Beruflich Dokumente
Kultur Dokumente
Ocean Engineering
journal homepage: www.elsevier.com/locate/oceaneng
art ic l e i nf o a b s t r a c t
Article history: The seabed response under wave loading is important for the stability of foundations of offshore
Received 7 May 2015 structures. Unlike previous studies, the wave-induced seabed response in a spatially random porous
Accepted 9 October 2015 seabed is investigated in this study. A stochastic finite element model which integrates random field
simulation of spatially varied soil properties and finite element modeling of wave-induced seabed
Keywords: response is established. Spatial variability of soil shear modulus, soil permeability, and degree of
Porous seabed saturation, are simulated using the covariance matrix decomposition method. The results indicate that
Wave the pore water pressure and stress distribution in the seabed are significantly affected by the spatial
Spatial variability variability of the shear modulus. The mean of maximum oscillating pore pressure in a randomly het-
Random field
erogeneous seabed is greater than that in a homogenous seabed. The uncertainty of the maximum
Stochastic
oscillating pore pressure first increases and then reduces with the correlation length. The effects of
Monte Carlo simulation
spatial variability of permeability and degree of saturation are less significant than that of shear modulus.
The spatial variability of soil permeability increases the uncertainty of oscillating pore pressure in the
seabed at shallow depths. The spatial variability of shear modulus and degree of saturation affects the
uncertainty of oscillating pore pressure of the whole seabed.
& 2015 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.oceaneng.2015.10.016
0029-8018/& 2015 Elsevier Ltd. All rights reserved.
L.L. Zhang et al. / Ocean Engineering 111 (2016) 116–127 117
2.1. Governing equations where G is the shear modulus, and μ is the Poisson's ratio. G is
related to Young's modulus E and the Poisson's ratio μ in the form
The problem considered in this paper is two-dimensional. of E/2(1 þ μ). In these formulations of the wave–seabed interaction
Assume a porous seabed of finite thickness h (Fig. 1). The x-axis is problem, the soil shear modulus G, soil permeability, and degree of
taken on the seabed surface and the z-axis is taken vertically saturation are expressed as constant parameters, but they will be
downward from the seabed surface. The waves travel from left to simulated as spatially random soil properties by random field
right along the positive direction of the x-axis. Based on Biot's simulation in this study.
consolidation theory (Biot, 1941), together with the storage
equation (Verruijt, 1969), the governing equation for the wave–soil 2.2. Boundary conditions
interaction problem within a compressible pore fluid in a com-
The governing equations, Eqs. (1) and (5), describing the
pressible porous seabed is given as:
water–soil interaction problem, can be solved by incorporating the
∂ ∂p ∂ ∂p ∂p ∂ε boundary conditions specified at the seabed surface and the
Kx þ Kz þ γ w nβ þ γ w ¼ 0 ð1Þ
∂x ∂x ∂z ∂z ∂t ∂t impermeable bottom, respectively. It is commonly accepted that
118 L.L. Zhang et al. / Ocean Engineering 111 (2016) 116–127
the vertical effective normal stresses and shear stresses vanish at al., 1997). The soil layer is a homogenous hydraulically placed sand.
the seabed surface, and the pore pressure is equal to the wave These six in situ cone penetration tests were performed at 9 m
pressure at the seabed surface (Jeng, 2013): intervals in a straight line. As shown in the figure, the mean values
σ 0z ¼ τxz ¼ 0 at z ¼ 0 ð6Þ of soil properties (systematic trends) vary in space. There are
always some fluctuations about mean values even within the
p ¼ p0 cos ðkx ωtÞ at z¼0 ð7Þ homogenous layer. Some degrees of coherence between the fluc-
tuations can be observed, which become stronger as the measur-
where p0 is the amplitude of the dynamic wave pressure based on ing points become closer. For even deeper depths of seabed, nat-
the short-crested wave theory to the first order, which can be ural spatial variability of the soil properties is also observed. Fig. 3
expressed as shows the soil property profiles of the Pleistocene clay deposit in
γw Hs Osaka Bay from the site investigation of Kansai International Air-
p0 ¼ ð8Þ
2 cosh kd port (Watabe et al., 2002). The Pleistocene clay deposit includes
in which Hs is the wave height; k is the wave number (k¼ 2π/L, in several sandy soil layers between the clay layers. Profiles of the
which L is the wave length); ω is the angular frequency of the liquid limit (wL), plastic limit (wP), natural water content (wn), void
wave (ω ¼ 2π/T, T is the wave period); and d is the water depth ratio e, consolidation yield stress (σ 0y ), and over-consolidation ratio
above the seabed surface. In this study, the linear wave theory is (OCR) for the Pleistocene clay deposit from 25 to 230 m below the
adopted to determine the wavelength with given wave period and seabed are presented on the graph. As shown on the graph, natural
water depth based on the wave dispersion equation (Phillips, spatial variability can be observed for different soil properties.
1977). A matlab code is compiled to determine the wavelength The spatial variation of soil properties can be described using
with given wave periods and water depth using the Newton's the geotechnical random field theory (Vanmarcke 1977; Fenton
numerical method. and Griffiths, 2008). The soil property in the field is modeled as
For the soil resting on an impermeable rigid bottom below a the sum of a trend component and a residual term. The residuals
finite thickness, as depicted in Fig. 1, zero displacements and no of the trend to exhibit spatial correlation, which can be expressed
vertical flow occur: through the autocovariance function and the autocorrelation
function. The autocovariance function and autocorrelation func-
∂p
u¼w¼0 and ¼0 at z ¼ h ð9Þ tion are usually estimated from soil samples from a site (Lloret-
∂z
Cabot et al., 2014). Analytical models can be used to fit the sample
When modeling an infinitely long seabed in a numerical model, autocorrelation functions using regression analysis. A commonly
the length of the seabed in the numerical model can be assumed used autocorrelation function is the exponential model (Van-
to be equal to multiple wavelengths. Since the boundary condition marke, 1977; Fenton and Griffiths, 2008):
of the lateral boundaries is periodic in both time and space, it is
2τ
reasonable to assume that u, w, and p are the same for the two ρðτÞ ¼ exp ð10Þ
lateral boundaries using the principle of repeatability.
δ
where τ is the separation distance between two points; δ is the
correlation length, which describes the distance within which the
3. Stochastic finite-element model of a heterogeneous seabed spatially random values will tend to be significantly correlated (i.e.,
by more than about 10%). Within a range of δ, soil property values
3.1. Spatial variability of marine soils are highly correlated, but in a range over δ, soil property values
can be considered as statistically independent.
Inherent spatial variability exists in seafloor sediments. Fig. 2 Research studies about the characterization of spatial varia-
presents the field-measured cone tip resistances up to 25 m depth bility of the geotechnical properties of marine soils are relatively
at one artificial island, which was later used as a drilling platform limited. Hoeg and Tang (1977) and Tang (1979) reported that the
for oil exploration on the Canadian Beaufort sea shelf (Popescu et horizontal correlation length of CPT cone tip resistance in a marine
0
MAC31 MAC04 MAC05 MAC06 MAC07 MAC08
-5
Elevation (m)
-10
-15
-20
0 10 20 0 10 20 0 10 20 0 10 20 0 10 20 0 10 20
Cone tip resistance (MPa)
Fig. 2. In situ cone tip resistances of a homogenous sand layer in the Canadian Beaufort Sea (modified from Popescu et al. 1997). The grey lines represent systematic trends.
L.L. Zhang et al. / Ocean Engineering 111 (2016) 116–127 119
Fig. 3. Profiles of physical soil properties of Pleistocene clay in Osaka Bay (modified from Watabe et al., 2002).
Table 1
Correlation length of geotechnical soil properties for marine soils.
clay was 55 m for the shallow soil layer within 3 m depth of the Table 2
seabed. For greater depths of soils, the correlation length in the List of deterministic input parameters.
horizontal direction varied from 35 to 60 m. Keaveny et al. (1990)
Parameter Unit Value
investigated the spatial variability of offshore soils and found that
the horizontal correlation length for the CPT cone tip resistance Seabed thickness h m 15
varies from 14 to 38 m. Valdez-Llamas et al. (2003) estimated the Model length l m 38.076
spatial variability of the marine soils in the Sonda de Campeche Water depth d m 5
Wave height Hs m 4
area of the Gulf of Mexico. The thickness of the soft cohesive Wave length L m 38.076
superficial soil layer and the soil water content of marine deposits Wave period T s 6
up to 100 m depth were investigated based on data from 89 Density of water ρw kg/m3 1000
boreholes. The correlation length for the thickness of the super- Density of seabed ρs kg/m3 2000
Modulus of volume Kw N/m2 2 109
ficial soft soil layer is estimated to be 22 m. The correlation length
Porosity n – 0.3
of the natural water content for the superficial soft soil layer is Poisson ratio ν – 0.333
0.8–2.0 m along the vertical direction. For deeper marine deposits,
the vertical correlation length of the natural water content is 21 m.
Table 1 summarizes the correlation lengths of geotechnical soil (Kleiber and Hien, 1992), each response quantity is represented
properties for marine soils reported in the literature. As shown in using a series of random Hermite polynomials. Due to its robust-
the table, the correlation length along the vertical direction varies ness and simplicity, Monte Carlo simulation (Papadrakakis and
from tens of centimeters to a few meters. There is no significant Papadopoulos, 1996; Fenton and Griffiths, 2008) is often used in
difference in spatial variability in the vertical direction between the literature as a reference method in order to check the accuracy
land and shallow marine sediments (Phoon and Kulhawy, 1999). of other approaches and is sometimes combined with other SFEM
However, for deeper marine soils, the vertical correlation length is methods. In the MCS, a deterministic problem is solved a number
usually larger than it is for land soils. The horizontal correlation of times and the response variability is calculated using simple
length of basic soil properties, for example water content and relationships of statistics. In this study, the deterministic model is
density, can be a few kilometers in the seafloor (Valdez-Llamas et a complex dynamic problem and hence the MCS method is
al., 2003; Goff et al., 2008; Verfaillie et al., 2009). adopted for simplicity and a feasible computation load.
Different algorithms are available for discretization of random
3.2. The stochastic finite element model field. The most popular methods include the spectral methods
based on Fourier transformation (Cooley and Tukey, 1965; Mejia
There are two main variants of Stochastic Finite Element and Rodriguez-Iturbe, 1974; Dietrich and Newsam, 1996), the
method (SFEM), i.e. the perturbation approach and the spectral moving average method (Matérn, 1986, Yaglom, 1987, Oliver, 1995;
approach (Stefanou, 2009). The perturbation approach (Ghanem Cressie and Pavlicová, 2002; Chiles and Delfiner, 1999), the local
and Spanos, 1991) is based on a Taylor series expansion of the average subdivision method (Fenton and Vanmarcke, 1990; Fenton
response vector. In the spectral stochastic finite element method and Griffths, 2008), the covariance matrix decomposition method
120 L.L. Zhang et al. / Ocean Engineering 111 (2016) 116–127
Table 3
List of stochastic model parameters.
Parameter Distribution Autocorrelation function Mean COV (%) Correlation length (m)
0 0.0
Present study
-0.1 Jeng (2013) T = 1.5 s
This study -0.2
d = 0.488 m
-0.2 L = 2.8 m
H = 0.056 m
-0.3 -0.4 h = 0.33 m
ν = 0.333
z/h
-0.4 n=1
|σz '|/p0 |p|/p0 -0.6 G = 5.0 10 Pa
z/h
Fig. 4. Vertical distributions of the maximum pore pressure |p|/p0 and effective
stresses |σz0 |/p0 versus z/h for a homogeneous porous seabed.
random fields can be obtained through a suitable transformation
of a normally distributed random field. For example, a log-
normally distributed random field can be obtained from:
X ¼ expðμlnX I þ σ lnX ZÞ ð13Þ
(Davis, 1987; Deutsch and Journel, 1997; Fenton and Griffths,
where μInX and σ ln X are the mean and standard variation of ln X,
2008; Jamshidi Chenari and Kamyab Farahbakhsh, 2015), the
respectively.
turning bands method (Matheron, 1973; Mantoglou and Wilson,
The advantage of the covariance matrix decomposition method
1982), the sequential simulation methods (Bellin and Rubin, 1996;
is that it is applicable for any type of covariance model and is not
Deutsch and Journel, 1997).
restricted to the geometric configuration of the simulated loca-
In this study, the covariance matrix decomposition method
tions. Compared with the turning bands method, the spectral fast
(Davis, 1987; Deutsch and Journel, 1997; Fenton and Griffths,
Fourier method, and the sequential Gaussian simulation method,
2008) is adopted to generate realizations of a random field. The
the covariance matrix decomposition method shows the least
basic idea of the covariance matrix decomposition method is to
assume that the parameters at different locations in the field are artificial bias (Harter, 1994). For random fields with many reali-
correlated random variables with a prescribed covariance matrix zations, this is indeed a very effective means of random field
and to produce a homogeneous random field which is the product generation, since the covariance matrix must be decomposed only
of the vector of independent standard normal random variables once for an entire simulation.
and the decomposed low triangular matrix of the covariance The stochastic finite element model for a porous seabed with
matrix. the spatially random soil property is established by integration of
The following is the brief description of the covariance matrix the finite element model and the random field simulation. First,
decomposition method. Assume that the prescribed covariance the coupled governing equations [Eqs. (1) and (5)] are incorpo-
matrix for these correlated random variables is C. If C is positive rated in a multiphysics (PDEsolver based on finite element
definite, then a correlated standard normal random field Z can be method, COMSOL (COMSOL, 2012). The Galerkin method is used to
produced using independent standard normal random variables discretize the PDEs. A nonlinear differential algebraic equation
according to the following equation: solver, IDA (Hindmarsh et al., 2005), is used as a time integrator for
the differential algebraic system, in which the backward differ-
Z ¼ LU ð11Þ entiation formulas (BDFs) are used to discretize the time
where L is a lower triangular matrix satisfying LL ¼ C; U is a vectorT derivative terms.
of n-independent standard normal random variables; Z is a vector The thickness of the seabed is 15 m and the length of the two-
of n-correlated standard normal random variables. The lower tri- dimensional finite element model is set to be equal to the wave-
angular matrix L is typically obtained using the Cholesky decom- length of 38.076 m. The soil within the seabed is assumed to be
position method. If a soil property X of the random field is nor- fine sand with the soil parameters presented in Table 2. The
mally distributed with a known mean and variance, then a reali- densities of the seabed and water, the porosity, and the Poisson's
zation for the soil property can be obtained using the realization of ratio are assumed to be constant. The shear modulus G, the per-
the correlated standard normal variate Z as follows: meability Kz, and the degree of saturation Sr are assumed to be
spatially random parameters, as listed in Table 3. The mean value
X ¼ μI þ σ Z ð12Þ
of G is 1 107 Pa. The mean of permeability in the z direction is
where I is the unit matrix; μ and σ are the mean value and assumed to be 5 10 4 m/s. The seabed is assumed to be
standard variation of the normally distributed X. Non-normal hydraulically isotropic. The seabed is unsaturated with a mean
L.L. Zhang et al. / Ocean Engineering 111 (2016) 116–127 121
-5 4 -5 4 -5 4 -5 4
z (m)
z (m)
z (m)
z (m)
-10 2 -10 2 -10 2 -10 2
0 0 0 0
-5 -5 -5 -5
z (m)
z (m)
z (m)
z (m)
-10 -10 -10 -10
δlnG = 2 δlnG = 2 δlnG = 2 δlnG = 2
deterministic case deterministic case deterministic case deterministic case
-15 -15 -15 -15
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
|p|/ p 0 |p|/ p 0 |p|/ p 0 |p|/ p 0
0 0 0 0
δlnG = 2 δlnG = 2 δlnG = 2
-5 deterministic case -5 deterministic case -5 -5 deterministic case
z (m)
z (m)
z (m)
z (m)
-10 -10 -10 -10
δlnG = 2
deterministic case
-15 -15 -15 -15
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
σz , / p 0 σz ,/p 0 σz ,/p 0 σz ,/p 0
Fig. 6. Effect of spatially random G (COVG ¼ 250%, δlnG ¼2 m) on seabed response: (a) realization of G ( 107 Pa), (b) |p|/p0 and (c) |σz0 |/p0. The dashed lines represent a
homogenous seabed with the mean values in Table 3 and are denoted as the deterministic case.
degree of saturation of 0.975. As the experimental data for per- of degree of saturation is also rarely reported in the literature.
meability and shear modulus are usually found to follow a log- However, the variation of natural water content is often reported
normal distribution (Phoon and Kulhawy, 1999; Baecher and (Phoon and Kulhawy, 1999), with the COV varying from 5% to 50%.
Christian, 2003; Carsel and Parrish, 1988), all three parameters are The reported vertical correlation length of natural water content
assumed to have a lognormal distribution in this study. The log- varies from 1 to 10 m in the field. For ground formed by natural
normal distribution is chosen because it is suitable for strictly non- decomposition, the horizontal correlation length is one order lar-
negative random variables, and enables the large variability of soil ger than the vertical one (Phoon and Kulhawy, 1999; Popescu et
properties to be analyzed. The autocorrelation function of the al., 1997). Based on the previously reported spatial variability of
spatially random soil parameters is assumed to be an exponential in situ soils, the COV values of shear modulus and saturated per-
function. The other empirical autocorrelation functions which are meability are assumed to be 250% and the COV of degree of
also commonly adopted in literature include Gaussian, Spherical saturation is assumed to be 50%. The correlation lengths of the
function. When field data of soil properties are available, different normalized parameters are 0.2, 2, and 10 m in order to compre-
empirical autocorrelation functions can be fitted and a most sui-
hensively investigate the possible scenarios of seabed response
table function can be chosen based on the fitness. In this study, as
due to the effect of spatial variability (Table 3). In the following
no field data available and the illustrative example is hypothetical,
sections, we denote the COVs of the three parameters as COVG,
the exponential autocorrelation function is assumed.
COVKz, and COVSr and the correlation lengths as δlnG, δln(Kz), and δln
Since the effect of waves on the seabed is important for soil
(Sr), respectively. Along the horizontal direction, the correlation
layers near the seabed surface and there is no significant differ-
lengths for all three parameters are assumed to be 100 m as the
ence in spatial variability between land and shallow marine soils
horizontal correlation length is at least one order of magnitude
(Table 1), the variability of ground soils on land in the literature
can be used as reference values. The coefficient of variation (COV) larger than the vertical one (Phoon and Kulhawy, 1999; Popescu et
of permeability for in situ soils ranges from 60% to more than 200% al., 1997).
(Carsel and Parrish, 1988; Duncan, 2000; Zhang et al., 2005; Bae- A finite element mesh of 2400 identical four-node rectangular
cher and Christian, 2003). The correlation length of log(perme- elements, each with four integration points and an area of
ability) in the vertical direction ranges from 0.2 to around 10 m 0.25 m 0.95 m, is employed in numerical modeling. A random
(Rehfeldt et al., 1992; Rackwitz et al., 2002). The COV of shear field of identical rectangular cells is generated and superimposed
modulus has rarely been reported in the literature. However, there on the finite element mesh. Each FE element is then assigned the
are quite a lot studies about the variation of soil stiffness or value of the random field cell using interpolation. In this study,
resistance measured by cone penetration tests, standard penetra- 500 random realizations are generated for each soil property. The
tion tests, pressuremeter, and so on (Phoon and Kulhawy, 1999; finite element model is then solved using each realization of the
Rackwitz et al., 2002). The reported COV values of soil resistance random field, and a population of the random response quantities
are usually from 20% to 100%. The vertical correlation length of soil is obtained. This population can then be used to obtain statistics of
resistance or stiffness varies from 0.1 to 10 m in the field. The COV the seabed responses.
122 L.L. Zhang et al. / Ocean Engineering 111 (2016) 116–127
0 4 0
-0.6 0.6 0.8 -0.-60.8
-0.
4
3 -0.4 0.4
-5 -5
0.4
z (m)
z (m)
2
-0.2 0 -0.2
0
-10 -10 0.2
1
-0.2 0.2
2
-0.
-15 0 -15
0 10 20 30 0 5 10 15 20 25 30 35
L (m) L (m)
0 4 0
-00..64
0.8 -0-0
.6.8
- .2 0.40.6 -0.4
-0 0.2 -0.2
3
-5 -5
0 0
z (m)
z (m)
-10 -10
1
-15 0 -15
0 10 20 30 0 5 10 15 20 25 30 35
L (m) L (m)
0 4 0
-0.6 0.60.8 -0-. 0.6
-0
4
.2
3 .4
-5 -5 -0
0.4
z (m)
z (m)
0.2
0.2
2
0
0
2
-0.
-10 -10
1
-15 0 -15
0 10 20 30 0 5 10 15 20 25 30 35
L (m)
L (m)
Fig. 8. Contours of p/p0 of random seabed (COVG ¼ 250%) (a) δlnG ¼ 0.2 m (b)
Fig. 7. Typical realizations of shear modulus G( 107 Pa) of random seabed
δlnG ¼2 m (c) δlnG ¼10 m.
(COVG ¼250%) (a) δlnG ¼ 0.2 m (b) δlnG ¼ 2 m (c) δlnG ¼ 10 m.
95% depth of the seabed than a homogenous seabed. For some extreme
cases (e.g., realization no. 109), high pore pressure close to 0.8p0
can be observed within 5 m depth of seabed.
-10
4.3. Effect of spatial variability of soil properties
Deterministic case
Five hundred random field realizations of stochastic soil para-
-15 meters are generated using the covariance matrix decomposition
0 0.2 0.4 0.6 0.8 1
method. Fig. 7 illustrates three typical realizations of the shear
|p|/p0 modulus G with a correlation length of ln(G), δlnG, varying between
0.2, 2, and 10 m. As shown on the graph, the heterogeneity of G
0 becomes more significant with smaller δlnG. Fig. 8 presents the
corresponding pore water pressure profiles for the three realiza-
5% tions of a spatially random seabed. When δlnG is only 0.2 m, the
soft layers and the stiff layers are intersliced and hence the wave
-5 mean cannot penetrate smoothly down to the seabed bottom. Therefore,
the high value zones of pore pressure are scattered. As δlnG
z (m)
95% increases, the random seabed is more homogenous and the pore
water pressure distribution is closer to the response of a uniform
-10 seabed.
Fig. 9 presents the mean and 90% uncertainty bounds of max-
Deterministic case imum |p|/p0 based on the 500 realizations with spatially varied
δlnG. Compared with the pore pressure response in a deterministic
-15 homogenous seabed, the mean |p|/p0 for a spatially random seabed
0 0.2 0.4 0.6 0.8 1 is higher. This increase in the mean |p|/p0 is more pronounced for
|p|/p0 small correlation lengths and decreases with increasing correla-
tion length. The dependence of the expected value on spatial
variability is described as "counter-intuitive" because the common
0 perception is that the performance will have a constant mean if
5% the soil property has a fixed mean value. However, this variation of
the expected performance with spatially varied soil properties
mean given a fixed mean has been observed in many other geotechnical
-5 problems. Fenton and Griffiths (1996) found that the mean flow
95% rate through an earth dam with spatially random permeability
z (m)
0 0.98 0 -0.8
0.6
--0.4.2 0.4 0.60.8 -0-.40.6
-0 -0.2
0.97 0.2
-5 -5
0.96 0 0
z /m
z /m
0.95
-10 -10
0.94
0 0.98 0
-0.4 0.6 0.8 -0 -0.6
-0.2 0.20.4 -0.2 .4
0.97 0 0
-5 0.2
-0.2
-5
-0
.2
0.96
z /m
z /m
0.95
-10 -10
0.94
0 0
5%
5% mean 95%
95%
-5 -5
z (m)
z /m
mean
-10 -10
Deterministic case
Deterministic case
-15 -15
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
|p|/p0 |p|/ p0
0 0
5%
5%
95% 95%
mean
-5
-5
z /m
z (m)
mean
-10
-10
Deterministic case
Deterministic case
-15
0 0.2 0.4 0.6 0.8 1
-15
|p|/p0 0 0.2 0.4 0.6 0.8 1
|p|/ p0
Fig. 11. Mean and confidence intervals of maximum |p|/p0 distribution
(COVSr ¼50%) (a) δln(Sr) ¼ 0.2 m (b) δln(Sr) ¼ 2 m.
Fig. 12. Mean and confidence intervals of maximum |p|/p0 distribution
(COVkz ¼250%) (a) δln(Kz) ¼ 0.2 m (b) δln(Kz) ¼2 m.
oscillating excess pore water pressure, as shown in Fig. 8(a). With
a smaller correlation length, it is more difficult for the wave- reduced. Therefore, the mean value of |p|/p0 decreases with the
induced excess pore pressure to propagate to a greater depth. As increase of correlation length.
the correlation length increases, this interference or obstruction is
L.L. Zhang et al. / Ocean Engineering 111 (2016) 116–127 125
As shown in Fig. 9, the uncertainty of pore water pressure first The mean of oscillating pore pressure in a heterogeneous sea-
increases and then decreases with the increase of correlation bed with spatially random shear modulus is greater than that in
length. It is generally believed that if the correlation length is less a homogenous seabed. The increase of mean pore pressure is
than the domain size or scale of interest, the variance of the output more pronounced for small correlation lengths and decreases
is reduced with the decrease of correlation length due to the space with increasing correlation length. The uncertainty of the
averaging effect (Fenton and Griffiths, 2008). However, this is not maximum pore pressure first increases and then decreases with
necessarily true for a flow-related problem. In order to obtain a the correlation length.
statistically meaningful realization of the hydraulic field, the The spatial variability of shear modulus and degree of saturation
domain size of flow should be at least 20–25 correlation lengths to affects the uncertainty of oscillating pore pressure of the whole
reach asymptotic behavior (Ababou et al., 1989). Elfeki et al. (2012) seabed while the random heterogeneity of permeability influ-
illustrated that for transport in aquifers with correlation lengths, ences the seabed response at shallow depths.
which are relatively large compared with the travel distance (i.e.,
the distance traveled is less than 20 correlation lengths), the
ergodic condition is not fulfilled and hence no trend is obtained in Acknowledgments
the results of spatial moments when the correlation length
increases. In this study, the propagation of wave-induced pore The work in this paper was substantially supported by the
pressure is confined within the 15 m thick seabed. Therefore, the National Basic Research Program of China (973 Program, Project
ergodic condition is not fulfilled for the cases with a correlation no. 2014CB049100), the National Natural Science Foundation of
length larger than 0.75 m. China (Project nos. 41172252, 41572243 and 51422905) and State
Assume that the scale of fluctuations for ln(Sr) along the ver- Key Laboratory of Ocean Engineering (GKZD010065). The authors
tical direction, δln(Sr), is 0.2 and 2 m. Fig. 10 illustrates the reali- are grateful for the supports from the Top Grade Young
zations of Sr of a spatially random porous seabed and the corre- Researchers (Ba Jian) Talent Program by the Organization
sponding oscillating pore pressures. Compared with Fig. 8, the Department of the Central Committee of the CPC.
oscillating pore pressure is less affected by the spatially random Sr.
Fig. 11 illustrates that the mean of maximum oscillating pore
pressure versus seabed depth is slightly greater than the deter- References
ministic value. The uncertainty bounds of pore pressure for the
case with a correlation length of 2 m are wider than those of δln Ababou, R., McLaughlin, D., Gelhar, L.W., Tompson, A.F.B., 1989. Numerical simu-
(Sr) ¼ 0.2 m. lation of three-dimensional saturated flow. Transp. Porous Media 4 (6),
Fig. 12 illustrates the mean of maximum oscillating pore pres- 549–565.
Ang, A.H.S., Tang, W.H., 2007. Probability Concepts in Engineering: Emphasis on
sure versus seabed depth for seabed with spatially varied per- Applications to Civil and Environmental Engineering, second ed. Wiley, New
meability. For fine sand seabed, the wave-induced pore pressure York.
attenuates rapidly within the top 3 m of the seabed thickness and Badaoui, M., Nour, A., Slimani, A., Berrah, M.K., 2007. Consolidation statistics
investigation via thin layer method analysis. Transp. Porous Media 67 (1),
finally decreases slightly with depth. The variance of the pore 69–91.
pressure is also concentrated in shallow depths. As shown in Baecher G.B. and Christian J.T., Reliability and Statistics in Geotechnical Engineer-
previous studies (Jeng and Seymour, 1997; Ulker et al., 2009; Jeng, ing, 2003, John Wiley and Sons, London and New York, pp 605.
Bari, M.W., Shahin, M.A., 2015. Three-dimensional finite element analysis of spa-
2013), the pore pressure is generally greater when the coefficient tially variable PVD improved ground. Georisk: Assess. Manag. Risk Eng. Syst.
of permeability is larger because the wave pressure can be easier Geohazards 9 (1), 37–48.
to transmit to deeper seabed. However, in a random hetero- Bellin, A., Rubin, Y., 1996. HYDRO GEN: a spatially distributed random field gen-
erator for correlated properties. Stoch. Hydrol. Hydraul. 10 (4), 253–278.
geneous seabed, the zones with different permeability are ran- Biot, M.A., 1941. General theory of three-dimensional consolidation. J. Appl. Phys.
domly intersliced and hence the pore pressure near the surface of 12 (2), 155–164.
the seabed cannot be smoothly transmitted to deeper depth of a Carsel, R.F., Parrish, R.S., 1988. Developing joint probability distributions of soil
water retention characteristics. Water Resour. Res. 24 (5), 755–769.
seabed due to the blockage of a possible low permeability zone. Cheon, J.Y., 2011. Analysis of spatial variability in geotechnical data for offshore
Therefore, the uncertainty of pore pressure is mainly concentrated foundations (Ph.D. dissertation). University of Texas at Austin, Austin, Texas.
in near the surface of a seabed as shown in Fig. 12. Chiles, J.P., Delfiner, P., 1999. Geostatistics: Modeling Spatial Uncertainty. Wiley,
New York.
Comparing Figs. 9, 11, and 12, the effect of spatial variability of Cho, S.E., 2007. Effects of spatial variability of soil properties on slope stability. Eng.
shear modulus is most significant. The spatial variability of shear Geol. 92 (3–4), 97–109.
modulus and degree of saturation affects the uncertainty of oscil- Cooley, J.W., Tukey, J.W., 1965. An algorithm for the machine calculation of complex
Fourier series. Math. Comput. 19 (90), 297–301.
lating pore pressure of the whole seabed. The effect of variable per- COMSOL, AB, 2012. COMSOL Multiphysics User Guide (Version 4.3 a). COMSOL, AB
meability is least significant and mostly in shallow seabed depths. Stockholm, Sweden.
Cressie, N., Pavlicová, M., 2002. Calibrated spatial moving average simulations. Stat.
Model. 2, 267–279.
Dasaka, S.M., Zhang, L.M., 2012. Spatial variability of in-situ weathered soils. Geo-
5. Conclusions technique 62 (5), 375–384.
Davis, M.W., 1987. Production of conditional simulations via the LU triangular
decomposition of the covariance matrix. Math. Geol. 19 (2), 91–98.
In this study, a stochastic finite element model is developed to Deutsch, C.V., Journel, A.G., 1997. GSLIB: Geostatistical Software Library and User’s
investigate the wave-induced seabed response in a spatially ran- Guide (Applied Geostatistics. Oxford University Press, Oxford, New York, p. 384.
Dietrich, C.R., Newsam, G.N., 1996. A fast and exact method for multidimensional
dom porous seabed and the effects of spatial variability of soil
Gaussian stochastic simulations: extension to realizations conditioned on direct
properties on seabed responses are investigated by Monte Carlo and indirect measurements. Water Resour. Res. 32 (6), 1643–1652.
simulation. The major conclusions are as follows: Duncan, J.M., 2000. Factors of safety and reliability in geotechnical engineering. J.
Geotech. Geoenviron. Eng. 126 (4), 307–316.
Elfeki, A.M., Uffink, G., Lebreton, S., 2012. Influence of temporal fluctuations and
The pore water pressure and stress distribution in a spatially spatial heterogeneity on pollution transport in porous media. Hydrogeol. J. 20
random heterogeneous seabed are significantly affected by (2), 283–297.
spatial variability of the shear modulus. The effects of spatial Fan, H., Huang, Q., Liang, R., 2014. Reliability analysis of piles in spatially varying
soils considering multiple failure modes. Comput. Geotech. 57, 97–104.
variability of permeability and degree of saturation are less Fan, H., Liang, R., 2015. Importance sampling based algorithm for efficient reliability
significant than the effect of spatial variability of shear modulus. analysis of axially loaded piles. Comput. Geotech. 65, 278–284.
126 L.L. Zhang et al. / Ocean Engineering 111 (2016) 116–127
Fenton, G.A., Griffiths, D.V., 2008. Risk Assessment in Geotechnical Engineering. Li, W., Lu, Z., Zhang, D., 2009. Stochastic analysis of unsaturated flow with prob-
Wiley, New York. abilistic collocation method. Water Resour. Res. 45, 8.
Fenton, G.A., Griffiths, D.V., 2002. Probabilistic foundation settlement on a spatially Liu, H.J., Jeng, D.S., 2007. A semi-analytical solution for random wave-induced soil
random soil. J. Geotech. Geoenviron. Eng. 128 (5), 381–390. response and seabed liquefaction in marine sediments. Ocean Eng. 34 (8–9),
Fenton, G.A., Griffiths, D.V., 1996. Statistics of free surface flow through stochastic 1211–1224.
earth dam. J. Geotech. Eng 122 (6), 427–436. Lloret-Cabot, M., Fenton, G.A., Hicks, M.A., 2014. On the estimation of scale of
Fenton, G.A., Vanmarcke, E.H., 1990. Simulation of random fields via local average fluctuation in geostatistics. Georisk: Assess. Manag. Risk Eng. Syst. Geohazards
subdivision. J. Eng. Mech 116 (8), 1733–1949. 8 (2), 129–140.
Gao, F.P., Jeng, D.S., Sekiguchi, H., 2003. Numerical study on the interaction between Madsen, O.S., 1978. Wave-induced pore pressure and effective stresses in a porous
non-linear wave, buried pipeline and non–homogeneous porous seabed. bed. Geotechnique 28 (4), 377–393.
Comput. Geotech. 30 (6), 535–547. Matérn, B., 1986. Spatial Variation, second ed. Springer-Verlag, Berlin.
Gatmiri, B., 1990. A simplified finite element analysis of wave-induced effective Matheron, G., 1973. The intrinsic random functions and their applications. Adv.
stresses and pore pressures in permeable seabed. Geotechnique 40 (1), 15–30. Appl. Probab. 5 (3), 439–468.
Ghanem, R.G., Spanos, P.D., 1991. Stochastic Finite Element-A Spectral Approach. Manjari, K.G., Rao, K.B., Babu, G.S., 2014. Stochastic model for settlement: footings
Springer, New York. on cohesionless soil. Georisk: Assess. Manag. Risk Eng. Syst. Geohazards 8 (4),
Gilbert, R.B., Lacasse, S., Nadim, F., 2014. Advances in geotechnical risk and relia- 269–283.
bility for offshore applications. In: Proceedings of the Fourth International Mantoglou, A., Wilson, J.L., 1982. The turning bands method for simulation of
Symposium Geotechnical Safety and Risk. Geotechnical Safety and Risk IV. random fields using line generation by a spectral method. Water Resour. Res. 18
Balkema, The Netherlands. (5), 1379–1394.
Goff, J.A., Jenkins, C.J., Williams, S.J., 2008. Seabed mapping and characterization of Mejia, J., Rodriguez-Iturbe, I., 1974. On the synthesis of random field sampling from
sediment variability using the usSEABED data base. Cont. Shelf. Res. 28 (4), the spectrum: an application to the generation of hydrologic spatial processes.
614–633. Water Resour. Res. 10 (4), 705–711.
Harter, T., 1994. Unconditional and Conditional Simulation of Flow and Transport in Mei, C.C., Foda, M.A., 1981. Wave-induced response in a fluid-filled poro-elastic
Heterogeneous, Variably Saturated Porous Media (PhD dissertation). University solid with a free surface - a boundary layer theory. Geophys. J. Int. 66 (3),
of Arizona, Tucson, Arizona. 597–631.
Hindmarsh, A.C., Brown, P.N., Grant, K.E., Lee, S.L., Serban, R., Shumaker, D.E., Moshagen, H., Torum, A., 1975. Wave-induced pressure in permeable seabeds. J.
Woodward, C.S., 2005. SUNDIALS: suite of nonlinear and differential/algebraic Waterw. Harb. Coast. Eng. 101 (1), 49–57.
equation solvers. ACM Trans. Math. Softw. 31 (3), 363–396. Mousavi Nezhad, M., Javadi, A.A., Al-Tabbaa, A., Abbasi, F., 2013. Numerical study of
Hoeg, K, Tang, W.H., 1977. Probabilistic considerations in the foundation engi- soil heterogeneity effects on contaminant transport in unsaturated soil (model
neering for offshore structures. In: Proceedings of the 2nd International Con- development and validation). Int. J. Numer. Anal. Methods Geomech. 37 (3),
ference on Structural Safety and Reliability. Munich, Germany, pp. 267–296. 278–298.
Hsu, J.R.C., Jeng, D.S., 1994. Wave-induced soil response in an unsaturated aniso- Myrhaug, D., Holmedal, L.E., Ong, M.C., 2014. Seepage effects on bedload sediment
tropic seabed of finite thickness. Int. J. Numer. Anal. Methods Geomech. 18 (11), transport rate by random waves. Ocean Eng. 82, 123–127.
785–807. Myrhaug, D., Ong, M.C., 2010. Random wave-induced onshore scour characteristics
Hsu, J.R.C., Jeng, D.S., Lee, C.P., 1995. Oscillatory soil response and liquefaction in an around submerged breakwaters using a stochastic method. Ocean Eng. 37 (13),
unsaturated layered seabed. Int. J. Numer. Anal. Methods Geomech. 19 (12), 1233–1238.
825–849. Myrhaug, D., Slaattelid, O.H., Lambrakos, K.F., 1998. Seabed shear stresses under
Hsu, J.R.C., Jeng, D.S., Tsai, C.P., 1993. Short crested wave induced soil response in a random waves: predictions vs estimates from field measurements. Ocean Eng.
porous seabed of infinite thickness. Int. J. Numer. Anal. Methods Geomech. 17 25 (10), 907–916.
(4), 553–576. Okusa, S., 1985. Wave-induced stress in unsaturated submarine sediments. Geo-
Huang, J.S., Griffiths, D.V., Fenton, G.A., 2010. Probabilistic analysis of coupled soil technique 35 (4), 517–532.
consolidation. J. Geotech. Geoenviron. Eng. 136 (3), 417–430. Oliver, D.S., 1995. Moving averages for Gaussian simulation in two and three
Jamshidi Chenari, R., Alaie, R., 2015. Effects of anisotropy in correlation structure on dimensions. Math. Geol. 27, 939–960.
the stability of an undrained clay slope. Georisk: Assess. Manag. Risk Eng. Syst. Papadrakakis, M., Papadopoulos, V., 1996. Robust and efficient methods for sto-
Geohazards 9 (2), 109–123. chastic finite element analysis using Monte Carlo simulation. Comput. Method
Jiang, S.H., Li, D.Q., Cao, Z.J., Zhou, C.B., Phoon, K.K., 2014a. Efficient system relia- Appl. Mech. Eng. 134 (3), 325–340.
bility analysis of slope stability in spatially variable soils using Monte Carlo Phillips, O.M., 1977. The Dynamics of the Upper Ocean. Cambridge University Press,
simulation. J. Geotech. Geoenviron. Eng. 141 (2), 04014096. Cambridge.
Jiang, S.H., Li, D.Q., Zhang, L.M., Zhou, C.B., 2014b. Slope reliability analysis con- Phoon, K.K., Kulhawy, F.H., 1999. Characterization of geotechnical variability. Can.
sidering spatially variable shear strength parameters using a non-intrusive Geotech. J. 36 (4), 612–624.
stochastic finite element method. Eng. Geol. 168, 120–128. Popescu, R., Prévost, J.H., Deodatis, G., 1997. Effects of spatial variability on soil
Jeng, D.S., 1997. Soil response in cross-anisotropic seabed due to standing waves. J. liquefaction: some design recommendations. Geotechnique 47 (5), 1019–1036.
Geotech. Geoenviron. Eng. 123 (1), 9–19. Rackwitz, R., Denver, H., Calle, E., 2002. JCSS Probabilistic Model Code. Joint Com-
Jeng, D.S., 2013. Porous Models for Wave-seabed Interactions. Shanghai Jiao Tong mittee on Structural Safety, Zurich, Switzerland 5th version.
University Press and Springer-Verlag, Berlin, Heidelberg. Rehfeldt, K.R., Boggs, J.M., Gelhar, L.W., 1992. Field study of dispersion in a het-
Jeng, D.S., Hsu, J.R.C., 1996. Wave-induced soil response in a nearly saturated seabed erogeneous aquifer: 3. Geostatistical analysis of hydraulic conductivity. Water
of finite thickness. Geotechnique 46 (3), 427–440. Resour. Res. 28 (12), 3309–3324.
Jeng, D.S., Lin, Y.S., 1996. Finite element modelling for water waves–seabed inter- Santoso, A.M., Phoon, K.K., Quek, S.T., 2011. Effects of soil spatial variability on
action. Soil Dyn. Earthq. Eng 15 (5), 283–300. rainfall-induced landslides. Comput. Struct. 89 (11–12), 893–900.
Jeng, D.S., Seymour, B.R., 1997. Response in seabed of finite depth with variable Sleath, J.F.A., 1970. Wave-induced pressures in beds of sand. J. Hydraul. Div. 96,
permeability. J. Geotech. Geoenviron. Eng. 123 (10), 902–911. 367–379.
Kanning, W., EOF, Calle, 2013. Derivation of a representative piping resistance Srivastava, A., Sivakumar, B.G.L., Haldar, S., 2009. Influence of spatial variability of
parameter based on random field modelling of erosion paths. Georisk: Assess. permeability property on steady state seepage flow and slope stability analysis.
Manag. Risk Eng. Syst. Geohazards 7 (2), 99–109. Eng. Geol. 110 (3–4), 93–101.
Keaveny, J.M., Nadim, F., Lacasse, S., 1990. Autocorrelation function data for offshore Stefanou, G., 2009. The stochastic finite element method: past, present and future.
geotechnical data. In: Proceedings of the International Conference on Structural Comput. Method Appl. Mech. Eng. 198 (9), 1031–1051.
Safety and Reliability. ICOSSAR 89’, San Francisco, California, pp. 263–270. Suh, K.D., Lee, C., Park, W.S., 1997. Time-dependent equations for wave propagation
Kitano, T., Mase, H., 1999. Boundary layer theory for anisotropic seabed response to on rapidly varying topography. Coast. Eng. 32 (2–3), 91–117.
sea waves. J. Waterw. Port Coast. Ocean Eng. 125 (4), 187–194. Tang, W.H., 1979. Probabilistic evaluation of penetration resistances. J. Geotech. Div
Kleiber, M., Hien, T.D., 1992. The stochastic finite element method: basic pertur- 105, 1173–1191, GT10.
bation technique and computer implementation. Wiley, Chichester, U.K. Tartakovsky, D.M., Lu, Z., Guadagnini, A., Tartakovsky, A.M., 2003. Unsaturated flow
Le, T.M.H., Gallipoli, D., Sanchez, M., Wheeler, S.J., 2012. Stochastic analysis of heterogeneous soils with spatially distributed uncertain hydraulic parameters.
unsaturated seepage through randomly heterogeneous earth embankments. J. Hydrol. 275 (3–4), 182–193.
Int. J. Numer. Anal. Methods Geomech 36 (8), 1056–1076. Thomas, S.D., 1989. A finite element model for the analysis of wave induced
Le, T.M.H., Gallipoli, D., Sanchez, M., Wheeler, S., 2013. Rainfall-induced differential stresses, displacements and pore pressures in anunsaturated seabed. I: theory.
settlements of foundations on heterogeneous unsaturated soils. Geotechnique Comput. Geotech. 8 (1), 1–38.
63 (15), 1346–1355. Thomas, S.D., 1995. A finite element model for the analysis of wave induced
Li, D.Q., Jiang, S.H., Cao, Z.J., Zhou, W., Zhou, C.B., Zhang, L.M., 2015a. A multiple stresses, displacements and pore pressures in an unsaturated seabed. II: model
response-surface method for slope reliability analysis considering spatial verification. Comput. Geotech. 17, 107–132.
variability of soil properties. Eng. Geol. 187, 60–72. Tsui, Y., Helfrich, S.C., 1983. Wave-induced pore pressures in submerged sand layer.
Li, D.Q., Qi, X.H., Phoon, K.K., Zhang, L.M., Zhou, C.B., 2014. Effect of spatially variable J. Geotech. Eng. 109 (4), 603–618.
shear strength parameters with linearly increasing mean trend on reliability of Ulker, M.B.C., Rahman, M.S., Jeng, D.S., 2009. Wave-induced response of seabed:
infinite slopes. Struct. Saf. 49, 45–55. various formulations and their applicability. Appl. Ocean Res. 31 (1), 12–24.
Li, J.H., Tian, Y., Cassidy, M., 2015b. Failure mechanism and bearing capacity of Valdez-Llamas, Y.P., Auvinet, G., Nunez, J., 2003. Spatial variability of the marine soil
footings buried at various depths in spatially random soil. J. Geotech. Geoen- in the Gulf of Mexico. In: Proceedings of the Offshore Technology Conference,
viron. Eng. 141 (2), 401–409. Houston, Texas, OTC 15266.
L.L. Zhang et al. / Ocean Engineering 111 (2016) 116–127 127
Vanmarcke, E.H., 1977. Probabilistic modeling of soil profiles. J. Geotech. Eng. Div, Zhang, L., Chen, J.J., 2012. Effect of spatial correlation of standard penetration test
103; , pp. 1227–1246. (SPT) data on bearing capacity of driven piles in sand. Can. Geotech. J. 49 (4),
Verfaillie, E., Du, Four, I., Van, Meirvenne, M., Van, Lancker, V., 2009. Geostatistical 394–402.
modeling of sedimentological parameters using multi-scale terrain variables: Zhang, J., Li, J.P., Zhang, L.M., Huang, H.W., 2014. Calibrating cross-site variability for
application along the Belgian Part of the North Sea. Int. J. Geogr. Inf. Sci. 23 (2), reliability-based design of pile foundations. Comput. Geotech. 62, 154–163.
135–150. Zhang, L.L., Zhang, L.M., Tang, W.H., 2005. Rainfall-induced slope failure considering
Verruijt, A., 1969. Elastic Storage of Aquifers, Flow through Porous Media. Academic variability of soil properties. Geotechnique 55 (2), 183–188.
Press, New York, pp. 331–376. Zhang, L.L., Zhang, L.M., Tang, W.H., 2008. Similarity of soil variability in centrifuge
Watabe, Y., Tsuchida, T., Adachi, K., 2002. Undrained shear strength of Pleistocene models. Can. Geotech. J. 45 (8), 1118–1129.
clay in Osaka Bay. J. Geotech. Geoenviron. Eng. 128 (3), 216–226. Zhou, X.L., Xu, B., Wang, J.H., 2011. An analytical solution for wave-induced seabed
Xu, H., Dong, P., 2011. A probabilistic analysis of random wave-induced liquefaction. response in a multi-layered poro-elastic seabed. Ocean Eng. 38 (1), 119–129.
Zhu, H., Griffiths, D.V., Fenton, G.A., Zhang, L.M., 2015. Undrained failure mechan-
Ocean Eng. 38 (7), 860–867.
isms of slopes in random soil. Eng. Geol. 191, 31–35.
Yaglom, A.M., 1987. Correlation Theory of Stationary and Related Random Func-
Zhu, H., Zhang, L.M., Zhang, L.L., Zhou, C.B., 2013. Two-dimensional probabilistic
tions. Volume I: Basic Results. Springer, Berlin.
infiltration analysis with a spatially varying permeability function. Comput.
Yamamoto, T., Koning, H.L., Sellmeijer, H., 1978. On the response of a poro-elastic
Geotech. 48, 249–259.
bed to water waves. J. Fluid. Mech. 87 (01), 193–206.