trev: reversal shear stress, gm: maximum shear strain, 4.1. Linear site response with frequency-independent
F(gm): reduction factor and G0: initial shear modulus. viscous damping
The main effect of having the damping reduction factor lower
than one in Eq. (19) is that the hysteretic paths do not follow the A set of four linear site response analyses with constant
Masing rules. When the reduction factor decreases, the tangent damping (x ¼ 5%) are presented to examine the influence of the
shear modulus gets closer to the secant shear modulus corre- proposed frequency-independent viscous damping. Simi Valley
sponding to the maximum shear strain gm, and resulting in a Knolls School (Fig. 12a), Treasure Island (Fig. 12b), Anchorage
decrease in the hysteric damping compared with the same result (Fig. 12c) and 1000 m soil column in the Mississippi Embayment
obtained if the model follows the Masing rules. If the reduction (Fig. 12d) profiles are analyzed to represent shallow stiff, soft clay
factor reaches a value of zero, the tangent shear modulus becomes medium depth, medium stiff glacial and deep sites, respectively.
equal to the secant shear modulus. Even though the model does The input motion in these analyses corresponds to the 1971 San
ARTICLE IN PRESS

C. Phillips, Y.M.A. Hashash / Soil Dynamics and Earthquake Engineering 29 (2009) 1143–1158 1151

Shear Stress - τ -

Shear Stress - τ -
Permanent strain Permanent strain

Shear Strain - γ - Shear Strain - γ -

Fig. 11. Computed stress–strain behavior for: (a) gm defined as maximum shear strain; (b) gm defined as maximum reversal shear strain.

Vs [m/s] Vs [m/s] Vs [m/s] Vs [m/s]


0 500 1000 0 500 1000 0 500 1000 0 500 1000
0 0 0 0
10 100
10
2 20
20 200
30
4 30 40 300

Depth [m]
40 50 400
6
60
50 500
70
8
60 80 600

10 70 90 700
100
80 800
12 110
90 900
120
14 100 130 1000

Fig. 12. Soil columns: (a) Simi Valley Knolls School; (b) Treasure Island; (c) Anchorage; (d) Mississippi Embayment.

0.2
Acceleration [g]

0.1

0.0

-0.1

-0.2
0 2 4 6 8 10 12 14 16 18 20
Time [sec]

Fig. 13. San Fernando record [30].

Fernando earthquake motion (Fig. 13) recorded at Pasadena analyses presented in this section are linear (no modulus
Seismo Laboratory station [obtained from PEER [30]] with reduction and constant damping during the entire time history)
amax ¼ 0.2 g. For highly nonlinear analyses the Gilroy Array record to establish a clear comparison between the results. The
[obtained from PEER [30]] of the Loma Prieta is used (Fig. 14); this frequency domain analyses use a frequency-independent damp-
motion has a PGA value of 0.60 g. ing complex shear modulus scheme [6] and provide the correct
Fig. 15 presents a comparison of the 5% damped elastic solution.
response spectra obtained for each one of the four selected For stiffer soil column (Fig. 15a) the difference between the two
profiles using (1) frequency domain analysis, (2) time-domain time-domain analyses is very small. The two time-domain
analysis using two-mode Rayleigh damping and (3) time-domain solutions are very similar to the frequency domain solution. For
analysis using the new small strain damping model. All the medium soft clay and medium stiff glacial soil profiles (Fig. 15b
ARTICLE IN PRESS

1152 C. Phillips, Y.M.A. Hashash / Soil Dynamics and Earthquake Engineering 29 (2009) 1143–1158

0.6

Acceleration [g]
0.3

0.0

-0.3

-0.6
0.0 2.5 5.0 7.5 10.0 12.5 15.0 17.5 20.0 22.5 25.0
Time [sec]

Fig. 14. Loma Prieta record [30].

1.8 1.8
Frequency
1.5 Domain 1.5
Time Domain
1.2 two modes 1.2
Time Domain
Sa [g]

Sa [g]
0.9 New Model 0.9

0.6 0.6

0.3 0.3

0.0 0.0

1.0 0.5

0.8 0.4

0.6 0.3
Sa [g]

Sa [g]

0.4 0.2

0.2 0.1

0.0 0.0
0.01 0.1 1 10 0.01 0.1 1 10
Period [sec] Period [sec]

Fig. 15. Surface response spectra comparison with constant damping x ¼ 5% profile and linear site response analysis: (a) Simi Valley Knolls School; (b) Treasure Island;
(c) Anchorage; (d) Mississippi Embayment.

and c) an increase in spectral acceleration values between 0.1 and The Anchorage soil profile (Fig. 12c) is used to determine the
0.3 s of period is obtained for the time-domain analysis using effectiveness of the model when the damping through the soil
Rayleigh damping. The overestimation of the seismic response profile is not constant (with small strain damping decreasing with
results from lower damping values for those frequencies between the increase of effective stress) as presented in Fig. 16a.
the two target frequencies (see Fig. 2). The proposed damping Fig. 16b presents the response spectra comparison whereby
formulation time-domain solution provides results that are very good agreement between the frequency-independent time-
similar to those from the frequency domain solutions. For the domain solution and the frequency domain solution. Although it
deep, long period, deposits the spectra (Fig. 15d) obtained using has been demonstrated that the model is able to capture the
the new approach is closer to the correct response (frequency damping independent of the frequency, it should be noted that
domain solution) compared to the one obtained using the two- there are still some differences between frequency domain and
mode Rayleigh damping. The observed difference of the time- time-domain propagation analyses caused by the frequency cut-
domain solutions for the 1000 m depth Mississippi Embayment off (highest frequency that a layer can propagate; greater than
column is a result of the important contribution of modes 50 Hz in all the examples presented in this paper) in the time-
different from the ones selected for the Rayleigh damping two- domain analyses and the issues related with the number of points
mode solution in the deposit’s dynamic response. Therefore, the of the input motion in the frequency domain analyses (i.e. if the
overestimation and underestimation of damping for these modes, number of points cannot be expressed as 2x the solution requires
in the two-mode Rayleigh scheme, generates a decrease and an addition of zeros to the motion changing the minimum frequency
increase for those frequencies, respectively. in the input motion). An important advantage of the proposed
ARTICLE IN PRESS

C. Phillips, Y.M.A. Hashash / Soil Dynamics and Earthquake Engineering 29 (2009) 1143–1158 1153

ξ Small Strain [%]


0 1 2
0 1.5
Frequency Domain
10 Time Domain two modes
Time Domain New Model
20
1.2
30
40
0.9
50
Depth [m]

Sa [g]
60
70 0.6
80
90
0.3
100
110
0.0
120
0.01 0.1 1 10
130 Period [sec]

Fig. 16. Anchorage profile linear site response analysis with variable damping (a) small strain damping profile Anchorage soil column. (b) Surface response spectra
comparison.

1.0 35
Target
MR 30
0.8 MRD
25
Damping [%]

MRDF
0.6 20
G/Go

0.4 15
10
0.2
5
0.0 0

1.0 35
30
0.8
25
Damping [%]

0.6 20
G/Go

0.4 15
10
0.2
5
0.0 0

1.0 35
30
0.8
25
Damping [%]

0.6 20
G/Go

0.4 15
10
0.2
5
0.0 0
0.0001 0.001 0.01 0.1 1 0.0001 0.001 0.01 0.1 1
Shear strain - γ - [%] Shear strain - γ - [%]

Fig. 17. Dynamic soil properties of Anchorage profile (a) modulus reduction and damping curves. (a) and (b) Middle point of Fill layer. (c) and (d) Middle point of BCF layer.
(e) and (f) Middle point of Glaciofluvial layer.
ARTICLE IN PRESS

1154 C. Phillips, Y.M.A. Hashash / Soil Dynamics and Earthquake Engineering 29 (2009) 1143–1158

Max Shear Strain [%]


0.00 0.05 0.10 0.15 0.20
0 1.0
10 MRDF
MR
20
0.8 MRD
30
EL
40
50 0.6
Depth [m]

Sa [g]
60
70
80 0.4

90
100
0.2
110
120
130 0.0
0.01 0.1 1 10
Period [sec]

Fig. 18. Anchorage profile nonlinear site response analysis with two-mode Rayleigh viscous damping: (a) maximum shear strain profile; (b) surface response spectra
comparison.

model is that it is no longer necessary to estimate the modes 4.3. Nonlinear site response with frequency-independent
needed in Rayleigh damping formulation. viscous damping

To test the complete damping scheme (frequency-independent


4.2. Nonlinear site response with two-mode Rayleigh viscous damping and new hysteretic damping formulation) a
viscous damping series of analyses using the Anchorage profile (Fig. 12c) and the
San Fernando earthquake as input motion (Fig. 13) were devel-
A set of nonlinear site response analyses, using the Anchorage oped. EL, MR, MR+D (Modulus Reduction Fit with frequency-
profile (Fig. 12c), are presented to evaluate the influence of the independent viscous damping), MRDF and MRDF+D (complete
proposed hysteretic damping reduction factor [MRDF, Eq. (17)]. model) analyses were developed. The computed maximum
The commonly used two-mode Rayleigh damping is employed to shear strain profile and response spectra obtained are presented
separately examine the effect of the new hysteretic damping in Fig. 19.
formulation. Fig. 17 shows the modulus reduction and damping MRDF+D analysis provides similar results to EL analysis in
curves used in the analyses. terms of PGA values and spectral accelerations. It should be noted
The analysis results are presented in Fig. 18 and include that although MR, MR+D, MRDF and MRDF+D procedures are
results using equivalent linear (EL, frequency domain), MR, MRD nonlinear time-domain analyses the results provided in this
and the proposed MRDF time-domain approaches. The EL results particular example differ greatly. Both maximum shears strain
do not represent the correct solution in the current set of in the profile and spectral acceleration are greater in the MRDF+D
analyses. In general, MR and MRD analyses give a surface response analysis compared with the obtained values for the MR and
that is lower than the equivalent linear approach in the short and MR+D analyses. This demonstrates that if the range of shear
longer period ranges but similar results in the mid-period strains is small to medium the increase of the damping
(0.1–0.4 s) range. The MRDF analysis provides a response that is values (viscous and hysteretic damping) in the MR and MR+D
significantly higher than both MR and MRD analyses. The MRDF models generate attenuation throughout the profile and a
spectrum is slightly lower than the equivalent linear spectrum in reduction of the maximum accelerations within the soil column.
the short and long period ranges but higher in the mid-period For nonlinear analysis the use of the frequency-independent small
(0.1–0.4 s) range. Interestingly, the maximum shear strains strain damping generates an increase in the peak ground
profiles for all analyses are similar, except between 30 and 40 m, acceleration and high-frequency spectral accelerations. This result
depths at which the maximum shear strains are higher for the could be observed for both MR and MRDF procedures, Fig. 19,
MRDF analysis. when are compared with the MR+D and MRDF+D results,
The result can be explained by the fact that in the MR analysis respectively.
damping is overestimated, which results in attenuation of the The last example presented in this paper corresponds to the
spectral accelerations and strains. In the case of the MRD analysis use of the two new damping formulations simultaneously for a
the mismatch in the stiffness and the overestimation of the strong input motion; using as an input motion the Gilroy record of
damping generates the observed decrease of the spectral accel- the Loma Prieta earthquake (Fig. 14) and the Anchorage profile
erations for periods lower than 0.60 s and a reduction of the (Fig. 12c). EL, MR and MRDF+D analyses were developed. The
maximum shear strains for the upper 40 m of the soil profile. computed maximum shear strain profile and response spectra
ARTICLE IN PRESS

C. Phillips, Y.M.A. Hashash / Soil Dynamics and Earthquake Engineering 29 (2009) 1143–1158 1155

Max Shear Strain [%]


0.00 0.05 0.10 0.15 0.20
0 1.0
MR DF+D
10
MR DF
20 MR +D
0.8 MR
30
EL
40
50 0.6
Depth [m]

Sa [g]
60
70
80 0.4

90
100
0.2
110
120
130 0.0
0.01 0.1 1 10
Period [sec]

Fig. 19. Anchorage profile nonlinear site response analysis with frequency-independent viscous damping; San Fernando earthquake: (a) maximum shear strain profile;
(b) surface response spectra comparisons.

Max Shear Strain [%]


0.00 1.00 2.00 3.00 4.00
0 1.2
10 MRDF+D
MR
20 1.0
EL
30

40
0.8
50
Depth [m]

Sa [g]

60
0.6
70

80
0.4
90

100
110
0.2

120

130 0.0
0.01 0.1 1 10
Period [sec]

Fig. 20. Anchorage profile nonlinear site response analysis with frequency-independent viscous damping; Loma Prieta earthquake: (a) maximum shear strain profile;
(b) surface response spectra comparisons.

obtained are presented in Fig. 20. Although the use of MRDF+D 5. Conclusions
method results in a lower PGA value than the one obtained in the
EL analysis, the spectral accelerations corresponding to periods In this paper, we proposed two new methodologies to model
between 0.03 and 0.3 s (correspondent to most of the 1, 2 and 3 energy dissipation in one-dimensional site response analysis. A
stories buildings constructed in the country) are higher when the series of time-domain elastic site response analyses (with and
MRDF+D model is used. The use of the MR model results in lower without constant damping) show that the proposed frequency-
spectral acceleration values; a result that could be explained by independent viscous damping provides results that more closely
the overestimation of the hysteretic damping in this model. The match the exact solution obtained for frequency domain analyses.
strain profiles are very similar for the two nonlinear time-domain The proposed methodology is an alternative to the commonly
analyses; however, the maximum strain for the upper 20 m of the used Rayleigh damping formulation and circumvents the need to
profile is higher when the MRDF+D model is used. identify the two modes needed in Rayleigh damping.
ARTICLE IN PRESS

1156 C. Phillips, Y.M.A. Hashash / Soil Dynamics and Earthquake Engineering 29 (2009) 1143–1158

A simplified constitutive model to simultaneously match Based on the results of Eq. (A.3) it can be obtained that
modulus reduction and damping soil curves for nonlinear site
2=k
response analysis is introduced. The proposed model uses the ð½M1 ½KÞ1=k ¼ ½Fdiagfoj g½F1 (A.7)
widely used modified hyperbolic model to represent the modulus
Damping matrix [C] is calculated using the following equation
reduction backbone curve. A damping reduction factor, that
modifies the Masing rule, is introduced such that hysteretic X
N 1
damping matches more closely the measured strain dependent ½C ¼ ½M ab ð½M1 ½KÞb (A.8)
damping. The model was tested for 50 different soil curves and a b¼0

good match for modulus reduction and damping curves is obtained. If the term b of Eq. (A.8) is equal only to 1
the damping matrix
2
The proposed models overcome some longstanding limitations can be calculated as
in nonlinear time-domain analysis regarding small and large
strain damping. All the analyses presented in this paper ½C ¼ ½Ma1=2 ½Fdiagfoj g½F1 (A.9)
correspond to total stress analysis. The implementation of the
new models to effective stress analyses is beyond the scope of this
paper and is currently under investigation.

Appendix B
Appendix A

If [B]k ¼ [A] and [A] has a matrix of similitude definite diagonal


matrix [D] ¼ diag{dj}
[A] matrix can be obtained as

½A ¼ ½P½D½P1 (A.1)


On the other hand [B] can be found to be
1=k
½B ¼ ½P½D1=k ½P1 ¼ ½Pfdj g½P1 (A.2)

where [P] is a transformation matrix of similitude.


The eigenvalues of matrix [A] are not affected by this
transformation. As a result of the previous statement the diagonal
[D] should contain all the eigenvalues of matrix [A].
Therefore,
1=k
½A1=k ¼ ½B ¼ ½P½D1=k ½P1 ¼ ½Pfdj g½P1 (A.3)

In the problem solved mass, damping and stiffness matrices are


symmetrical and positive definite. The dynamical system has
eigenvalues o and normalized eigenvectors F. Eigenvalues and
normalized eigenvectors have the following relationships with the
mass and stiffness matrices, respectively

½FT ½M½F ¼ ½I; ½FT ½K½F ¼ diagfo2j g (A.4)

Because real modes of the system are orthogonal

½M1 ¼ ½F½FT ; ½K ¼ ½FT diagfo2j g½F1 (A.5)

As a result

½M1 ½K ¼ ½Fdiagfo2j g½F1 (A.6)


ARTICLE IN PRESS

C. Phillips, Y.M.A. Hashash / Soil Dynamics and Earthquake Engineering 29 (2009) 1143–1158 1157

Appendix C

No. Reference Year Description b s gr xsmall (%) p1 p2 p3 R2 G/Go R2 x

1 Vucetic and Dobry [31] 1991 PI ¼ 0 1.000 0.870 0.031 0.882 0.940 0.380 0.550 0.999 1.000
2 Vucetic and Dobry [31] 1991 PI ¼ 15 1.000 0.810 0.070 0.913 0.980 0.380 0.700 0.999 0.997
3 Vucetic and Dobry [31] 1991 PI ¼ 30 1.000 0.810 0.139 0.953 0.960 0.360 0.500 0.999 0.993
4 Vucetic and Dobry [31] 1991 PI ¼ 50 1.000 0.840 0.276 0.979 0.920 0.380 0.500 0.999 0.990
5 Vucetic and Dobry [31] 1991 PI ¼ 100 1.000 0.930 0.562 0.996 0.660 0.280 1.000 0.998 0.991
6 Vucetic and Dobry [31] 1991 PI ¼ 200 1.000 1.140 0.897 1.000 0.600 0.360 0.950 0.999 0.986
7 Boulanger [32] 1998 Sherman Island peat 0.330 0.840 0.278 1.994 0.820 0.380 1.400 0.995 0.981
8 Seed & Idriss [33] 1970 Upper 1.000 0.930 0.060 0.204 0.700 0.200 0.600 0.999 0.998
9 Seed & Idriss [33] 1970 Mean 1.000 0.870 0.040 0.381 0.980 0.380 1.850 0.999 0.998
10 Seed & Idriss [33] 1970 Lower 1.060 0.810 0.024 0.533 0.980 0.280 3.200 0.999 0.993
11 Zhang et al. [34] 2005 Quaternary soil PI ¼ 0 1.560 0.810 0.139 0.239 0.780 0.280 3.100 0.999 1.000
12 Zhang et al. [34] 2005 Tertiary soil PI ¼ 0 1.558 0.960 0.053 0.251 0.660 0.260 3.200 0.999 0.999
13 Zhang et al. [34] 2005 Residual soil PI ¼ 0 1.570 0.810 0.076 0.190 0.860 0.320 1.600 0.999 1.000
14 Zhang et al. [34] 2005 Quaternary PI ¼ 15 1.570 0.780 0.170 1.440 0.740 0.200 2.150 1.000 0.997
15 Zhang et al. [34] 2005 Quaternary PI ¼ 30 1.570 0.810 0.194 1.458 0.700 0.200 2.800 1.000 0.996
16 Zhang et al. [34] 2005 Quaternary PI ¼ 50 1.564 0.900 0.214 1.481 0.640 0.240 3.100 1.000 0.998
17 Zhang et al. [34] 2005 Quaternary PI ¼ 100 1.570 0.930 0.318 1.489 0.680 0.300 3.050 1.000 0.998
18 Zekkos et al. [29] 2006 100%o20 mm 0.750 0.870 0.095 2.977 0.640 0.200 3.200 1.000 0.999
19 Zekkos et al. [29] 2006 62–76%o20 mm 0.750 0.720 0.222 3.966 0.720 0.200 0.650 0.998 0.996
20 Zekkos et al. [29] 2006 8–25%o20 mm 0.750 0.780 0.642 3.991 0.640 0.200 0.500 0.999 1.000
21 Borja et al. [35] 2000 6m 1.320 0.990 0.048 3.450 0.840 0.380 3.000 0.996 0.979
22 Borja et al. [35] 2000 11 m 1.300 0.870 0.045 1.385 0.980 0.380 2.300 0.998 0.988
23 Borja et al. [35] 2000 17 m 1.270 0.720 0.046 1.261 0.980 0.200 4.000 0.999 0.986
24 Darendeli [26] 2001 s0 0 ¼ 0.25 atm-SM 0.980 0.870 0.019 0.977 0.660 0.240 3.000 1.000 0.999
25 Darendeli [26] 2001 s0 0 ¼ 1.0 atm-SM 0.980 0.840 0.036 0.717 0.660 0.220 3.000 1.000 0.999
26 Darendeli [26] 2001 s0 0 ¼ 4.0 atm-SM 1.000 0.840 0.072 0.552 0.660 0.220 3.000 0.999 0.998
27 Darendeli [26] 2001 s0 0 ¼ 16.0 atm-SM 1.000 0.840 0.151 0.575 0.660 0.220 3.000 1.000 0.999
28 Yamada et al. [28] 2008 Dejima clay Fc ¼ 99.1% 1.000 0.990 0.202 1.093 0.660 0.380 1.100 0.999 0.998
29 Yamada et al. [28] 2008 Dejima clay Fc ¼ 43.0% 1.000 1.020 0.066 1.484 0.600 0.320 1.650 1.000 0.998
30 Okur & Ansal [36] 2007 PI ¼ 12% 0.750 0.990 0.043 2.878 0.600 0.280 0.500 1.000 0.964
31 Okur & Ansal [36] 2007 PI ¼ 27% 0.756 0.960 0.083 1.986 0.600 0.280 0.550 1.000 0.972
32 Okur & Ansal [36] 2007 PI ¼ 43% 0.748 0.990 0.150 1.393 0.620 0.340 1.500 0.998 0.978
33 Kramer [37] 2000 Mercer Slough peat* 1.020 0.840 0.027 2.857 0.900 0.380 0.500 0.994 0.998
34 Bozzano et al. [38] 2008 Roma’s clay 0.900 1.290 0.138 1.898 0.980 0.300 4.000 1.000 0.989
35 Crespellani et al. [39] 2001 Roma’s clay 0.930 1.200 0.041 4.986 0.920 0.380 3.100 1.000 0.995
36 Carrubba & Maugeri [40] 1999 Roma’s clay 0.750 1.230 0.153 1.999 0.600 0.300 1.300 1.000 0.990
37 Cardona & Yamin [41] 1997 PI ¼ 20 1.560 0.660 0.225 2.891 0.760 0.200 2.600 1.000 0.989
38 Cardona & Yamin [41] 1997 PI ¼ 40 1.500 0.750 0.338 2.963 0.660 0.200 3.250 1.000 0.993
39 Cardona & Yamin [41] 1997 PI ¼ 60 1.560 0.840 0.442 2.985 0.600 0.200 2.350 1.000 0.993
40 Cardona & Yamin [41] 1997 PI ¼ 80 1.580 0.870 0.533 2.990 0.600 0.240 3.050 1.000 0.990
41 Cardona & Yamin [41] 1997 PI ¼ 100 1.600 0.900 0.608 2.993 0.600 0.240 2.200 1.000 0.989
42 Cardona & Yamin [41] 1997 PI ¼ 150 1.460 0.960 0.649 2.996 0.620 0.280 1.350 1.000 0.988
43 Darendeli [26] 2001 PI ¼ 0/s0 v ¼ 50 kPa 1.000 0.930 0.025 1.031 0.700 0.320 2.550 1.000 0.992
44 Darendeli [26] 2001 PI ¼ 0/s0 v ¼ 100 kPa 1.000 0.930 0.031 0.848 0.680 0.300 2.800 1.000 0.994
45 Darendeli [26] 2001 PI ¼ 0/s0 v ¼ 500 kPa 1.000 0.900 0.057 0.528 0.660 0.280 3.000 1.000 0.996
46 Darendeli [26] 2001 PI ¼ 0/s0 v ¼ 1000 kPa 1.000 0.900 0.073 0.436 0.660 0.240 3.100 1.000 0.997
47 Darendeli [26] 2001 PI ¼ 20 1.000 0.930 0.050 1.132 0.640 0.240 3.500 1.000 0.996
48 Darendeli [26] 2001 PI ¼ 50 1.000 0.930 0.078 1.554 0.640 0.240 3.500 1.000 0.997
49 Darendeli [26] 2001 PI ¼ 100 1.000 0.900 0.124 2.260 0.640 0.220 3.500 1.000 0.998
50 Darendeli [26] 2001 PI ¼ 200 1.000 0.900 0.215 3.658 0.640 0.200 3.500 1.000 0.999

References [8] Park D, Hashash YMA. Soil damping formulation in nonlinear time domain
site response analysis. Journal of Earthquake Engineering 2004;8(2):
249–74.
[1] Schnabel PB, Lysmer JL, Seed HB. SHAKE: a computer program for earthquake [9] Kwok AOL, Stewart JP, Hashash YMA, Matasovic N, Pyke R, Wang Z, Yang Z.
response analysis of horizontally layered sites. In: Earthquake Engineering Use of exact solutions of wave propagation problems to guide implementa-
Research Center, Berkeley, CA, 1972. tion of nonlinear, time-domain ground response analysis routines. ASCE
[2] Kramer SL. Geotechnical earthquake engineering, Vol. xviii. Upper Saddle Journal of Geotechnical and Geoenvironmental Engineering 2007;133(11):
River, NJ: Prentice Hall; 1996. p. 653. 1337–481.
[3] Newmark NM. A method of computation for structural dynamics. Journal of [10] Gerolymos N, Gazetas G. Constitutive model for 1-D cyclic soil behaviour
the Engineering Mechanics Division 1959;85:67–94. applied to seismic analysis of layered deposits. Soils and Foundation 2005;
[4] Lee MK, Finn WDL. DESRA-2, Dynamic effective stress response analysis 45(3):147–59.
of soil deposits with energy transmitting boundary including assessment [11] Pyke RM. Nonlinear soil models for irregular cyclic loadings. Journal of the
of liquefaction potential. In: Soil Mechanics Series, No. 36, Vol. 36, Geotechnical Engineering Division 1979;105(GT6):715–26.
Department of Civil Engineering, University of British Columbia, Vancouver, [12] Duncan JM, Chang C-Y. Nonlinear analysis of stress and strain in soils.
Canada, 1978. Journal of the Soil Mechanics and Foundations Division 1970;96(SM5):
[5] Matasovic N. Seismic response of composite horizontally-layered soil 1629–53.
deposits, Ph.D. Thesis, University of California, Los Angeles, 1993. [13] Pyke R. TESS: a computer program for nonlinear ground response analyses.
[6] Hashash YMA. DEEPSOIL V 2.6, tutorial and user manual. 2002–2005. In: In: TAGA Engineering Systems and Software, Lafayette, California, 2000.
University of Illinois at Urbana-Champaign, Urbana Illinois, 2005, p. 49. [14] Muravskii G. On description of hysteretic behaviour of materials. Interna-
[7] Masing G. Eignespannungen und verfestigung beim messing. In: Second tional Journal of Solids and Structures 2005;42:2625–44.
International Congress on Applied Mechanics, Zurich, Switzerland, 1926, [15] Davidenkov NN. Energy dissipation in vibrations. Journal of Technical Physics
pp. 332–335. 1938;8(6).
ARTICLE IN PRESS

1158 C. Phillips, Y.M.A. Hashash / Soil Dynamics and Earthquake Engineering 29 (2009) 1143–1158

[16] Puzrin AM, Burland JB. A logarithmic stress–strain function for rocks and [29] Zekkos D, Bray JD, Riemer MF. Shear modulus and material damping of
soils. Geotechnique 1996;46(1):157–64. municipal solid waste based on large-scale cyclic triaxial testing. Canadian
[17] Muravskii G, Frydman S. Site response analysis using a nonlinear Geotechnical Journal 2006;45(1):45–58.
hysteretic model. Soil Dynamics and Earthquake Engineering 1998;17: [30] PEER, P.E.E.R.C., PEER Strong Database, in, Vol. 2007 (PEER, 2000).
227–38. [31] Vucetic M, Dobry R. Effect of soil plasticity on cyclic response. Journal of
[18] Rayleigh JWS, Lindsay RB. The theory of sound, Vol. 2(1). New York: Dover Geotechnical Engineering 1991;117(1):87–107.
Publications; 1945. [32] Boulanger RW, Arulnathan R, Harder LFJ, Torres RA, Driller MW. Dynamic
[19] Lai CG, Rix GJ. Simultaneous inversion of Rayleigh phase velocity and properties of Sherman Island peat. Journal of Geotechnical and Geoenviron-
attenuation for near-surface site characterization. In: Georgia Institute of mental Engineering 1998;124(1):12–20.
Technology, 1998. [33] Seed HB, Idriss IM. Soil moduli and damping factors for dynamic response
[20] Hudson M, Idriss IM, Beikae M. 1994. QUAD4 M-A computer program to analyses. College of Engineering University of California Berkeley, Berkeley,
evaluate the seismic response of soil structures using finite element 1970, p. 41.
procedures and incorporating a compliant base. In: Center for Geotechnical [34] Zhang J, Andrus RD, Juang CH. Normalized shear modulus and material
Modeling, Department of Civil and Environmental Engineering, University of damping ratio relationships. Journal of Geotechnical and Geoenvironmental
California, Davis, CA., Davis, CA, 1994. Engineering 2005;131(4):453–60.
[21] Wilson E. Dynamic analysis by numerical integration. In: Wilson E, editor. [35] Borja RI, Lin C-H, Sama KM, Masada GM. Modelling non-linear ground
Technical reports. San Francisco, California: Computers and Structures Inc.; response of non-liquefiable soils. Earthquake Engineering and Structural
2005. Dynamics 2000;29(1):63–83.
[22] Clough RW, Penzien J. Dynamics of structures, Vol. xxii. New York: McGraw- [36] Okur DV, Ansal A. Stiffness degradation of natural fine grained soils during
Hill; 1993. 738 p. cyclic loading. Soil Dynamics and Earthquake Engineering 2007;27(9):
[23] Liu M, Gorman DG. Formulation of Rayleigh damping and its extensions. 843–54.
Computers and Structures 1995;57(2):277–85. [37] Kramer SL. Dynamic response of Mercer Slough peat. Journal of Geotechnical
[24] Press WH, Teukolsky SA, Vetterling WT, Flannery BP. Numerical recipes in and Geoenvironmental Engineering 2000;126(6):504–10.
FORTRAN 77: the art of scientific computing. Cambridge, UK: Cambridge [38] Bozzano F, Caserta A, Govoni A, Marra F, Martino S. Static and dynamic
University Press; 1992. p. 963. characterization of alluvial deposits in the Tiber River Valley: new data for
[25] Park D, Hashash YM. Estimation of non-linear seismic site effects for deep assessing potential ground motion in the City of Rome. Journal of Geophysical
deposits of the Mississippi Embayment. In: Mid America Earthquake Center, Research B: Solid Earth 2008;113(1) [art. no. B01303].
Urbana, 2004, p. 276. [39] Crespellani T, Madiai C, Simoni G, Vanucchi G. Dynamic geotechnical testing
[26] Darendeli MB. Development of a new family of normalized modulus and seismic response analyses in two sites of the Commune of Nocera Umbra,
reduction and material damping curves, Ph. D., Civil Engineering, University Italy. Rivista Italiana di Geotecnica 2001;35(4):39–52.
of Texas at Austin, 2001. [40] Carrubba P, Maugeri M. Dynamic soil properties from full scale tests
[27] Seed HB, Wong RT, Idriss IM, Tokimatsu K. Moduli and damping factors for with a vibrating foundation. In: 2nd international conference on earth-
dynamic analyses of cohesionless soils. Journal of the Soil Mechanics and quake geotechnical engineering, Vol. 1. Lisbon, Portugal: Balkema; 1999
Foundations Division, ASCE 1986;112(SM11). pp. 35–40.
[28] Yamada S, Hyodo M, Orense RVDS, Hyodo T. Strain-dependent dynamic [41] Cardona OD, Yamin LE. Seismic microzonation and estimation of earthquake
properties of remolded sand–clay mixtures. Journal of Geotechnical and loss scenarios: integrated risk mitigation project of Bogota, Colombia.
Geoenvironmental Engineering 2008;134(7):972–81. Earthquake Spectra 1997;13(4):795–813.

Das könnte Ihnen auch gefallen