Environmental Engineering Reference
In-Depth Information
of the random field. For a two-dimensional (2D) Gaussian random field, this function is
given as follows:
xx
l
yy
l
1
2
1
2
2
Cx yxy
[(
,
), (
,
)]
=
σ
exp
(15.1)
11
22
x
y
in which σ is the standard deviation of the random field, and l x and l y are, respectively, the
horizontal and vertical autocorrelation lengths. An autocorrelation length is the distance
within which the values of the uncertain property are significantly correlated. It is the dis-
tance required for the autocorrelation function to decay from 1 to e −1 (i.e., 0.3679). Finally,
( x 1 , y 1 ) and ( x 2 , y 2 ) are the coordinates of two arbitrary points in the space.
Let R ( x , y , θ) be a Gaussian random field where x and y denote, respectively, the horizontal
and vertical coordinates and θ indicates the stochastic character of the random field. The
random field R can be approximated by the K-L expansion as follows (Spanos and Ghanem
1989; Ghanem and Spanos 1991):
M
1
Rx y
(, ,
θµ λφ
)
≈+
=
(
xy
,
)
ξθ
()
(15.2)
i
i
i
i
where μ is the mean value of the random field, M is the size of the series expansion, λ i and
ϕ i are the eigenvalues and eigenfunctions of the covariance function, respectively, and ξ i (θ)
(for i = 1, …, M ) is a vector of standard uncorrelated random variables. For the exponential
covariance function used in this chapter, Ghanem and Spanos (1991) provided an analytical
solution for the eigenvalues and eigenfunctions as follows:
a. For a one-dimensional (1D) horizontal random field generated in the interval [− a x , a x ],
the eigenvalues can be calculated as follows:
2
c
λ
=
(15.3)
i
ω
2
+
c
2
i
where
1
c
=
(15.4)
l x
and
c
ωω
tan(
a x
)0
=
for oddi
i
i
i
and
(15.5)
ω
+
c
tan(
ω
a
)0
=
for eveni
i
i
i
x
The eigenfunctions are calculated as follows:
cos(
ω
ωω
x
)
i
φ
=
for oddi
i
(15.6)
i
a
+
(sin(
2
a
)
/
2
)
x
i
x
i
sin(
ω
ωω
x
)
i
φ
=
for eveni
i
(15.7)
i
a
(sin(
2
a
)
/
2
)
x
i
x
i
 
 
Search WWH ::




Custom Search