Sunteți pe pagina 1din 7

Engineering Geology 234 (2018) 167–173

Contents lists available at ScienceDirect

Engineering Geology
journal homepage: www.elsevier.com/locate/enggeo

Probability density evolution method for seismic displacement-based T


assessment of earth retaining structures

Yu Huang , Hongqiang Hu, Min Xiong
Department of Geotechnical Engineering, College of Civil Engineering, Tongji University, Shanghai 200092, China

A R T I C L E I N F O A B S T R A C T

Keywords: Earthquakes have damaged many earth-retaining structures worldwide, causing much loss of life and property as
Earth retaining structures well as disrupting traffic. To assess the seismic performance of earth-retaining structures in earthquake-prone
Stochastic ground motions areas, a new probabilistic methodology is proposed, namely the probability density evolution method (PDEM).
Probabilistic methodology As a novel dynamic reliability method based on random vibration theory, PDEM combines stochastic dynamic
Dynamic reliability
analysis with current advanced deterministic methods. Not only does PDEM give the dynamic reliability of the
Seismic performance
global system of a structure, it also gives the probability density function and its evolution characteristics of the
seismic response of interest. A tie-back wall is chosen as the type of earth retaining structures with which to show
the feasibility of the method. After stochastic ground motions are generated by stochastic ground motion model,
a series of deterministic seismic responses of the tie-back wall are obtained by the finite-element method. The
PDEM is then used to obtain the stochastic seismic responses and the dynamic reliability of the system. Finally,
the seismic performance of the tie-back wall is assessed using performance indices and displacement-based
criteria selected from standards and the literature. High efficiency and accuracy of the PDEM is verified by
comparing its results with those from Monte Carlo simulations. Not only can this novel method give more
objective indices than can traditional evaluation methods, it can also provide a reference for the performance-
based seismic analysis and design of geotechnical engineering structures.

1. Introduction unrealistically large reinforcement length (Ling et al., 1997). Likewise,


Al Atik and Sitar (2010) have stated that the current seismic design of
Earth-retaining structures (ERSs), one kind of geotechnical en- earth-retaining walls is overly conservative. All these demonstrate that
gineering structure, are used extensively in highway and railway sys- ERS seismic design and analysis using the pseudo-static method is un-
tems, slope reinforcement, landslide stabilization and geological dis- reasonable to some extent. In recent decades, some displacement-based
aster prevention. However, ERSs can be seriously damaged or destroyed methods for ERS seismic analysis and design have been proposed (Ling
by earthquakes, leading to much loss of life and property. Such damage et al., 1997), as have been seismic displacement charts to assess ERS
can also hamper disaster relief work by affecting safe personnel eva- seismic performance (Huang et al., 2009). Methods have also been
cuation and delivery of materials to the disaster areas, as were the cases presented to calculate permanent ERS displacements (Michalowski and
in the 1999 Chi-Chi earthquake in Taiwan, China (Fang et al., 2003) You, 2000; Meehan and Vahedifard, 2013). Additionally, Trandafir
and the 2008 Wenchuan earthquake in Sichuan, China. In the 41 roads et al. (2009) showed that the 2004 Chuetsu earthquake in Japan led to
investigated after Wenchuan earthquake in the disaster areas, there are slope failure due to excessive ERS displacement. Ichii (2004) have
over 347 ERSs were damaged, and most of them had to be removed and pointed out that the extent of horizontal displacement might be related
reconstructed (Zhang et al., 2012). to the type of damage done to retaining structures and could be con-
Traditional ERS stability analysis and design involve the limit sidered as a damage criterion from the perspective of performance-
equilibrium method with safety factors. For seismic conditions, the based seismic design. Furthermore, Huang et al. (2009) have stated that
pseudo-static method is used with Mononobe–Okabe theory. However, performance-based seismic design of retaining walls is required to mi-
even though these approaches are still widely used today, some scholars tigate disasters and reduce design costs in earthquake-prone areas.
and engineers have identified certain deficiencies. Using the pseudo- Therefore, displacement is a reasonable performance index with which
static method for geosynthetically reinforced earth structures can cause to assess ERS seismic performance.


Corresponding author.
E-mail addresses: yhuang@tongji.edu.cn (Y. Huang), 1510265@tongji.edu.cn (H. Hu), 1310298@tongji.edu.cn (M. Xiong).

https://doi.org/10.1016/j.enggeo.2018.01.019
Received 10 October 2017; Received in revised form 25 December 2017; Accepted 21 January 2018
0013-7952/ © 2018 Elsevier B.V. All rights reserved.
Y. Huang et al. Engineering Geology 234 (2018) 167–173

Traditional ERS stability analysis and design involves a determi- the performance function. But it is very difficult to acquire the explicit
nistic approach. However, seismic ERS stability analysis involves con- expression of performance function of nonlinear seismic response of
siderable variability and uncertainty, such as the seismic load and the ERS. MCS is considered to be effective for the study and reliability
properties of backfill soil and retaining structures. These uncertainties analysis of stochastic dynamic systems but is difficult to apply in en-
mean that the ERS displacement is also uncertain, and hence such major gineering practice because of the large number of simulations required
uncertainties should be taken into consideration when we assess the (Schuëller, 2006).
seismic performance of ERS. In recent studies, reliability theory, which Integrating deterministic and probabilistic analyses in a com-
can quantify these uncertainties (Wang et al., 2013), has been applied plementary manner can bring together the best of our knowledge
to the ERS analysis and design. Yuan et al. (2003) used the limit (Lacasse, 2016). The probability density evolution method (PDEM) is a
equilibrium method to analyze the reliability of soil-nailed walls. As novel probabilistic method based on random vibration theory for sol-
random variables, Yuan et al. considered cohesion, the internal friction ving stochastic dynamic systems (Li and Chen, 2009). Combines sto-
angle of the soil, and the mobilization degree of friction resistance, and chastic and deterministic dynamic time history analyses, PDEM can also
they conducted a parametric study into the influence of these variables decouple randomness and nonlinearity. By deterministic dynamic time
on the safety factor and reliability index. Zevgolis and Bourdeau (2010) history analysis, the seismic ground motion characteristics, magnifica-
used Monte Carlo simulation (MCS) to compute reliability under static tion effect and dynamic nonlinear characteristics of soil can be taken
conditions by assuming probability distributions for the backfill and into account, which are the parts that most other reliability method
foundation material properties. For seismic conditions, Shinoda et al. cannot realize. In addition, not only can PDEM obtain the second-order
(2006) used a low-discrepancy-sequence Monte Carlo method and an statistics of seismic responses, it can also obtain the probability density
important sampling method to carry out a probabilistic seismic de- function (PDF) at different time points and PDF evolution character-
formation analysis by considering the variability of soil properties, istics of stochastic seismic responses, while most other reliability
structural height, and slope height. Sayed et al. (2008) used the first- methods can only obtain PDF and reliability by supposing the seismic
order reliability method (FORM), the first-order second-moment response obey some kind of distributions. Furthermore, PDEM has been
method (FOSM), and the point estimate method (PEM) to conduct re- shown to be feasible for geotechnical problems (Huang et al., 2015;
liability analyses of reinforced soil walls under both static and seismic Huang and Xiong, 2017).
conditions. They considered soil and reinforcement-parameter un- In the present paper, we use PDEM to analyze stochastic dynamic
certainties and evaluated the seismic force by pseudo-static method. system of an ERS. This leads to the PDFs of the relevant physical
Basha and Babu (2009) presented a reliability-based way of analyzing parameters and the dynamic reliability of the global ERS system. As our
the internal and external seismic stability of reinforced soil structures. ERS, we chose a tie-back wall with which to assess the validity of PDEM
They treated soil properties and geosynthetic reinforcement as random for modeling a stochastic dynamic system of ERS. We used a stochastic
variables and calculated the horizontal seismic acceleration based on ground motion model to produce a set of earthquake ground motions of
pseudo-static approach or pseudo-dynamic approach. Kramer and a given site firstly, which is the way we considered the randomness of
Paulsen (2004) selected 20 recorded earthquake motions to evaluate earthquake ground motions. We then used a deterministic finite-ele-
the seismic performance and parametric sensitivity of a reinforced slope ment method to obtain a series of seismic responses of the tie-back wall,
by the Monte Carlo method, and they used the modified Newmark and applied PDEM to obtain the stochastic seismic responses (PDFs of
method to calculate the permanent displacement. displacement and dynamic reliability) of tie-back wall. Finally, we as-
From the above literature review, although many studies have sessed its seismic performance based on displacement-based seismic
proposed several methods for ERS reliability analysis that can consider criteria. Considering the MCS is an accurate enough reliability method,
the uncertainties of soil and structural properties under both static and we verified the high efficiency and accuracy of PDEM by comparing its
seismic conditions, there are several aspects that limit the extensive use results with those of MCS. It is shown that PDEM is more than 10 times
and effectiveness of these methods. Firstly, as pointed out by Bray and than traditional MSC. PDEM have the ability to conduct refined relia-
Travasarou (2007), the primary source of uncertainty is the input bility analysis of retaining structures, which is very useful for reliability
ground motion when assessing the seismic performance of earth sys- analysis, design and reliability-based optimization of geotechnical
tems. But few considered the uncertainty of the seismic ground motions structures in engineering practice. To a certain extent, it can also ensure
in the seismic performance assessment of retaining structures. Earth- the seismic safety of the geotechnical structures and reduce the oc-
quake ground motion is intrinsically random (Housner, 1947), so is the currence of geological disasters in earthquake prone areas.
seismic response of ERSs under the random seismic excitation. Ad-
ditionally, the pseudo-static approach and Newmark's method some-
times neglect certain characteristics of seismic ground motion (e.g., its 2. Seismic displacement criteria for earth retaining structures
time-history and frequency characteristics) and fail to consider both the
magnification effect and dynamic nonlinear characteristics of soil. As mentioned above, displacement is a reliable performance index
Using dynamic time-history analysis can solve these problems to some with which to assess the seismic performance of ERSs. Many scholars
extent. Hence, it is necessary to assess ERS seismic performance from and some national specifications have specified seismic displacement
the perspective of dynamic reliability and dynamic time-history ana- criteria for evaluating and designing earth-retaining walls (Table 1).
lysis. In another aspect, some reliability methods, like first-order re- These criteria differ somewhat because of different requirements re-
liability method, the first-order second-moment method, need to obtain garding traffic engineering and slope engineering. Matthewson et al.
(1980) proposed that the top displacement of the retaining wall should

Table 1
Seismic displacement criteria proposed for seismic performance assessment of earth retaining structures.

Criterion Matthewson et al. Commission of the European Wu and Prakash AASHTO (2002) Huang et al.
(1980) Communities (1994) (1996) (2009)

Permissible horizontal 3% of wall height 300amax (mm); amax is maximum design 2% of wall height 250amax (mm); amax is maximum 2% of wall
displacement ground acceleration (g) design ground acceleration (g) height
Failure horizontal 10% of wall 5% of wall
displacement height height

168
Y. Huang et al. Engineering Geology 234 (2018) 167–173

m
Table 2 ∂ρzΘ (z , θ, t ) ∂ρzΘ (z , θ, t )
Three different seismic performance levels and displacement criteria of earth retaining ∂t
+ ∑ ZJ̇ (θ, t ) ∂zj
=0
structures (Zhang and Han, 2012). j=1 (5)

Performance level Criterion Failure state Corresponding with the initial condition
seismic level
ρzΘ (z , θ, t0) = ρΘ (θ, t ) δ (z − z 0) (6)
Level 1 1% Slight displacement; does Low-level
not affect normal use earthquake where ρzΘ(z, θ, t0) is the joint probability function of (Z, Θ)(Θ = θ), z0 is
Level 2 3.5% Obvious displacement; Design earthquake the initial value of Z(t), and δ is the Dirac function. The PDF of Z(t) can
normal use can be restored be expressed as
after repair
Level 3 6% Large deformation; does
not collapse and can be
High-level
earthquake
ρZ (Z , t ) = ∫ΩΘ ρZΘ (z, θ, t ) dθ (7)
used in emergence
situations In Eq. (5), the dimensions of the PDEM equation depend on those of
the physical parameters being evaluated. To assess the seismic perfor-
mance of an ERS, the parameter of interest is the horizontal displace-
not exceed 3% of the wall height. Eurocode (1994) advised 300amax ment. Hence, we choose the displacement time history as the random
(mm) as the permissible horizontal displacement of soil retaining walls, process in the PDEM equation, whereupon the following one-dimen-
where amax is the maximum design ground acceleration (g). Wu and sional PDEM equation can be obtained:
Prakash (1996) suggested 2% of wall height as the permissible hor-
∂ρDs Θ (Ds , θ, t) ∂ρzΘ (Ds , θ, t)
izontal displacement and 10% of wall height as the failure horizontal + Ḋs (θ, t) =0
displacement. The Guide (2002) gave 250amax (mm) as the permissible ∂t ∂Ds (8)
horizontal displacement. Huang et al. (2009) also proposed 2% of wall
where Ds is the seismic time history of the ERS top displacement.
height as the permissible horizontal displacement of ERSs based on a
Eq. (8) can be solved using the following numerical method (Li and
series of shaking table tests and seismic displacement analysis using
Chen, 2009):
Newmark's method. Many ERSs were damaged in the Wenchuan
earthquake, and Zhang and Han (2012) proposed performance criteria
1) Select representative discretized points θq(q = 1, 2, 3, ⋯, nsel) in the
for earth-retaining walls by combining field investigations of these
space of the basic random variable ΩΘ, whereupon the corre-
damaged ERSs and laboratory model tests. They gave three different
sponding probability can be determined; nsel is the number of dis-
seismic performance levels (Table 2) that they pointed out correspond
cretized points.
to the three different earthquake levels in the Chinese seismic code
2) For every determined θq(q = 1, 2, 3, ⋯, nsel), solve the dynamic bal-
(Ministry of Housing and Urban-Rural of People's Republic of China,
ance equation (Eq. (1)) by dynamic time-history analysis under the
2010), namely a low-level earthquake, a design earthquake, and a high-
given earthquake excitations respectively and obtain a set of seismic
level earthquake.
responses.
3) Substitute a series of the displacement time histories into the PDEM
3. PDEM method and dynamic reliability
equation (Eq. (8)). Use the finite-difference method to solve the
PDEM equation and hence obtain ρDsΘ(Ds, θ, t)(q = 1, 2, 3, ⋯, nsel).
In general, the dynamic balance equation of an ERS under earth-
4) Sum the results of ρDsΘ(Ds, θ, t)(q = 1, 2, 3, ⋯, nsel) to obtain the PDFs
quake excitation can be written as
of displacement.
¨ + CẊ + f (X) = −MIχ̈ g (Θ, t)
MX (1)
As described in the solution process, the PDEM solution is based on
with the deterministic initial conditions a series of deterministic dynamic time-history analysis which means
X (t 0) = ẋ 0, X (t 0) = x 0 (2) using the PDEM to solve stochastic dynamic system of ERSs can make
good use of current advanced deterministic numerical techniques. To
where M is the mass matrix, C is the damping matrix, Ẍ , X, and X are obtain dynamic reliability for a global ERS system, a virtual stochastic
the acceleration, velocity, and displacement vectors, respectively, f is process can be constructed by using an equivalent extreme value event
the nonlinear restoring-force vector, I is a unit vector, χ̈ g is the earth- (Li et al., 2007). The time history of the extreme value of each seismic
quake ground motion, and Θ is a random vector that reflects the ran- horizontal displacement can constitute a virtual stochastic process. By
domness of χ̈ g . The physical solutions of Eq. (1) exists and depend on Θ. substituting this stochastic process into the PDEM equation and solving
For convenience, the solutions of Eq. (1) can be written as it, the dynamic reliability of ERS can be obtained.
X = H (Θ, t), Ẋ = h (Θ, t) (3)
4. Stochastic seismic ground motions
where H = (H1, H2, ⋯Hm) and h = (h1, h2, ⋯hm) .
T T

In general, the interest in a structural system such as Eq. (1) is in its


The first and significant procedure of evaluating the seismic per-
physical parameters, such as displacement, velocity, acceleration, and
formance of engineering structures is to obtain a set of time histories of
strain. These attracting parameters can be obtained from constitutive
relations and deformation compatibility relations in terms of X and Ẋ . If the earthquake ground motions; as mentioned in Section 3, this is also
the parameter of interest is Z, then it can also be expressed as functions the key step in solving the PDEM equation. In the present study, we
of Θ: obtain the acceleration time histories of seismic ground motions by
spectral representation and the stochastic function method (Liu et al.,
Z = HZ (Θ, t) Ż = hZ (Θ, t) (4) 2016), so as to consider the randomness of the ground motion. Earth-
Eq. (4) can be regarded as a stochastic dynamic process whose quake ground motion is assumed to be a zero-mean-value stochastic
randomness all arise from Θ. Throughout the evolutionary process of process. According to spectral representation method, a zero-mean and
stochastic system, if no random factors are added and disappeared in real non-stationary stochastic seismic ground motion can be written as
the extended stochastic system (Z(t), Θ), it is a conserved probabilistic N
system. Based on this, Li and Chen (2009) established a generalized X¨ g (t ) = ∑ 2S X¨g (t , ωk )Δω [cos(ωk ) Xk + sin(ωk ) Yk ]
probability evolution equation (PDEM equation), namely k=1 (9)

169
Y. Huang et al. Engineering Geology 234 (2018) 167–173

ωk = k Δω (10) 250

where S X¨g (t , ωk ) is the improved evolutionary power spectral density 200


function of the non-stationary seismic stochastic process X¨ g (t ) , {Xk, Yk}
150
(k = 1, 2, 3…N) are standard orthogonal random variables, and
Δω = 0.2 rad/s is the frequency interval.

Acceleration [cm/s2]
100
Based on the idea of stochastic functions (Liu et al., 2016), the ex-
pression of random function of {Xk, Yk}(k = 1, 2, 3…N) can be con- 50
structed. Assume that any two sets of standard orthogonal random
0
variables are functions of basic random variables θ, respectively. Then
the random functions can be established as -50

π π -100
Xk = 2 cos ⎛kθ + ⎞ Yk = 2 sin ⎛kθ + ⎞
⎝ 4⎠ ⎝ 4⎠ (11)
-150
in which θ basic random variables with uniform distributions in the
interval [− π,π]. -200
Based on the Clough–Penzien spectrum model of stationary ground
-250
motion, Deodatis (1996) proposed an evolutionary power-spectrum 0 5 10 15 20 25 30
model of non-stationary ground motion. Cacciola and Deodatis (2011) Time [s]
improved it, giving the evolutionary power spectral density function as
(a)
ωg4 (t ) + 4ξg2 (t ) ωg2 (t ) ω2 4
S X¨g (t , ω) = A2 (t ) ∙ ·
[ω2 − ωg2 (t )]2 + 4ξg2 (t ) ωg2 (t ) ω2 Samples

Mean [cm/s2]
2
Target
ω4
·S0 (t ) 0
[ω2 − ωf2 (t )]2 + 4ξ f2 (t ) ωf2 (t ) ω2 (12)
-2
d
t t
A2 (t ) = ⎡ exp ⎛1 − ⎞ ⎤ -4
⎣ c ⎝ c ⎠⎦ (13) 0 5 10 15 20 25 30
Time [s]
where A(t) is the intensity modulation function, S0(t) is the spectral
intensity factor, c is the time of the peak value, and d is the shape 60
control index of A(t). The parameters that control the non-stationary Samples
Std.D [cm/s2]

characteristics of the earthquake ground motions are listed as follows 40 Target


(Cacciola and Deodatis, 2011):
t t 20
ωt (t ) = ω0 − a ξg (t ) = ξ0 + b
T T (14)
0
ωf (t ) = 0.1ωg (t ) ξ f (t ) = ξg (t ) (15) 0 5 10 15 20 25 30
Time [s]
a b
ω0 = ωg + ξ 0 = ξg − (b)
2 2 (16)
where ω0, ξ0, a, and b can be determined by site classification, ωg(t) and Fig. 1. (a) A typical simulated time history of earthquake ground motion. (b) Second-
ξg(t) are the angular and damping ratios of the site soil, respectively, order statistics between 610 earthquake ground motion samples and stochastic ground
motion model.
and ωf(t) and ξf(t) are the corresponding filtering parameters.
In Eq. (12), the parameter S0(t) that reflects the intensity of the
earthquake is To generate a set of non-stationary stochastic seismic ground mo-
2
tions, the representative sample points of uniformly distributed random
2amax
S0 (t ) = variables θ are selected firstly by
1
r 2πωg (t ) ⎛2ξg (t ) + ⎞
θ = −π + 2(j − 0.5)π/nsel
⎝ 2ξg (t )
⎠ (17) (18)
2
where is the average value of the peak ground acceleration (which
amax Where nsel is the total number of representative sample points, j = 1, 2,
can be determined from the earthquake time history) and r is the ef- 3…nsel. Then substitute the every θ into Eqs. (11) and (9), and we can
fective peak factor (which can be determined by optimally fitting the obtain a set of time histories of earthquake ground motions by the
earthquake response spectrum). stochastic earthquake ground-motion model and the spectral re-
In the evolutionary power spectral density function (12), c and d presentation method. In present study, the 610 samples of stochastic
reflect the earthquake intensity characteristics, ω0, ξ0, a, and b re- earthquake ground motions are generated. And the second-order sta-
2
present non-stationary frequency characteristics, amax and r reflect the tistics between our 610 samples and the stochastic earthquake ground-
intensity of the seismic ground motion, and T is the duration of the motion model are shown in Fig. 1, which verify the accuracy of spectral
earthquake. All these earthquake ground-motion parameters are de- representation and the stochastic function method in generating a set of
termined by seismic zoning and site grouping according to the Chinese stochastic acceleration time histories of ground motions.
seismic code (Ministry of Housing and Urban-Rural of People's Republic
of China, 2010). The parameters we used in present study are as fol-
5. Numerical study
lows:

ω0 = 11 s−1; ξ0 = 0.7; a = 8 s−1; b = 0.15; c = 7 s; d = 2; r = 2.65; T We choose a tie-back wall as the type of ERS with which to de-
monstrate the feasibility, accuracy, and efficiency of the PDEM. The
= 30 s
configuration is shown in Fig. 2. It is a sheet-pile wall tied back with

170
Y. Huang et al. Engineering Geology 234 (2018) 167–173

3 Bonded portion of anchors 60


Free portion of anchors PDF at 2 sec
3 Sheet pile wall
PDF at 2.5 sec
50 PDF at 3 sec
3
3
40
11
8

PDF
30

12 8 20
20
Fig. 2. Geometry of tie-back wall (unit: m).

two rows of anchors that have a bonded length and a free length (GEO- 10
SLOPE International Ltd., 2010). The height of sheet pile wall is 9 m,
and there are also 3 m under the ground. The sheet pile wall is modeled
0
by beam elements. The ties (anchors) are 8.4 m in length, and the in- -0.1 -0.05 0 0.05 0.1
clination angle is about 17.35°.The bonded portion is modeled as a Displacement(m)
beam that can be pulled and compressed, and the free length is modeled
as a bar. There is no need to set the friction parameters for anchors in Fig. 4. Probability density functions (PDFs) of displacement at three different time points.

GEO-SLOPE. The main physical parameters of soil are as follows: nat-


ural unit weight γ = 20 kN/m3, cohesion force c = 20 kPa, internal
friction angle φ = 25°, and Poisson ratio μ = 0.334. We choose the
Mohr–Coulomb model with which to characterize the soil properties. As
60
stated earlier, the PDEM can transform the stochastic dynamic analysis
of a system into a series deterministic seismic time-history analysis and 50
the solution of the PDEM equation. We used the finite-element method
40
to obtain a series of seismic responses of a tie-back wall.
PDF

Displacement is reliable performance index with which to assess the 30


seismic performance of an ERS. After we have finished a set of de-
20
terministic dynamic time-history analyses, we have acquired a set of
displacement–time curves of the tie-back wall. We then input these data 10
regarding the time histories of top displacement into the PDEM equa- 0
tion and solve the PDEM equation by the finite-difference method. This 0.05
gives the stochastic seismic response and dynamic reliability of the 3
global system of the structure. Dis 0
plac 2.5
The second-order statistics (mean and standard deviation) of the top e me
nt(m
displacement are shown in Fig. 3, which show great variability. The -0.05 2 sec)
standard deviation of displacement increases with time, partly because
) Time(
the cumulative displacement increases with time under stochastic Fig. 5. PDF surface of displacement.
seismic loading, thus increasing the variability of the displacement. The
PDFs of top displacement at three typical different time points are also
PDFs using a conventional regular distribution model such as a normal
illustrated in Fig. 4, from which it is quite clear that the PDFs change
distribution or a logarithmic normal distribution. The three-dimen-
with time and show great variability. It is impossible to describe the
sional graph (Fig. 5) is PDF surface of displacement at time interval of
2–3 s, and it shows that the PDFs evolve with time and also shows great
2 variability; it has many peaks and troughs that appear like mountains
Mean(10 -6m)

1 and valleys. The peaks of the PDFs mean large PDF values and high
probability within this time interval. The displacement is distributed
0 mainly between ± 0.01 m within the time interval of 2–3 s.
One significant advantage of the PDEM is that it can acquire the
-1
actual PDFs of the seismic response of the structural system at different
-2 time points, which provides the basis for obtaining an insight into the
0 5 10 15 20 25 30
stochastic dynamic system and conducting a refined dynamic reliability
Time(s) analysis. Considering that the PDFs in our study show great variability
0.8 and evolution characteristics, further study is needed to understand the
physical nature of the process and the mechanism governing the evo-
0.6
Std.D(m)

lution of the displacement PDFs in the stochastic dynamic system of


0.4 ERS.
We used MCS to verify the accuracy and efficiency of the PDEM.
0.2 MCS is based on stochastic simulation of large numbers of samples and
is appropriate for stochastic analysis of a system with an arbitrary
0
0 5 10 15 20 25 30 number of degrees of freedom and different sources of randomness. In
Time(s) addition, it is widely accepted that MCS is an accurate enough sto-
chastic method that is generally used to check and verify the accuracy
Fig. 3. Mean and standard deviation of top displacement of tie-back wall.

171
Y. Huang et al. Engineering Geology 234 (2018) 167–173

60
1
MCS: PDF at 2 sec
PDEM: PDF at 2 sec 0.9
50
0.8

0.7
40
0.6

CDF
PDF

30 0.5

0.4
20 0.3
0.2
10
0.1

0
0 -0.1 0 0.1 0.2 0.3 0.4 0.5 0.6
-0.1 -0.05 0 0.05 0.1
Displacement(m) Displacement(m)
(a) Fig. 7. Cumulative distribution function (CDF) of equivalent extreme-value event.

20 PDEM equation gives the PDF of the maximum displacement. The cu-
MCS: PDF at 3 sec mulative distribution function (CDF) (Fig. 7) is obtained by integrating
18 PDEM: PDF at 3 sec the PDF of the maximum displacement.
16 In this study, we select the displacement criterion proposed by
Zhang and Han (2012) to assess the seismic performance of a tie-back
14 wall. If we choose 3.5% of the wall height as the threshold between
12 satisfactory and unsatisfactory seismic performance of the tie-back wall,
its dynamic reliability is 0.9375, which is considered as safe according
PDF

10 to the reliability criterion proposed by Huang and Xiong (2017).


8

6 6. Case analysis of practical engineering

4 To illustrate the usefulness of our approach in practical engineering,


2 a sheet pile wall, which is damaged in Wenchuan earthquake, is se-
lected. The sheet pile wall is located at the section of
0 DK1016 + 800–DK1016 + 835 in 213 State Road in Sichuan province
-0.1 -0.05 0 0.05 0.1 0.15
of China. It is a wall of 10 m high with 5 m embedded in the ground.
Displacement(m)
The width of the pile is 1.5 m. The height of slope crest to the pavement
(b) is about 71 m, and the degree of slope is about 41°. The geometry of
anti-slide pile wall is illustrated in Fig. 8. The main properties of soil
Fig. 6. Comparison of displacement PDFs of MCS and PDEM at two different time points.
and bedrock are as follows (Zhang and Han, 2012): soil: natural unit
weight γ1 = 20 kN/m3, cohesion c1 = 20 kPa, friction angle φ1 = 35°;
of other stochastic and reliability methods under both static and seismic bedrock: natural unit weight γ2 = 26 kN/m3, cohesion c2 = 50 kPa,
conditions (Schuëller, 2006). In our study, MCS has simulated 10,000 friction angle φ2 = 40°. After Wenchuan earthquake, Zhang et al.
times dynamic time history analysis, and we compared the results of (2012) carried out the systematic post-earthquake investigation and
MCS and PDEM in relation to the PDF level (Fig. 6(a) and (b)). The MCS found out that there was about 3° tilt of the sheet pile wall and the
and PDEM PDFs at 2 s and 3 s agree relatively well with each other permanent displacement of the pile top reaches about 20 cm, which can
which has shown the accuracy of PDEM. To better show the efficiency
of PDEM, the calculation times of two methods using the same com-
puter (3.40 GHz CPU and 12GB RAM) are compared. The MCS took Bed rock
Accumulation body
about 416.67 h, while PDEM took only 25.42 h approximately. It has to
Pile
be mentioned that a series of seismic responses of dynamic time history 66
analysis should be substituted into generalized probability evolution
equation to get stochastic seismic response in PDEM method, but this 213 State Road
time is negligible comparing with time of dynamic time history ana-
lysis. Our comparison results demonstrate that the PDEM is much more 5
efficient (more than 10 times) than MCS with an acceptable degree of 5
precision. Hence, the PDEM can be applied in practice for dynamic
reliability assessment and design of ERSs because of its high efficiency. 45
Considering that the maximum top displacement is not a time his-
tory, to obtain the PDF of maximum displacement, the virtual stochastic
process was constructed by using the equivalent extreme value event (Li 50 2 60
et al., 2007). Taking this virtual stochastic process as the input into the
Fig. 8. Geometry of sheet pile wall (unit: m).

172
Y. Huang et al. Engineering Geology 234 (2018) 167–173

be considered as safe in performance level 2 according the performance Geotech. Geoenviron. 136 (10), 1324–1333.
criteria mentioned in Section 2. Basha, B.M., Babu, G.S., 2009. Optimum design for external seismic stability of geosyn-
thetic reinforced soil walls: reliability based approach. J. Geotech. Geoenviron. 136
According to the process of PDEM method, we firstly used spectral (6), 797–812.
representation and stochastic method (in Section 4) to obtain a set of Bray, J.D., Travasarou, T., 2007. Simplified procedure for estimating earthquake-induced
stochastic ground motions. The parameters in stochastic ground motion deviatoric slope displacements. J. Geotech. Geoenviron. 133 (4), 381–392.
Cacciola, P., Deodatis, G., 2011. A method for generating fully non-stationary and spec-
model are determined by seismic zoning and site grouping according to trum-compatible ground motion vector processes. Soil Dyn. Earthq. Eng. 31 (3),
Chinese seismic code (Ministry of Housing and Urban-Rural of People's 351–360.
Republic of China, 2010). Then PDEM, combining with deterministic Deodatis, G., 1996. Non-stationary stochastic vector processes: seismic ground motion
applications. Probabilistic Eng. Mech. 11 (3), 149–167.
dynamic time history analysis, was applied to acquire the dynamic Eurocode, E., 1994. Structures in seismic regions. In: Part 5: Foundations, Retaining
reliability of sheet pile wall. If 3.5% of the wall height is also chosen as Structures, and Geotechnical Aspects. Commission of the European Communities,
the threshold for the seismic performance of retaining wall, then the Brussels.
Fang, Y.S., Yang, Y.C., Chen, T.J., 2003. Retaining walls damaged in the Chi-Chi earth-
dynamic reliability of sheet pile wall corresponding to performance
quake. Can. Geotech. J. 40 (6), 1142–1153.
level 2 is 0.9663, which also represent safe. Our result shows agreement GEO-SLOPE International Ltd, 2010. Dynamic Modeling with QUAKE/W.
with the result of post-earthquake investigation. Further studies are Guide, R.D., 2002. American Association of State Highway and Transportation Officials.
needed to research the performance-based design and dynamic relia- Washington, DC.
Housner, G.W., 1947. Characteristics of strong-motion earthquakes. Bull. Seismol. Soc.
bility-based optimal design for the improvement of the seismic perfor- Am. 37 (1), 19–31.
mance of retaining structures in earthquake-prone areas. Huang, Y., Xiong, M., 2017. Probability density evolution method for seismic liquefaction
performance analysis of earth dam. Earthq. Eng. Struct. Dyn. 46 (6), 925–943.
Huang, C.C., Wu, S.H., Wu, H.J., 2009. Seismic displacement criterion for soil retaining
7. Conclusions walls based on soil strength mobilization. J. Geotech. Geoenviron. 135 (1), 74–83.
Huang, Y., Xiong, M., Zhou, H., 2015. Ground seismic response analysis based on the
In this study, the PDEM method is proposed for assessing the seismic probability density evolution method. Eng. Geol. 198, 30–39.
Ichii, K., 2004. Fragility curves for gravity-type quay walls based on effective stress
performance of ERSs from the perspective of stochastic dynamics. The analyses. In: 13th World Conference on Earthquake Engineering, Vancouver, BC,
following conclusions can be drawn: Paper No. 3040.
Kramer, S.L., Paulsen, S.B., 2004. Seismic performance evaluation of reinforced slopes.
Geosynth. Int. 11 (6), 429–438.
(1) Comparing with traditional ERS analysis and design methods, we Lacasse, S., 2016. Hazard, Reliability and Risk Assessment-Research and Practice for
selected displacement with which to assess the seismic performance Increased Safety. In: Proceedings of the 17th Nordic Geotechnical Meeting Challenges
of an ERS, thereby facilitating design cost effectiveness and re- in Nordic Geotechnic, Icelandic Geotechnical Society, Reykjavik, pp. 17–42.
Li, J., Chen, J., 2009. Stochastic Dynamics of Structures. John Wiley & Sons.
vealing the nature of the destruction to some extent.
Li, J., Chen, J.B., Fan, W.L., 2007. The equivalent extreme-value event and evaluation of
(2) An earthquake ground motion has random characteristic and is a the structural system reliability. Struct. Saf. 29 (2), 112–131.
non-stationary process. Under excitation of stochastic earthquake Ling, H.I., Leshchinsky, D., Perry, E.B., 1997. Seismic design and performance of geo-
ground motions, the ERS displacement is also a random process. It synthetic-reinforced soil structures. Geotechnique 47 (5), 933–952.
Liu, Z., Liu, W., Peng, Y., 2016. Random function based spectral representation of sta-
is of great significance to analyze the seismic performance of such tionary and non-stationary stochastic processes. Probabilistic Eng. Mech. 45,
structures from the perspective of reliability. Using the dynamic 115–126.
reliability method, we can quantify the uncertainties of earthquake Matthewson, M.B., Wood, J.B., Berril, J.B., 1980. Seismic design of bridges; section 9:
earth retaining structures. Bull. NZ Natl. Soc. Earthq. Eng. 13 (3), 3.
ground motions. Meehan, C.L., Vahedifard, F., 2013. Evaluation of simplified methods for predicting
(3) The PDEM method is a novel dynamic reliability method based on earthquake-induced slope displacements in earth dams and embankments. Eng. Geol.
random vibration theory. By using the PDEM method, we not only 152 (1), 180–193.
Michalowski, R.L., You, L., 2000. Displacements of reinforced slopes subjected to seismic
obtain the second-order statistics of displacement but also the PDFs. loads. J. Geotech. Geoenviron. 126 (8), 685–694.
The top-displacement PDFs of a retaining structure under random Ministry of Housing and Urban-Rural of People's Republic of China, 2010. Code for
earthquake excitation is complex and varies considerably with time. Seismic Design of Buildings (GB50011-2010). (in Chinese).
Sayed, S., Dodagoudar, G.R., Rajagopal, K., 2008. Reliability analysis of reinforced soil
Based on the reliability and performance-based seismic criteria, the
walls under static and seismic forces. Geosynth. Int. 15 (4), 246–257.
dynamic reliability of a tie-back wall under stochastic earthquake Schuëller, G.I., 2006. Developments in stochastic structural mechanics. Arch. Appl. Mech.
ground motions is 0.9375, which can be regard as safe. 75 (10), 755–773.
Shinoda, M., Horii, K., Yonezawa, T., Tateyama, M., Koseki, J., 2006. Reliability-based
(4) This study demonstrates the advantages and validity of assessing
seismic deformation analysis of reinforced soil slopes. Soils Found. 46 (4), 477–490.
the seismic performance of retaining structures using dynamic re- Trandafir, A.C., Kamai, T., Sidle, R.C., 2009. Earthquake-induced displacements of
liability and stochastic dynamics, which can provide a more ob- gravity retaining walls and anchor-reinforced slopes. Soil Dyn. Earthq. Eng. 29 (3),
jective index than can traditional evaluation methods. In addition, 428–437.
Wang, L., Hwang, J.H., Juang, C.H., Atamturktur, S., 2013. Reliability-based design of
this study can serve as a reference for the performance-based rock slopes—a new perspective on design robustness. Eng. Geol. 154, 56–63.
seismic design of geotechnical engineering structures, so as to en- Wu, Y., Prakash, S., 1996. On seismic displacements of rigid retaining walls. In: Analysis
sure the safety and seismic performance of geotechnical structures and Design of Retaining Structures Against Earthquakes, pp. 21–37.
Yuan, J.X., Yang, Y., Tham, L.G., Lee, P.K.K., Tsui, Y., 2003. New approach to limit
and reduce the earthquake-induced geological disaster in earth- equilibrium and reliability analysis of soil nailed walls. Int. J. Geomech. 3 (2),
quake-prone zones. 145–151.
Zevgolis, I.E., Bourdeau, P.L., 2010. Probabilistic analysis of retaining walls. Comput.
Geotech. 37 (3), 359–373.
Acknowledgments Zhang, J., Han, P., 2012. Displacement-based aseismic design method for gravity re-
taining walls—large scale shaking table tests. Chin. J. Geotech. Eng. 34 (3), 417–423
This work was supported by the National Science Fund for (in Chinese).
Zhang, J., Qu, H., Liao, Y., Ma, Y., 2012. Seismic damage of earth structures of road
Distinguished Young Scholars of China [Grant No. 41625011].
engineering in the 2008 Wenchuan Earthquake. Environ. Earth Sci. 65 (4), 987–993.

References

Al Atik, L., Sitar, N., 2010. Seismic earth pressures on cantilever retaining structures. J.

173

S-ar putea să vă placă și