Environmental Engineering Reference
In-Depth Information
∑
all θ
PDecision ConsequenceInformation
(
)
=
PConsequen
(
ce
θ
,
…
,
θ ΘΘ
)( ,
P
…
,
ε
)
1
n
1
n
,,
…
θ
1
n
(13.11)
When working with model parameters, it can be insightful to express the likelihood func-
tion in terms of the decision consequences (e.g.,
Figure 13.11
)
since it is the relationship
between decision consequences and information that governs the value of information:
PInformation Decision Consequence
(
)
∑
all θ
∑
=
PConsequen
(
ce
θ
,
…
,
θ ΘΘ Εθ θΘ Θ
)( ,
P
…
,
ε
)
×
P
(
=
ε
,
…
,
)( ,
P
…
,
)
1
n
1
n
1
n
1
n
all
,,
…
θ
1
n
θ
,,
…
θ
1
n
∑
PConsequence
(
θ
,
…
,
θ ΘΘ
)(,,
P
…
)
(13.12)
1
n
1
n
all θ
,,
…
θ
1
n
13.3.3 Illustrative example: Design of pile foundation
An example illustrating insights from Bayes' Theorem is for the design of a pile foundation
for axial loading (
Figure 13.12
).
The decision is between two different pile lengths. Spatial
variability in axial pile capacities across the site is modeled with a normal distribution,
Construction
cost
(per pile)
Failure
cost
(per pile)
R
= Capacity
S
= Load
R > S
Θ = Gradient of average unit
side shear versus depth
$4500
$0
A: 30 m pile
P
(Θ)
Do not test any
pile
E
(
C
) = $6870
R < S
$4500
$45,000
R > S
$6125
$0
B: 35 m pile
P
(Θ)
E
(
C
) = $6550
R < S
$6125
$61,250
R > S
$4500
$0
30 m pile
P
(Θ|
)
R < S
$4500
$45,000
R > S
$6125
$0
Te st a number
of piles
35 m pile
P
(Θ| )
R < S
$6125
$61,250
.
Uncertain test
pile capacities
p
E
( )
R > S
$4500
$0
30 m pile
P
(Θ| )
R < S
$4500
$45,000
R > S
$6125
$0
35 m pile
P
(Θ| )
R < S
$6125
$61,250
Figure 13.12
Decision tree to assess value of information for load tests in pile foundation design.
Search WWH ::
Custom Search