Geology Reference
In-Depth Information
Figure 4. Neural networks agreement for the response DIST
RLIM RLIM
=
( .
1
+
COVRL X
N
)
(45)
obtained from the corresponding neural networks.
The Beta distribution for each of the damage
indices requires the introduction of an additional
random variable,
U
1
, uniformly distributed be-
tween 0 and 1.
The uncertainty in each of the capacities
RLIM
is modeled with a normal distribution, introduc-
ing an additional Standard Normal variable
X
N
3
:
3
A summary of the random variables, with
their distribution types and parameters, is shown
in Table 2. The statistics for the peak ground ac-
celerations
a
G
correspond to the seismicity for
the city of Mendoza, Argentina, as provided by
INPRES (1995).
Table 2. Random variables
σ
X
σ
X
X
X
Variable
Type
Variable
Type
X
(10) =
a
G
X
(1) =
m
5.1x10
-4
5.1x10
-5
Normal
94
cm
/
s
2
130
cm
/
s
2
Lognormal
X
(2) =
b
b
20
cm
1
cm
Normal
X
(11) =
f
g
2.50
Hz
0.375
Hz
Normal
X
(12) =
σ
a
G
0.05
X
Normal
X
(3) =
h
b
?
cm
0
0.25
Normal
X
(4) =
b
c
30
cm
1.5
cm
Normal
X
(13) =
a
G
X
(13) =
X
(10) [1.0+
X
(12)]
0.05
X
X
(5) =
h
c
?
cm
Normal
X
(14) =
R
N
1
0
1
Normal
X
(6) =
ρ
span
0.10
X
Lognormal
?
X
(15) =
U
1
0
1
Uniform
X
(7) =
ρ
end
0.10
X
?
Lognormal
X
(16) =
X
N
1
0
1
Normal
X
(8) =
ρ
col
0.10
X
?
Lognormal
X
(17) =
X
N
2
0
1
Normal
/
′
0
X
(9) =
f
f
0.10
0.01
Normal
X
(18) =
X
N
3
0
1
Normal
r
Search WWH ::
Custom Search