Documente Academic
Documente Profesional
Documente Cultură
DEPARTMENT OF CIVIL
Wave Loading
ENGINEERING
on Offshore
A Probabilistic
Structures:
Approach
Thesis submitted
in accordance. with
the requirements
of the, University
of Liverpool
for the degree of Doctor in Philosophy
by
R. Burrows,
B. Eng.
May, 1982
ST
COPY
AVAILA
i
SUMMARY
The objective
the structural
type,
response
to the loading
lifetime
here is
a description
of the steel
lattice
during
structures,
of offshore
induced
to develop
of
in the ocean.
of exposure
is restricted
to structures
which do not respond dynamically
loading.
The main emphasis of the approach is the
to the induced fluid
Application
necessity
for
retention
of the non-linearity
analysis
for
these
drag are
where the effects
of fluid
theory is adopted as the most appropriate
and the description
of loading is developed
to the problem
approach
of loading
is
Morison's
using
climate,
representing
of interest.
the complete
of the probabilistic
The long-term
Equation.
by convolution
then obtained
(representing
wave field
of the short-term
stationary
The behaviour
in the
structures
Probability
significant.
of the loading
of this
population
a
description
the wave
model with
of sea states
at the site
model,
from
for
the
intermittency
in
loading
the
currents
or
of
of
uni-directional
presence
in terms of the probability
distributions
the splash zone, is investigated
of load,
extreme
which
in form,
is non-Gaussian
Various
peak variates.
produced
by spectral
estimate
extreme
chosen for
description
loading
description
conditions.
term properties.
It
is
a model employing
to those
equivalent
shown to considerably
elevation
are also
with
of the spectral
undermodel
demonstrated
of
of structural
response, based on linear
is developed from that of individual
member
of the'statistics
Convolution
peak and
is discussed.
of the structure,
in terms
results
The significance
description
The probabilistic
yielding
of the surface
requirements
computational
by comparison
which
wave loads.
behaviour
mechanism,
analysis,
narrow-band
are
member and spa state conditions
drag
of the non-linear
of full
retention
considered
a linearised
and its
with
is the probabilistic
for
short-term
again, produces
properties
the long-
of response,
either
ii
Or internal
displacements
stresses,
than individual
rather
in the design
value
member loading
and first
fatigue
against
failure.
excursion
I
In the model it
is necessary
to idealise
the distributed
the structure
loading
on
number of locations.
are distributed
These elemental
random loads
each other
varying degrees
with
fluid
The computational
of this
of correlation.
requirements
to investigate
simple
and it has proved practicable
model are considerable
systems of only two load components due to runtime limitationsv
However,
from this
response
is
statistical
response
variate
is
distribution
the probability
of
form as that
defined
of the underlying
by its second and fourth
a computational
still
on system complexity
ceiling
shown that
for
the
of the
wave field.
short-term
There is
it
loading.,
order
study
imposed.
Using 30-minute
problem
dependent
run times
with
this
approach
imposing
limitations
at
However,
systems of up to 35 load components can be analysed.
to the model has yet to be realised
and as a
refinement
an important
for
the
the
systems exceeding 12
of
present
model
accuracy
consequence
load components is uncertain.
Liverpool,
Finally,
test
application
structures
the results
under
of short-term
those
obtained
for
a number of simple
sea state
using
conditions
time-series
and
simulation.
iii
ACKNOWLEDGMENTS
The author
would like
Mr. R. G. Tickell,
and co-investigators
guidance'throughout
The contribution
two years
Finally,
their
to thank his
Professor
P. Holmes and
supervisor
both colleagues
in the Department of Civil
Engineering
in the OSFLAG 5 Research Project,
for their help and
the course
of study
of the Department
leading
of Industry
thanks
diligence
are also
due to Christine
in producing
the typescript
is
Thesis.
to this
in funding
greatly
the author
for
appreciated.
Cotgreave
for
and drawings.
tI
-1
iv
CONTENTS
Page
SUMMARY
ACKNOWLEDGEMENTS
vi
NOTATION
CHAPTERONE
INTRODUCTION
CHAPTERTWO
CHAPTERTHREE
CHAPTERFOUR
2.1
on Submerged Rigid
10
2.2
Theories
Waves
14
2.3
Prediction
Waves
of Periodic
of Loading
Induced
by Random
18
3.1
The Effect
of Non-linearity
Loading on its Probabilistic
3.2
Effect
of the Choice of Wave Spectra
Load Prediction
3.3
66
3.4
Prediction
73
3.5
Computational
Considerations;
Use of
Quadrature Methods for Run-time
Minimisation
and a Technique of Hand
Calculation
on
48
55
80
Suppression
the Data
4.2
4.3
A Method
4.4
Estimation
Distributions
of Long-term
Covering
Of 111/3 from Measurements
a
One-year
Period
at the Famita Location
of
of Seasonal
Wave Climate
Fluctuations
in
118
Wave Climate
120
Extrapolation
123
124
Page
CHAPTERFOUR
CHAPTER FIVE
4. S
The Effect
of Wave Climate
on Long-term Distributions
Height and Wave Loading
4.6
Limitations
of the Wave Climate
Techniques
polation
Extrapolation
of Wave
Extra-
133
138
PROBABILISTIC
158
5.1
Introduction
5.2
Multi-variate
5.3
Response of
Components
5.4
Second
5.5
Long-term
Probability
Distributions
Peak and Extreme Response
5.6
Application
Structures
pdf.
of
Systems
and Fourth
of
the
159
Response
of
162
Two Load
Moments
Methods
of
Response
to
of
Typical
173
193
194
CHAPTERSIX
CONCLUSIONS
236
CHAPTERSEVEN
RECOMMENDATIONS
240
REFERENCES
243
APPENDIX ONE
ANALYSIS OF RANDOMVARIABLES
250
APPENDIX TWO
APPENDIX THREE
PAPERS
269
278
WAVE
CLIMATE
APPENDIX FOUR
287
APPENDIX FIVE
IX.
IX
IX.
EXPANSION OF E{X.
jlXk3
APPENDIX SIX
LIST
COMPUTERPROGRAMS
XL31
291
295
"I
vi
NOTATION
The following
differs
notation
parameters
comprehensive for
in the Appendices (notably
notation
are defined
in isolation
is
where they
first
I and II)
may be read
appear
Where
only.
Eq. (5.4.23)
of conditional
expectations,
in Gaussian Quadrature, Eq. (5.6.1)
abscissa coefficients
in Pierson-Moskowitz spectrum,
constant coefficient
Eq. (2.3.23)
aij, a2
coefficients
a
A
or constant in Weibull
Appendix Two only
A'
standardised
[A]
transformation
Ak
coefficients
A
r
distribution,
Eq. (2.3.13)
wave amplitude,
matrix,
Eq. (5.2.7)
of Chebyshev-Hermite
quadrature,
to flow,
perpendicular
of conditional
coefficients
constant coefficient
Eq. (2.3.23)
B
Je
C.
ii
{c
coefficients
expectations,
in Pierson-Moskowitz
of Chebyshev-Laguerre
of [M], Eq. (5.3.2)
co-factors
Eq. (2-1-1)
Eq. (5.4.37)
spectrum,
quadrature
of conditional
vector of coefficients
Eq. (5.4.42)
Eq. (3-5-1)
expectations,
function
cdf
cumulative
distribution
cD
'Morisons'
drag coefficient
cm
'Morisons'
inertia
water
diameter
structural
Det
determinant
exp
exponential
Ef I
E{N+ (0)
E{n+(O)
coefficient
depth
of cylindrical
member
type 1
IH1/3il
abbreviations
for
expectations,
5.4
vii
f
linearised
Y
f(
'Morison'
wave force,
function
wave force
FD
F
DN
F
Dx
{F
e
F
i
F
F
F
Dz
FD
y
normal
component of force
inertia
component of force
Eq. (2.1.7)
Fx, Fyj
Fz
normal
components of F in co-ordinate
g(
gravitational
function
functions
wave height
Hi/3
height
Hs ; 'significant'
wave
-;
(1/3) of waves
largest
H
rms
H
j
Hn
'root
weighting
variable
annual
to member axis,
directions
components of FDN in co-ordinate
Eq. (5.2.7)
vector of nodal forces,
force components
inertia
IN
to member axis,
directions
acceleration
of d, L and H in Stokes
mean square'
Chebyshev-Hermite
Wave Theories
- average
wave number,
polynomial
kD
27r/L or variable,
drag coefficient,
Eq. (2.3.46)
kj
inertia
[K]
structural
variable
log
coefficient,
natural
stiffness
period
Section
SA
Eq. (2.3.46)
Eq. (5.2.8)
matrix,
logarithm
wave length
LD
time
duration
(years)
wave length
Lo
deep water
integer
mi
of
= YffH--21
Eq. (5.6.1)
Gaussian Quadrature,
variable
height
of wave height
value
coefficients,
max
IT
r
i
kn;
Eq. (3.1.2)
/27r
gT
=
constant
-patrJLx of crost-covariances
Eq. (3.3.6)
of surface
elevation
, (cross- correlations).
viii
M
n-th
n
M(t)
order
bending
statistical
moment
moment, Section
5.6.1
MPPV
most probable
integer
N,
N+ (Y)
rate of up-crossing
Eq. (2.3.58)
N(HI/3i,
Tzj)
peak (extreme)value,
Section
3.4.1
constant
of level
y with
positive
total
number of peaks in sea state
of exposure, te, Eq. (3.4.12)
P(
probability
density
function
pdf.
probability
density
function
PH(
Pierson-Holmes
P(
cumulative
P
H(
P
Pi
P
P2
P
EP(
P/H
Pierson-Holmes
(i,
j)
slope,
during
Eq. (2.3.47)
pdf,
distribution
function,
c df,
see Appendix
1, Eq. (2.3.58)
cdf.
of positive
peak values,
type
cdf.
of positive
peak values,
cdf.
of extreme values,
Eq. (2.3.70)
'Pierson-Holmes'
Pierson-Moskowitz
spectrum
defined
through
HI/3
P-M(T )
Z
P-M(DNV)
Pierson-Moskowitz
spectrum
defined
through
Q(a)
CO
Z (X) dX
Q(t)
shear force,
r;
correlation
XIX2
RXiXj (T)
cross-correlation
and xj with time
function
lag T
between
cross-correlation
wave steepness
swl.
still
S(
spectral
IS]
'influence'
S
S
1P ij
t
influence
time
te
period
wave period
mean zero-upcrossing
water
One
see Appendix
coefficient,
random variable
; Rij
1j
ixjl
modification
z
to Pierson-Moskowitz
Section 5.6.1
r. v.
Rx
One
cdf.
P-M(HI/3)
period
parameter,
and xj
(T = 0),
Section
(3.2.2)
level
density
function
or flexibility
or flexibility
matrix,
Eq. (5.2.9)
coefficients
of exposure
or time
duration,
Appendix
wave period,
One
Eq. (2.3.22)
xi
2.3.5
ix
Tz-1
'long-term
Tr
'return
horizontal
direction
mean period
component of water
of wave advance
water
Uw
wind speed
vertical
,= dv/dt
vertical
particle
component of water
number of occurrences
instantaneous
EEW..
all ij
all
W(HI/si)
in
velocity
particle
velocity
member volume
particle
acceleration
V_
0
VR
Ij
Eq. (2.3.32)
period'
du/dt,
VN
factor',
resultant
velocity
vector
in scatter
diagram
_+V2
/U-2
class
(i,
j)
1]
TZ
E
Eq. (4.6)
wij,
horizontal
cartesian
Section
or variable,
3.4
Appendix
or random variable
X(t)
random variable
Xf
fetch
X1
mean-zero Gaussian
Eq. (5.2.2)
horizontal
vertical
sea bed
1
':
co-ordinate
L
(y2l
2,
{_
XP
in Jonswap Spectrum,
coefficient
Eq. (3.2.12)
Eq. (2.3.53)
kurtosis,
yp
random variable
in Jonswap Spectrum,
coefficient
Z(. )
dirac
spectral
water
of u, fi,
One
responses,
cartesian
functions
co-ordinate
Appendix
of nodal
Eq. (3.2.1S)
random variables,
cartesian
or random variable,
vector
One
in Jonswap Spectrum,
length
{yj
co-ordinate
delta
function
band width,
surface
Eq. (3.2.12)
elevation
Appendix
relative
One, Section
to still
1.2.2-3
water
level
:I
ol
x
of advance of wave component in 3-D random sea
vector
angle of inclination
of instantaneous
velocity
in x-z plane, Section 2.1.
direction
v
11
skewness
fluid
standard
a2
variance
ab
density
deviation
in Jonswap Spectrum,
coefficients
Eq. (3.2.12)
time-lag
random phase,
P
veld6ity
angle
X2
chi-square
angle
lj
x2
Eq. (2.3.1)
Section
potential,
of inclination,
of inclination,
mean-zero
value
frequency,
W
WO
W
P
at
C)
2.1.2
random variables,
Eq. (2.3.54)
of r. v. x
peak
of
Symbols
C)
Four
27r/T
wave
frequency
Appendix
Fig.
'Gaussian'
mean square
2.1.2
Fig.
distribution,
2.2
mean value
time derivative
second time derivative
Jonswap
Spectrum,
Eq.
(3.2.12)
-1CHAPTER0N
E-
INTRODUCTION
Current
exploitation
of oil
is
throughout
shelves
locations
and offshore
The North
the world.
Sea not
represents
respect
hostile
the available
technology
associated
with
in many areas
and gas has been found to be lacking
design of production
structural
platforms.
a result,
at the
aimed primarily
in coastal
the continental
within
only
development
emphasis on offshore
The existing
of fixed
generation
production
the extraction
including
the
of oil
are founded
structures
on
are either
lattice
are steel
bracing
of the
'jacket'
structures,
members of tubular
or
consisting
section,
'gravity'
types.
legs
of vertical
supported
by piles
The former
and cross-
driven
into
the
sea bed.
large
large
It
is
The gravity
structures
are usually
of reinforced
concrete,
diameter legs and a base designed to spread the loading over
for stability.
areas of the sea bed, relying
on bearing pressures
with
likely
of
production
facilities
into
even after
the development
including
structures,
water
structures
that
tethered
in relatively
shallower
The major
loads
upon these
motions
exerted
associated
(working)
oscillatory
buoyant
'new generation'
platforms
and sub-sea
inevitably
be required
which will
as'exploration
ventures
depths in excess of 200m, the present generation
of fixed
be employed as new fields
will
still
are discovered
and
exploited
particle
of the
these
loadings.
wave loading
by semi-empirical
with
being
waters.
fixed
structures
from the
in excess of potential
formulae.
result
These differ'in
rise
to this
waves
wind or
random
for
the generally
small diameter members associated
with
both viscous drag and inertia
steel platforms
effects
are significant
whilst
on the larger diameter legs of concrete gravity
the
structures,
of structure,
drag effects
are negligible.
installed
Structures
exposure
offshore
30 or more years
of possibly
the facility.
loading
encountered
resulting
reversals
for
Gulf
However,
depths
these
hurricanes
tropical
five
seas of only
conditions
or six
design
current
methods for
shallower
intensity,
water
North
the success
limited
design
of the structural
to the adequacy of
assurance
at Northern
structures
distress.
resulting
area offers
over
of structural
at generally
moderate
Consequently,
in this
of platforms
in the
in the oilfields
signs
years.
the
million
many years
of
fatigue
against
with
to wind generated
dangerous loading
of five
are located
structures
wave likely
associated
life
to resist
to be
strong
integrity
their
retain
of
periods
on the design
be sufficiently
depending
exerted
for
loads
these
must resist
North
Sea locations
with
the possibility
severity,
gas fields
of*their
already
the
design
of the Southern
design
against
on subjective
a single
Sea for
longer
calmer
has been
installed
periods,
and shallower
in the
but still
waters
short
have
shown signs
typical'procedure
a number of'structures
North
for
been installed
of damage resulting
and no evidence
in relatively
lives,
lives
In contrast,
have only
structures
of individual
loading
to date.
reported
Sea fixed
of their
a small-part
from either
North
design
procedures
employed to date
for
the estimation
for
are largely
deterministic
of wave loading..
strength
The
and stability,
loading
anticipated.
condition
bears
sinusoidal,
little
lifetime
the structure's
even if
the risk
fully
envisaged
Fatigue
problems
current
techniques
were not
for
to those
assumptions
relationship
assumed to represent
in the real
fairly
until
and
recently
the
is
based
height
wave
range and
regular
under several idealised
stress
of the structure
height
waves of different
with
associated
the entire
frequency
frequency
of occurrence
of wave heights
distribution
sea.
to be the case that
is proved
Sea fulfill
their
functions
more thorough
of wave load
techniques
realistic
and physically
of damage no quantifiable
be feasible
until
will
overdesign
structural
of possible
North
of the Northern
waters
any sign
without
structures
in the hostile
indication
of the associated
of exceedence
load prediction.
to extreme
applied
design
exceeded during
being
loading
of fatigue
the estimation
it
no reliable
permits
of this
waves in
large
is known.
wave height
Even if
such a technique
assessment
approximately
resemblance
Furthermore,
random seas.
wave, being
The information
and its
failure
excursion
range,
analyses
tions
associated
the realistic
resulting
However,
cannot
in terms
sinusoidal
waves.
preserve
regular
would satisfy
favourable
with
first
these
since
it
can fully
ascertained
of a discrete
A thorough
analysis
responses,
Both probabilistic
account
for
from idealisa-
must
the
internal
as
such
method is
analyses
the most,
the non-linearities
of
to the description
and spectral
The former
each
number of
structural
requirements.
distribution
transformations
stresses
wave
with
associated
necessary in fatigue
probability
be reliably
approximately
for
required
of exceedence,
risks
of extreme
estimation
may be provided
wave loading.
for
essential
loading,
stress
This
are developed.
prediction
of viscous
inherent
drag,,
-4-
yielding
the complete
Spectral
loading
thus
probability
requiring
which cannot
a limited
yield
an approximation
the full
retain
of loading
structure
description
statistical
for
or response.
the probability
of the
structure
of the non-linearities.
effects
The behaviour
probability
there
although
still
many
and
some significant
remain
in the mathematical
descriptions
of the phenomenon. The
in this field
most notable contributions
are summarised by Draperl.
limitations
Significant
various
Bretschneider
The probability
of loading
structure
members submerged beneath ocean waves has been developed
2,
3
by
Pierson
Holmes
Borgman
techniques
and
and
last
fifteen
have been
years.
The allied
of the superimposition
effects
of
currents
onto the wave motion and the intermittency
of
loading experienced
by members located near the water surface within
506
investigated
by
However,
have
been
Tung
the splash zone
more recently
.
in
have
been
loading
limited
to
the
these
of
prediction
all
applications
unidirectional
of the water
An extension
the short-term,
was proposed
domain proposed
loading
on marine
Even incorporating
distributions
value.
response
vessels
The probability
at critical
at offshore
in the wave
in the prediction
of wave
by S6dinge.
of loading
in the sea
the long-term
probability
on single
distributions
of wave loading
locations
on complete
structures
practical
or structural
is
required
by the
account
for
-5-
as discussed
However,
for
previously,
of the loading
the non-linearity
descriptions
required
for
mechanism,
fully
to the
which may be of
In these
account
when transformed
domain,
structures
viscous
methods cannot
in the probability
importance
considerable
spectral
of
effects
or response.
In view of the above comments the major
improvement of current
i)
loading
the extension
in general,
research
Fluid
Department
of Industry
Loading
(later
(formerly
the National
lines,
from applications
and into
wave loading
topics
Physical
to the Department
and the National
Laboratory)
although
and
transferred
ten research
Advisory
to
structures.
these
along
of
wave climate
techniques
probabilistic
by the British
was identified
Structures
Institute
of the long-term
members to complete
further
distributions
probability
of these
on single
The need for
descriptions
with
the long-term
yield
ii)
of the short-term
the
design are:
methods of structural
the combination
for
pre-requisites
of Energy).
Maritime
were appointed
as
by
were undertaken
organisations.
In October
OSFLAG 5 project
Liverpool,
forming
awarded to Professor
the research
the bulk
undertaken
of the contents
OSFLAG 5 research
members of offshore
probabilistic
by the author
of this
description
structures
descriptions
was stated
as a Research Assistant
Thesis.
or probabilistic
statistical
problem
was appointed
under
of
contract
The objective
of the
a method of calculating
of
long-term
based on short-term
of wave conditions.
in the original
this
on the
wave loading
a
on
and long-term
The background
OSFLAG 5 proposal,
as follows:
to the
-6'Regulations
integrity
concerning the long-term
of offshore
be subjected
to wave action over several
structures
which will
tens of years are of obvious importance.
For such regulations
,
in
design computations
and
corresponding
a unified
method is
imposed
loading conditions
to estimate the lifetime
required
by waves.
For particular
the load on a member can be
wave conditions
based on a single wave - said to
estimated by computations
based on the
represent
a given sea state - or by calculations
probabilistic
properties
of that sea state,
using known
between wave kinematics
relationships
and load and accepting
the limitations
thereof.
The former method leads to a deterministic
description
of the
load, the latter
distribution
to a probability
of the load
which has been shown to give good agreement with measurements
in the Gulf of Mexico and the North Sea.
descriptions
It is then possible
to accumulate 'short-term'
based on the second method given above according
to 'long-term'
is in need of further
Although this behaviour
wave behaviour.
it is considered
that the Weibull probability
research,
distribution
for wave heights
forms a viable basis on which
predictions
of long-term
wave loading can be based.
There is an urgent need to formalise
this computational
loading on a single member in terms of magnitude
process giving
and number of cycles per annum.
This computation
then needs to be extended to entire
structures.
This is a significantly-more-complex
problem because of the
between wave lengths and structural
spatial
relationships
for dealing with such
Techniques are available
geometry.
loading of complex
and the problem of cyclic
correlations
is of such importance as to warrant an investigation
structures
design procedures. '
of their use, leading to viable
This
theoretical,
computer
in the drag/inertia
two parts;
single
regime
elements
Thesis
on steel
involving
the first
structural
based project
describes
lattice
estimation
at loading
aimed primarily
was split
structures
of the long-term
involving
loading
into
on
an extension
of
complete structures.
the results
of work undertaken
by the author
on
reference
-7Two develops
Chapter
description
the background
of long-term
mode121 which
is
accepted,
It
sections
with
proceeds
an
wave load
'Morison'
semi-empirical,
to short
applicable
to the probabilistic
relevant
of the, generally
outline
theory
with
of submerged, objects
dimensions
is
This
theories
for
sites
Airy
for
offshore
or first
installations
the description
of the particle
summarised,
the particle
through
'Morison'
kinematics
model.
reviewed
Both short
and long-term
beneath
estimation
is discussed.
identified
techniques.
loading
for
is
in
prediction
long-term
first
values
field
through
first
order
design
methods are
for
in this
Models yielding
analyses.
the
structure
the Morison
of load
the purposes
involving
transformations
wave load
are
non-deterministic
more rigorous
and fatigue
distributions
probability
shown to be favourable
excursion
of
deterministic
in these
and probabilistic
respect
of the waves is
descriptions
Limitations
waves are
regular
of the orientation
to the direction
theories
amplitude
of which define
statistical
(or
The linear
finite
motions
to the Morison
and modifications
Sea.
The significance
members in relation
cylindrical
and Stokes'
wave theory
order)
the
in the North
loading
model and
wave theory.
The theoretical,
procedures
described
in Chapter
of
within
Three describes
Chapter
in-the
of. the
loading
and as proposed
leads
distributions
probability
mechanism,
diameter
members, typiqal
for
and it
inherent
in some probabilistic,
to a considerable
of work carried
long-term
of loading
and discussed
are presented
values
tion
of the techniques
application
Long-term
the results
and its
demonstrated
linearisa-
that
speqtral
techniques,
approaches'to
simplify
analysis'
extreme
structures,,
in existing
underestimation-of
of steel
is
loading
for
small
of
viscous
used to describe
models,
spectral
is
load prediction
implementation
these
investigated.
also
hand-computation
wave field,
Computing
techniques
on the wave
for
requirements
are considered
the
and since
in most design
available
is
procedure
theoretical
of various.
the short-term
of the probabilistic
are likel*Y
a useful
The significance
offices
outlined.
Five
members to entire
structural
large
the probability
extends
are idealised
number of individual,
extension,
limitations,
therefore,
members.
supported
are developed
tions,
structure.
the
long-term
wave climate
The results
of these
procedures
of structural
load or structural
total
response
loading,
as the elemental
data,
Mathematical
probability
is
due to computer
all
expressions
short-term
resulting
statistical
This
structure
from analysis
of a
shown to be impractical,
systems of greater
inadequate
for modelling
However, by assuming that
postulation
or associated
conditions-where
for
possess
with
is
stationary,
loading
on single
to result
correlated,
to-short-term
applied
statistically
this
structures
responses
from application
theory
distributions
to yield
are applied
only
definition
these
of
moments
and the
of the wave field
of load may be convoluted
long-term
to loading
for
probability
distribu-
-9Chapter
Six contains
and Chapter
Seven presents
recommendations
for
work.
lo
-
CHAPTERTW0
MEORY
PREDICTION
2.1
structural
structures
against
Application
for
wave loading,
which
the diameters
of the majority
of the cylindrical
for the
is intended primarily
or jacket, type of offshore
wavelengths
regime
on vertical
exerted
water
particle
in the
of loading
into
1.
by
Morison
et af
members
ocean wave
The method
the total
that
force
is equal
fluid
velocity
particle
the estimation
from an investigation
has developed
cylindrical
semi-empirical
by the
method for
accepted
acceleration,
interaction.
force
The horizontal
F-F+F=
D12Dr0
ulul
by 'Morison's
Equation'
as:
+ CM PV fl
inertia
drag
FI
FD
the
are
and
components, respectively;
and
where
velocity;
component of the water particle
u is the horizontal
iuis the horizontal
A=
the water particle
of
component
dt
acceleration;
V0 is
Ar is
p is
the fluid
density;
-It
is
perpendicular
to u;
and
drag and inertia
coefficients.
would be present
its
in
location
the
the
the
at
of
centroid
of
member
in these,
that spatial
assumed in Eq. (2.1.1)
variations
kinematics
are negligible
that
up by-the
member.
The
11
-
direction
to
is
included
the
preserve.
of the drag component
modulus sign
of loading which is a function of velocity
squared.
The absolute validity
of the linear
is open to question
assumptions
action
being
therefore,
effects,
as discussed by
the derivation
of the inertia
for
by the presence-of
component are-violated
combination
since,
However,
significant.
inter-
of drag,
the effects
from an
in
In their
point
on vertical
Little
fluid
strictly
(2.1.2)
to the loading
around non-vertical
to a short
only
some writers
section
mechanism resulting
members, Eq. (2.1.1)
of a vertical
in any particular
the load
members lying
author
oil
in a plane
in association
production
principle,
platform.
parallel
with
his
This
colleagues
representation
for
follows
in
by the
to a North
Sea
the cross-flow
in agreement
orii entation
may
inclined
was developed
application
motions
mechanism for
of the loading
representation
being
direction
A more thorough
from
member, although
be derived
that
load per
+p
motion
applicable
the loading
only with
horizontal
2C6
attention
wave-induced
7rD
jul
CpDu
D4m
et al were concerned
members for
cylindrical
is:
length
unit
work Morison
original
with
with
respect
previous
to the
waves.
Consider
the water
a unit
surface
length
section
at location
of cylindrical
(xO, yo,
zo),
2.1.2.
The'
is horizontal
to the y
the member is inclined
plane
and
-y
angle
at
an
.;
Regular long-crested
axis in the y-z plane.
waves are assumed to advance',
,
in the positive
x-direction.
12 istant
At a particular
of time the resultant
vector at location
velocity
(xo, yo, zo) will
be inclined
at an angle 0v to the horiz. ontal,
as
in Fig. 2.1.2,
indicated
where E) varies between 0 and 27Tduring the
v
Furthermore,
the resultant
giving
passage of each wave.
velocity
vector,
is out of phase with the resultant
rise to the drag component of loading,
component of loading,
giving rise to the inertial
force at any instant
must be evaluated using vector
vector,
acceleration
addition
two components.
friction
drag will
to the total
essentially
by the pressure
the direction
of VN (the
acts
normal
difference
in
normal
may,,
water-particle
velocity
F=
DN
[U2
pCD
-1
2D
(V
F=F
Dy
sin
DN
in the co-ordinate
loading
only
significant
pCD
2D
sin =-F
in a plane
2
7TD
(CI) 2+(,
c
4
M[
Fp
IN
Reducing this
pC
2D4m
Fy =-F
P-
way if
it
normal
COS ) 2] 1/2
Z"
is
assumed
are
(2.1.7)
Du I V(U2
bvl 1/42
tan
(2.1.5)
(2.1.6)
in a similar
as:
(2.1.4)
force to its
are:
tan
Dy
may be estimated
accelerations
29 yielding:
,
directions
ulv"u2 + (v cosfl
cos
F
sin
'
-F
DN
Dz
The inertia
(2.1.3)
Cos
pCDD
F
Cos
=F
Dx
DN
that
it
Consequently,
the cylinder.
of skin
force
be negligible
27
the contribution
regimes
flow
(V
(V
COS
COS. )21
+p
2C
7TD
2
7rD
P4-
CL
*]
CM
and
force
(2.1.8)
Cos
(2.3.9)
(2,140)
13
member =
For a vertical
FY = Fz = 0.
still
plane
perpendicular
the
condition
of the cross-flow
is rotating
velocity-vector
normal
with
to the application
water-particle
in a vertical
to Eq. (2.1.2)
special
subject
the resultant
since
principle
in this
Note that
member is
vertical
is
and, therefore,
and troughs
only.
a horizontal
to the wave
member with its axis parallel
crests represents a condition in which the resultant
rotating velocity
vector is always perpendicular to the member axis, obviating the need to
0 and Fx
For such a member
apply the cross-flow principle.
to:
simplifies
In contrast,
1: - -1
x2D4m
A similar
Du IVu-2--+--v-yI+p
expression
Eqs.
the values
if
equation
Although
in this
a larger
Fy is
the roles
zero.
a horizontal
of CD is
value
but with
loading
situation
member using
former
in
the
required
are to be predicted.
U. S. Army Coastal
the vertical
Obviously,
and (2.1.11)
(2.1.1)
for
exists
of u and v reversed.
Whilst
2C
7rD
Engineering
28
27
by
Hogben
Borgman
the
and
supported
,
Research Centre 30 indicate
that from
is
cylindrical
vertical
equivalent
It
tion
is probably
outlined
by appropriate
re-emphasised
developed
loading
so far
would prove
acceptable
of CM and CD*
the basic loading
as combinations
are essentially
mechanism relying
per unit
length
on an
to say that
choice
that
force
justifiable
above,
interaction,
pile
as the horizontal
At this
for
engineering
stage
it
should
applications
be
forms
model and the modified
of drag and inertia
components without
semi-empirical
on the choice
approximations
of CM and CD values
to the actual
to yield
a.
'14
reasonably
of this
by virtue
form,
engneering
behaviour.
loading
In view
available
the
of the empirical
coefficients,
to be of the
made by assuming the loading in any direction
approximations
Morison
flexibility
and the
uncertainty
considered,
to the actual
approximation
close
of the presence
developed
may well
for
be acceptable
purposes.
Much effort
IInadequacies
in order
However,
thereby
and inaccuracies
reduce
models.
the uncertainties,
with
and
correctly,
chosen in any
the coefficients
be compatible
should
application
adopted.
values
from vertical
the coefficients
selected
not values
for
herein
In the work described
in detail,
the interpretation
for
loading
2.2
only
loading
being
in any direction
that
investiga-
of statistical
analyses
waves.
on the appropriate
conditional
deterministic
use with
members using
in non-deterministic
Similarly
on horizontal
load analysis
choice
of empirical
coefficients
CM and C
D*
There exist
a large
of periodic
motion
particular
water
depth
waves,
some being
to their
Deepwater
than
others
more suited
regions.
which describe
d,,
is
-
Since,
and into,
towards,
tending
height
tion
finite
here,
motion
('long-crested'
extent
on conditions
32
wave theories
Fig.
2.2.1
non-viscous
34
1880).
describing
water
the
waves of constant
to the direction
perpendicular
(Lamb",
wave theory
(Stokes
theories
amplitude
of oscillatory
and of infinite
be centred
will
system shown in
the co-ordinate
irrotational
attention
deep water
ones considered
here,
of propaga-
waves).
4p
,
Pand
v
Fx
(2.2.1)
az
the requirement'that:
(Laplace)
the continuity
equation;
(dynamical,
the energy
and
'Bernoulli')
equation,
for
are satisfied
the
In addition,
i)
ii)
particles
it
(first
that
order
the surface
where
surface
'goodness
of the theory
Wave Theory
Linear
Linear
the free
to satisfy
possible
the
exactly,
conditions
surface,
must be atmospheric.
the pressure
boundary
are:
particles
In practice
to be satisfied
conditions
or Airy)
boundary
strictly
and of-small
at'location
wave theory
condition
being
applies
applicable
steepness.
x and time
makes the
For this
t is
at z=d
linearising
rather
assumption
than
to waves of infinitesimal
theory
the water
giVen by24:
surface
zd+
16
n (X, t)
(kx - wt)
cos
2
(2.2.2)
wave height
where H-
gk tanh
w2=
(2.2.3)
d= water depth
g= gravitational
The water
acceleration.
kinematics
particle
The component
are:
velocities
U(x,
Z" t)
v (x 0z
t)
!i
=w 2
kz
cosh
Z:-- sEiU
(x,
,(X,
Z,
t)
Zu
=
-i -t
11
=w2
-!
=_W2
H
I
LV
Z,
t)
at
- wt)
(2.2.4)
s in (kx - wt)
(2.2.5)
cos(kx
kz
kd
sinh
.w!! 2 [
sinh
which
as follows.
are expressed
are:
kz
cosh
sinK kdl
sinh
sinh
kz
0
their
sin(kx
- wt)
(2.2.6)
cos(kx
- wt)
(2.2.7)
respective
velocity
components.
I
In deep water
the depth.
(d/L
For first
and symmetrical
shown that
2.2.2
about
the water
becoming circular
their
> 0.5)
original
Stokes!
order
wave theory
still
water
particles
in deep water,
position
Finite
become independent
the surface
motion
Furthermore,
move in closed
as indicated
it
elliptical
in Fig-2.2.2,
is
of
sinusoidal
can be easily
orbits,
returning
to
Wave Theories
Iy
is expanded;
theories
the
amplitude wave
velocit
potential
24 which yields
described
by
form with
Wiegel
a
symmetrical
wave
as
.
larger
to the
and more peaky crests than troughs conforming more closely
In the finite
17
form of steep periodic
observed
gravity
expressions
= Gi cos(kx
n(x,
t)
u(x.
z, t)
- wt)*G6
of d,
Sth order
2.2.3
Utility
of Linear
motions
of offshore
order
Linear
wave theory
to describe
of wave loading
estimation
values
of wave amplitude
theories
provide
on members
more accurate
acceleration
it
However,
theory.
and period
and
is
leads
30
suggested
from empirical
are obtained
of parameters
and
to greater
the non-linearities,
involved
in the
theories.
of large
only
steep waves,
be applied
kinematics
estimated_by
derived
for
It
to still-water-level.
particle
for
the British
strictly
the final
about
uncertainty
hence questioning
- wt)
(2.2.8)
(2,2.9)
- wt)
of wave height
surface
36
(non-linear)
higher
cos 3(kx
With precise
order
when estimates
+ G3 cos 3(kx
L and H.
deterministic
structures.
fields,
- wt)
Wave Theory
pressure
effect.
+ G5 cosh(2kz),
- wt)
cosh(3kz)
solutions
for
the higher
estimates
cos(kx
approximation
+ G2 cos 2(kx
form involving
of a similar
Stokes'
period
- wt)
G4
cosh(kz)
--
cos 2(kx
particle
resulting
35
3rd
the
in the following
of
order approximation
results
for surface elevation
and horizontal
velocity:
The solution
Expressions
orbits
closed,
exhibit
that
Particle
waves.
and total
linear
theory
linear
theory
from the
the
although
the deviation
wave induced
differed
being
force
should
and at elevations
between the
on a typical
member,
by
up
18
Milgram 26 has stated
that
waves of realistic
steepnesses the first
order theory
has become the basis for much of the wave loading work in ocean engineering.
The major advantage of linear
theory is that it is wholly linear
of regular
so that
of any height,
can be
or direction,
period
Also, any
to form a realistic
added together
sea condition.
wave forces which are expressible
as linear
operations
on the sea
These properties
surface are given by the sum of the component forces.
form the basis of the non-deterministic
methods of wave load prediction
simply
discussed
2.3
later.
The theory
outlined
structural
members by particle
height
in various
Under these
value
of the surface
loading
of
exerted
on
waves of constant
form of the ocean surface is
regular
of differing
the variation
height
and
and period
of the surface
elevation
in
of a random nature.
it
conditions,
beneath
the actual
directions,
the estimation
motions
confused.
both
However,
and period.
travel
enables
is no longer
possible
to predict
accurately
the
elevation,
or associated
particle
motions,
at any
instant
point in space at a particular
of time, as is the case for
regular
waves, and the most complete description
of the phenomenon is
in terms of its
provided
2.3.11
Description
The usual
is
With reference
and phase.
expressed
waves,
as
properties.
method of describing
to consider
linear
statistical
of an infinite
a random height,
to Eq. (2.2.2)
period,
sum of infinitesimal
direction
the surface
elevation
of advance
may be
37.
H
n(x, y, t), =
En
cos(k
2nnnnnn
n1
x cos 0+ky
sin. 0
-wtt
(2.3.. l
I
where x, y are orthogonal co-ordinate axes*in the. horizontal
-j',
plane-,,-,
is a-random phase-uniformly distributed
between 0 and 27r; and
0 is the direction
advance measured relative
to the x-axis.
of
n-
19
The wave height'
in each small
0, space:
w, and direction,
such that
element
rectangular
components,
E
(AWAG)
(2.3.2)
s (W, 0) AwAG
nil
rectangle.
where the summation is over the whole of the infinitesimal
density
Snn (w, 0) is the bivariate
spectral
of the sea surface
0. The summation may be interpreted
at frequency w(> 0) and direction
the variance
of all
addition
in the rectangle
describes,
therefore,
the distribution
of the total
surface
unit
the spectral
that
and noting
Eq. (2.3.1)
of wave
a particular
sea state,
Eq. (2.2.3),
the time
y) can be described
(x,
21T
W
(w,
O)AwAO
EE
cos(kx
vf2STITI
W=o 0=0
for
In a similar
manner expressions
to Eqs. (2.2.4)
reference
Although
then
using
as:
n (X, Y, t)
with
through
at location
elevation
is known for
function
density
Also
of the
variance
be considered
in surface
variation
as
ky
sin
0-
wt
(2.3.3)
the particle
can be developed
motions
(2.2.7).
to
distribution
the directional
cos 0+
has
directional
description
the
the
spectrum
of
ocean
waves,
of
of the
as a result
partly
not been widely used in design computations,
37
Theoretical
lack of reliable
models have been developed
measurements.
feature
since
is
most measurements
in the 0 domain,
spectrum
much attention
receiving
it
is
S (w) in engineering
nTI
at the present
of wave climate
common practice
applications,
do not
to consider
time.
include
informa-
the uni-directional
where:
21r
(W) =ES
Snn
0=0
(w,
TITI
0)
AO
(2,3,4. )
20
-
-
is considered herein.
situation
may be expressed
as
co
V2S (w) Aw cos(kx
=E
Tin
W=o
00
kz
cash
Ew
U(XO Z, t)
U i(2S
n(x,
and
t)
sinh
W=O
S (W)
uu I
is
(w)
nn
Aw cos(kx
E- v12S (W)
cos(kx
uu
W=U
2
kz
cosh
(W)
Sylyl
sinh kdl
and
the horizontal
velocity
particle
(2.3.5)
- wt
- wt +
(2.3.6)
wt +
(2.3.7)
density.
spectral
Similarly:
00
'(Xlzit)
W=0
kz
cosh
2
V2S
(uj) Aw sin(kx
63 'Iln-hkd
TITI
(2.3.8)
- wt +
00
=E
v2S66(w) AW sin(kx
W=0
(w)
S
where 61a
Expressions
.2
cosh kz
sinh kdl12
kinematics
for
particle
the same way with (cosh kz)
2.3.2
Statistical
2.3.2.1
STITI (W)
- wt +
= W2 S
uu
in the vertical
replaced
and Probabilistic
by (sinh
(2.3.9)
(W)
direction
follow
in
kz).
Properties
Short-term
In the short-term,
over a period
of several
hours,
intensity
be
Under
be
these conditions
the surface
to
constant.
assumed
can
in
form
it
is
be
the
that
and
shown
of constant
can
spectrum
elevation
as Aw --O, by the central
is statistically
surface elevation
limit
limit
Gaussian probability
following
stationary
a mean-zero
38.
(A summary of statistical
and*
in Appendix One).
probabilistic
distribution
concepts
is
given
theorem,
the randomly
varying
21
density
The probability
Appendix
function
(pdf. ) Of rl is,,
One) :
f-
(TO =1
P
-xp
Yr2-TrcrT,
Ebil
where
11
("
2
-7A
(2.3.10)
cin
value
mean
-0=
of
(2.3,11)
Ti
GO
2=
(W) dw
Tlyl
EQTJ -S
ti
- variance
of
and
is
Furthermore,
the water
the expected
value
operator
reference
to Eqs.
(2.3.6)
with
with
random variables
respective
kinematics
particle
distributed
kinematics
Gaussian
of the linear
as a result
output,
Gaussian
that
given
input
a property
r2T
to linear
proved.
above still-
_C2)1/2
x2
-
+
_C
Y)
2
--P(-
X(1_C2)1/2
a2
'K2
exp(-
exp(- -) da
(2.3.13)
9=
nmax/anor nmin/aTI
where
and
c=
bandwidth
spectral
For a narrow
band spectrum,
(see Section
c=0
1.2.2.3,
Appendix
and:
X2
exp (which
is
a Rayleigh
For a wideband
(see Section
pdf.
c=1
spectrum,
X2
p(X) =1
1
/27
exp(-
to
transformations
that
clear
systems produces
are also
is
variances
density
spectral
(2.3.9),
it
mean-zero Gaussian
and:
1.3.1.6,
Appendix
One).
One).
22
-
frequency width
less
substantially
most ocean wave
a narrow band
applicable
in the North
'Famital
wave environment
Saetre 42
41
Driver
and
and
Sea by Draper
Employing the narrow band model, the pdf. of wave height, H, defined as
between two successive upcrossings of the mean-water-level,
Nax
11
mid
is a Rayleigh pdf. given by
H
4a
(H)
exp[-
H.
-8
(2.3.14)
Ti
is:
Iff = 42-Trcr
TI
also
wave height
the root-mean-square
VE-THT' =2
H=
rms
height
p(H)
1/3
4H
HI/ 37
4a
(2.3.16)
of the highest
logarithm;
- erf
(1/n)-th
of the waves,
H
1/n'
rx-nn)]
is 39 :
(2.3.17)
and
function
wave height
The significant
HH
ZI a
TI
-(l
[41--nn + 11 r7,
2
rms
corresponds
to a value
3, hence:
(2.3.18).
TI
)t42
(2.3.19),
23
f
distribution
P(H) =1-
Hl'/s
(see Section
cdf.
1.3.1.2,
2H2
31
Rj
exp[-
Eq. (2.3.12)
Substituting
function,
(2.3.20)
into
(2.3.18):
rmo
(2.3.21)
The significant
is
one parameter
short-term
behaviour
of surface
elevation
records.
is
The other
Tz = 21r
the
used to characterise
and is obtained
from analysis
defining
spectrum (see
elevation
period,
crossings
1.3.4.2,
time with
per unit
Appendix
One):
A2 a
12
m
(2.3.22)
(W).
S
is
the
of
second
moment
m2
where
nn
The probabilistic
those
for
of wave period,
properties
wave height,
of T2 being
the pdf.
and
used in practice.
are little
HI/3
The parameters
and spectral
descriptions
of short-term
to define
both
sea states
using
the probabilistic
the most
description
The
of TI
models.
probabilistic
existing
employed
commonly
is
(2.3.10)-(2.3.12)
Eqs.
the one parameter mean-zero
through
obtained
Gaussian process
Moskowitz
may usually
from analysis
The spectral
For example,
particular
(2.3.18).
sea surface
or Tz or both parameters.
of HI/3
speed,
by an using
functions
large
defined
may be expressed
be defined
in terms
by curve
for-fullydeveloped
density
fitting
sea states
as a function
of a
over
of wind
24
A-f!
A24
= -,
(w)
S
TITI
B[(40) 1
expl-
(2.3.23)
where A=0.0081
B=0.74
WO = g/u
w
(2.3.21)
Eqs.
and applying
is
parameters
sea state
2U
w2A
and
27rU
w
z9
2.3.2.2
formed.
(2.3.24)
1/4
(2.3.25)
description
of wave intensities
statistical
Long-term
'one-year'
for
taken
Each record
values
yielding
analysis
hours.
the distribution
months or
are often
elevation
three
every
six
to represent
considered
of typically
of surface
durations
15 minute
describes
over periods
Recordings
durations.
can thereby
other
be
sea state,
stationary
with
approxi-
significant
parameters.
data
is often
represented
histogram,
or bivariate
of HI/3
values
Long-term
The long-term
mately
link
the
1/2
(k-d
and (2.3.22)
giving
and Tz fell
in the
form of a scatter
the proportion
within
certain
of the time
classes,
diagram,
for
which
the
2.3.1.
as shown in Fig.
taken in the
obtained from wave recordings,
shows results
between
from
October,
1969 and
operation,
one
winter's
manner,
above
March, 1970, by shipborne wave recorder placed on m. v. 'Famital,
a
This
figure
Norwegian
rescue
and meteorological
North
2.3.1
is
a prediction
of a corresponding
extension
Four,
Also
a predicted
Famita
computations
scatter
described
included
diagram
scatter
the one-year
long-term
as shown in Fig.
stationed.
in
Chapter
expounded
in
input
all
used as
vessel
in
for
a six-
one-year
scatter
description
is
diagram-being
herein.
25 2.3.2.3
distribution
The long-term
(see Appendix
distribution
The marginal
the bivariate
of HI/3
pdf.
wave height
of significant
One) of HI/3
is obtained
from
and Tz as:
co
p(HI/3)
and
HI/3
f
0
P(HI/3)
dT
Tz)
p(HI/3,
(2.3.26)
d1lip
p(HI/3)
(2.3.27)
diagram,
of the i-th
AHI/3 = class width of HI/3
I/si
= mid-point
- mid-point
zj
AT
z=
w
ij
class
of the j-th
class
j=1,
...,
diagram
class
(i,
j)
as:
Imnm
2=ZZw..
this
interval,
in the scatter
since
in
of Tz
width
AHI/3
However,
class interval,
= number of occurrences
Eq. (2.3.27)
we have:
i=l
/ ZSw..
13 i=l j=i
j=i
expression
(2.3.28)
13
an incorrect
assigns
of 1.0 to
probability
data
the upper-most
class
is
modification
Ali 1/31mnm
wij /(
P(Hl/
EZ
i=l j=j
j=i
This
enables
data points
all
effect
on the cumulative
points
to be plotted
probabilities
w
ij
associated
number of occurrences
(2.3.29)
with
in the
the other
scatter
data
is
diagram
large.
2.3.2.4
Eq. (2.3.20)
but
through
as a conditional
P(HIHI/3,
since
Hi/3
is
this
and T
it
conditional
exp[-
2-
wave height
wave height,
H, is
is more correct
(See Appendix
cdf.
Tz)
of individual
of individual
distribution
The short-term
defined
distribution
The long-term
to re-write
by
given
intensityl
the equation
One);
H2
3
(2.3. -O)
26
-
distribution
The long-term
7
derived
by
Battjes
as
as a
of the above short-term Rayleigh distributions:
may be obtained,
ff
P (H) -1
by Tz for
the transformation
into
diagram,
TZ)
}
2
dIIi/3
expf:-_
Hi,,
/ ay
d1ij/3
dT
dTz
(2.3.31)
(2.3.32)
each short-term
Of 111/3, Tz reflects
condition
in converting
necessary
time,
by the scatter
given
number of waves.
form,
In discrete
2
11
Tz
..
p(HI'/3.
__
T
'
0z
The division
p(III/3,
_7T
T
where Tz-
Tz)
Eq. (2.31)
becomes:
nH2MW
[exp{
P(H)
2 ----i
HI/311
; l
Tzj
(2.3.33)
nmW.
.
13
ET
i=1 zj
where TZ-1
(2.3.34)
The
This cdf. is plotted for the Famita wave climate in Fig. 2.3.3.
plot is made on Weibull paper and it is clear that the tail of the
indicating
is reasonably straight
distribution
that the cdf. of H may be
approximated quite well by the theoretical
follows with reference to Appendix Two.
The 'design
design
wave' approach
The objective
failure,
to ensure
to occur
is
during
structural
to design
waves is
passage of the
of a single
largest
wave.
This
load is
wave is
ocean
first
excursion
the greatest
of the loading
assumed to result
load
induced
by
from the
'regular'
the
of exposure
a period
against
In deterministic
ignored
Limitations
and their
of the method is
that
which
and structural
of structures
waves.
likely
distribution,
Deterministic
2.3.3
Weibull
the form
Wave kinematics'
motion of regular
developed using an appropriate wave theory', as outlined"
are, therefore,
in Section 2.2, yield ing wave loading on structural
elements using
,
If the structure is modelled as comprising of. a
Morison's equation.
large number of loaded elements a quasistatic
structural
analysis may'be
27
-
imposed
loads and moments, internal
total
to
estimate
performed
deformations, etc.
and structural
'design'
wave is
of the design
The height
Tr,
period,
return
of a wave which is
*Since wave data
in Tr years.
magnitude
50 or 100 years.
typically
the height
chosen to correspond
often
equalled
wave height
covering
Long-term
'one-year'
represents
from measurements
distribution
the long-term
using
in the previous
Fig.
using
2.3.3
section.
be estimated
could
This
possibly
covering much shorter periods,
heights
introduced
individual
wave
as
of
from the
to a certain
must be extrapolated
For example,
stresses
as follows:
T
lSll
Z_
23.8 x 106
1x
0*9999999958
Prob (H 4 Hso) = P(Hso) =1
23.8
105'
from
figure,
26m.
H50
the
=
extrapolating
and
It
risk
in the design
must be appreciated
of any design
the adoption
with
associated
there
wave height
is
always
(or design
is
larger
than the
there
that
a
a
wave
possibility
as
return
design wave will
occur during the period of exposure of the structure.
It can be shown that if all waves are assumed to be independent and if N
for
design
in
then
the
structure
a
of
one
year
of
waves
number
represents
Tr)
period,
life
L years,
that
the probability
the largest
wave will
exceed the
is:
H
Tr ,
design'wave,
LXNI
Prob(H
and since
(Tr x Ni)
Prob (H
max
be large:
will
[expi-
H1Tr
1-
the design
For example,
if
the structure
Is design
of the design
(2.3.35)
(1 -1
(Tr--xNi))
>H)=1Tr
max
expf-
return
lifetime
wave height
being
1
(Tr- -xNI)
LxNi
11
(2.3.36)
Tr
period
than
to,
a 63% chance
28
-
the height
Having determined
to be associated
wave period
linear
Theoretically,
wave theory.
1/7 although
approximately
analysis
of 1/12.
limit
As a general
height,
it
is often
in wave period
reduction
is only
this
in design
rule
true
strictly
for
for
assumed that,
increases
the induced
members located
close
wave
a given
loading.
However,
to the water
surface
as a result
depth.
deep water
design
it
on all
members.
exceedencel
It does not
the
regarding
follow
of assessing
taking
into
sufficiently.
100-year
the
and structural
account
all
the current
strong
existing
must
of wave height
yield
a more acceptable
required.
method
behaviour
probabilistic
is
response
design
to resist
To obtain
structure
mechanism there
wave will
long-term
non-linearities
gained with
or
design
load or stress.
From experience
appear that
the
of failure
risks
that
design
1100-yearl
of failure'
the loading
of the form of
from a 'risk
the transformation
to a 'risk
criterion
to maximise
loading
the
on some members of
idealisation
of the
be some doubt
loading
a structure,
Hence, although
wave theory.
other
linear
and using
conditions
This
of
must be
offshore
methods yield
the extreme
installations,
structures
loads
imposed.
it
would
which are
Unfortunately,,
the
in the absence of the more rigorous
probabilistic
analyses,
by
departure
designs,
between
the
the"
the
measured
as
of
efficiency
'- 29 design
accepted
failure,
associated
the
with
assessment, is unknown.
members, following
of structural
periods
to
of exposure
induced
the oscillating
concern
great
risk
loading
deterministic
Fatigue
z
of failure
description
for transformaof the peak loading distribution,
probabilistic
is essential
for fatigue
tion into the stress range distribution,
analysis,
45
by
described
Lin
later.
In the absence of probabilistic
and outlined
as
deterministic
design rules,
descriptions
of load or stress,
and more
methodS46 , have been developed,
by the development
of which need to be investigated
more thorough
recently
performance
for
analyses
probabilistic
in the previous
loading
the non-linear
of
situation.
Non-deterministic
2.3.4
the
spectral
it
Section
is
clear
deterministic
that
their
statistical
(in
analyses
simulation
domain)
frequency
in association
properties
the time
domain),
or probabilistic
summarised in Appendix
Simulation
considered
techniques,
demanding in computer
for
run times
to short-term
application
analysis
of structures
although
for
rigid
it
(in
with
in detail
briefly
which
and consequently
such effects
respond
many marine
are thought
(in
the amplitude
the
domain)
Five,
are only
Spectral
are extremely
feasible
really
techniques
have the
can be applied
they
to
structures
are relatively
to be negligible.
of spectral
parameters
However,
descriptions
of load or response.
analyses
must be transposed
Consequently,
to yield
probability
necessary to define the relevant
(linear)
techniques
spectral
can only
existing
as
in Chapter
dynamically
time history
either
analyses
analyses
methods in that
spectral
conditions.
over probabilistic
advantage
with
and
the
the statistical
distributions.
develop
the
30
-I
form' requires
later.
moments higher
Statistical
techniques
50.
492
For example, the bispectrum would be related to the
order spectra
third order statistical
to the fourth moment.
moment and the trispectrum
in
However, although higher order spectra have received some attention
52
behaviours',
far
from
being
to
they
ship
are
applications
,
well understood for inclusion in design procedures.
sufficiently
using spectral
Consequently,
Chapter to considerably
under conditions
or response,
2.3.5
Statistical
Kinematics
The short-term
locations
loading,
are required
Morison's
In Section
developed
properties
from a description
Properties
and Probabilistic
of water
particle
as an intermediate
earlier
kinematics
step
at particular
in the transformation
to the description
using
linear
of wave
wave theory
and
equation.
2.3.2.1
it
was demonstrated
that
the particle
kinematics,
u,
31
The relationship
for
between certain
the development
of the kinematics
pairs
of the probability
More generally,
p4irs
of the kinematics
at different
considering
domain,
is
when
required
in Chapter
covered
of
Five,
properties
is under
load components
'Morison'
the behaviour
space locations
loading
the multi-variate
in the following
describing
the relationship
for
of wave load,
structure
are required
The cross-covariance
expressions are of prime importance in this respect
degenerate
for
to
the
these
cross-correlation
meanexpressions
although
functions relating
zero variables (see Appendix One). The correlation
kinematics at location (xi, zj) to those at
the value of particle
location (X2j Z2) measured T. time units later have been derived by
53
from expressions of the form of Eqs. (2.3.6) and (2.3.8), as
Foster
,
follows:
(T)
UIU2
(-
=R
u2ul
Co
2
=fwS
0
(w)
liT1
dw.
U2(t
kz,
cosh
kZ2
cosh
sinh? kd
COS(k(XI
- X2)
+ WT)
(2.3.37)
Co
f w2 S
(T)
R
U1V2
= E{ul(t)
T)
(w)
cosh
dw.
kzi
kz2
sinh
sinh2 kd
sin(k(xl
+ WT)
X2)
(2.3.38)
RV1U2
(T)
W2
(T)
R
V1V2
inh
kz,
s
cosh
S
(w)
dw.
1111
sinh2 kd
Co
f W2 S (w)
1111
0
sinh
dw.
kz, sinh
sinh2 kd
kZ2
sin(k(xl
X2)
WT)
(Z. 3.39)
kz2
cos(k(xi
- X2)
+ WT)
(2.3.40)
Also:
RU114
(T)
R.
(T)
UIV2
and
a2
=3T
(T)
R.
UI12
5-T
[ Ru
(T)
(2.3.41)
IV2
[R
(2.3.42)
u IV2
etc.
The cross-correlation,
or values
are required
in probability
T=0,
For kinematics
correlations
at a single
degenerate
point
as follows,
function
of the cross-correlation
at
analysis.
(i. e.,
xi
omitting
the subscripts:
32
R2;
uu
Rv=
uvvu,
2;
R2
a.
u
cf
(2.3.43)
and
RR.
w
uv
-R.
uu
Methods for
Probabilistic
Structural
Associated
at a point
(see Appendix
kinematics
and acceleration
velocity
components
independent
2.3.6
vv
(2.3.44)
in either
are uncorrelated
co-ordinate
the Prediction
Response
The prediction
of wave loading considered herein is restricted,
as
form
in
Section
2.1,
loading
'Morison'
to
the
mechanisms
on
of
mentioned
do
dynamically
and will be limited for
not
respond
which
members
Eq. (2.1.2),
Thus, restating
in the horizontal
of loading,
convenienceto a treatment
from long-crested waves.
the horizontal
direction
only,
length
of
is:
member
kI ii +kD
2
7TD
CM P
4D
where k
The effects
(2.3.45)
ulul
currents
of uni-directional
in the splash
(2.3.46)
kICpD
- 2D
and
included
here,
but
are discussed
of loading
in the
OSFLAG reportslop".
Short-term
2.3.6.1
the statistical
In the short-term
Section
2.3.2.1,
particle
Probability
all
distribution
follow
6.
and
u
mean-zero
of the basic
to Eq. (2.3.4S)
memory non-linear
process involving
variables
remain
kinematics
With reference
on the values
conditional
constant,
properties
it
transformation
the statistically
The probability
is
Gaussian
force
the
of the mean-zero
structure
distributions.
probability
variate
seen that
stationary
and
loading
is
bivariate
and independent
a zeroGau3sian
random
of F may be*obtained
from
I-
33 -
the biVariate
(u. A) by applying
of
pdf.
(see Sections
random variables
1.3.2.9
Appendix
of
II
One),
yielding:
00
PH (F)
where
6=
- 21T K
(F -kD
1uu
-CO
from
ulul)/k,
.2
[--T
2- CT +I c.
uu
exp[Iuu
(2.3.47)
du
(2.3.45).
Eq.
The above expr ession was developed by Holmes and Tickel 154, the pdf.
,
in a different
form by Pierson and
having been derived previously
'HolmeS2 both expressions
integration
for solution.
requiring
numerical
,
3
by
developed
Borgman
An alternative
was
which involves
expression
functions'.
'parabolic
tabulated
cylinder
The pdf.
by Eq. (2.3.47)
given
or P/H, distribution.
Pierson/Holmes,
and is
herein
work reported
moments being
statistical
order
a mean-zero
Solving
lie,
= 3k I
function
symmetrical
with
odd
all
is
fully
(2.3.48)
odd integer
defined
by its
2+
3k
D2au4
l8k
%.
(2.3.49)
2k2a.
2a4+
105k 4a8
IDuuDu
(2.3.50)
78CFU8
D.
M2 -
k2
expressions
and inertia
records
2.
Holmes
of
n=
44+
M4
These
of much of the
moments:
=k12
M4
to as the
zero:
be shown later,
statistical
M2
It
Mn = E{F"I I=0
and, as will
be referred
hereafter
will
3k
D2
1/2
CrU4]
(2.3.52)
can be used
coefficients
of
(2.3.51)
surface
in
for
the, statistical
Morison's
elevation
evaluation
of
the
drag
from analysis
CM-and C
DI
by Pierson
as described
and
equation,
and force,
III
34
2
is
M2
M4
to
Of
The ratio
M4/M2
Gaussian form,
There
is
(2.3.53)
This parameter
Standardised
loading
or
further
on the
in
since
should
is,
where
T)---1[1 27T
for
applicable
the
applies
pdf.
be considered
E{y
Tz}
2 JHI/3,
describes
that
2-
cy
to
TzI
be conditional
to
, Ij
a2
(2.3.54)
dj
14
is plotted
17, prepared
(2.3.56)
(2.3. S7)
Gaussian
rise
Clearly,
random variables.
pdf.
from Reference
either
short-term
Tz J)2 11/4
by its
of y is defined
1"1/3,
{y4
TzI,
E
respectively.
and
the
The P/H'distiibution
taken
-00
3cr
when considering
E{y2lHi/3,
22
(Y2 }HI/3,
3(E
78
Tz}-
independent
statistically
apparent
to
14117
becomes
y
:
random variable
f exp{
67
cr.,
2=
ll
1
Eq. (2.3.54)
appropriate
(2.3.55)
E{y4 I'l/s,
cr2
therefore,
111
2 '2 Y-
2=
a proposition
directly
co
PHz(yIHI/3,
load5s,
wave
It
show a
response
and TZ.
In standardised
the
of
form
Also,
it
structural
Five.
a general
conditions
Of H1/3
that
pdf.
of
Chapter
variables.
sea state
values
to
in
P/H pdf.
response
stationary
distributions
behaviour
the
represent
the Pierson/Holmes
the
that
evidence
to be investigated
to
for
notation
probabilistic
similar
0, and measures
of the pdf.
the peakiness
0'
to as the kurtosis,
referred
second
in the standardised
by the author
It
and fourth
is
also
moments,
form in Fig.
on Gaussian paper
2.3.4,
such
(note
3.0)
line
that
Gaussian
plots
as
a
straight
the
pdf.
that
Ifil/3,
2
demonstrates
how
Tz').
This
the probability
E{y
clearly
E
graph
a2
,
be
by
Gaussian
high,
levels
y
can
underestimated
of
of,
of occurrence
assumptions.
35
distribution
Probability
of peak values
following
of random variables
the
P/H 2df.
it
In design applications
is the probabilistic
descriptions
of the peak
and extreme values of either load or response, rather than the basic
Pistributions
that are of most importance.
variite,
of peak values may
be transformed into the pdf. s of stress range required in fatigue analyses
to check the long-term integrity
of structures
and they also form the
input to the distribution
of extreme values necessary for the assessment
excursion failure.
of risk of first
distributions
generally
loading
where the
inertial,
wave loading
In the past
process
to facilitate
extreme
loading.
However,
under-estimationIz,
as will
variables
Appendix
y having
investigations
these
into
cross-sectional
wave
area
by S6dinge.
conditions
form,
where drag
to an equivalent
the distribution
for
Gaussian
of peak and
assumptions
be demonstrated
the probability
structure
of the peaks of random
is given by Lin 45 and Bendat 56 and is summarised in Section 1.3.4
being presented below for variables
One, the important
results
a narrow band spectrum.
covering
The narrow
band assumption
be reasonably
well
from analysis
For the narrow
wave loading
y is
often
considered
2.3.2.1.
This
as the input
to
data.
band (type-1)
process,
peak values
Tz)=,
acting
would appear to
to be narrow banded, as
5
is
by
115
Ticke
proposition
supported
(see Section
I -
1.3.4,
Appendix
One)
(maxima) is:
E[N+(Y)IHI/3,,
Ppl(YIHI/3,
or response
since
of prototype
of positive
for
founded
the cdf.
Morison
are significant,
linear
long-term
of the non-linear
effects
The theory
2.1.1,
see Fig.
by Ochi4 0 and for
conditions
short-term
distributed
to members of large
applicable
is
of Gaussian
E{N+(O)IHI/3,
Tz
(2.3.58)
36
-I
co
TZI -f
E{N+'(Y)IHI/3,
JHI/3,
E{N+(O)
and
level,
If
y=0,
TzI
with
is
JH1/3,
p(y,
0z
the expected
T)
rate
termed
the type-2
PH(YIHI/3,
p2
(YIHI/s,
TZ)
(1.82),
from
Eq.
and,
The great
Appendix
a simplified
peak cdf:
T
Z)
TZ)
Tz)
(2.3.60)
Co
f
2 approximation
is
effort
for
Fig.
2.3.5
plot
which
2.3.5
of Fig.
show similar
its
(0 = 3.0)
fatigue
The expected
are discussed
expressed,
E{OIHI/3,
following
in
2.3.4.
since
In the case
the narrow-band
the Rayleigh
pdf.
damage
value
for
paper
The
of y is made in
to Fig.
characteristics
1 approach,
evaluation.
peak distributions
the plots
Gaussian process
Expected
of the type-2
represents
the type
over
densities
it
that
(2.3.61)
Tz)
P(IHip,
in computational
only
independent
one:
IHI/3,
('y
PH
reducti6
considerable
PH(o IHI/3,
of the type
advantage
requiring
=1-
E{N+(*)1H1/s,
y
of the mean
slope.
positive
results,
(2.3.59)
of up-crossing
expression
the type
Tz
time,
2 peak distribution,
c-1
+
E[N (0) Ifil/3,
D', associated
6P45
as
:
with
y may be
CO
TzIfYb
-00
dy
(2.3.62)
b
the characteristics
representing
of the
are constants
and
c-1
where
fatigue
to the Palmgremmaterial
according
of the structural
properties
57
for the expected damage represents
Miner criterion
of
unity
value
.A
the failure
condition.
i
37
Long-term
2.3.6.2
Basic variate
The long-term
cdf.
of y is
distributions
with
the long-term
bivariate
of
pdf.
(
p (y)
(H
of the short-tirm
data,
wave climate
in the
form of a
Tz):
1/3'
00
(ylill/3,
=ffPH
00
nmw
P(Y) -EE
i=l j=i
nm
where w=EEW
i=l j=l
TZ)
diagram
scatter
becomes, with
equation
by convolution
obtained
reference
[PH(YIH1/3i,
P(Iil/j,
Tz)
d111/3 dT
representation
to Section
Tzj)
(2.3.63)
of wave climatethis
2.3.2.3:
(2.3.64)
ij
Peak variate
Convolution
of short-term
Section
into
In this
follows,
the type
domains is provided
E{N+(0)1H1/3,
ff
(y1H1/3,
Pp2
Tz)
'2
00
CoCo
E{N+(0)IHI/3,
T1
where E(N+(0)1 =ff
00ZZZ
supported
surface
by S6dinge,
elevation
E[N+(O)IHI/3,
by E{N+(O)IHI/3,
domain
Tz) as
2 process:
p
p2(Y)
by assuming that
00 00
An approximation,
TZ)
TZ1
E(N+(0)1
= 1/Tz
dT
(2.3.65)
reduces
by the reciprocal
TZ) dHI/3
T )-dH, /3 dT
p(H1/3,
rate
p(H1/3.
the computational
of y is
of Tz,
(2.3.66)
effort
for
thus:
(2.3.67)
II
and
E{N+(O)}
=Tz -1
weightings
those
the convolution
applied'in
are,
to
assumed to be identical
of wave heights
and. for discrete
wavetherefore,
data: '
climate
Ii
$1.4
$
$ $.
- 38 -
nmW.
E
P
p2(y)
The implications
of this
El
j=
a
T ].:
T
zj
zj
IPp2 (yIHI/3i,
(2.3.69)
are discussed
approximation
in Chapter
Three.
Extreme variate
The distribution
of extreme
values
may be derived
of Gumbe158.
by
the theory
the
applying
peaks
of
independent,
be
the cdf.
to
assumed
N peaks is
p
This
given
peaks of y are
or extreme value In
all
of the greatest
by:
[pp
(Y) ]N
=
EP(Y)
may be used to assess the risk
distribution
of y during
level
particular
passage of N cycles
Expected
for
in
short-term
fatigue
to the
equivalent
failure
analyses.
excursion
first
time
to produce
given
by Eq. (2.3.62)
the estimate
CO00
E{D'I=
E{DjIII/3,
tff
TzI p(HI/3,
TZ) dIII/3
00
A less
using
strictly
of a
of exposure
a period
application
of exceedence
Damage
Fatigue
The expected
(2.3.70)
dT
of mean
(2.3.71)
justifiable
the
applicable
to the non-stationary
long-term
behaviour:
00
EJD'J= t E[N+(O)l
c-1
YbP
p2(y)
dy
(2.3.72)
FIG. 2.1.1.
20
\. I#,
I,
WAVE
HEIGRY 3
METERS2
1.0
L
1.1
BRACING-
JACKET STRUCTURES
LEGS
8-5
"23&1
10
20 )o #ASO 100
DIAMETER METERS
FIG. 2.1.2.,
ZA
v -16stantansous
vertical velocity
component.
%kb
VR \FU2'.v2-1
Instantaneous
000000'O'/ resultant water
particle velocity
vector,
(,3;0rY0.
Zo)
V
I\N
u- Instantaneous
horizontal
velocity
component
Aj-
I.
II...
I,
,:-;."
41
I
I
Direction
Propagation
of
Length
Crest
st ill Wter,
L'evel
--F-
S.
W.
*
*'<Trough
V,
Water
Depth
d
U,U
(X, Z)
NOTE
I: uv- Water particle velocity components
Water particle acceleration
I
components
ed
-x
FIG. 2.2.2.
WATER PARTICLE
DISPLACEMENTS
FROM
MEAN POSITION
FOR SHALLOW-WATER
AND
DEEPWATER
WAVES FOR FIRST ORDER WAVE
TH50RY.
2A
S.W.L.
S.WL.
I ---
KoSitioh
PA
D,
Orbits
1 Elliptical
I
A0B
-,
Z,
0,
CircularOrbits
A
u
if
Bed z =o
Bed z0
0o
0
u0uo
Shallow-water
wave
or
Transitional
wave
-water
d<I
V=o
Deepwater
d
;;,
L2
wave
I
42
-
II
a
wa
I.u0
IL
CY
c4
'A
2i u
0:
tn
th
*0
in
m
PN Co
e4 0
Ix
W0
30-
rx
An
Z
CO
, .
CO
in
Ia:
0q
WW
U.
In
in
>i
cr
M,
%0
,q
C)
W
z
0up
w
CL
04
ef
z
C4
CO
IN
at
cc
w
(L
CV in
N
co
tot In
m -
r%
04
p: cr
,a
IN
Cb
U?
>-
C4
wo
in
10
",
"
kI
1
W.
Ol
w
ul
U)
cv
"I
(y),
z0
=)
w
w
W9
C01
-3
40
-n
0
I
C4
1 01
I ol 'n
C7
0 45
qb
w
IL
-1
C4
i...
0,
"1 1 iz r%
It
911 co
I
in 'T X
in mto
47610 -
lw
<
C4
CM
cc
AI
0
'o
in
in 1-I
I-I
N -M a U?
16
I, le
43-
FIG. 2.3-2.
60'
59
Wick
58*
7Aba-,
FAM
FAMITA
57
rdoon
KI
DK
D
je
ja
EE
bb
a
55
53
58*
GB
54*
Traschalling Blk
t .04
52*
1*
VS
S
45j5
VS
ISV
VSV
ViVV*
"V
V
p "V'
VV
"5
Sh0S$'V
5
? VV
VVVSVt
VVVVS
VS
I
SV
N,
44,
woo
go
II
0
z
k-
RLW
I -- -I-
D
m
CL
I. -
J
u
Ljr
-,
(TIVOS 001)
40 to
Wt
co ID
%A6
'A
4S
CUMULATIVE
PIERSON/.
HOLMES
DISTRIBUTION
(c df 's), PM
FUNCTIONS
N
6.0
% " , Ll-,
5- 5
o
.,
5-0
4-5.
4-0
3-5
b
b
%. 0
ca
3.-0
2-5
10-3
10-4
(a)
8.0
- 1111
10-5
0?
O)t\(o 46
2-0
i-5
7-5
1-0
70
0-5
-5 -4
,
-3
-2
ti I It HJ.1
10-6
7Y.
10-5
10-4
104
-02
-002
-01
-05
-1
PROBABILITY OF EXCEEDENCE
ON
46
7rrrriTr.!l
CL
1111i1111111il
III I iijill 1HUM: 9!
: Go
0
C3
CE
u
CAI
Z0
ck:
site
VL
w
w
D -. i
U-
rsi
--
-d3S
>
-
I--j
Z,
0z
PO
V)
cr.
>
I
=I
(A
Z
a
W
w0
a:
'0
-0
IN
c
LLJ
W
41
4
,)
><
w
W
> LL
0
4,n
fIfII
Ii
-I
Co
<
0
7: :i 2
- CO 0
E
IiII
:,
-1
In
tl:l
4w,
CP
C\i
r-C\j
13 U
C
Cl
to >,
CL
4., 0
to
in
C3a-
Wp-
%-
2
U
C
3-
w
W>
. 0. 0
-D
=,
-j
2w
LL
Lq
0 C L
'n
-it $fill
-,
Ln
cm.
if,
Cm
allVaNVIS
4, -on
-'r;
sm
Ln
9,. do NOUVIA30
cm
' TISVIM
(13SIOUVONVIS
o
21
0.! 0
>&n
47
969,
roll 1qC96: IIIm
4
C9,9.
J.
-L
LL
LO
0-
LO
is-sgo
tail,
1z
-,
13
",
-I:
La
111
U")
Lc)
LO
04
..
ol(,
*C
ql.
e
O-VO.
E%
110000
0
1111
IN in niq M,
111190.
.........
ea9z
111111111 III
lift lll!lfl 11111
U. 0o
Lo
d)
o
6)
U) o
Cb Cb
to
r.
0
t
U)
tb
48 CHAPT
PROBABILISTIC
E'R
investigation
In the following
into
of cylindrical
members application
'splash-zone',
or region
is,
therefore,
The effects
of surface
and is
by the author
restricted
to locations
wave activity.
Within
that
This
of uni-directionl
is
considered
loading
in Reference 11.
reported
a number of inadequacies
were found in existing
study
by the author
Before
it
proceeding
this
be emphasised
should
made in this
predictions
Chapter
is
restriction
and his
are for
in all
that,
the
the splash
and the
has been
effect
A further
to act
on still
currents
fluid
11.
sections
beneath
of the time
in Reference
reported
characteristics
for
is
length
on unit
a proportion
of an intermittent
nature.
wave loading
submerged for
THREE
colleagues
that
the
In this
latter
which have
59
Liverpool.
at
theory
long-term
comparison
the computations
wave load
purposes
only.
The reason
herein,
performed
is assumed to be represented
by the Famita
the long-, term wave climate
diagrams of Fig. 2.3.1.
In Chapter Four it will
be shown that
scatter
for the prediction
of extreme events it is necessary to extrapolate
the
into long-term
descriptions
to convolution
wave climate prior
of wave
and wave load.
height
For this
in the
'one-yeariprobability
reason
following
(the
level
or is
exceeded,
distributions
probability
on average
will
be truncated
associated
with
at
a wave
3.1
DDnDPDTTPC
The objective
the
of this
loading,
given
is
Section
to investigate
distributions
loading
mechanism,
typical.
which yields
the departures
between
'Morison'
of the non-linear
wave.
Eq. (2.3-47),
and a linearisation
of the
for
a Gaussian pdf. of wave loading,
conditions.
49
Linearised
past
methods of analysis
for
by Naval Architects
the response
as indicated
are also
applied,
analysis
techniques9
925.
in Section
Procedures
for
of ships
2.3.4,
in the
satisfactorily
the prediction
of the extreme
values
of including
or otherwise
certain
conditions.
3.1.1
The Linearised
in the procedures
the non-linearities
The basic
under
may be linearised
/58
uJul = U(al i7)
u
which
results
tical
sense.
from minimisation
The force
kIA+kD
Clearly,
of the
this
as:
(3.1.2)
over-estimates
it
In the short-term,
f follows
when u is
of the Gaussian
7r
(T
u
4+k
the Gaussian
variables
12
CrCL2
is:
Z2-7r a
xp{-
f2
,
fl
the
when u is
loading
small
large.
a2=8
f
in a statis-
au u
approximation
kb
error
/8
and underestimates
combination
'mean-square'
pdf.
since
Also,
u and 6.
independent,
for
it
is now a linear
since
u and d are
a sea state
(3.1.3)
so
-
distributions
The long-term
linearised
of this
loading
manner as described for the non-linear
Eqs. (3.1.3) and (3.1.4) are strictly
conditional
and Tz.
Methods of Computation
3.1.2
was developed-for
distributions
tion
The program is
Six
for
of both
the computation
of non-linear
wave loading
In all
for
essentially-a
in the same
in Section 2.3.6.2 where
on the values of HI/3
a list
using
of computer
programs)
in Fig.
representa-
herein
2.3.1
of an existing
probability
diagram
a scatter
considered
given
refinement
Considered
and long-term
short
applications
Famita
follow
loading
the
been used.
has
developed
program
by others
similar
replaced
In both procedures
by the Gaussian
the surface
linearised
the
for
the algorithm
has also
being
of the short-
estimation
Eq. (3.1.4).
pdf,
it
loading,
This program
10.
spectrum
elevation
describing
the sea
intensity
state
for
defined
in
different
The
the
current
practice.
effects
of
use
of
adopted
often
distributions
forms on the long-term
of loading are investigated
spectral
in the following
section.
Computations
in 150m. of water.
surface
the locations
of existing
in water
To avoid
the effects
This
here.
surface
data
offshore
of intermittency
a minimum depth
represents
variation
depth
water
structures
increase
lsplash-zonel,
This
greater
although
with
that
the effects
at
of an
considered.
3 standard
comparable
of immersion
than
is
deviations
condition
be submerged for
present
greater
than
99.7% of the.
51
-
sea state
computations
these
of CM - 2.0 and CD "0.
to in Section 2.1.
range referred
coefficients
acceptable
being
values
The effect
with
within
Morison
the
of change in the
been investigated.
has also
coefficients
Conditions
Short-term,
3.1.3
loading
of the
3.1.1
and 3.1.2.
With reference
linearised,
procedure
2]
[aF2_af
and (3.1.3)
= (37T - 8) KD2
clu
it
is
the
by:
4 /7T
(3.1.5)
loading.
of the non-linear
variance
.
appare nt that
of loading
the variance
underestimates
M2 is ihe
where aF2
(2.3.49)
to Eqs.
results
in the loading
loading
dependent
of 10%.
is only
equations
However,
Cr
of 3.0.
long-term
4+
l8k
J2
2 cr
k2
{78/(3
+
-D
is
from Fig.
3k
deviation
of load
3.1.1.
loading
(2.3.53),
4+
(1
k22
Cy.
U
standard
The non-linear
by Eqs.
given
(k
3+
circumstances.
has a kurtosis
a-z2
in the
variation
of kurtosis
3k J4
the underestimation
considered
pdf.
these
as may be inferred
marginal
by a value
under
attached
weighting
is
represented
loading
(2.3.49)
105k
in Fig.
following
and'(2.3.50)
3.1.2
the P/H
as:
48
CY 4)2
)21
(3.1.6)
52 -
cr
0
and
-+ 3.0
For
drag
dominated
kD2au4
loadinj
>> k12a62
and a -o-,11.67
it
can be seen that for the sea state and member conditions
in excess of eight are obtained.
The
considered here values of kurtosis
kurtosis
measures the importance of the drag component of loading and
increases
under the same conditions
which were shown above to increase the
3.1.2
From Fig.
of load variance
underestimation
of the differences
The significance
reference
to Fig.
2.3.4
loading
may be represented
deviation
For example,
for
same conditions
The departures
Fig.
to 0=3.0
3.1.1,
the appropriate
with
line
with
is
reduced
ordinates
by an amount given
of
value
distribution
loading
wave
of non-linear
by the square
the underestimation
representing
of
of load.
a 0.5m.
with
diameter
member at a depth
= 9.3m.
HI/3
of immersion
the linearisation
of 7.5m.
underestimates
41% and
of exceedence of 10-3 by approximately
increasing
of 10-4 by 50%, the underestimate
at
rapidly
lower probabilities
basic
obtained
at a probability
at a probability
linear
the cdf.
by the straight
from Fig.
the standard
loading
The linearised
corresponding
of the values
figure
by the curve
figure
on this
represented
and kurtosis
of wave loading
On this
3.1.2.
model.
may be appreciated
P/H cdf. for
shows the standardised
which
of kurtosis.
values
from Fig.
in variance
on predictions
various
kurtosis
in the linearised
the underestimate
is
13% at the
and linearised
loading
For a 2.0m.
of exceedence.
procedures
characteristics
of exceedence
level.
10-4 probability
of peak loading
show similar
at low probabilities
diameter
with
to the
reference
to
L3.5.
However,
at probabilities
linearised
procedures
of exceedence
predict
loads
greater
in excess
than
about 0.02
of those
the
53
-
This behaviour
method, the effect increasing as the kurtosis increases.
implies that linearisation
of the loading overestimates fatigue damage
associated with the small waves.
Conditions
Long-term
3.1.4
The long-term
cdf-s
for
variate
the basic
is,
-attention
therefore,
P(F)
P (F)
p2
For the
s'ince
the conditions
present
is the long-term
on the figure
(3.1.8)
for
being
in the
long-term
different
given
observed
long-term
for
are plotted
resulting
in Reference
10.
in the short-term
properties.
by
are replaced
loading
show the
(3.1.7)
E
I j=l
values
to:
reduce
densities
method the P/H probability
values in Eqs. (3.1.7)
and (3.1.8).
short-term
by HI/3
Z-
the characteristics
In Fig.
and
a more
These plots
demonstrate
properties
of loading
for
the smallest
3.1.4
there is only a
are most pronounced,
member where drag effects
between the peak loads predicted
difference
at low levels of
diameter
PH (OIHI/3i)
presentation
marginal
are defined
conditions
IT
analyses-an4
variate.
m W..
nm
Z -LIP *WZW.
PH (FI111/3j)
peak loading
are retained
load
to 3.1.6
3.1.4
diameter
on this
(2.3.64)
Eqs.
[P (F1H1/3i)
of the short-term
linearised
that
here,
member for
Superimposed
detailed
fatigue
in both
concentrated
=E1
i=l
3.1.3
a typical
Figs.
'type 21 peak
the narrow-band
is of-most practical
value in
linearised
the Gaussian
In Fig.
for
as considered
only,
(2.3.69)
are obtained
of peak values
input
forming-the
design,
tion
and Eq.
The distribution
variate.
loading
of non-linear
the differences
exceedence the
For example,
of exceedence).
For members of
larger
from Figs.
3.1.5
linearised
at a probability
significant.
method considerably
of 2.2 x 10-7,
At low probabilities
underestimates
corresponding
to 3.1.6
of
the loading..
to the
lone-,,
54
-
is
the
underestimation
approximately 48%, 40% and 15%
condition,
year,
for the O. S, 1.0 and 2.0m. diameter members, there being negligible
differences in the case of the S. 0m. diameter member.
of the loading
Linearisation
costs
by virtue
of run-time
results
in considerable
requirements
in computing
saving
10% of those
of approximately
for
follows
because
loading
This
the
computations.
non-linear
required
the basic loading pdf. is Gaussian in the short-term
and hence its values
from Eq. (3.1.4),
loadwhereas for non-linear
can be obtained directly,
Eq. (2.3.47),
requires
integration
numerical
its
for
solution.
Variation
3.1.5
Fig. 3.1.7
in Morison Coefficients
of changes in the Morison
member at an immersion
1.0m. diameter,
of 15.0m.
for
coefficients
The pairs
of coefficients
(2.0, O. S), (1.5,
(2.0,1.0),
C
CM
are
and
respectively,
considered
D'
these values being representative
1.0) and (1.5,0.5),
of a range of
From the distributions
design values.
method it is
of the non-linear
in
is
drag
the
that
critical,
changes
coefficient
at extreme values
clear
for
the inertia
linearised
being
coefficient
method,
significant
in the
lower
range.
are less
as
critical,
the large
waves
in Water Depth
Variation
3.1.6
In the
for
An increase
at depths
respectively,
indicates
water'
that
the majority
and suggests
presented
of immersion
also
in subsequent
of
(7.5,1S.
of sea conditions
that
the results
sections
for
considered
presented
).
0,22.5m.
This
were in
here,
depth,
level
'deep-
and those
are applicable
55
1
3.2
Spectral
3.2.1
investigated
fitting
in Eq. (2.3.23),
defined
(P-M) spectrum
shown in Fig.
the
3.2.1.
This
measured spectra
was developed
spectrum
from fully
developed
as an 'average'
wind generated
curve
sea states
basic
P-M
In
limited
the
wind
speeds.
existing
range
of
practise
a
over
for
description
is
the
the
accepted
model
of shortmost
widely
spectrum
in non-deterministic
Although other
analyses.
term sea conditions
62
for
by
Neumann",
Darbyshire
forms
exist,
example those given
spectral
63
in engineering
little
they
and
applications
are
used
and Bretschneider
,
been omitted here.
have, therefore,
Pierson-Moskowitz
Basic
3.2.1.1
in its
The spectrum,
in Fig.
sea state,
reference
a scatter
is
a function
moments, Uw
the short-term
parameters used to describe
With
Hi/3 and Tz, through Eqs. (2.3.24)
and (2.3.25).
it is evident that Uw may be derived from
to these equations
to the statistical
Consequently,
or Tz.
Hi/3
either
form,
However,
3.2.1.
is related
basic
spectra
diagram,
for
any particular
class in this
a single
representing
PM spectra
distinct
the values
of Hi/3
(2.3.25).
For this
with
developed
and Tz yield
condition,
wave climate
data
in the
form of
short-term
diagram,
parameters
unless
(2.3.24)
and
as follows:
21T2 HI/3
2
zg
Tz
This
(A7r) 1/2
A113
3-55
relationship,
sea states
In practice,
characteristic,
if
superimposed
the actual
sea states
scatter
(3.2.1)
in S. I. units
surface
deviate
on Fig.
2.3.1, 'should
be satisfied
is
all
elevation
spectrum
from this
theoretical
form yielding
relationship
between HI/3
in the bi-variate
for
the
"
Tj.
and
56 -
Definink
to
'characteristic'
the
a deep water
wave length,
HI/3/1.56T
Eq. (3.2.1),
wave steepness,
s, to be the
2/27r,
we obtain:
-Lo = gT z
z2
represents
present
of seven stations
the steepness
limit
to lie
majority
will
Eq.
steepnesses
(3.2.1).
diagram
much lower
based
classes
wind
on Tz will
whilst
1/11.6.
Battjes
to Fig.
reference
the observation
that
diagrams from any offshore
than
speed
those
required
predictions
on average
7 found
infrequent
only
is
in scatter
Consequently,
with
These constants
represent
Isles,
with
is
diagram
scatter
the British
around
between
of 1/19.67
a wave steepness
in the Famita
From analysis
The significance
lil/3
of
(3.2.2)
therefore,
ratio
exceed
the vast
location
to satisfy
for
short-term
the
values
scatter
predicted
by HI/3*
purposes it
is
to consider
only
sea state
severity
relative
to the
latter
procedure,
as
An alternative
to the above procedures,
the basic P-M model,
which retains
(2.3.24)
(2.3.25)
function
to
Eqs.
the
yield
and
wind
speed
a
of
as
uses
the ratio
of Hi/3
1/4
3B
IT
U=-=
and Tz
TTz
IA
The surface
which permits
spectrum
HI/3
Hi/a
24.32
(3.2.3)
a more complete
description
function
of HI/s
and Tz
to be
57
t
developed from the scatter diagram. This approach, would appear to have
ion since the same spectrum is predicted from,
lit tle physical justificat
'
for example, a sea state with HI/3 = 4.0m. and Tz = S. Osecs. as from
HI/a = 8.0m, TZ = 10secs. Consequently, it is not pursued further here.
of the adoption of any of the above P-M models in
in Fig. 3.2.2 showing the cdf.
to the others is illustrated
The implication
preference
tion
is
it
the P-M (T
for
that
clear
diagrams
of
approach.
z)
descripwave climate
by the other
wind speeds in excess of those predicted
Furthermore,
sea state intensity
may be measured by the
predicts
methods.
given
of surface elevation
the P-M spectrum is obtained from Eqs.
variances
cr
fill
procedures
as:
ir-
magnified
are considerably
descriptions
for
(3.2.4)
departures
Consequently,
and (2.3.23)
which
uW
Ag 2Z=0.00274
4B wo
2=
(2.3.12)
(w),
intensity
of sea state
in the resulting
which
statistical
into
wave loading
computations.
to the Pierson-Moskowitz
DNV modification
3.2.1.2
the complete
utilise
design
in their
wave climate
information
HI/3
(to)
Snn
It
is
(w)
11TI
where'a
---87rl-
-5
&
27-T
-]
exp{-
to compare this
instructive
by expressing
[.
TZ T?
both
spectra
ab
= 7, - expl-
for
structures
the description
of wave
referred
f. Tzw.1
(3.2.5)
IT L2 iT j
function
with
the basic
P-M expression
in the form:
(3.2.6)
-,,,41
offshore
by
recommending a modification
conditions
by:
(DNV),
P-M
here
given
to
as
2
for
rules
spectrum
thus
demonstrating
that, both
spectra
are-of
sa
-
firstly
Concentrating
expression:
4
B wo4=B
(3.2.7)
-UFL--r
w
and substituting
41
B, B7r
P]
E27f
2jr
UP]
:F
(3.2.8)
zz
This
expression
Constant
identical
is
to the value
given
in P-M (DNV).
coefficient,
2
3
a=[1.4Tr
T4
z
Substituting
again
2
HI/3
(3.2.9)
Tz:
B4
(3.2.10)
U-,
w
for
and substituting
for
HI/3
yields:
Ag 2
which
It
is
(3.2.11)
the coefficient
be inferred
may, therefore,
P-M spectrum
that
P-M (DNV) is
identical
to the basic
Eq. (3.2.1).
It
can
densities,
be easily
shown that the spectral
and hence the values
',
(DNV)
be
from
less
P-M
than those derived from P-M (Tz) for
a,,
will
of
diagram
for
the
scatter
within
which the significant
wave height
classes
Eq. (3.2.1).
Since this condition
is less than that satisfying
on the
also
value
of H1/3 is
earlier,
it
for
satisfied
follows
that
the majority
sea state
intensities
less
3.2.1.3,
JONSWAPspectrum
severe
as discussed
of the classes,
will
on average
greater
predict
than P-M
(lil/3)-
65,66
demonstrated
the existance
The JONSWAPprogram of research
of a
in
from
departure
the
the P-M form for fetch
shape
spectral
significan
,t
from situations
This conditi. on results
limited
sea states.
where the
generating
distance
for
the'fully
59
to be reached.
developed
state
contained
in each frequency
further
to the equilibrium
return
fetch
a maximum 'equilibrium'
wave steepnesses
causing
value,
wave breaking
and
condition.
dependence a modification
To account
for
developed,
in
Fig.
shown
as
developed
band possesses
increases
input
energy
In fully
of which
The JONSWAPspectrum
was
an enhanced spectral
is
a five
peak,
function,,
parameter
as:
expressible
ag
(W) =
TIT1
WS
-4
exp
I]y
(w-wD)
exp [- 2a26)p
(3.2.12)
aa if W4 wp
7 ab 'fW " bjp
Wp is the frequency at the spectral
where a=
and
In Eq. (3.2.12)
by the inclusion
describe
of a variable
is
a replacing
equilibrium
is unchanged
function
the
exponential
of
density:
P-M spectral
4B
wp=W0 (t=)
parameter
in Phillips
'constant'
the
peak
since
0.0081
dependent,
of the
1/4
[= 0.88wo)
(3.2.13)
The remaining
to
The argument
by differentiation
Parameters
cr = 0.07;
3.3;
modified
the best
estimates
the JONSWAPspectrum
being 65
gXf -0*33
(2iTg
U
with
-21
variation
(3.2.14)
Crb '3()'09
defining
parameters
peak as shown
are fetch
(3.2.1S)
.pWw
and'
Xf
9.076{g
-0.22
-1
U1117,
w
'distance
in
direction
fetc,
the
sea
open
of
x
where
=
f,
.
available for wave generation.
(3.2.116)
of the wind
60 -
spectral
W, 22.036
5.78(HI/3[
Using this
2ff
IPT-1
the value
through
'referred
value of HI/3
for,, a given
Eq. (3.2.17)
satisfy
is
(2.3.18))
which
developed
in
this
HI/3
is
value
i)
of
initial
ii)
iii)
V)
if
if
11-
agreement
for
study
for
below.
as explained
is
iterative
beneath
with
the
specification
(w),
TITI
original
of
producing
using
Eqs.
since
which
an
(2.3.12)
and
The algorithm
value.
the
each short-
spectrum
for
llpp_M.
here
a given
as follows:
of peak frequency,
from Eqs.
s ctrum
4cYTIT,
(3.2.13)
and integrate
4 0.01,
, pp_M is
0Wp=1.
and (2.3.24);
spectrum
to
2.
Cr
nn I
obtain
satisfactory
since
it
yields
an
HI/31
1% of the true value;
estimate of Hi/3 within
(iv) not satisfied
obtain a second estimate
of peak frequency
(assuming a to remain un-changed):
from Eq. (3.2.17)
ww[
repeat
4crTITI
4.072
(v)
until
P1
p2
vi)
can be generated
as JONSWAP(HI/3),
specify
area
(3.2.17)
calculate
calculate
iv)
in
the
estimate
obtained
values
and which
(from
HI/s
of
estimate
of HI/3
to hereafter
The procedure,
(yp))
0.2981n
-
the JONSWAPspectrum
expression
(1.0
(i)
to
convergence
is
achieved.
developed
the additional
of a departs
directly
approach,
by virtue
beneath
the spectral
peak, unless the value
contained
area
from A =, 0.0081'when the reverse may be true.
considerably
is,
to show that, the JONSWAPspectral,,,
the
comment
above
of
frequency
in
domain
bodily
the
to, the
be,
relative
shifted
may
The purpose
densities
significant
of
frequency,
P-M (HI/3)
spectra.
61
-
By analogy to the comments made with respect to the P-M spectrum a similar
formulated
for
be
JONSWAP
based
the
spectrum
could
on Tz
procedure
in
However,
the absence of an equivalent expression of
rather
Tz to a and wP based on data analysis
the form of Eq. (3.2-17) relating
impractical
is
is
and
not considered further.
such an approach
than Hip.
from Surface
Transformatibns
3.2.2
input
Wave' field
in the Section
beneath
the respective
of Eqs. (2.3.7)
tions
'deep-water'
depth
conditions.
15m/s. for
With reference
P-M spectra
Furthermore,
are likely
Although
considered
wave properties
for
water
Fig.
the splash
zone,
of this
Chapter,
shifted
to a higher
since
for
of
portion
a negligible
frequency
'deep-water,
'deep-water'
3.1.6.
'deep-water'
transfer
functions.
sketched
purposes
and their
use in
for
demonstration
provides
an appreciation
all
inline
with
will
correspond
of the process
of
are convergent.
frequency
range than'those
decay in magnitude
are greater
approximately
are below'w'=
than
for
functions
for
elevations
functions
velocity
the
constraint
the
from consideration
functions
(or
transformation
Excluding
3.2.2
'deep-water,
energy
outside
of w=0.45
representing
for
of
conditions
depths,
3.2.1
this
limit
'deep-water'
are sufficient
with
conjunction
within
this
here are
3.2.3
little
from Fig.
since
to satisfy
to intermediate
in Fig.
is
there
the
by a frequency
3.2.1,
to Fig.
of n) contained
variance
conditions
than
of w greater
3.2.3.
herein
in the figure
radians/sec,
as the area
are obtained
in Fig.
considered
values
states
the
are sketched
represented
range.
takes
developed
and (2.3.9).
d/L = 0.5 is
the total
loading
of
These variances
3.1.
spectra,
conditions
150m. water
For the
Kinematics
2,
the variances
of %2 and a,,
estimates
of water
in Eq. (2.3.47)
as described
and
and acceleration,
velocity
discussed
to Particle
description
to the probabilistic
form of short-term
particle,
Elevation
Those for
for
velocity.
exponentially,
with
acceleration
are
acceleration
The
depth
and
of the peaks
k.
62
-
This departure
made in Section
From thecomments
and with
of u and fj cause
in the variances
variation
3.1,
reference
to Eq. (3.1.6),
drag
inertia
loading
the
and
components thereby modifying
of
magnitudes
A
through change in kurtosis.
the form of the loading distribution
into the significance
investigation
detailed
of changes in wind speed and
,
between
drag
inertia
loading
the
relationship
on
and
elevation
member
here.
illustrated
However,
is
these
pursued
are
not
effects
components
in the following
forms on long-term
spectral
from Different
3.2.3
The results
described
P-M (HI/3),
OSF4 for
of the consequences
account
here
were
P-M (Tz),
obtained
OSF8 for
Spectral
from
computer
P-M (DNV),
Models
programs
OS10 for
OSF2 for
P-M (Hi/3,
Tz)
JONSWAP (111/3)-
OSF2
in
is
based
Section
3.1
on
which
was
used
and
are
programs
being in the algorithms
documented in Reference 10, the only differences
All
defining
ings
term predictions.
Program OSF2 in its
short-term
for
being
conditions
wave loading
identical
convol ution
P (F),
for
is
P-M (T
z)
m
E , [{l
j=j
PH(FIT
and long-term
the JONSWAPOil/3)
approach.
approach,
convolution
An
The
as:
11
w
Z3 11E.
Tzj
p701T
H
zj
only
by analogy
follows
[P (FIT
)E
H
zj
i=l
and
p, (F)
p2
for
followed
mnw..
E
-a
by HI/3
taking
procedure
defined
(3.2.18)
'n
W'.
11]
'w
(3.2.19)
63
conditional
distribution
of
the value
of T
loading
is defined
by,
and hence
of velocity
at a 7.5m. depth
and acceleration
of immersion
is not unexpected
ing to Y=3.3,
takes
variation
a value
HI/3.
with
following
Therefore,
spectrum
is
equivalent
magnitude
similar
Fig.
3.2.1,
and both
implies
that
frequency
for
the
exists
and 3.2.3.
transfer
both
of the transfer
density
variations
spectra
variances
the resulting
spectral
of wp for
this
will
in area is of a
reduction
Fig.
of u and CLwill
falls
are,
frequencies.
3.2.3,
be small
within
therefore,
in
the peak
unless
the region
This latter
close to the peak of S (w).
nTI
iower intensity
sea states,
with reference
The transformations
for
to that
cover
functions,
functions
from the
cover
approaches
the JONSWAP
to 0.005
JONSWAP(111/3) will
will
in the transfer
frequencies
if
little
of A=0.0081
However,
correspond-
showing
shape generating
the value
P-M (HI/3)
This
by the ratio
P-M spectrum.
0.005
of approximately
3.2.1.3,
reduced
the value
since
of significant
condition
to Fig s. 3.2.1
more sensitive
to
departures
2
in
the estimates
The measure of agreement,
Of (J and' a CI from P-M (HI/3)
U2
in
the
JONSWAP'(HI/3)-was
also'reflected
values of variance
and
and
,
'diameter
both
loading
O.
Sm.
S.
0m.
kurtosis
on
wave
and
of short-term
loading
in
Figs. 3.2.4 and
long-term
'The
of
peak
plotted
s
cdf.
members.
3.2.5
also
retain
the agreement
-resulting
with
load predictions
in the higher'values.
differing
-
by less
64
(HI/3)
Having compared -JONSWAP.,
P-M (HI/3)
with
appropriate
it
is
convenient
from
the results
with
and
firstly
on P-M (T ) an impression
of the
.
z
in place of P-M (HI/3)
of its, adoption in wave load prediction
effects
In
of the wind, speed predictions
may be obtained from the distribution
The plots in this figure
Fig. 3.2.3.
show that more severe conditions
P-M (HI/a).
for
Concentrating
the long-term
loading
to the waves.
since
-This
is
Use of
P-M (DNV) in
values
of
HI/3
in
force
the
deviation
between
The departure
and 3.2.5.
for large diameter members as a result
For the 0.5m. member
Of drag effects.
is reduced
place
P-M (111/3)
of
and higher
variance
diagram.
force,
included
in
estimate
from
the
underestimating
level
of
in
the S. 0m.
to 24%.
scatter
of
probability
peak forces at the 'one-year'
by more than 77% whilst
from P-M (Hi/1)
for
excess of those
figure
this
member
to a
3.2.4
insignificancq
) predicts
z
leads
load
the rIelative
This
result
larger
of generally
prediction
dependent
from P-M (T ).
z
waves yielding
greater
wave climate
is
found
values
This
results
Figs.
3.2.4
to
lower
produce
of
kurtosis
in
the
for
most
long-term
of
classes
standard
P-M (DNV)
and 3.2.5,
from
by about
9% in
P-M (HI/3)
mean
both
cases
of
diameter.
member
distribution
long-term
diameter
member, Fig.
'one-year'
drag dominated
Gaussian
loading
probability
level.
and inertia
dominated
conditions.
limit,
Small
is more significant
12% greater
approximately
primarily
loading
about
in kurtosis
takes
for
the 5.0m.
significant
high
kurtosis,
tail
of the'distribution.,
on the other
hand,
in the. region
3.0,
the value
shortrterm
skewness of-the
in behaviour
at
between
is probably
caused
in the kurtosis
of the short-term
here by all
sea states
P-M (HI/3)
conditions
increases
difference
This
than
with
of the.
represented'
diameter
iesulting
member, -cause-a,,
in an increas; ed'
Tor, conditions
of initially
65
Assessment
Critical
Of the'models
most reliable,
design.
This
are favoured
only
they
since
2.3.4.
Models
considered
in terms Of HI/3
of the Spectral
magnitude
sea state
description
are considered
to be the
it
load
to
the
the
may.
results
of
wave
prediction
with reference
be presumed that such a technique would considerably increase the
in
description
heights.
However,
the
wave
climate
predicted wave
little
from
has
domain
Tz
the
received
attention
and
marginal
for
the
techniques
performance
of
existing
apparently satisfactory
based
description
HI/3
height-prediction
the
of
on
marginal
wave
be
it
loading
that
the
surmised
predictions
of
wave
may
climate
wave
based on P-M (Tz) are probably
2.
over-conservative.
any degree
of confidence
has,
of HI/3
from extensive
for
is
for
Much effort
tion
in the bivariate
Such an extrapolation
with
using
attempts
model its
to mathematically
As a result,
extrapolation
in the HI/3
wave climate
developed,
As discussed
in detail
of knowledge.
state
to the long-term
analysis.
data
Tz domain,
HI/3,
the present
been directed
however,
of
acceptable
distribubehaviour
procedures
in Chapter
Four.
No such
long-term
description
Tz
been
has
the
and
of
on
placed.
emphasis
be
little
placed on data extrapolationsin
could
thus
reliability
in
the
of
wave
analysis
program
climate
the absence of, an extensive
Tz domain..
66
Additional
I
computational
required
the greater
a result'of
effort
number of short-term
sea state
conditions
in the
long-term
convolutions,
considered
as may be deduced from
in comparison with Eqs. (3.1.7)
the form of Eqs. (2.3.64)
and (2.3.69)
'(3.1.8).
The computer run-times
and
required
are directly
proporto the number of short-term
conditions
and thus if P-M M1/3)
(HI/3)
16 units
JONSWAP
72
then P-M (DNV) requires
represent
and
This considerable
only 10,, units.
units while P-M (Tz) requires
tional
doubts
on the practicality
such
circumstances.
From the results
presented
it
load predictions
resulting
for
sufficiently
small
acceptable.
However,
engineering
practice
both
since
approaches
this
in wave
and JONSWAP(HI/3)
to be considered
at present,
the differences
that
would appear
as being
are
equally
in
of all
remaining computations.
3.3
The 'type
11 and 'type
were introduced
Section 1.3.4.
exist
ftype
21 models
in Section
for
2.3.6.1
narrow-band
in Appendix
One,
variables.
The great
disadvantage
far
demands
effort
on
computational
makes
for
impractical
itself
extensive
yielding
tions.
However,
results
predicted
the distributions'of
1.3.4.3,
Section
covering
11 model is
in excess
of the
it
that
'type
21 model
usage in engineering
applicait is instructive
between
to investigate
the deviations
Tickel 15 5 has recently
from the two models.
compared
a practical
'type
21 model predicts
range of loading
From these
lower
non-linearity,
results
forces
than
it
the
described
by kurtosis,
the simple
11 model,
which
in
'
67
-
turn was in close agreement with the wide-band model. The level of
kurtosis,
increases
being negligible
in the lower
with
underestimation
range but representing values of up to approximately 30% at some
for
Nevertheless,
the most non-linear
situations.
probability
Tickell shows that the 'type 21 model for structural
response resulting
from wave loading compares favourably with observed values, as will be
further demonstrated in Chapter Five.
levels
of this
The purpose
is to extend
section
All
it
applications
in this
simplifications
the
the implications
conditions.
'type
21 model and
of the inherent
long-term
model on the
preferred
of the narrow
to long-term
wave environment
here to illustrate
is intended
the comparison
distributions.
of the Short-term
Derivation
3.3.1
Iheshort-term
Eq. (2.3.58)
of Peak Loading
of narrow-band
cdf.
Type 1 cdf.
from
as:
E{N+(F)IHI/3,
p (FIHI/3,
pl
TZ
(3.3.1)
TZ)1
E{N+(0)1H1/39
and
Co
Tzl,
E{N+(F)jH1/3,
=f
fp(F,
PIHI/3,
zi 1
TZ)
dP
(3.3.2)
is
of
of F and
The bi-variate
pdf.
From Morison's
equation,
it +kD
of the force
upcrossing
may be derived
level,
F.
as follows:
Eq. (2.3.45):
ulul
hence:
4-F
dt
k+
I
2k
(3.3.3)
Dlull
U=
dti/dt.
where
Using the method of Appendix
and defining
arbitrary
known probqbility,
Appendix
One)
variable
One, Section.
ys = u),
domain of u, 6'and
obtain:
we
,
1,3.2.9
to enable
u (equivalent
(with
yj
E F; Y2
transformation
to xj,
from the
X2. X3 in
68
-*
-
(3.3.4)
U) = P(U'p A, u)/k I
p(F.
and integrating
out u:
00
p (F,
1
fp
(u,
= -C j1 -00
t)
for
Substituting
p (F,
--
3/2
(21T)
with
R
uu
= cr
u
R..
a.
uu
R
= cy
u
uu
(w)
=fS
u
2
[M]-l
(U) ] du
(3.3.6)
_go
R....
uu
00
-f
ful
-R..
uu
0
R.
uu
0
uu
fexp[-
of cross-covariances
R
uu
0
kinematics
of particle
00
k12 Vt-e-tTM]
M=
a)
(u.
Q.
where
[M] is the matrix
pdf.
1.3.3.1:
11T
t)
(3.3.5)
Gaussian
the trivariate
One, Section
from Appendix
Ct, U) du
dw
0 uu
00
1S
(w) dw
co
fW2.
0
co ,
0
00
a6
(w) dw
ijil
W4
uu
(w) dw
S (w) dw
uu
(3.3.7)
and
S (W) is
uu
Det [M] =
is
obtained
2
2 ((1
cl.uuu
the determinant
Expanding
Cr
4)
[M].
of
the argument
substituting
of the exponential
function
in Eq.
(3.3.6)
and
in Eq. (3.3.2):
IHI/3,
EW(F)
Co
fFf
1
zi 1
(21T)3/2 kI 2VbetTMI
00
exp[-
-co
(F
it
=
where
-kD ulul)/k,
2kD[ulfi)/k,
u=-
2.
(kl,
U2
+ 213 U +
k22
k33
2)
du
69
jtl,
2/(Cr
Cr..
uuIuu
2/(a
y 13=
Cj.
14
P-22
1/CTjt 2
'
Cy
33ua
uuu
2/(CrU2
l
CT..
Cr..
4,).
Cr.
_ Cf.
%2
cr
t
4)
4)
are strictly
of Hip
on the values
conditional
(w).
TITI
Long-term
IType
scatter
3.3.2
2.3.6.2,
description
P (F) =EE
P1
i=l
may be obtained,
[P
j=l
P1
(FIHI/3i,
to the procedure
according
zj
). E{N+(O)IHI/3i,
of force,
of zero-upcrossings
nm+
[E{N (0)IHI/3i,
=ZZ
of
between the
TII
zj
+
E[N (0)11,
IW
is
E{N+(O)ll
(3.3.9)
given. by:
(3.3.10)
zj
the departures
.1
lj]
T)
i=l j=l
Consequently,
the long-
of wave climate
as:
nmW.
E(N (0,
of
diagram
of peak loading
term cdf.
Section
11 cdf.
1
Peak Loading
'type
11 and 'type
21 models
long-term
the difference
here,
concern
to the short-term
attached
latter
force
rates
for
given
by Eq. (3.3.8)
the rate
applied
herein
for
for
surface
for
the
in the procedures,
the
'type
'type
TZ).
between these
that
upcrossings,
It
The
of zero-upcrossing
namely E{N+(O)lHi/3,
Z)
21 model,
elevation
to here as E{n+(O)IHI/3,
the relationship
conditions
arises
applied
resulting
is,
E{N+(O)IH1/3,
therefore,
two estimates
Tz ))is
important
as well
distribution'from.
21
for
'type
the
upcrossing_. -rate
+
tO)
Hip,
TI
E{N
to
as
referred
Z
equal
refered
to investigate.
as the force
Eq. (2.3.61).
to
70
-,
Method of Computation
3.3.3
Eq. (3.3.8)
in the computer
solved
In line
integration.
deviations
step numerical
OSF2, a range of force
used for
the procedure
with
by constant
routine
short-term
condition
To ensure stability
with 10 steps per standard deviation.
P
it
in
domains
found
integrations
the
the
u
and
was
numerical
of
limits
former
between
integrate
the
to
of :tSau with steps of
necessary
was considered
4k
CIP
Cf
Duu
cr.
O. lap
step width,
where:
independent.
In consistency
the splash
immersion
with
as 7.5m.
frequency
However,
the estimate
conditions
the upper
here,
short-term
Sections
truncation
in the transformation
at this
as demonstrated
beneath
with
location
of aU2 is
applied
member elevations
of
under
sensitive
3.3.1.
This
arises
to the spectra
truncation
than
for
convergence
from Fig.
3.2-3,
negligible
sea state
with
of 8wo yields
a truncation
the rapid
kinematics
do not
HI/3
= 9.3m,
it
this
is
behaviour.
found that
using
14%
of cr
approximately
u2
At greater
depths of immersion
estimates
at 2.75wo.
of the depth
restrict
transformations,
of the spectral
reduces the effect
in the computed variances.
difference
and there
is'
71
3.3.1
Fig.
also
shows that
the spectral
truncation
has a marginal
effect
by a
on the estimate of 0.2 at the 7. Sm. depth of immersion, represented
u
from the above truncation
3% variation
for 111/3 = 9.3m.
Similarly,
points
in aU2 was found to be less than 0.5%.
departure
the corresponding
2P
departures
in
However, these
au
002 have no discernible
on the
effect
truncation
are not.
results
given in the preceding Sections and spectral
therefore,
for
significant
those
situations,
and in this
occurs
Section
on which to truncate
an upper
frequency
S (w) when
nn
at
cut-off
This value
models for
theoretical
tions
in tile high
3.3.4
Results
order
wave theory
transforma-
of Computations
Fig.
it
is
seen that
a high
levels
become
departure
in
between
load
the
predictions,
most
extreme
and
parallel
for
in
0.95
0.9
to
the
terms,
probabilities
occurs
region
of
percentage
behaviour
is
identical
by
30%.
This
to
that
the
of
shown
order
andis of
Tickell's
results".
are plotted
on Fig. 3.3.3 for both 0.5m.
depth
immersion.
7.5m.
For
diameter
2.0m.
the small
of
a
at
members
and
between
'type
21
11
'type
departures
the
short-term
cdf. s
the
and
member
in the long-term
descripfor the high intensity
sea states are retained
The long-term
tions.
loading,
indicated
peak distributions
decrease
The departures
there
being
only
and negligible
depth of immersion
small
with
differences
difference
reduces
reduction
for
the departures.
in the non-linearity
for the 2.0m. diameter
a 5.0m. member.
Increase
of
member
in the
72
-
for
The absence of any apparent departures between the two distributions
P
implies
dominated
loading
inertia
F
that
the
and
are statistically
independent which is to be expected since in the limit F is a function
t
U,
ii
6
function
independent.
is
6
of
and
and
a
are
statistically
only,
of
in the shortthe agreement also implies that the variations
have
term mean zero-upcrossing rates, E[N+(O)IH1/3)
and E(n+(O)IIII/3),
I
effect on the cdf. s.
negligible
Furthermore,
Short-term
mean zero-upcrossing
considered
tion
used in the
whilst
the values
values
of HI/3
values
The short-term
spectrum
can be obtained
eleva-
surface
by E{n+(O)II11/3)
by Efn+(O)IHI/3)p_M.
(or H1/3)
intensity
for
The values
are represented
for
FAMITA
given
The differences
from
of EIN+(O)IHI/3)
values
range of conditions
arise
I
and decrease
This
behaviour
of force
since
depth of immersion.
frequency
3.3.4.
the theoretical
the entire
obtained
are represented
between these
in sea state
on Fig.
21 procedures
'type
for
rates
can be understood
the estimates
Appendix
increase
with
by considering
of zero-upcrossing
and
the
rate
One, as:
F22
+1
E{N (O)IHI/31
-- TrTV
mo
mn
force
the
the
spectrum
and
second
of
zeroth
moments
m2
are
and
where mo
It is not necessary to develop the
Appendix One).
1.2.2.2,
(Section
as it
force
spectrum,
frequency
significant
shift
a positive
thereby
increasing
(to
is
to appreciate
sufficient
that
the zero-upcrossing
a shift
affects
rates
from Section
E[N+(O)IHI/3)
22
for
directly.
force
rate.
based on the
An estimate
of the loading
1.3.4.2,
7r crF
since
Appendix
E{N+(O)IHI/3).
the ratio
M2/mO and
frequency
The relevant
ii,
from
fj
deduced
be
that
the
of
spectra
of
and
u,
range can
from Figs. 3.2.1 and 3.2.3.
turn can be inferred
The zero-upcrossing
in the
'type
which
21 approach
in
have
such a Gaussian
process
One:
(3.3.12)
73
has been evaluated
This expression
condition
marginal.
between F and
'type
the
has negligible
on the
effect
value
might be expected that the correct
Eq. (2.3.61),
the non-linearity
which retains
it
to Fig.
Returning
considered
it
is
zero-upcrossing
rates
of E{N+(O)j111/3)
in the loading,
since
from
2,
lie
would
above.
that
evident
lies
zero-upcrossings
elevation
surface
loading
1, estimates
in Table 3.3.1.
it is seen that the difference
between the estimates
is
It may, therefore,
be inferred
dependence
that statistical
for
the Famita
the
within
estimate
'band'
of
of the force
21 approach.
intensity.
conditions
between the
'type
insensitive
to moderate
on the
effect
the
long-term
lowest
long-term
load
model for
are
zero-upcrossing
practical
requiring
are in the
and, therefore,
effort,
rates
is
application
namely
IHI/31-
EfTl+(O)
It
that
departures
previously,
11 and 'type
can be concluded
that
significant
which have negligible
As mentioned
distributions.
it
The only
should
be appreciated
that
interpretation
vary
3.4
Design against
excursion
probability
distributions
experienced
during
extreme
positive
the positive
(F)
EP
requires
of the extreme
the period
value
of exposure
peak loading
failure
the consideration
may be obtained
to be
of wave loading
of the structure.
under
directly
bf
The cdf.
of
from that
of
the assumption
that
are independent:
[P (F) 1
p
(3.4.1)
74
force
is
N
the number of
where
during
peaks expected
the period
of
exposure.
of exposure, te, we haVe, for a narrow-band
For a period
(3.4.2)
mean rate
where the long-term
(3.3.10).
from
Eq.
obtained
21 processes
For the
'type
surface
elevation
is
Tz
-1
where
distributions
probability
to those
of wave loading
of the positive
values
is
real
in the previous
proceeding
are for
here is
in the
it
should
comparison
values
again
purposes
wave climate
in
excess of the period
period
in Chapter
discussed in detail
distributions
loading
of positive
Sections
application'the
the long-term
and negative
peaks are
number of extrema.
of peak loading
study
of
value
(3.4.4)
the total
the extreme
be
will
= [P (F) 12N
p
(IFI)
EP
Section
Before
the
In many
the absolute
is
investigated
values
herein.
considered
it
analysis
excursion
2N
now represents
where
In this
and thus:
are symmetrical
of
rates
(2.3.34).
of negative
of first
applications
the zero-upcrossing
as discussed in Section 3.3,
is
(3.4.3)
distributions
herein,
Tz- 1
in Eq.
defined
+
E[N (0)),
of force,
of zero-upcrossings
applied
+
Eiii (0) 1=te
N=te
force
11 process:
E{N+(O)li
Nte
identical
'type
derived
of this
be emphasised
that
the results
Wave climate
description
only.
diagrams
scatter
should
under
In this
of peak force
derived
it
used
In any
to cover
investigation,
study
of this
2.3.1.
of Fig.
be extrapolated
of exposure
Four.
are developed
is
previously
as
assumed that
represent
75
-
Behaviour
3.4.1
3.1 to 3.3 in
of Extremes
the Prediction
distributions
The probability
Models of Sections
of extreme
A summary of
a range of member diameters and depths of immersions.
diameter
5.0m.
Sm.
for
O.
the
and
members at a 7. Sm. depth of
the results
immersion is presented in Tables 3.4.1 and 3.4.2 and the pdf. s for the
in the
Variation
drag dominated condition
on Figs. 3.4.1.
are plotted
for
loading
of extreme
properties
of immersion
or
behaviour
in
high
load
diameter
the
the
observed
regions
reflect
member
discussed
distributions
the
the
peak
variate
previously
and are,
of
of
omitted here.
therefore,
three
In the tables
of the properties
force
is associated
(a)
probability
distribution.
with
a different
probability
d PEP(F)
dF
expression
Each estimate
of being
of
exceeded in a
(MPPV) representing
the force at the
From Eq. (3.4.1)
the pdf. of extreme
(F)
as the solution
obtained
to:
d {p
(F) 1=
dF P
= 0, yielding
cannot
description
are as follows:
The estimates
dP (F)
EP
N-1
1pp
(F)
N[P
=
dF
p
is
the
mode
and
for
Peak Value
(F)
PEP
are included
of its
(or
maxima) of the pdf.
mode
force is given by:
This
force
of extreme
of exposure.
period
partigular
levels
be solved
since
analytically
the mathematical
form
it
is
found
known
However,
(F)
that
Pp
(F)
explicitly.
not
are
and
pp
of
is,
in
force
surprisingly,
agreement with the
not
the mot probable peak
load prediction
abstracted
level.
probability
is given by'P
Substitution
This
(F - MPPV) -- 1
PN
into Eq. (3.4.1)
since
of peak loading
for
where N is
yields
F-
at the relevant,
P (F = MPPV) = exp(EP
1) = 0.3679
76
-
from
discrete
is
herein
for
both
the
this
satisfied
cdf.
s
computed
and
0.5 and 5.0m. diameter members at depths of immersion of 7.5,1S. 0 and
be
22.5m. With reference to Eq. (TI. 4), Appendix Two, it may, therefore,
implied
follow
t{F
(b)
E{F
Co
f Fp.
MAX
and, therefore,
p(F)
therefore,
is
member in
taken
The periods
a load
In Table's
3.4.1
asithe
than
the value
lsix
that
'identical
environments,
quoted.
and previously
reported
months
summerland'six
months winter',
in conjunction
of
one
here,
considered
'one-yearl,
a level
'and 3.4.2
probability
exposed to statistically
greater
was applied
exposure
skewness of the
is
may be described
of exposure
each of which
Fig.
diagrams1in
it
exceeded.
which
12, are
(3.4.1)
to Fig.
sustain
Reference
to a
3.4.1.
a, of being
probability,
will
of the pdf,
I
corresponding
I
there is a high
apparent that
iIi
It is,
that the MPPV and E{F
estimates
willbe
exceeded.
MAX
i
desirable
to predict
an extreme value with only a small
probability
a=0.01
as:
(3.4.5)
With reference
(c)
MAX
the centroid
demonstrated in Fig.
pdf.
is obtained
dF
represents
in excess
force
with
the relevant
in
scatter
As a further1consideration
a period of
ased
i
has been investigated
on the 'one-year'
2.3.1.
of 50 years
data
set.
The distributions
of
to
conditions
I (HI/3
to
cause
for
those
identical
the
force-for
of
the
winter
six
forces
peak
'extreme
sea
s ate
comparable
data
b ecause
monthperiod
reach
a magnitude
I
'one-year'
set
summer
intensities
with
are
sufficient
in
those
the
winter
distributions.
In Tables
3.4.1
and 3.4.2
'type
11 and 'type
force.
levels
correlation
it
21 narrow
is
band models
-mentioned
above.
s
as
cdf.
seen that
This
is
between
te'departures
Ire! turn-period'
I
view
estimates
the
at high
of the
77
Instability
in the extreme
was experienced
force
'type
in the computations
covered
deviations
integrations
in Section
employed for
increase
required
deviations
as opposed to only
condition
the basic
of the integration
in the precision
range
Unfortunately
3.3.
short-term
eight standard
in the computations
inaccuracy
numerical
from the
The force
range considered.
extended
for
of force
pdf. s derived
(through
in step
reduction
covered) due to
and expansion of the range of the variables
limitations
thus further
emphasising the impracticality
computer run-time
of. the 'type 11 model.
lengths
The linearised
3.1 underestimates
extreme
values
with
I
by
MPPV
E{F
levels
the
the
to
procedure
at
and
non-linear
respect
MAX
in
Section 3.1.4.
in
the
quoted
with
values
close
agreement
amounts
These values were measured from the peak cdf. s and correspond to
approximately
However, it is
density
48% for
seen in Fig.
functions
the conditions
is
that
3.4.1
that
the
'one-year'
the linearised
wave climate.
method gives
method for
which are much narrower than the non-linear
The change in shape of the pdf.
where drag is important.
in the consideration
important
force
estimates
at the a=0.01
is 58%, a further
21% on
the under-estimation
3.4.1
From Table
level.
of
in the non-linear
tail
distributions.
3.4.2
Long-term
year'
effect
Force Distributions
between extreme
The agreement
data
Extreme
sets
implies
in the prediction
that
load estimates
lower
intensity
of extreme
(3.3.1)
into
narrow
(3.4.1):
band condition,
and 'one-
have negligible
values.
short-term
conditions
The significance
of particular
be assessed by expansion of Eq. (3.4.1)
as follows:
For the general
Extremes
substituting
on the extremes
from Eqs.
(3.3.9)
may
and
nm
p
78
E{N+(')IHI/3i,
Tzjl
E{N+(0)IHI/3i,
81
E{N+(0)1"1/3iTzj_I
IIN
w..
E{N+(0)}
71
(F)
EP
})
zj
N
1nmW.
E[N+(F) 1111/3i,T
1.0 -EE
E[N+(O)l i=l
In the study
high
of extremes
is much less
and is, therefore,
than the
in the pdf.
of extremes
long-term
mean
it
Consequently,
(3.4.6)
sea state
be much less
will
range of forces
of threshold
rate
conditions.
in Eq. (3.4.6)
of force
levels
+
EIN (0)1,
rate,
zero-upcrossing
lower intensity
j=l
I _a
zj
w
for
the significant
is
of the
form:
[1.0
(F)
EP
(3.4.7)
X]
by:
- X]
N-
[expf-
xl]
N.
expf-
(3.4.8)
Nxl
p (F) = exp
EP
tei:
i=l
W.
[E{N+(F)IH1/3i,
zj
j=I
represents
(3.4.9)
i, jw
[exp - t E{N+(F)IHI/3i,
H
e
all
all ij
I
JI[
where
nm
PW*
i*
TIw
zj
short-term
(3.4.10)
values of the
argument.
Rearranging
this
equation
all
(F) =
EP
where N(HI/3j,
using
Tt
zi
ij
11
N(lil/3i,
E{N+(F)IHI/3i
1.0 --
Eq. (3.4.8):
(3.4.11)
E{N+(O)IHI/3i,
E{N+(O)IHI/3i,
total
T}
zj
T
w'j
I-I
zj
zj
) during
(3.4.12)
by (HI/3i,
Tzj)
period
in
sea state
oUexposure.
given
79
Hence:
, all
(F)
.p EP
ij
11
(FIHI/si.
EP
Z]
(3.4.13)
and
p
(FIHI/si,
EP
= [p (FIlli/3j,
p
zj)
distribution
The probability
zj)
with
Tzj)
value
of force
of the extreme
deviation
force
of
for
(3.4.14)
may thus
be
distributions.
extreme
expressed
the standard
N(Hj/sj,
I
are of interest
the highest
(large
intensity
compared
sea state)
be unity
for many of the lower intensity
Tz) will
of PEP(FIHI/3,
be
for
these
the
all
conditions
short-term
will
peaks
since
states
sea
For example, suppose in
lower than the thresholds
of force considered.
the value
the investigation
denoted by (Hi/3j,
pp (F
EMIN'
Fli/3j.
EP
(F)
zj
2.3.5
EP
21 process
'type
For the
11
for
and similarly
sea state
a particular
sea states
lesser
of
becomes:
(FIHI/3i,
extreme
(3.4.1S)
zj)
considered
the short-term
approach,
for
)=1.0
intensity,
then
EMIN
Tzj):
from Fig.
as observed
F>F
of extremes,
distribution,
Eq.
(3.4.14),
simplifies
to:
p
where
1-0
(FIHI/3i)
EP
N(HI/3i)
3.3.4.
Fig.
from
and
for the Famita data
PH
period
of exposure.
13
W
(3.4.17)
rate
associated
HI/3i
with
is:
=E
for
according
W**
j=l
the
m
Ew
W..
(y)/
j=l
Table
IE
E[Tl+(O)IHI/3i)
3.4.3
(3.4.16)
-(OIHI/3i)
Efn+(O)IHI/3i
Values. of N(HI/3)
P,H (F 111/3i)
N(HI/3i)
ij
i=l
zj
Famita, data
'one-year'
to Eqs.
(3.4-17)
The resulting
(3.4.18)
are derived
and (3.4.18)
short-term
witha,
in
one-year
pdf, s of extreme
peak force
on a 0.5m. diameter
for
HI/3
long-term
'one-year'
force
is
force
level
for
probability
.
of the short-term
on Fig.
for
cumulative
represents
force
the significant
the cumulative
considered,
3.4.2
and 3.4.3
it
distribution
long-term
of effectively
is
probability
level
of Figs.
range of the
probabilities
up to that
in
conditions
the short-term
From inspection
of
the pdf.
at each force
values
3.4.3.
at any level
beneath
distributions
For long-term
also plotted
is seen that
for
unity
sea states
of Hi/a
on Gumbel paper
is plotted
3.4.3
implies
forces
the extreme
that
for
extreme
cdf. s since
addition
required
for
properties
which is not
fully
always
for
to estimate
derivation
intensity
does
this
of the
values
in
3.4.2
extreme
conditions
However,
of extreme
the highest
of only
Two).
inadequate
is
of the cdf. s
and short-term
specification
observation
it
that
long
of the variance
an estimate
An important
both
in the procedure
a simplification
not permit
(see Appendix
follow
3.5
based on the
period
distributions
these
the product
fact
a one-year
represented
-3.4.3
Fig.
for
wave climate.
The cumulative
Fig.
Also
peak force
of extreme
pdf.
in Fig. 3.4.2
are plotted
included
in the figuie, is the
the
from the
values
short-term
is
and 3.4.3
condition,
a concept
appreciated.
3.5.1
Sections
In the preceding
retention
considerable
integration
narrow
integration
of this
increase
required
in computation
for
(Eq.
solution
(2.3.47)),
method numerical
required
instability
effort
Methods
of the probability
Furthermore,
if
is
considered
a further
(Eqs.
(3.3.1)
and (3.3.8)).
it
Chapter
of the Morison
of the non-linearity
Using Quadrature
Integration
in the solutions
densities
the
level
'type
implies
of short11 model
of numerical
For this
are experienced
latter
in the
$1
-
prediction
integration
being limited
by computer run-time
As an attempt
to reduce
linear
limitations.
probabilistic
the computer
the
'type
'Gaussian-Laguerrel,
type S69.
Quadrature
For Gaussian-liermite
00
2n
f ef (x)
dx =ZAkf
the integral
is
the function
point
f(x)
is
approximated
of degree n.
for
Co
fx
0
Chebyshev-Laguerre
3.5.1.1
Eq.
polynomials
yields
of degree
an exact
PH (F)
where x=
can be expressed
1
=EAkf
fir
ulr2a
whenever
4(2n - 1).
(X
point
(3.5.2)
x
and coefficient,
respectively,
of
of degree n.
for
Application
solution
of Pierson-Holmes
of wave load, Eq. (2.3.47)
(2.3.47)
solution
Quadrature
dx =ZBZf
w-1
where (x,,
The quadrature
Gaussian-Laguerre
e-
(3.5.1)
respectively
as a polynomial
f (x)
(n)
by a summation of terms
and coefficient
expressible
Similarly
(X
k=l
_Co
polynomial
(n)
in the
(X
ci. k=l
probability
densities
yielding:
(3.5.3)
k)
1u
and
(x) - exp f-
(F
2k
2k Iu2 (r. 2D
%2
xllxl)21
(3.5.4)
For the computations performed herein, using the constant step procedure
developed by Pierson and Holmes and'summarised in Reference 2, a total of
82
-'
-
The values
n=
(x
of
for
k'
20 and some examples are given
of this
purposes
using
the, following
in Table
3. S. 1.
values
study
of degree up to
polynomials
for
However,
the
up to n=
polynomials
42
relationships.
Chebyshev-Hermite
of degree n, "n(x)l
polynomials
are derived
by
from:
induction
Ho(x) =1
and
2xH
n-1
The function
points
coefficients
are given
{H
-d dx
n1_
(x)
xk
(n)
(3.5.5)
to the roots
correspond
of Hn (x)
and the
=0
by:
n+l
(n)
nIrTr
_2
(n)
M2
[- {H (x
dx
nk
(3.5.6)
from inclusion
The cdf. s of peak loading resulting
of the quadrature,
dragin
for
Fig.
3.
S.
OSF7,..
1
the
shown
are
most
program
using
computed
investigated.
The oscillations
dependent short-term
present
conditions
in the plots
illustrate
the numerical
integration.
Although
approximate
n it is evident
that
f(x)
cannot
to n=
of degree 83 corresponding
load kurtosis,
are reduced
immersion
H1/3,
conditions
effects
negligible,
in which
the
it
case
loading
42.
This
significance
of the exponent
may be exactly
wholly
fitted
behaviour
in the
increase
decays with
in
by a polynomial
approximated
As drag effects,
and hence wave
in member diameter and depth of
intensity
is not
surprising
inertial,
these
since
In the
limit
and f(x)
by a polynomial
by considering
taken
in
decrease
through
attached
in Eq. (3.5.4).
being
inherent
be, accurately
by increase
diminish.
the argument
this
in sea state
or by reduction
these
instability
this
term is
becomes invariant
of degree
various
as an integer
force
constant.
1.
levels
83
Du
k
cr,,z
(3.5.7)
=8
and substituting
['l
f (x)
= expl-
For inertial,
is
yl
2
distributed,
for
therefore,
linearised
f(x)
significant
is
behaviour
over only
for
amplified
are distributed,
reasonable
3.5.2
values
3.5.2
of
reasonably
for
at x=0
of degree up to 83.
nomials
a narrow
to suppose that
in Fig.
f(x)
functions
larger
for
0=5.0
in the.
uniformly,
quadrature
for
However,
peaky functions
force
over
with
levels,
a range. of x
18.
n>
low levels
it
levels
at various
This
region of the x axis.
(n)
a.
for
Function points xk
is unlikely
that
is,
It
of force
the
I
app roximated
be accurately
might
is,
The quadrature
considered.
on Fig.
are plotted
values
the quadratures
and f(x)
0=3.0
(j)
of force
(3. S. 8)
probabilistic
The functions
XJX
as represented
analytically,
performed
(j)
1) 211
the kurtosis
the integration
since
unnecessary
loading
of force
each level
[ 8
-2
/X-3
-
v/ 8-1
Gaussian
constant
taking
(j
:S] 1/4
by poly-
this
will
be
of the narrow
increased
yielding
the divergent
remote
follows
in
distributions,
significant
long-term
for
the
s
cdf.
and
sufficiently
leads
to computer
about 70%.
oscillations
drag diminish
costs,
accurate
for
run-time
in
oscillations
that
quadrature
over
over that
quoted
above,
permit
for
are less
as demonstrated
larger
levels
in Fig.
3.5.3,
the constant
3. S. 1.
the instabilities
most applications.
savings
Fig.
step procedure
necessary
reduces
additional
diameter
to contain
as the effects
savings
members.
of
the
of
in computer
84
!-,
for
Application
3.5.1.2
Eqs.
wave loading,
'type
of the
solution
(3.3.1)
11 distribution
of peak
and (3.3.8
3.2.
into
in terms
be
a form expressible
rearranged
may
Gaussian-Hermite
quadrature,
and Gaussian-Laguerre
as:
Eq. (5.3.8)
mn
TZ11 2'
E{N+(F)IHI/3,
3/2 Y-3 3
(2Tr)
where x
(m)
(n)
(iq) (n)
f(xi
Aj.
Yg
t
(3.5.9)
EB
k-1
Z=1
Det [M]
of both
V2u
P-3 32
-2
"kEI
!
2-A]
f(x, y) = exp[-
2k
F2
2kl
A=
9,22
1/2
3xy2
33
2k
(F
ij
2kD2F2
D xlxll
kil
89-3 3[
DF2
x3kD
VPP3-3 1311
1XI +
k1211k
XII
Fix
4k
x2
[k12
89-33
1/2
2k
D2y
k2k2k,
2k
lyl/2
kll
tll
IfZ33
33
4k
D2
3/2
X3)
D4x6
kJ4
-F371
IXI
D34Fx3
X
In the constant
step
a=0.
quadrature
for
and
the Gaussian-Laguerre
nxm
then
savings
in computer
run-times
should
result.
85
by Hand Computation
3.5.2
of this
The object
is to illustrate
Section
and its
distributions
how the
long-term
peak ('type
developed
be
without
recourse to the complex computer
can
values
for
herein.
developed
The procedure,
the solution
routines,
reported
Reference 17, is based on the standardised
cdf. s of Pierson-Holmes
distributed
essential
although
in Figs.
random variables
computing facility
the derivation
for
each short-term
and 2.3.5.
a standard
and integration
condition
The only
hand calculator
scientific
a programmable
for
is
2.3.4
in
be required
kinematics
of particle
would normally
of the spectra
The method is
consideTed.
by
described
follows.
as
an
example
of
means
Problem
To derive
the
'one-year'
wave climate
the surface
for
given
of basic
a one-year
in Fig.
in a 150m. depth
load,
period
peak load
in the
positive
of exposure
2.3.1,
member, 0.5m.
cylindrical
of a vertical
beneath
load,
extreme
and positive
distributions
long-term
of water.
Assume
the wave force
with
CM=2.0;
to be given
by Morison's
Equation,
defined
by 111/3 only
Eq. (2.3.4S),
form given
by
approach).
Applying
and (ii)
(i)
in Section
2.3.6
for
each
condition
The long-term
produces
distributions
of basic
can be built
up
force
and carrying
of
values
out the summations
a range of
considering
by hand.
The wave climate weightings
for
in Eqs. (3.1.7)
and (3.1.8)
in
Table
3.4.3
in
these
given
and the short-term
are
equations
use
(FIHI/3)
Pp
(FIHI/3)
P
and
probabilities,,
H
2.3.5 for each force level considered.
are taken
In practice
from Figs.
2.3.4
and
86
is
if
less
n,
[[I
P(F) =E
j=l
mw
Ew
j-1
(Filli/si)).
PH
-
and likewise
The probabilities
in Table 3.5.2
are computed,
of non-exceedence
by computer.
obtained
P(F)
The probability
This
is
further
of non-exceedence
2.3.5,
at three
agreement
illustrated
of three
levels
with
on Fig.
force
levels,
the values
3.1.3.
force
of extreme
are
Eq. (3.4.16)
from Table
and values of N(HI/3)
The computations
3.4.3.
are performed in Table 3.5.3 and again the
are in close agreement with those from the computer solution.
estimates
derived
using
Fig.
The precision of the hand computation procedure decreases for low levels
in
increase
force
the number of short-term
the
of
result
as
a
of
conditions
making significant
of which represents
2.3.4 and 2.3.5.
It
is anticipated
procedures
a possible
that
contributions
error
the primary
in
be
the prediction
would
to the probabilities,
in reading
application'of
of extreme
off
values
these
values
each
from Figs.
hand computation
for
use in first
A
the application
reservation
possible
regarding
analysis.
excursion
fatigue
is that it is the low
for
analysis
subsequent
to peak values
damage. In this range
levels of force which cause most fatigue
is
high
as most short-term
conditions
contribute
effort
computational
with
the possibility
of error
accumulation.
87
-
Ln
0
V-4
"4
W4
-4
00
P-4
CA
r4
P-4
Li
0
Z
0
Ln
z
0
ith
\0
00
m
qzr
Ln
Ln
Ln
r-A
C4
P.
t, )
0
P.
C4
0
P.
Lrb
44
t-
z
0
C4
u
00
(7b
r4
CN
88
6
TABLE
LEVELS OF EXTREME PEAK FORCE PER UNIT
CYLINDRICAL MEMBER;FAMITA
DEPTU OF
Period of
Exposure
One year
or
Si x Mont h s
Winter
Six Months
Summer
50 Year s
.
WAVE
LENGTH OF A
VERTICAL
Most
probable
peak
force
kN/m.
Expected
p eak
force
kN/m.
Peak Force
with 1%
of
probability
exceedance,
U/m.
Type 1
P-M (111/3)
4.05
4.1S
S. 80
Type 2
P-M (HI/3)
3.22
3AS
S. 02
Type 2: linearised
(HI/3)
P-M
1.70
1 74
.
2.09
Type 2,
P-M (DNV)
3.28
3.49
5.03
Type 1
P-M (HI/3)
1.50
1.60
2.20
Type 2
P-M (H1/3)
1.35
1.43
2.02
Type 2: linearised
P-M (HI/3)
0.87
0.90
1.07
Type 2
P-M (DNV)
1.33
1.42
2.01
Type 2
(HI/3)
P-M
4.71
4.95
6.57
Type 2: linearised
P-M (HI/3)
2.03
2.07
2.38
Type 2
P-M (DNV)
4.7S
4.95
6.57
Type of
probability
distribution
89
-
TABLE3.4.2
LEVELS OF EXTREME PEAK FORCE PER UNIT
LENGTH OF A
VERTICAL
Period of
Exposure
one Year
or
Six Months
Winter
Most
probable
peak
force
kN/m.
Expected
peak
force
kN/m.
Peak force
with 1%
of
probability
exceedance,
kN/m.
Type 1
P-M (HI/3)
94.9
98.1
115.7
Type 2
(HI/3)
P-M
94.9
97.6
11S. S
Type 2: linearised
(HI/3)
P-M
94.7
97.0
114.1
106.9
109.1
130.0
Type 1
(HI/3)
P-M
69.9
72.1
84.9
Type 2.
(H1/3)
P-M
69.9
71.9
84.6
Type 2: linearised
(HI/3)
P-M
69.8
71.8
84.6
Type 2
P-M (DNV)
70.9
73.1
87.2
Type 2
(HI/3)
P-M
112.5
114.7
131.2
Type 2 linearised
P-M (Hi/3)
112.3
113.5
128.9
Type 2
P-M (DNV)
127.3
129.3
147.8
Type of
probability
distribution
Type 2
P-M (DNV)
Six Months
Summer
So
Years
90
-
Ln
t ca
C
C
00%.
01
t-I
C4 tn
80
tn(14
01CA
-1V)601.0
%D
P-4
Ln
%D
00m%0
Ln
r-4 M
cn
wt
NO
t4l
Ln
44
t4t t4 t t4
"4
L4
L4
00
%0
tr
%o "4
"4
111r r
8
)-4
E-
0-%
tn
0
m
lt
kn
Rb
w
u
cz
00
rl%
14,00
-4 00
r94
eq
"
10 "
V)
%0
C4 00
-4
Ln c, 8P -4 MT
r4 Ln
"
%0 00 V-4 tn
0m
ko Go V-4 Itt
%0
4
.
CN
e4 eq V) tn m Irt "T
-4 -4
C; (S (S C; (S C; cs (S 6 C;
0000C; C; C; C;
t-I
r- 0M
Im
to.
4)
cm m
c"
on
cn
tn
en
cn
Ln
%-d
r-
tp
%-.
%D m
xxxxxx
P-4C4 CI Nil 00
tLn
eq
C7,
Coft
Lnft
"'*ft
0
oft
"
* Ln t.- zr r, 4Ln m 6 :rc!eq qT 0 co ft
"
W) -.0 C4
t4
t"s
% r
11-T 0
00
C14 r-4
t-
0;
%0 'D
t4
00
Lf)
(4
01
-e
'1*
C14 4
%D
m
tri
V-1
V-4
1-4
11
14
L.
0
Ell
r.u
00
te)
cn
If)
e-,-
00
r-
m
-4
0-4
(14
110 0
Ln U)
t-I
Ln
IIt
Lr)
m
t4l
ItT
ql*
M
00
r-
C14 -4
fl.
00
00
t1f)
1-4
Lr)
P,(71
IM%0
ChcnCt c"')8"8,, 8 c)
C31
-41*
o0
.
C,
6
6
C;
C;
C;
%-.
o
C;
C;
C;
c;
C3
C;
C;
C;
C;
(s(s
0+
F-'
Ln
1-4
r4
C4
%0
r-4
-1 -4
r-
Ln
"
000
-4
$7
%-.o
0
0
0
LL.
44
0
(.13
10
0
U)
U
HI
t4l ) C
00
ao%
('n'
Ch
00 "
? * 00
1-4
Itt
00
It
m
%0 r-
00
00
%a
P-4
%D
t-)
00
%D C*4 00%0
F-4
(71
M"
Ln
C%
IRt
"
000
"
Colo
%D
t-.
Ln mt P-4eq -.4
88
r-I
000000800
C14
q
r-q
Mw11 0000000000000000
................
F-
P4
Ln
r-4
(n
C;
00
e--,
00
P--%
C)
r-
="
,o
rm
r"
M C14 V-4 o CA CA 00 00
M 00 r*- co -4 -. 4 mm
t1r) r- 00 r-4 00 r-4 r'l "
r"
mt "4
r4 r-4 Q
C71 Ln Q
(n Ln Ln c)r-
Ch D t4
co r,
Rt
Mm
Ln
"T
IRT
r4 e4 r-4 00
(71 "4
rl
14
r-4 "I V-4 r-4 "4 -4 4 r-I V-4 r-I 0 P-4 -4
C; C; C; C;
C; C; C; C; C; C; C;
V)
CZ
(S
CS
Ln 4
_;
C4
t') al Ln r-4t, m
CA
C4
t4
ti;
Ln 4 r, m
9
0;
C)
eml%
%-.o b4
-H
LL.
pi
4)
4)
ul
91
TABLE
GAUSSIAN-HERMITE QUADRATURE FUNCTION POINTS AND COEFFICIENTS
xk
(n)
Ak
(n)
n1
0.00000
00000
00000
1.77245
38509
055
0.88622
692S4
S28
1.18163
0.29S40
S9006
897S1
037
S09
0.80491
(-1)0.81312
40900
83544
OSS
72S
0.94S30
0.39361
(-1)0.199S3
87204
93231
2420S
829
522
90S
0.61086
0.24013
(-1)0.33874
(-2)0.13436
(-S)0.76404
26337
86110
3944S
4S746
328SS
353
823
S48
781
233
O. S6410
0.41202
0.15848
(-1)0.30780
(-2)0.27780
(-3)0.10000
(-S)O. los9l
(-8)O. lS224
03087
86874
891S7
03387
68842
44412
lSS47
7S804
264
989
9S9
2SS
913
32S
711
2S4
n2
0.70710
67811
86548
n3
0.00000
1.22474
00000
48713
00000
91S89
n4
O. S2464
1.6SO68
76232
01238
7S290
8578S
n5
0.00000
0.95857
2.02018
00000
24646
28704
00000
13819
56086
n
0.34290
1.03661
1.7S668
2. S3273
3.4361S
13272
08297
36492
16742
91188
23705
89514
99882
32790
37738
n
0.00000
O. S6506
1.13611
1.71999
2.32S73
2.96716
3.6699S
4.49999
00000
95832
SS852
25751
24861
69279
03734
07073
(n)
(n)
The Ak
give only
10
0000
SSS8
1092
8649
7386
OS60
044S
0939
are symmetric
and xk
the values corresponding
IS
is the power of
before a value of a coefficient
A number in parenthesis
for example,
value must be multiplied;
10 by which the tabulated
is 0.08131
that the coefficient
(-1)0.8131
means
...
...
92 -
I-
TABLE2
LONG-TERM PROBABILITIES
Force = O. SkN.
HI/3
0.3
0.9
1. S
2.1
2.7
3.3
(a)
F
CrF
(b)
(c)
(d)
{1-(b))
P (FlHl/s)
ti
x(3) t
8.052
S. 6S6
4.433
3.65S'
1.0
9999999
.
0.999977
1.67xlO-e
2.93x
2.39xlO-
0.9971
2.22x
10-4
4.18xlO-
0.988
7. lSxlO-
0.972
8.2 SX10-4
4.5
2.678
0.9933
2.44 X10-4
S. 1
2.342
0.986
3.35xlO-
S. 7
2.069
0.978
2.63x
10-4
0.968
2.00xlO-
4
4
6.9
1.6S3
0.9S4
2.63xlO-
7.5
1.493
0.944
2.33x
10-4
8.1
1.355
0.932
7.07xlO-
8.7
1.237
0.922
1.62x
=2.08xlO-
10-4
9.36xlO-
0.910
J-P(F)=
COMP**
0.9998
0.9995
0.997S
1.13S
9.8SxlO-
0.9999994
10-6
3.102
1.842
=1.0
3.9'
9.3
x( 4 ) *
15.528
1.72 X10-4
6.3
(FIHI/3)
(e)
{1.0x(d))
0.9s
9.68xlO-
0.92
7. S4x 10-4
0.888
4.86x
0.86
5.2 9XIO-4
0.84
4.3SxlO-4
0.82S
1.42x
10-4
0.80
2.76x
10-4
0.77
1.17x
10-4
l-P
(F)= =5.493xlOp
P(F) = 99.79
P (F) = 99.45
p
P(F) = 99.80
Pp(F)
= 99.46
10-4
-93
0
TABLE
LONG-TERM PROBABILITIES
Force = 1.5kN.
(d)
P (FIHI/3)
(e)
{1. Ox(d)1
x(4) *
!21.0
HI/3
F
CYF
0.3
0. >9
1.5
2.1
16.968
2.7
13.298
3.3
10.965
3.9
9.305
0.99999983
1.01
4.5
8.034
0.9999952
1.415X10-7
7.026
0.999955
8.708x
5.7
6.206
0.9998
1.886x1 0-6
6.3
5.527
0.9993s
2.82
6.9
4.959
0.9986
5.292x10-6
7.5
4.478
0.9977
6.256x
8.1
4.065
0.995
4.05
8.7
3.711
0.9921
1. o9 x10-5
9.3
3.404
0.989
5.61
1-p
COMP**
(F)
E=3.784x10-5
F (F)
- P
0.999962
Pp(F)
= 0.999961
x10-8
10-7
1X10-6
1()-6
Xl()-6
X, 0-6
94
-
TABLE3.5.2C
LONG-TERM PROBABILITIES
Force = 3. OkN.
HI/3
(a)
F
aF
(c)
(1 (b))
(b)
Jill
F
PH(
/ 3)
x(3)t
(d)
(e)
'x(4)*
{1.0-(d))
P
PI
0.3
0.9
1.5
2.1
2.7
3.3
3.9
4.5
16.069
5.1
14.052
5.7
12.412
6.3
11.053
6.9
9.917
7.5
8.95S
8.1
8.130
8.7
7.422
9.3
6.809
.
0
=1.0
.
.
.
.
.
.
99999987
8.0
X10-11
9999982
1.03
X10-8
999992
3.33
X10-8
999978
2.29
X10-8
99994
1.25
X10-7
99987
1.352x
1-P(F)=E=3.273
COMP**
=1.0.
X10-7
P(F)
= 0.99999967
P(F)
= 0.99999967
10-7
.
.
.
.
.
.
9999999973
1.0 X10-11
99999994
6.0 x10-10
9999993
3.0 xlO-9
999996
i. sixio-8
999984
4.3SxlO-e
99995
99987
4. OSX10-8
7
1.79xlO-
9974
1.33x
l-P
p
(F)=E=4. lSxlO
2p(F)
Pp(F)
-7
= 0.999999S9
= 0.99999960
10-7
95
-
P.4 in
(4
en Co M
cyb 00 "0
-4
m Ch cyb 0% (71
g ! g ! Il 12! c !
.
r4
(D
cl
P-f
=
0--%
u%-d
0% 0% M
%D Ln
m al
ch cl
Cw
',,,
1;
L
0
934
ei
-1
0
00
Ch in
0-%
CY)
r-%
Co
Ob 00 et
al 00
m 0% Ch cyb m
Ch
d
11
11
0--%
U.
0-%
%..d
%-0
93.
m
Ow
93.
LU
Ci.
%D
r4
l4
c Cho
00
8x
a-
0-N
t"
lI
J..4
ch
l
Li
%912!
Il
0 oo %0 -e in r4
%..o
00
ch
%D
in
"0
r4 r- 00 "
94 %D
al -4 0m
Co 8
,---% =:
Mt
%D
Ch M \O NI 0
c M 0%COin
in
Ln
Co
Lo
c11
u
MX
CQ
44
U%-d
1.
424
w
F.,
0
e-%
en
%.-4
co e r ri en
en
et
C:)
CD
;
c11
t-%
in
11
01-%
P..
Qw
CY%Ci r P-i
Cn al al Ch al m 00 %o
01 mmm
ch ab cn m
Lt.
%-. 0
0.
%D r
m 00 Ln in
mm
00 f4 0
P.
in
rA
Ch M
C4
te %D in
ab %t Ln en qt Ln Co
Ln
00
c71
0mmm
......
';
21
"
P-4
Kt
124
m0
rq et Co
%i
li
l in"i ell l9
lli
3
Il
1:
1
Il
,
C:) rc)
ch
rCo
%0 \O
". t
m
nt
Co tj
Me
"O-s
%--#
in
Ln CD en in
r4 00 r- 00 r- Ln
r- (n r-. -4 e40
mm
88
Ln
m)
r_
let
%D
12!12!l oLi li %-
..
8
C:
)
0%
80
in
v>
let
t(D
in
le %D
Igt r4
Co
en
9-1
rC)
1
I
cyb cyb %D 00 in
0
rq
Ln
le
P-1
C:)
"0 00
MM
ab 00 %D bl
0% ch ch -e Ln
Ch Ch 0% r_
e0r.
c;
-4
00 in
ab al
Chmm
C%al 0%
0%aa
r, CY%
Ln Ch
01 C%
m al mmm
cm crt
.0-
0%m cyt
21
%D%0r- A %00%Ln C) 00
ce
P-40
A in in -e.ir, enr4 en Ln c l l
01 Cor %D%0i n
1L4
m
......
C) -' "N
en
Ln,
..........
Iml
F3
0
$4
44
Cd
P-f r- cyl "
\o
41
:s
P4
14
0
u
m o m 0% m 0% 0% Ch
c c c c c ! c a l C !
W-4
ei
0--%
P-1 r-
ch
Ln
-4
r-
en
e in Ln %0.D r- oo Co ch
FIG. 3.1-1.
RATIO OF VARIANCE
OF LINEARISED GAUSSIAN
WAVE LOAD TO
OE PlERSQN/jjQQdM
--VAR16MC&
LOADING. (0?f / M2) FOR DIFFERENT
MEMBER &
(DEFINED
SEA STATE CONDITIONS
BY HI/3), '
If
Cr2
DIAMETERI
ELEVATION
P2
50/142-5
0-992-01127-5
2-01135-0
2-01142-5
0-971.01112-5
i-o/ i2o-O
0.96
0/127-5
0-95
01135-0
0-94
0/142-5
0-93
0-92
1-51112-5
091
-5112(>O
-511275
0.90
-T
0-89
%Plvr-r, 4 OT
r-WUAI
IUN
0-5/135-0
2 . 3-49. )
OZ/142-5
(m units)
0.9
2.1
3-3
6-o
5-7
4-5
0.1
SIGNIFICANT WAVEHEIGHT H1/3 (M)
N'l
1,
lip
; 04
FIG
3.1-2,
KURTOSIS OF PIERSONIHOLMES
LOADING
FOR DIFFERENT MEMBER AND SEA STATE
CONDITIONS (DEFINED BY HI/3)
DIAMETER/
ELEVATION
0-51150-0
4.5
8.0
7-5
7-0
0-51,142-5
ONDI TIONS
,
S0
MOSKOWITZ
SPECTRUM
PIER
WITH FREQUENCY RANGE.
(0-4Wo- 2-75Wo)
(O'4Uh)-
d
6-5 -1
6-0 a
0 -5115CH)
B-OUk))
150 m
3
10 kgIM3
0.51127.5
Cm z 2-0 Cd z 1-0
KURTOSIS GIVEN BY
EQUATION (2.3.53)
0-5/120.0
V)
55 -
0-5/112-5
1-01150-0
(NOTE: LINEARISED
lit
5-0 - GAUSSIAN LOADING
TAKES THE VALUE
0z 3-0 FOR ALI,
4-5 -CONDITIONS. )
W1142-5
1-0115CX)
1-01135-0
4-0
M127-5
. 110
1.011200
101112-5
2-01142-5
1:81il'3291
3-5
3-0
0-3
2-7
3-9
5-1
6.3
1.5
7.5
SIGNIFICANT WAVEHEIGHY Hj/3 (! I%)
8-7
5-01142-5
(9
z
25
0)
4
0
to
Lb
th
Cv
Lo
N60
to
w
0tL
0
LL
co
U)
Z
0
LL
wo
3:
%.o
0- ,
La
Lf-
25
w
Lli
w
U
>
X
w
p
u
LL
0
CL
Ul
w
Ix
<
LL
w
U
cl
z
2 =) X(n
%..0
_j
<
(r
2
0
Ll
<
LL
CL
(L
-: to
m
E0
W
!UJ 0Z ,
V) LL E E
0W LO (n u
u
6
0
v)
z
if 0
if
0 1- 0 N CW&
V) U
cr w
E
CL
V)
LL CL
it
Ln
z
0
E
z
0
u
CL
I--
<
Z
-< n
(L
0 0
LLJ
>
!IW?
Q)
LL
(LU/N)I)'. d HIONYI
IINn
-=
to
99
E
Z
Je
%_o
LL
0
rs
w
CL
u0
Z
0
F3
to
w
cr_
uw
L,J
E 0o
0 LD
LE
ZO
Z,
uua:
wl
ui
Z
0
ILL
000
Lb
(LWN) HIONYI
IINn
83 d 3080=1
104
CP
'4
10J.
0
z
0
to
LL
A
w
u
0000
M'o
LL
co
0
ir
IL
UU
Co
w
z
Co
2
CL
xz
z
0
w
k-
0
Z
0
-i
m
. 161
-o'-
;-
0
(UINI)
A 4HICH371
11 Nf U34
30HOd
'101
1%
'. 1"
E w
Z
jg IL
aoc
3e.
LL
0
LL
61
co$2
R
IL
cr.
I--
75
U.
A
w
u0
01 U
Im ir
0
LL
L.j
to
0
Ir
CL
0
CC
W<
z
co
2
01
(L
W,
z
0
J
(j
LL
I,
0.
01 0
A 'HION31
M.
C4
'r
,-
,.,
102
'E
z4
eg,
2
I-
"
U)
0
LL
A
w
75
0
w
(
-t
0
In
(a
a:. *I..
.
I
-i
'
E
<
0
CL
N0
V,
q:-
Z
0
u iu
a'
44
.9
kh
.9
ITA,
k,
el
-10
,.
11 . ",,.
103
-
FIG. 3.2.1.
PIERSON -MOSKOWITZ
AND JONSWAP
SPECTRAL SHAPES.,
kuuj
,DT171
2s)
(m
22
Jonswap
S-q-n (w P) j
20
S-q-q(Lip) J
M
P(wp
I
cr
P-M
(tip)
18
P-M
I
Ws
16
Faqucncy
wp
THE JONSWAP
14
SPECTRUM:
luw= 19-91M/S
12
10
Uw 18%43nVs
uw=16-82m/s
UW=15-05m/s2
PIERSON MOSKOWITZ(P-M)
SPECTRA
J* [LJ5)14
Jgj
El
sjj (w) =A
exp
0-0081
A
*
0-74
B
wo != g/Uw ,
Uw
Uw--13-03rrV
WIND SPEED
UW=10.64
r
%.f
r% c- Wv
F- 114
r% 0%LJ aIQ
r-
104
-
F-
LO
C\J C\l
LL
C\J
2:
0
CLI
0+
z
C!r
WW
<
(n
LL
w
w
CL
+
.0
z
w
(n
zz
00
0a
in
La
WW
W V) V)
+
LL
0
ca I co co
AD
(n
Z
0
0
+0
Co
(n
cr_
w
1-Z
0
,;r
0
(slw)
LL
(0
(133d
Cy
(INIM
,
Mtl
CD'
FIG3.2-3.
TRANSFER
FIJNCTIONS
APPLIED TO SURFACE,
SPECTIRUMTO PRODUCE SPECTRA
Et. EVAT
TION
OF PARTICLE 'KINEMATICS
lNeDEEP WATER'
CONDITIONS.
0-24-
7-5m
7.5M
022
Suu(w)
TFU(ti) cm
0-0 m
0-20-.
T F(j (W)
%.41
04P
-u
12
I
I
z 0.14
0
u
z
U- 0-12- t.
Ix
w
LL
z
a0b -
sobw
SIM Jw)
w
w
0.
Ln
11X1
4 0.10-
cc
w
-20
-10
5n4
.
xI
w
w
00-06
15rom
Ix1
it
G04
22-5m
002-
Jo
06
1-6'40
2-2 24 24 2-8
.
FRE qu ENcY W,(RADSISIEC)
46
CL
Ln
Ir
cr.
'o
w
u
lz
U.
W
LL
w
CL
in
z
LL 0
0
LA 0
ZZ
00
u
I
co
w
Ix
z
co
00 w
in
CL
I
Ix
w
<
co
cr
Q)
40
UJINA
114'
rv
,. ;l jN
"L ,
3ovoj,
.,,
.*, -
- 0,
J,
zj
1
11
-, t,
cc
jl
1,14
IIIf,
II
Uce
w b.Iu
wo
19
40
Ito
IL
U.
%. #
U. 6-,
4nt
za
06
u
EE
o in 0o
cc
x0
(A
,,
-a
48
cc
ZS
1-WW
Ico
cc
0Ww0
IL
U
cc
CA
't
fn
fo
.
40
'.
-xII
ol
CK IX
C4
Iu<I
IE '34:
U,
cr.
%
OW4%
-i
t- a
uo
In
_j
W
%..
-. 0 '10
4,4
,
.:
10
'.
IFIG3.3.1.
OF CHANGE IN FREQUENCY
RANGE OF Sjj (W) ON VARIANCES
EFFECT
AND
cq
FOR
P-MH
113)SPECTRA's
CONDITIONS.
150 m
da
0.1%
0 46
04,14-
P%34
(c).4
18-0 wo)
--l
X,
000'
.
x
. 010,032
x
-0-02133
000,0
1
0-30
3
cxd2/
0.12 -x
(0-4 Wo/8-0 00
Wo)
NV
.x .90-10
-028
u 10 - '.
0
0
%.
PL*
xr om -
pr
93
-024
0
'13
u.2
0.06
0-22
b
x
ON
-020
0-02vIttI
0-3
1-5
-2-7
3-9
sil
6-3,8-7
7-5
H113(M)
rv
-it
90"
w
U
z
w
W
4( IA
In
:6 cc
CK
190,
25
to
0
)VV
LL
0:
0.
IL
IL w
ow
W
en P0
S
4(
xW & u
".
UZ
.. -.
>.
U
E
VIP.
Or
(L
W.6
U.
C4
di
00
iMINI)
HION41
'o
66
'Nv4d ".A
JINg 43d, 3: )uojj.
.
-110-
ltwtiolth
paper
21
,
24
Diameter.
-, -,,
LQ,l; ' - " ltlar d davlatio
f.
P,6rq" la'Ientth kN/4
14r, I LT
Approximation type I
Approximation type 2
10
'.
05. to. -I
01,16
IS178
I
'
c7
,,
774
);,>,,
>4,
FIG. 3.3.3.
bo
-1 ,
LONG-TERM
OF PEAK
PROBABILITY
DISTRIBUTIONS
APPROXIMATIONS
BAND
-.
Cr4
001
'S
TYPES 1 AND 2. -
II-
ZERO-UP
MEAN
LONG-TERM
AND
SHORT
3.3.4.
'FIG.
CROSSING RATES'OF SURFACE ELEVATION
AND FORCE FOR VARIOUS MEMBER
CONDITIONS BASED ON P-M (H 113)
Famita
LONG -TERM
0.30
12
-A4 (4
we
DEPTWOF
ONE-YEAR
PROENABILITY
Til
-EIN-* (OT),
Ways Climate)
t 0. le)46,
7-5
1-72X 10-;
1.98 X109;
pr
--a-
" 0.147
2.18 X 10-1
22-5
0.140
2-27 X 10",
30-0
xProbability
exceedence
024
with one
associated
per year 2 II(No peaks
per year)
0.22
z
W
0 020
0
z
E Jq
V)
0 Gle ce
U
(o)
H1131p_MzI
from
equation
5-5 5 IH-1
(3-2-1. )
j3
mm
E
046-
(o) IN li 31
FAMITA
0
or
W
IE
044-
Tjj
[N
)
4 (0) 114,,
diaI immer'sion)
W
ix
5-M
U ffi
012
i
W
0-5/7-5
S-015-0
Oil
0.10-
ILL
fbi,
11110%.
.116
IIL.
1.15
AIIt,
IIIt,
6.3
I CAN
T
,
112
-
4%
,
Ii"I 1
-4
'. (
NP
.%-
I.P4wj 1,
1'141.*
Plawisewoveclimate
I=
M kowits spectromdell nedH
I"
Deptha( Immersion74 M,
4".
",
..
04 -1-0
deviatioll
force
tons-term It"dar4
per unit
of
I&PIth 0-112 Mae
IiIierrow hand 6pproxignatiotiO,type 2.
Linearized
a.
9.
'.
" II
.II
1-
"0
'r
V),
(F
O.S.,
.5
.0
1-0
.0-.
FIG. 3.4.1.
PROBABILITY
LONG-TERM
FUNCTIONS
DENSITY
OF FORCE P(F).
FORCE PP2(F)
AND
FORCE
EXTREME
PEP (F).
PEAK
PEAK
65
1-0 0.
t
its,
-
r.
z
0
qy u
lb
CL
4
I
.t-W 0 U
IL
w
I
w
lz
x
w
LL.
0
op
50
. U.
go
%0
o
w
cc
LL
>-
<
w
I1
CL
9
11
Z
b-
U W
W Z
a.
W 0 0
z
0
p
u
z
U.
z
w
in
%.o LA.
0
0
in
w
Z
w
0
Z (L
x
W W-
('I
IA
W
co
to
0
cc
(L
''
o
t.
A:
,""
"t. k
ri
V Gf
3.4-3.
BABIL
R
IT Y DISTRIBUTIONS OF EXTREME PEAK
RCE ON 0.5m DIA. MEMBER (UNIT LENGTH)
R
L
MER'S&Q 7.5 rn BE60WS-W. L [CM w 2-0
M
1-0
MM
Da
I
4, 'k
6-0-
LONG - TERM
#ONE YEAR'
5-0jig
0
%.
x
4-54-0-
cr
8-3
3-5-
5-1
w
u 3-01x
0
LL
3-5
w
w
2-0x
HV
3
(M)
1.0Q-5016
14 31;0 41NO
.
to w
04
PROSABIL ITY OF I4ON"ME9DEHC9'
)1$
:
1II
CL
Wz
IL
I%I
z4c
I.. W
ON
\.
w
u<
lz
00
%, 01f,
k%
st
- >- x
Xa
co -
\6
LL
LL <w0
01
4,9
Dz
a-z
<<0W
Ww
CjL
CID
0<w
U-cr.
(go
WW
CL
to
mX
LL Iz0
OZX
F-
<W
D I-
cy <
U<
jL
w
10
Da
W<W
WZz
2
u.
00
WE
W
0
0
T%
V)
lz
'C
zw
0
x
DuU CL
'E
ES E
Z kn 1
'n
In
Xf
in 0.
Xi o
CL
W X
u
tII
C4
J4
FIG. 3.50
.
OF fW
VARIATION
FORCE
f (m),
WITH
CHANGE IN
FOR A KURTOSIS OF 5-0
12345
10
101
10-
10"
I
I10,
I.
I.
I.
I
I Or
la
i"lI 1
10
2.0
.
'vs
a
. 1-0 -0.1
' 1-5 - ZO
2-S
3,0'- -
!:.
:.
'
CONDI TIONS
DIA a 0.5m
150M
d
FAMIYA ONE-YEAR WAVE
CLIMATE P-M (Hill)
DEPTH OF IMMERSION. &7-5m
4-0
3-8
2*0
CM
II
Mms 35
MWB27
CD 0 1'0
MM a 20
CONSTANT STEP
INTEGRATION
3-4
3-2
144M is
03-0
z
w 2-8
i
2-6
2-4
MM a 5-1
2-2
w
U 2-0
ix
0
4- i-S
a.
1-4
1-2
I-C
0.6
06
0-4
0.2
PROBABILITY
'
"1
4.
''
-.
.
OF NON**F.XCIEF-DENCf. ':', -
118
CHAPTERF0UR
THE ROLE OF WAVE CLIMATE
IN
THE
PREDICTION
of the long-term
Formulation
descriptions
of wave height
in
of wave loading was described
and the basic and peak variates
have been applied
for wave loading in
Chapter Two and the techniques
distributions
The long-term
form a convolution
Chapter Three.,
result
of
probabilistic
of the stationary,
the properties
wave field
short-term,
pdf.
of
Eqs.
(2.3.31),
' (2.3.63)
the bivariate
pdf.
(HI/3,
of
as described
model for
No theoretical
exists
by the bivariate
and in real
by
applications
by a histogram
of
in the form of a
the measured probability
structure
of these parameter*
in
diagram.
Typically
these measurements cover periods rarely
scatter
excess of one year
as described
Winter
months,
herein
and presented
only
and often
covering
in Section
in Fig.
2.3.2.2
for
period
the Famita
of six
data used
2.3.1.
descriptions
of the long-term
of wave height and the
dependent upon the accuracy of the measured
induced loading is critically
in
its
long-term
climate
representing
of
wave
characteristics.
sample
Accurate
assessment
to identify
the primary
are firstly
data,
describe
to
climate
secondly
of
wave
a method for the
requirements
data
fall
the
the
when
periods
covered
of
short of the desired
extension
investigate
finally
the effects
to
and
of such extrapolations.
requirements
of this
The contents
in the final
OSFLAG 5 report,
The primary
procedures
single,
are included
Chapter
11.
Reference
4.1
Chapter
of this
The objectives
requirement
for
or multiple,
complete
12-month periods.,
in sea conditions.
to avoid
bias, resulting,
119
-.
In fatigue
of peak loads
dependent
induced
loads
low intensity
the generally
In fact it may be shown 46 that
waves during
sea states
periods.
fatigue
the major
investigation
excursion
case prediction
scatter
methods of extending
diagrams
in Appendix
Fig.
for
III.
seasonally
procedure
The results
of this
first
may be
covering
only
data
Winter
the
to
of
2.3.1
which
by the prediction
periods,
latter
for
sea states
weighted
Summer periods,
associated
alternative
and a preferred
to
contributions
'one-year'
covering
conditions
simulate
six
usually
in Summer
present
necessary
intensity
where lower
analyses,
In this
or wave height,
loading,
of extreme
neglected.
by the passage of
throughout
failure
and is,
damage result
frequently
occur
to Eq. (2,3.71)
according
influenced
therefore,
damage is
the accumulative
analyses
as summarised
in
input
in this
cdf. s of significant
from the Winter and one-year data
and individual
sets
are plotted
wave height
on Fig.
developed
4.1,
the
Corresponding
computed in Table 4.1.
Both figures
distributions
of wave load are given in Fig. 4.2.
clearly
indicate
the greater weighting
attached to high intensity
sea conditions
of HI/3
distribution
in the Winter
exceedence
for
distribution
that
by the increased
of wave height,
or load,
should
probability
of
from this data.
comparable.
The basic
load
the one-year
a certain
any period
data
levels
the distributions
Howqver,
of time
as demonstrated
data
of all
being
be restricted
to only
Winter
application
periods.
Although
for
the Winter
it
is
easily
Summer periods),,
in the middle
shown that
estimates
the time
times of exceedence if
and lower
data
over-
120 -
1-
4.2
Estimation
is a fundamental
there
the design
in applications
error
along
lines
these
the prediction
of the
29
(as
in
by
100
Det
Norske
Veritas
the
years,
recommended
extreme value
,
for offshore
installations),
is extrapolated
Norwegian certifying
authority
data base.
beyond the duration
This effect
covered by the wave climate
since
condition,
approach.
scatter
diagram
appears
for
reference
Although
it
of Battjes
to that
diagram
little
cover
which
in the extreme
to account
of time
weightings,
followed
approach
for
a similar
the necessary
to proportion
of the number of
as described
in Section
2.3.2.4
used.
is
procedure
the complete
Pedersen's
failed
from proportions
transformation
his
and others applying
between the typical
the discrepancy
making specific
without
wave height prediction.
procedure
by Batties
to extrapolate
feasible
range of physically
data
sea states,
many of
range will
to
in the short
likely
which
are
and
weightings
sample
justification
for
The
be
inferred
from
if
this
may
observed.
even
(2.3.65)
(2.3.63)
(2.3.31),
and
where the integrals
the sea
Eqs.
covering
infinity.
have
limits
of
upper
state parameters
data
There will
location
the fetch
be a physical
which
limit
to the values
available
Clearly,
for
the generation
extrapolation
severity".
domain beyond such an upper
truncation
of HI/3
wind speed,
recorded
and its
at any
duration,
of extreme
in the HI/s
would be meaningless.
and
121
-
Tr
intensity
sea state
[function
that
Probability
Prob [I
(HI/3,
Tr
is
of return
period
Tr
TzA
equalled
l I]=1
TrTr
max
hence:
PI(I
Prob(I
and probability
that
(4.2)
<Ii-1TrTr
MAX
is not
IT
exceeded in
LD years
is:
r
LD
PL (I
T)=
Drr
(l
-)
-
(4.3)
years
exceeded
of intensity
IT
is
equalled
or
LD
(4.4)
>I)=i
Prob (I
T
1. MAX
r
D,
for
in
4.2
between
1
Table
Tr
is
and
values
of
evaluated
This expression
100 years and for durations of 1,20 and 50 years which shows the
intensities
state
or
sea
greater'than
of
occurrence
of
probability
during periods
IT
to
equal
r
of exposure LD years.
122 -
Relating
information
this
to the prediction
of wave height and wave
data it can be seen that LD -1 year would
procedure applied previously using the basic
conditions
if
In contrast,
of a SO year
equivalent
that
this
conditions
A one-year
scatter
in this
to be encountered
diagram
data
to be representative
considered
applied in extreme
then there
extrapolated
likely
record
set
of the
long-term
of exposure.
be
therefore,
not,
of conditions
analysis
value
should
20 year period
wave climate
when
periods in
covering
by
Wave
omitted
such a procedure would
conditions
year.
one
of
excess
in all probability
to the convolution
contribution
represent
a significant
(2.3.63)
in the extreme
period)
return
where omission
regions
sea states
on the long-term
The effects
and (2.3.65).
pronounced
(high
of wave height
descriptions
of short-term
in
'one year'
in Eqs.
distributions
of the high
(2.3.31),
will
intensity
data
heights
of
wave
considerable
demonstrated
for
the Famita data later.
first
as
analyses,
excursion
high
in
data
the
the
of
an
event
of
return
period
presence
contrast
underestimation
to this
attached
weighting
In the
to considerable
lower
wave height
In
overestim4tion
event.
samples will
be the
few severe
range is negligible.
influence
in this
analysis,
the effect
be
of wave climate
wave conditions
Consequently,
will
fatigue
46
range of the cdf. S.
probably be insignificant.
for
whose
lower
4.3
Extrapolation
representation
of the
long-term wave climate requires extension of the scatter diagram.
domain of HI/s and Tz is complex and
However, the bivariate
probability
In the absence of extensive
cannot be defined in mathematical terms.
a reasonable
description
long-term
diagram.
the scatter
Using
E{No+IHI/3i)
where
This
the
from
obtained
of Eq. (2.3.28)the
wave height
1:j
/TZ)/W(HI/3i)
(4.5)
Ew
(4.6)
ij
does not
a loss
represent
of wave height
with
TZ
i
procedure
of zero-crossings
significant
W(HI/3j)
together
of
TZ
all
all
the notation
with
associated
rate
distribution
the marginal
only
values
corresponding
crossing
is to consider
approach
in the prediction
of accuracy
when, in developing
the statistics
field,
the
the surface elevation
of
wave
the
parameters
required
of
in
is
defined
form
in
terms of HI/3 only as demonstrated
spectral
because the most commonly
This assumption is justified
Chapter Three.
is
Pierson-Moskowitz
the
that
of
spectrum
employed
(2.3.24)
(2.3.2
Eqs.
3)
through
and
way,
.
form defined
in this
the scatter
a reasonable
until
follow
assumed to
fit
this
over
an extended
distribution
of H1/3.
tion
be performed
entails
points,
graphically
is
attained.
involve
methods
by eye,
from which
an estimate
range of H1/3 yields
The data fitting
to theoretical
paper which
'best-fit'
yields
the estimation
linearises
straight
the required
line
extension.
of the parameters
a discretisa-,
of the marginal
distributions
The former
or mathematically.
judged
distributions
probability
The distribution
of H1/3 is then
theoretical
distribution
theoretical
on probability
a plot
distribution
various
with
technique
the theoretical
through
the data
Mathematical
which define
the
may
124
-
distribution
by the 'method of
intervals
may be discretised
distribution
'fitted'
The resulting
class
increased
range,
long-term
wave height
into
data,
as the original
but
a number of
covering
an
of occurrence
and the probability
associated
with each
For classes in the extended range of H1/3 an estimate
is then computed.
is
to provide
the correct
required
the
rate
mean zero-crossing
of
in
for
the
of
short-term
convolution
conditions
of
prediction
weightings
Pierson -Moskowitz
yield
spectrum,
Eqs. (2,3.24)
In applications
using the
and (2.3.. 25) may be combined to
these estimates:
E{No+IHI/sil
4.4
-11Tz3.
f
(4.7)
SS
clear
Hi/3.
of
population
complete
the distribution
to estimate
of this
The objective
that
of
measurements
is,
therefore,
from analysis
of the
in the previous
The procedure,
as outlined
section,
bne-yearldata
set.
follows
is to assume that the complete population
a theoretical
'reasonable'
fit
data
distribution
to
the
shows
a
which
probability
bearing
in mind that
infinite
populations
Scatter
Interpretation
data
samples
from
errors.
of cdf. s. of HI/3
Obtained
From the
Diagrams
scatter
diagrams
proportion
of time
for
certain
of such short
to sampling
are subject
The Famita
classes
population
the distributions
A Note on the
4.4.1
of this
during
presented
the values
which
the period
in Fig.
2.3.1
indicate
covered
the
within
by the measurements.
The
representing
short-term
conditions
recordings,
number of three-hour
by
HI/s
Tz,
the
by
the
of
and
are
values
not
given
mid-class
classified
numbers assigned
period
covered
to each class
by the basic
in the diagrams.
the total
number of occurrences
similarly
The six
month winter
1456 recordings
is
962 and
of 1924
scatter
diagram
whereas
The original
presented
per
parts
occurrences-as
were analysed
diagram,
as a scatter
included
here
in
Fig.
4.3,
representing
in the consolidation.
is
71
Driver
and
and
by Draper
with
single
class
recordings
occurrences
indicates
that
which
in
the scatter, diagram since
three-hour
were
not
entered
recordings
some,
histogram
factorisation
the
the
on
marginal
of H1/3 and adding
reversing
It was stated by Draper and
the crosses accounts for only 1427 events.
Driver that low intensity
sea states (small values of HI/3 and Tz) could
instruments
it
by
be inferred
be
the
and
may, therefore,
recorded
not
that
histogram
of H1/3 for
causes for
possible
the
1456 actual
way as no
such
omission,
measurements
may,
be obtained by factoring
the marginal histogram
of HI/3 from
14S6
Fig. 4.3 by (it01-;) , adding the crosses and assuming the missing events
(HI/3 < 0.6m).
In this way the
the first
class interval
within
to fall
therefore,
distribution
cumulative
shown in Fig.
P(Hi/3
P(Hi/3
The Winter
for
4.1,
ft.
31
=
=2
ft.
scatter
of the three-hour
may be plotted
recordings
as
example:
1456
= 9.45m. ) = 1457 = 0.99931
)
0.61m.
=
diagram
given
1456
(5 x
29)/1457
+
1000
used asIthe
in Fig.
input
basic
2.3.1.
is
= 0.025
inthis
a modification
study,
and in
of Fig.
4.3,
due
to metrication
was
necessary
of the
redistribution
where a slight
in each class are assumed to represent
diagram and occurrences
parts
ignores single
(crosses)
latter
This
9,62.
simplification
events
per
diagram and the-missing
events and this attaches a
from the original,
iincreased
to
the
of
occurrence
all
probability
classes retained.
slightly
126
-
had a negligible
that
that
apparent
on the distribution
effect
be appreciated
is
4.1 it
on Fig.
the plots
Of HI/3-
it
However,
three-hourly
represent
have
the modifications
must
short-term
of non-exceedence
and the probability
assigned to a
wave height of one-year return period is approximately
significant
962
1456
1
ff47),
to Eq. (2.3.29) and not (t=) .
according
3
conditions
Similar
return period
1924
91 ) not (
to a probability
of non-exceedence
of (2920
1925)'
return period is associated
with the probability
an In'-ye'ar
corresponds
Furthermore,
of
[1 - 1/2920n].
Data Fitting
Graphical
4.4.2
Probability
Weibull
for
papers
theoretical
Probability
using
the Gaussian,
in Appendix
outlined
the behaviour
the surface
Gaussian
distribution
from past
distribution
is
approximated
by a particular
well
distributions
compared with
the Rayleigh
Each of these
to reasonably
analyses
distributed.
distribution
theoretical
probability
The long-term
Two.
Rayleigh
approximately
on the relevant
Gumbel and
approximate
elevation
quite
Rayleigh,
distributions
example,
Papers
log-Gaussian,
distributions
probability
to the procedure
according
line
Of 111/3 based on
for
the
wave height
Data which
plots
is well.
as a straight
paper.
of Hj/3P
obtained
and Gaussian
distributions
in
data,
Figs.
are
4.1 and
distributions
None of the plotted
are well
respectively.
1
distribution
by
implied
by the
theoretical
either
as
approximated
4.4,
significant
curvature
of all
plots
in the
by
'best-fitting,
line
those
given
a
exceed
would
straight
range
upper
It was not anticipated
that either
of these
through the data plot.
fits
to the data but the
distributions
would provide acceptable
probability
probability
comparison
4.4,
with
plots
papers
here
for
of the other
differing
iliustrates
the
probability
distributions.
probability
characteristics
These
completeness.
in Chapter
plots,
paper plots,
Three and
from Figs.
discussed
of the various
4.1. and
below,
theoretical
127
It
by Jasper 72 that
height
line
as a straight
plot
of significant
wave
and hence H1/3 should
distribution
log-Gaussian
on the
indicates
4. S which
in Fig.
the Gaussian
follow
should
logarithm
the
the inclusion
that
Such a plot
paper.
is made
logarithmic
of the
height
for the 'positivethe
wave
scale overcompensates
of
compression
by
Fig. 4.4.
The result
is a negative
exhibited
skewness'
skewness of
log-Gaussian
the data plot in the upper range for which the theoretical
distribution,
'best-fit'
as the
wave height
predictions
in this
The Weibull
distribution
73 has been
as demonstrated
data
In this
significant
wave heights
the
by the
of
the
line
A.
of H, defined
value
to
wave
wave height.
by parameter
A of
line,
therefore,
of
adoption
of
The Weibull
graphically.
yields
effect
leads
is
that
to significant
of wave height/probability
lower
limiting
distribution
a satisfactory
distribution,
techniques,
the
from
given
estimate
wave height.
significant
of the Weibull
to the
departure
when used in
it
to the
a more accurate display
by virtue
of the logarithmic
nature
This
axis.
by the
eliminated
behaviour
values
could
is
straight
and gives
extraction
individual
disadvantage
data
range
for
based on graphical
heights
2.3.3
Fig.
unacceptable.
applied7,70,74,75
in Fig.
plot
long-term
analyses
extensively
deviation
determined
-37m,
best-fit
The great
height
for
of Hi/3
plot
straight
value
therefore,
excessive
in Appendix
This
results.
of the value
A=O.
would yield
construction
Weibull
limiting
described
the distribution
choice
in
earlier
lower
the
plot
line,
straight
of the wave
uncertainty
information
in the
paper presentation.
A better
graphical
technique
should
include
a linear
wave height
scale
in
the
same
accuracy
to all
plotting
the probability
58
has this proper ty, as illustrated
distribution
Gumbel
The
data points.
to a
in Fig. 4.7, and the plot of HI/2 shows an extremely
good fit
paper which
line
straight
wave height
distribution
the long-term
'
approximated by this distribution.
indicating
is well
gives
that
of significant,
on
128
forms an ideal
being
in Table
conditions
of H1/3 is
'best
through
eye fit'
from the Weibull
demonstrating
for
4.7,
that
plot
in Fig.
4.6
of Fig.
is
information
data
as the
obtained
4.7.
The corresponding
generally
distribution,
The limiting
period.
14.0m.
approximately
class
of the one-year
1100-yearl
with
the abstracted
an extrapolation
of up to
of the long-
associated
in Fig.
heights
the discretisation
return
15-9M- from the Gumbel distribution,
value
value
4.3
for
probabilities
is demonstrated
presented
to represent
into
of HI/3
This
intervals.
basis
predicts
wave
by Saetre 75
as confirmed
In view of this
this
does not
wave height,
of individual
by the significant
curvature
4.4.3
CTHI/31 for
the
lower
as demonstrated
4.7 and in this case
range,
be preferable.
from the
'Method
distribution
a two parameter
deviation,
of Moments'
and is
as demonstrated
are given
the Gumbel distribution
to as the
and referred
fully
in
and standard
in Table
deviation,
of Hip
'theoretical'
on Fig.. 4-7-
plotted
the graphical
deviates
distribution
slightly
lower
slope
significant
graphically
to be insignificant
are likely
If
by 'eye-fitting'
data obtained
the
it
figure
From the
long-term
description
and either
of significant
'one-yearldata
to generate
applied
set.
The estimates
Two.
Appendix
to the, distribution
on Fig.
plot
tO 111/3 Obtained
defined
for
of its
may well
is
in the
especially
distribution
the Weibull
a good fit
yield
is
be plotted
reasonable.
on probability
from the
15.9m.
obtained
a 1100-year,
terms
from the
these
differences
be acceptable
wave height
paper
'theoretical'
and predicts
method should
should
of the
extrapolation
for
this
be stressed-that
to confirm
that
data
the
Gumbel
129
is
for
favoured,
will
which
reasons
be used in all
remaining
in Section
4.4.5,
and
computations.
'Goodness of Fit'
4.4.4
be discussed
will
to HI/3
to HI/3 is further
The goodness of fit of the Gumbel distribution
density
histogram
in the probability
demonstrated,
plots of Fig. 4.8.
is extremely
The fit
good in the middle range of the data values and
in the extreme range of the
the oscillation
of the histogram densities
distribudata, as displayed
plots
on the probability
of the cumulative
is probably
tions,
the result
set
is
slightly
to probability
correspond
probabilities
rise
of Fig.
be assigned
4.8,
there
greater
goodness of fit
Gumbel distribution
which
that
terms
there
sugge!tts that
111/3 follows
intervals
event
the
fit
must
The erroneous
per year.
can
conditions
of probability
paper plots
difference
is no strong
the slight
this
in the shape
and the
first
two
of the
'theoretical'
over
over
reason
the entire
to question
over this
distribution
have an insignificant
of HI/3
effect
the hypothesis
In practical
range.
first
two class
and the
on the
fitted
Gumbel,,
long-term
distribution
will
distributions
Consequently
the Gumbel
or wave load.
be
for the description
be
to
satisfactory
considered
may
distribution
HI/3
the
H1/3 range,
However, over the range of H1/3 in
Four.
is satisfied
at the 0.05 level of significance
which rejects
in
Appendix
described
as
1.2m.
test
the
of
excess
test'
although
in the lower
is also
This
data
paper
to be a notable
appears
histograms
density
the
of
intervals.
class
'chi-square
also
of
to the extreme
tail
of less
10.2m. is
However,
in the one-year
investigation
the
under
phenomenon
even when
distribution
postulated.
probability
In Fig.
4.7.
case
period
in excess
on probability
in the upper
to distortions
a return
of Hi/3
of one exceedence
In this
wave climate.
or considerably
covered
give
or conditions,
condition,
period
a return
likely
the extreme
in the one-year
to include
extreme
variability
which depart
conditions
data
Any one-year
wave climate.
of sampling
of wave height
range of values.
of'
Variations
4.4.5
,
The validity
wave climate extrapolation
of the above procedures'for
when
based on a single year's measurements would be adversely affected by
in
annual wave climate.
variations
significant
Saetre75 has analysed wave measurements covering three successive Winter periods at Famita, including the Winter of 1969-1970 used in this
is
there
that
no strong evidence of any
study, and concluded
in
differences
between
the
the
extreme
wave
conditions
significant
for
his
'Gumbell
HI/s,
data
the
three
plots
of
sets
although
years
in
departures
here
in
the
range
extreme
as
reproduced
notable
showed
4,9.
Fig.
The HI/s
predictions
for
plots
in Fig.
1969-1970 data
the
set
apart
in Saetre's
those
with
point
representing
HI/S
with
occurrence
a storm
from
Draper
here,
taken
used
for
the one-year
data
three-year
Period
Return
3-year
data:
Saetre's
1-year
data:
Graphical
'Gumbell
'Gumbelf
Theoretical
It
Gumbel plots
for
sets
of estimates
likely
by virtue
differences
plotting
position
be exaggerated
of
applied
of these
way a graphical
height
wave
13.4m.
15.4m.
13.0m.
14.9m.
and hence no
sometimes
procedure
16.3m.
fitting,
14.4m.
was applied
set
even further
data points
demonstrate
extreme
if,
fitting.
graphical
in
for
could
is much closer
yield
line
with
of the
earlier.
a 110-yearl
to the value
The two
These
in the
were neglected,
to his
the variability
the uncertainty
76
points
as discussed
to account
extrapolation
14.1m, which
data
methods of data
of the different
could
50 years
fitting
which method of
the generation
conclusions
reliable
this
by Saetre
indicated
is not
10 year
'In
significant'
quoted
by Saetre. -
On the basis-of
deviations
that
in the
from
the use of one-year's
reme
conditions
will
ext
result
of
predictions
,
data. Clearly the uncertainties
cannot be assessed from a single year's
be
data
is
in
likely
the
to
only
available
which
many
recordings,
in
Reliability
the prediction
applications.
of long-term
practical
of HI/s can only be significantly
distributions
then
by performing
sets
and its
climate
Consequently
here,
which
less
produce
between the
emphasis
It
is,
tion
variation
years
of Ht/3
theoretical
lower
the
conditions
in the upper
that
for
of the method of
the
fitting
of the
would
Gumbel distributions
theoretical
fitting,
graphical
believed
techniques
squares',
in the resulting
sea state
are less
than
'least
towards
are weighted
therefore,
extreme
or
may
exercise
from year
conditions
wind
of the annual
as a result
moments, considered
data,
storm
'short,
of HI/3-
in the middle
of this
to the. distribution
in the extremes
significantly,
variations
The results
adjustments
4.9
in Fig.
The plots
behaviour.
long-term
lead to intuitive
deviate
should
attempts
However, whenever
tends
which
to place
more
range.
the effects
variations
in the
of the long-term
distribu-
of annual
on the estimation
be minimised by fitting
the data to the relevant
distribution
using mathematical
methods which
probbility
will
sensitive
to the uncertainties
associated
the occurrence
with
of
extreme events.
Physical_ Assessment
4.4.6
Before
that
an extrapolation
employing
representative
of the
of the long-term
the extreme
conditions
1100-yearl
of a recorded
climate,
are realistic.
steps
Extrapolation
wave climate
should
'To this
of HI/3
as being
be taken
to ensure
imposed by
limit
at the physical
of H1/3 must be truncated
fetch
limitations.
duration
In
this
the
study
and
wind
speed,
wind
initially'arbitraril
1100-yearl
the
y
chosen
at
v'alue. of
was
truncation
distributions
H1/3
132 .
'indicated
in Fig. 2.3.2
At the Famita'locatlon
the fetch available for
is
high
intensity
distances
limited
generation'9f
between
to
of
seas
the
for
in
directions
400
2SO
and
miles
winds
other than
miles
approximately
These distances
from the North where the fetch is effectively
unlimited.
discount
likelihood
directional
to
any
of
magnitude
any
sufficient,.
of
are
bias in the wave climate since the vast majority of storm conditions may
from any direction.
be generated
Bretschneider's
Employing
curves
empirical
6m.
15;
HI/s
would require
of
=
Value
from the North-and
45 hour duration
for
30
wave forecasting
,a
29m/s. for
be
it
presumed that the extreme
may
comments
direction,
from
blowing
from
any
storms
result
From these
condition
feasibly
although
most likely
to occur
could
is
it
from winds
75
by
Saetre
Famita
analysed
a maximum value of
In actual recordings at
H1/3 = 12.3m. was observed from only three years of measurements. The
from
North-West,
the
these
to
over a
conditions
was
rise
storm giving
fetch of approximately 250 miles, having a peak wind speed of 24.2m/s.
fairly
the
was
constant blowing from the Norththe
wind
During
storm,
West and North, with an average speed of approximately 20m/s. over a period
On Bretschneider's
curves this wave height would
in excess of 24 hours.
for
hour
duration
14
24m/s.
to
a
corresponding
of
speed
wind
a
require
From this comparison it
hurricane conditions on the Beaufort scale.
for
location
Bretschneider's
Famita
the
curves predict
appear'that
would
for
those
the
actually
over
required
generation
conditions
wind
excessive
seas
of
If
of
HI/3
given
the extreme
sea conditions,
the 24.2m/s.
measured-value
27m/s.
for
fully
feasible
or a Force
developed
under
Similarly
form with
Pierson-Moskowitz
developed
and described
by the
Eq. (2.3.23)
using
recorded
of 12.3m.
are fully
spectrum
Pierson-Moskowitz
with
values.
10 storm,
conditions
the, 1100-yeart
sea state
of, the.,
Hip-
to exist.
This
be
133
to
in view of the-above discussion,, the wave climate extrapolation
1S. 6m. for Famita is not considered to be physically
heights up o HI/s
unrealistic.
THE EFFECT OF WAVECLIMATE EXTRAPOLATIONON LONG-TERMDISTRIBUTIONS
4.5,
,
LOADING
WAVE
HEIGHT'AND
WAVE
OF
and wave load have been derived
of wave height
distributions
The long-term
4.10
the cdf.
for
is plotted
data
various
H, 'derived
wave height,
of individual
sets.
observation
to be
from
basic
'one-year'
distribution
the
is
scatter
the
obtained
that
made
skewness' in the upper range resulting
diagram data exhibits
a 'negative
heights
in
this region
of
wave
in a considerable
under-estimation
compared with
to include
extrapolated
long-term
for
If,
storm with
a severe
present
in
those
of
excess
predictions
illustrated
be
will
as
climate
The cdf. s
Gumbel extrapolations
(B) and (C),
that
wave climate
than one-year,
a probability
of occurrence
long-term
The resulting
skewness'
wave height
with
obtained
wave
later.
100-year
the
given
in Fig.
respectively
too high
a 'positive
from both
of HI/3,
result
be appreciated
in under-
much greater
period
in the storm.
of H derived
should
the one-year
example,
a return
It
wave climate.
here to
taken
data
that
of the climate
H1/3,
1100-yearl
up to the
the complete
represent
based on an estimate
the predictions
in Table
4.10.
graphical
4.3,
theoretical'
and
as curves
are plotted
it
can be seen
more
slightly
extrapolation
predicts
example the graphical
from
long-term
the
the
s
cdf.
comparison
of
as
expected
conditions,
in this
severe
of HI/3
in Fig.
of wave climate
the remaining
As mentioned
using
is
preferred
of
moments
method
earlier,
the theoretical
extrapolation
The significance
of the choice of short-termapplications.
'assigned
in,
be,
HI/3
E{No+IHI/3).
to
to
of
class
each
rates,
zero-crossing
the data is also
on values
4.7.
reflected
in Fig.
from'the'Pierson-Moskowitz
4,10,
Curves
relationship
given
in,,
zero-crossing
in
Table 4.1),
computed
and.,
the departure
differences
of some of the actual short-term
reflect
from
Pierson-Moskowitz
form
discussed
the
as
spectra
elevation
surface
Eq. (4.5)
on the distribution
The effect
Three.
in Chapter
in
the lower
the
cdf.
s
of,
between curves (B) and (D).
departure
1100-yearl
the
Gumbel distribution
theoretical
obtained
rates
zero-crossing
of H is a slight
by the departure
range as indicated
for
for
within
classes
data
higher
given
values
of wave height
than curves
probability
levels
the entire
however,
estimates
value
directly
not
are
over
curves
Plots
comparable.
at
are predicted
range.
a more severe
represent
than
condition
using
for
levels
probability
the Pierson-
mean zero-upcrossing
Moskowitz relationship,
correspond to a long-term
Tz -T,
(A)
0.22
and
second
cycles
per
whilst
curves
of approximately
rate,
diagram values correspond to a value of
(B) which use the actual scatter
in the differences
between the
0.151 cycles per second which results
indicated
, return-period'scales
of a certain
return
period
on Fig.
for
curve
4.10.
1
in
T
departure
the short-term
In view of the
Z_
(4.5),
Eq.
diagram,
using
obtained
the sctter
the description
where E(No+IHI/3}
of the extrapolated
decreases
with
from
zero-upcrossings
should
wave climate.
increase
wave height
of curve (D).
Consequently,
in H1/3,
be included
To preserve
in
the trend
4.11
(4.5)
have
data
from
Eq.
only
values
this particular
(4.7)
for
Eq.
5.7m,
Hi/3
used
all other class
with
to
been used up
=
the remaining computations.
This procedure is favoured since
in
values,
follows.
long-term
R
the
still
cdf.
of
which
has
on
effect
it
negligible
TrAs
being
long-term
the
(D)
zero-crossing
rate,
unaffected,
and z
curve
in the zero-, crossing
insensitive
to
changes
rates of'shortrelatively
low
of occurrence.
probabilities
only
with
term conditions
for
application,
135
-
better
it
and
rate
been considered
here.
The 1.ong-term
Fig.
vertical
The results
to those
individual
short-term
have not been considered
distribution.
Figs.
WHI/Ap
weightings,
these
scatter
return
one-year
periods
cases
form is
the Rayleigh
wave height
and,
therefore,
have
possess
values
with
associated
Curve 1 represents
sets.
than
return
sea
Accounting
one year.
period
the
From Fig.
no extreme
conditions
loading
of this
sample contains
greater
by extrapolating
states
by changes
representation
this
that
with
extreme
present
the distribu-
long-term.
on the extreme
diagram
apparent
(HI/3)
simulate
illustrate
dependent
are critically
in the short-term
in the
and 4.13
4.12
states
loading
less
exhibiting
dominated
its
structure
characteristics
original
4.7 it is
because short-term
with
Inertia
following
peak variate
Hence, the peak loading and individual
distributed
Gaussian
although
of,
rates.
zero-upcrossing
to,
characteristics
Furthermore,
range.
are independent
force
of basic
lower
still-water-level.
showed similar
height,
wave
in the
deviations
significant
tions
for
observed
here as they
included
are not
length
member of unit
(HI/3)
using
for
to
curves
10.2m. with
The effepi
3,4
return
produced
and 5 incorporate-classes
periods
up to 13.8,12.0
and,
of 20.2,4,
by increasing
and-0.8
years, respectively).
-0
data
the length of the extrapolated
set,
136 -
high return
distribu-
of long-term
of the tail
of
samples.
of these
The trend
an effective
represent
purposes
data
climate
from
would
set
'complete'
the
100-year
by the
increase
the
earlier
return
period
single
wave climate
is
than
greater
be incorrect.
description
the
of
data
and even
In
absence
underestimate
extreme
the
these
is
This
could
by curves
illustrated
one-year
wave climate
presence
of a single
three-hour
curve
7, and a hypothetical
curve
8.
of HI/3
complete
population
almost
height
wave
Uncertainties
of magnitude
greater
the
from
to
set.
with
sea state
in Fig. 4.14
of
a probability
to the extreme
in the
sample whilst
applied
its
In other
distribution,
in
applied
sketched
associated
4.3.
weightings
words,
weighting
is
the weighting
= 15.3m. ) would be
Consequently,
should be.
P(FIHI/3
than
it
between values
short
storm
Gumbel distribution)
see Table
to the conditional
two orders
using
(1100-year
0.000003,
approximately
applied
1100-yearl
is
4.12
any
were contained
weightings
of the
observation
each observation
events
5) is
curve
with
in
unlikely
conditions
W(Hi/3),
1/2920 the weighting,
be 0.00034 in thelone-yearl
will
of
Ioccurrence
value
Since
storm
7 and 8 in Figs.
(associated
is
Of'
be overestimated
(111/3),
associated
extreme
rare
would
sea states
if
present
(predicted
was demonstrated
It
raw wave
be represented
to
period,
if
the
design
of 20 to
wave load
collection
of
However,
of
assumed
prediction.
a range
for
of the order
use
may
extrapolation
adequate
'one-year,
the
sample
the
the
by 9% and the
of
presence
that
seen
1100-year
the
set underestimates
data.
the
wave
raw
use of
in this short record then,
for
it
underestimate
data
for
are only
extrapolation)
that
will
lifetimes
by the
covered
1100-yearl
the
condition
wave climate
22% for
to
4.13
Fig.
Using
30 years.
that
saturation
design
when typical
to the duration
respect
suggests
extrapolation
wave-climate
well
with
curves
from curves
(single
year)
1 and 7 must,
therefore,
be expected
without
when
extrapolation,
137
-
vary
will
inevitable
is
it
Although
the higher
that
from year
considerably
intensity
to year,
sea state
occurrences
based on
the extrapolation
tting',
probability
of the HI/3 data histogram to a theoretical
,
from
depart
long-term
the correct
distribution-may,
markedly
-not
in
4.14.
illustrated
In the Figure a range of
Fig.
distribution
as
'moments-fi
hypothetical
(one-year)
short
the
considered:
'one-year'
Gumbel climate;
its
plus
of Hi/3
are
enhancement associated
1100-yearl
H1/3
1.100-yearl
the
the
and
the
storm,
of
presence
with
4.13,
Gumbel
includes
in
Fig.
a
wave
and
climate
which
only
considered
be
6.6m.
These
thus
HI/3
to
to represent
considered
may
up
of
values
'high
'average',
biased'
climates
intensity
For all
respectively.
of HI/s
distributions
tion,
sufficiently
distributions
height)
wave
to the extreme
is not
consistent
sea states
importance
in comparison
it must be
However,
the distribution,
the
lower
length
the
the
then
H1/3
of
wave climate
values),
of
in
degrees
high
predictions
of
uncertainty
short and
as discussed
it
(or
wave load
variability
is not restricted
distribu-
parent
long-term
in the procedures.
uncertainties
that if the wave climate
appreciated
1100-yearl
on the
sea state
the fitted
conditions,
be of secondary
will
the other
with
these
the effect
that
intensity
and 'low
to the
close
are very
close
biased'
sea state
P(HI/3),
and middle
is
record
range
too
are inevitable,
earlier.
interesting
is
climate
of extreme
techniques
value
From Fig.
prediction.
4.12
the
150-year,
extrapolation
eye with
adjuStMentOf
straighten
the plot.
compression
subjective.
tail
for
as plotted
truncated
only
is
the Famita
at the
either
fitted
dati
'one-year'
the entire
curve is
alternatively,
in
A
the Weibull distribution
parameter
of the
as a result
return
by eye to a straight
or,
Consequently,,
on
fitted
by
to
logarithmic
highly
height
the
abstracted
the
scale,
values
are
wave
of
For example, from F.ig. 2.3-3,, the 150-yearl
wave could lie
the Tange 26m, to 28m.
anywhere within
data
set only
winter
Draper
Using
this
71
Driver
predict
and
technique
for
the
a Value of 27.4m.
138
It
is also
basis
for
interesting
If
three-hour
that
1100-yearl
the
wave is
this
1100-yearl
of
sea state
severity
the stationary
by 111/3 - 15.6m.
characterised
to the
corresponds
derived
the relationship
using
the
is 30m. from
during
assumed to occur
forms
wave, which
design
Fig., 4.12.
to note
by
kn NJ 1/2
Hi/3
%ax
where N is
Using
height
the zero-uperossing,
for
the Famita
to determine
N, the wave
as 28.4m.
in a less
4.6
is predicted
location
method is unable
sea state,
rate
between these
The difference
latter
(4.8)
which might
occur
for
periods
well
1100-year,
than the
condition
in excess
hours.
of three
of this
variability
associated
a one-year
period
Chapter
with
wave climate
in approximating
measurements
the characteristics
the sampling
typically
covering
of the
long-term
wave climate.
diagram representaproposed here are based on the scatter
description
since this is the form most often
climate
tion of wave
designer.
As
form
the
this
to
a
consequence
of
of wave
available
in
intensity
the
sequential
variations
sea
state
presentation,
climate
The procedures
which prevents
interpretation
extreme
74#75.
of the long-term
waves,
observed
or load cycles,
in reality.
occur
r,orresponds to a probability
or wave load
cdf. s of Figs.
no account
is
often
which
wave height
It
of
large
4.10,4.12
waves,
and 4.13
a phenomenon'
of non-exceedence of (I
return
if
period
Although
13.9-
this
is
over a long-term.
correct
out that
between occurrences
of amplitude
associated
be greater
than T-years
because
period
return
average,
will
with
large
in
For example, if a IT-year'
together
time.
to
tend
group
wave
waves
height is exceeded twice in &,,.singlestorm.
between
then the time interval
'true'
the corrected
occurrences must be 2T, thus representing
return
for
period
wave height.
this
of this
Similar
reducing to approximately 2% at a 50-year return period.
but
reductions would also be applicable for the peak load distribution
in extreme value
in view of their magnitude in the region of interest
the effects
analyses,
may be considered
of grouping
to be of second order
of importance.
The techniques
herein
applied
bias
from no directional
suffer
extreme
are strictly
in wave generation
may result
wave conditions
to locations
applicable
from winds
which
in any direction.
For
by
where wave generation
may be severely
restricted
it is essential
description
that the wave climate
fetch limitations
in addition
to recordings
includes
wind direction
of significant
wave
The above techniques
height and mean zero-upcrossing
period.
could
locations
coastal
then be applied
distributions
to discrete
of HI/3
classes
being
available
distributions
according
in each direction.
of wave height
to the probability
the complete
truncated
of wind direction,
long-term
of occurrence
the
fitted
to the constraints
The resulting
long-term
be convoluted
band to
descriptions.
"1
'140
TABLE
DISTRIBUTIONS
PROBABILITY
46
OF H
1/3
'One Year'
Upper
Limit
(u)
MidClass
(m)
I
EIN+IH
0 1/3
W(H
1/3m
'0.6
1.2
1.8
2.4
3.0
3.6
4.2
4.8
5.4
6.0
6.6
7.2
7.8
8.4
9.0
9.6
0.3
0.9
1.5
'2.1
2.7
3.3
3.9
4. S
S. 1
5.7
6.3
6.9
7.5
8.1
8.7
9.3
0.1738
0.1734
0.1578
0,1487
0.1419
0.1381
0.1312
0.1224
0.1223
0.1192
0.10S1
0.1000
0.09899
0.1177
0.1003
0.07407
96
402
389
'321
24S
161
132
70
46
23
12
11
8
2
4
2
Data
E=1924
%.-o
ac
5
104
172
1S2
133
117
ill
63
43
23
12
11
8
2
4
2
OH1/31 . 4240m
E962
%0
>
cd
k
r:
in
0
$4
u
u
10
44
0
4)
bo
0
4j
M
in
cd
"-I
u
a,
93
4-j
rj
r-4
r4
.0w
cd
w
Q
u
r4
.0
cd
44
0
>
rj
4J
co
r-4
1-4
a
E {H
classes
1/3
21
914
>
-r4
4J
cd
r-4
.0,
ZI
classes
4J
r4
F-4
r4
0
0
1.4
1=EH
1/3
all
E {H 21
1/3
all
0
4-b
0
f. 4
.4
.0
.0
*E {H
k
bo
.0
u
0
tA
F1
14
(H, in) AH
P
/,
1/3n,
1/3
'p(H
H2
1/314,1/31n)
AHj/.
O. OOS2
0.1132
0.2918
0.4496.
O. S877
0.7092
0.824S
0.8899
0.9346
O. 9S8S
0.9709
0.9823
0.9907
0.9927
0.9969
0.9990
cd
>
$4
1/3m
P(II /3u
cd
>
4--b
1
9
W(II
Data
th
(D
0
r-4
cd
0
4J
tf)
H
tn
P(111/3u)
0
(D
in
co
4-J
cd
Winter
0.0499
0.2S87
0.4608
0.627S
0.7S48
0.8384
0.9070
0.9434
0.9673
0.9792
0.98SS
0.9912
0.9953
0.9964
0.9984
0.999S
EIH1/3}=2.2441m;
Li
P,
141
TABLE4,
.2
OF EXTREME CONDITIONS
PROBABILITIES
TO Tr
YEARS,
YEAR PERIOD
_LD
Period LD
(Years)
Return Period
Tr
(Years)
1
20
so
1.0
1.0
1.0
0.5
0.999999
1.0
0.2
0.9885
0.999986
10
0.1
0.878
0.9948
20
O. OS
0.641S
0.923
so
0.02
0.3324
0.636
100
0.01
0.1821
0.39S
142
4)
W
***
d-1
oo Ln cn oo mM
in
ad
r-4itnitni
"
kD V-4 M
00
-vt
4 t',
%D 0)
0 r-4r-46r-4r-400000
C;
C; (S C;
Cd
4J
(S
C;
(S
t-n M at
C14 v-4
%D 10
M eq
88
-.0
V-4
C!
r-4
%10
Ln
11
0866C;
C; C; (S C;
04
4J
r4
1eq 100
"
Lf)
00
r4
Ln
C)
%0
CA
-4
00
CA
M
"**
cl
V)
"
NO%0
C14 %0
MmM
Ln tln
r-4
-4
0000
I'll
6
66C;
6
6
V-4
(SC;(SC;C;C;(SC; C;
>9 cd
(SC;
-4'
r4
4)
0-43
IV)
41
x
C.
$4
0
z
0
pi
%.-0
P,
z
0
---
ao
%o "4
**
018'
01
01
m %.
o t-) " r-4
(71 Itt
r-4
Ln 4m co tn
r-4 v-4 P-4 r-4 00
I-
tn
C)
0)
006c;
?)
(A
Ln
01
r-4
*
1
*1
88
v-4 -4
0
10
t*f) (3) V)
"
P-4
8,
o
%0
"
o 01
P-4
NO
Ln
"4
")
00
mr
0
%.
D V-4 C14
%.
V)
"
P-4
V- 4 ". 4
<
cd
U
.rj
m
t--)
o-%
Ln "
"D
r-4 I-,
r-4 4m .*
F!
tn 00 Ln (7100 V) m o C14-4
.I --, C)C;
(D
1-4
0
Q
V)
0
4-1
cd
r-4
o
P
;
1-4
x
%..
0
C14
66
00000
C; C; C;
C;
11
F)
0
m
t4)
C-4 %
19) o
4
cs cs C; (s
8
CA Ln V)
%0 W-4 00
C; C; C;
V)
C,4 -4
C)
8 8
0 C; C;
C;
C; C; C; o
., -j
z
0
10
r4
". 4 4
000000000000666666666
-4
%D (71 U)
%D
6
6
C; C; C;C;
41
Ln C4 %D V-4 cyl T
00 Ln M
00 tn cn
NO
o V-4 00 0
%.
m
8888
c;lc; C;C;
l
E-
Cl.
94
-4
-t:p 0
e-N
:3
-,t1f)
%.
"4-0
4)
a
-4
No
-Rt .o
"D r, Ln Ln -4 Ln r- co (A (n at
C4 00 t- " Ln f- co C76
cn m
m C) C71
C71
%0
0m
trt4)
0) (A cn
t-1)
00
Ln
00
Ln
*0 t- DoCACAch CAcn cn CAC71
,
t
-4
m4
oo in
o
Ln
*a,
in
ci
oo
(n
mm
oi
oi
ci
ci
in
in
cn
cn
cn
m
--t
-3
0"
C71M C710*6 m C) cra Mmm
CA C71Clb C74
ICT %0 r- 00 01 M C71(7) mM
C;
C;
x
0-4 M0
"
11
C;
C;
C;
C;
C;
C;
C;
(S
Ln
C;
6000
4J
4
Cd
4)
91.
0-%
V-4
>co
I
8
11
r- Ln co
r-4
a
t4l
0
C)
"'
044
.*
0
L"
t4l
"I
o8
888C
A
Ln
%0
'
,
P-4t--O
r-l 0000
CA
00C;
1(C)
-4 I
tr)
r--4
%D -,I
r4
C;
C;
C;
C;
C;
co
0
"
r-4
r .'
ko
r -M
v r4
eq
4
Ln
r"
1
C;
C;
C;
C)
-4
r-1
t-)
C;
C;
P-4
C) C'
0
6
C;
C;
C;
$34
C)
%f
Q
In
cc
P4
%-.0
a.
te)
r% Ca C) 4m "
o c,4 -tr tf) r- oo co in
8686C;
.*
oo6C;
P, 0
+0
z
,:r
:
. I.
M
04
'it
4 Ln ")
tn M rC14 r-4 4
V) 00 C14tl-
mm
Ln
4 %--o 0,6
-4 r-
t-l
1; C Ct4 t
14 41
(D -H
%D N
00
94
C;.
1:
r4%-o
6-3
%o
No
66C;
C;C;C;C;
10 (A e-%
"Ln
o
cn
m
rte)
oo
oo
co
%o
0 0) t4l Iq co
0 Ln
m
al
(3)
-:r %0 00 00 mmm
ci
m
c,
4
req
rm
in
m
oo
Ln
oo
cy,
m
(m
Ln
(oN
(m
im
%o
m
r
C%
m
c)
ab
co
mmmmmm
0)
al
CA
cl m
(3)
%0
E-
6
Z
0
vi
Ln
e-%
14
'14TO
C4
v;
ko N
44
t4
C;
Ln
9-4 r-
t-1) M
L4 L4 14;1
00 -1 0
kn
,I*1
%0
cn c7i c7i
6
C; CL
ca
0
0
0
0
V)
M
fLf)
U)
"ll
Ln
1t
-R*
0
(n Ln C14(3) f- Irt eq
co I'D Ln M C14
C71 M CA 00
1-1 0000000000000
00
00
00
00
r,
r-
I-
r-
J-1%
WC
6
6
,
C)C;C;C;C;C6C;C;cs C;(s C;
Ln
-4 t-
0 Q
v)
4m Ln
r-4 r-
01
Ln
P-t t-
4
4
4
4
C3
C
t
C,;
"T 1-4t Ln
.
'.
V-4 q-4 r-4 4 0-4 7-4 q q:
r-4
0 04 00 It C) %-DN
%D eq 00 qj. (D N.
9r l*
NO
cz C; C C; C; 14 C r t4
00 Itt
%9 I
*6
te
)VP-4 Ln
W4 r-4 Ln.
1
0
%
4.
l b..
o
143
TABLE4
(Continued)
Notes
(A)
Estimate
(B)
(C)
of mean-zero
crossing
Cumulative
probability
values
read off
Cumulative
probability
values
calculated
from Table
(. l)
Figures
ensure
in Eq. (11.4),
underlined
are slight
summation to unity
figures,
as required
periods
of H1/3 in extrapolation
0.80 yrs]
using
Appendix
in numerical
Approximate
(4.7)
Fig.
E{H
to probability
routines,
associated
using
with
with
upper
theoretical
1/3
and aH
1/3
II.
distributions.
return
Eq. (4.7)
using
adjustments
on the probability
class
rate
values
to four
negligible
limit
to
significant
effect
of largest
Gumbel distribution
10.2m
144
II
lit
39
A
Ck
0
.<
cr0>
w
z a.
4
Zx
.40
UW
ri n
. I-
-i
E.
vi
<
(3<
s6 CL
ui
a
ri W tn
e0;;
0
N
V)
UJ - ;-/i zA
0:
Ix
.14
UJ UJ
'I
Of I.- W0x
w a: Mz
oll.
$0
qa
0
lp
13
_
w
d:
2511.
W
r_
Io
<I
cl,
-
0x..z0WX
to
42
>_
kn
lir
U.
cr.
w
er
W
Zia
gX0-
M0
oIm
0
3t 3t
C4
xa
044
d0
U.
4-
43W
WN
to
1HO13H
3AVM
-4
lqr
fm
0-
145 -
FIG. 4.2.
COMPARISON
2-5CONDITIONS
DIA
-0
5m
-. 7.*5m
DEPTH: 150m
IMMERSION
2 ) t-
Cm
2-0
Cd
1-0
I
I
zi
(LI
-j
Iz
D
r-
a
ui
IL
SIX MONTHS
WINTER
tj
u
ix
I.
$4-
ONE-YEAR
DATA
0.
Gaussian Scale
II
.0
.1
.0v
PR.013-(FORC E >F1 10- 1
'vz)
FORCE kF3
1
10-3
10-4
10-5
10-6
.I,
I,,
Z,
146
0..
I"
I.
a:
w
CL
U.
L1
cr
N I.
Ix
w
>
lp4,
it fn + In q1t 40 +
In
I+
+
C41
W
U
(A
+-
C-4
40=0
cr.
w
z
!q
N9,
'CIO
r
Qo
1
-
I
]
LL
H
R
133J 'NI IHSIIH
_1 1-1
-.
.1
.
'JAWAk INVOIJINSIS
147
-
ts
a
u
V)
3:
(D
w
m
3:
X
in
x
w
CL
co
2
<
X
CL
CL
<
X
z<
X
(D
cp
<
A
-V)
W
D
L9
uj
Q)
c6
z
I-z0
Lo <
LL
o
z
0
; -- 0
W.
D LL
ro
UJ
<
x
(D
ui <
I<
co
o
0:
cy-
-r
w
ui o
UJ
>
<
z<
u
w
ui
Z
w
I--
:: <
w
<
-J
D
W
Z
an
6
%0
0
IT
0
LL
ou
rl%
(W)
U)
W 'IH013H
3 AVM
CY
446
ice
y
0
Ui
x
uj
04
IL
LX
0
w
ui
.4
w0
W
414
CID
CL
.4
LL.
-a
12
10W
cow
9
1w I
x
Im-m-
Irk
16
.I...
NO
I
C,4
,
IIIII
mr, ao in v
0-,
.0
I'
149
e
. -.
,01
w
Z
0
re
w
w
Z
0
CK
in
0,4
"
't"
'. (! -4H013H3AYM)
:q
Ii IGRAPHICAL EXTRAPOLAT*N
OF OEST FIT TO Hit, DATA, POINTS
I
(11Y EYE)
A
I
)11
0-%
-. 0
HI/3
7-
5-
M//
/H
THEORETICAL!
.
GUMBEL DISTRIBUTION
FOR Hy, BY METHOD
5-
OF MOMENTS.
(SEE
4-
TABLE
4.1. )
3-
t-t tf
10
20
1-"
-1
50 100-
1-IA
Gumbel Pppfr
I (WAVE HEIGHT 'Ch I
10-1
10-8
10
10-S
P0106 EWAVE' HEIWT
FIG. 4.8.
0.20 1
151
PROBABILITY
1--0
-)L
0-07 -0
006 V0,05 -.
0-04 003 JL
002 -
0-010 -
y-
0005 0004
0-003 CL
0002
GUMBEL 'THEORETICAL:
DISTRIBUTION.
(FIT BY METHOD OF MOMENTS)
GUMBEL GRAPHICAL
00010 -0
0-0009 00008 00007
c=06
I
FAMITA ONE-YEAR I DATA
06
1-2 1-8 2-4 3-0 3f5 4-2 4-8 5-4 60 &6 712 7.8 8-4 9O 9-6
SIGNIFICANTWAVE HEIGHT,H,/3 (m)
.4
X0
UD
w
I-
. -01co
ed
#a
Iq
ow
C4
V-
40 "
V.
cid
";
""
';
:i
U.!
'U'
'
U.
'',
"
U. '
'"
H.
U,
lii
'U
'
%.,!
'.
'
U,
'-UU
U.
''
".
U'
'
IL
. ',,
qojq-.
,
w
xu
w
o
in
ui
cf.
al
K
ix
w
Lj-
64 wW0
oz
Ir tD
I-- to
ITt4
<
6-;,,
"-
w Znx0
0wzw
cr
P,
C.4
N
3: zZ062<j-t
g -0i
u
(L
w
a:
wu4i
;:
rw
%.. 0
ot <aa
%ft.0
z
w
WEtIAA
cr.
wW
ITin
zEw
1,
lx
a: ;:
ll
CD
WI
in F.
(L
-j
%A
n=
Xw
uj
z
0
-j
ik
67-710
tv
"
01
11
'.
m 'I. IN
44
IN, i
'"
-"
-4013H
410
I'm
40
PW
,
3AYM
IV
VM
aaa
r4 ... -6-Vh.
,IP-
woi
iwna
go. - 10,
I
qq - '14 ,
'U,,
154
0
Z in
0 ix
w
CL
at
+
10
Z
-J
W
#-
w
M
E
CV
u 1%z
< 'I
IL li ix
p0z z
4A- 0
LL
'41
0-) -
in
2
1
()=)
.
. rW
U)
CAHION
a)
I
Iq
II
In
0
LL
0
.14
0z
C)
n
-1
in
-
co
N
Pl
+0
z
w
ir
td
-/
I,:
ZW0
0
N-
ul
<
C)r
o
q Cc :3
ui 6a
aa
ul
0z
0:
02
Z
w;
0w
7t"( r
=)
0
:)
I< -<z
0:
:3" Ld
or
<u
U.L A ui
In
JJ 0
7i
.
Ull
in
Ml)/
7F
LA-
r-
f/1H
*ON
ui
uj >
(L
cr
I.
X-
-6
g
-1
CL >
in
0
T
cm
LLI
0W
z
q
u
liw
rr
2
cn
(V
I. -
ix
0 LL
0
jj a
w
>
Gow
J (L
0u <
ui W IX
Ld > x
V)
uj
:)9
P
.f.
>
9
3:
0
+-.
cm
I
z
0
I <0
cr- u eU) o 0vl_
4
U.
mIm
-1
z Iz
<0
(A ix!
q
< -jEw
cr
di E
a-
IXNUJ W-
I
.;.
44
155
-,
w
Z
w0
Co
w
Z 0<
cr.
Z
0
0-
Z
tn
_j
00X
0:
, --Z
CL
X
wwN
Ui
w
YE
0
W
iL
X
w0
CL
90
Ld
r
tL
UJ
0J
crw
CO(!)
E
W,
Ir
Z gL
UJ
z
0
t4
Z)
(' to
,
<Z<
Lj
ir
>w>w
:ww
w
>-
-1
CL
V)
-i
Z:
z5
00>>ZZ
).-
ri
cr
w<
w'
ir
rr
0w
ia
CC %A
w<ww
<
ZWJ
Z
0:
U (D
ir
w
(D
0
lw
uj
al
-1
CL
LL
C)
ir
w
V)
C)LL
iL
V)
KA W
b-- Q0
a0ui4ZD
>
Z
0
vi
w
LL
w
LL
LLI
w
Z,<
b- r,
-J
<
Z
12
Co
ZW
(13
dz w
w
m
ww
C)
C)
rx
w
a-
S2
co
0
>;
Ww
iL
Z00
01
-0X00
(98
17-1
0
m
LL
co
CM
IV
N
E)
N
cm
(W) IHV13H
co
IT
-3Avm -ivnalAI(3NI
04
'H
co
156
eN
CY.
0
Z
\@
-0-0
me-to
Z0
Z
CY
N
02
8
rr
W
0:
CL
4
zw
5 t;
ta
Ix
w
o
b- 1.t
X CiW0
X
w V)
1--
Z
wW
JE
r*:-
<
b<
Z
0
w
cr
tii
a.
WW
<0
ir
in
:
:E
LLW
0Jcil
Z
01<
wi
U)
CD -4
Ir
- w
LA
w
tz
vi
--k- ).
<
ei
w
>
I.
to
10
0 ;1
E dr
Ili
Z
-<
Z0
_j
KLI
LL
< ir
C) tr
W
-i,
EEE:
mUw0Z
c;
0
tn in 0.0
r-;.
cm -Z tn Co
<w0 Q.
t1
Lii
i2 :
w le
CLin
LL 00
LL
0 12L
u
wz
LL 0
LL
w
ca1
<4q
(1)
b- Z
4 tLj rx
wx>
ow
:
Z,
0
ei)
ww<<
w>>:
00:
MO
kU00
0-
-i
_j
WJ0:
Ww
i
(DW
)Z
DO
00
LAJ
38
:r:
ir
<
w
WW0
>
jc
Co
:ZZ
ir
I
tn
(WIN!
) HION31
I'Nn
(x
c
>-0
ir
D
-3
CL
lj3, j
.3: )80j
UV3d
#4
157
-
z
0
LD
0
CL
I
E
Lo
, ,
8
0 in !2.F
av
-c 3
lj
IL
LL
0
C. E
EW
%. 0
T----j
i -
z
ww
F- I
0
2
U)
Z
0
Znz
LLIZ
in
z
00
.0
LL
0
:7 -0 0
al
0
:r
w
Ix
LL z
0
(1)
z z
0
ro
".
\
\
LD
z
c
9L
00
F-
00
LL
0
0
I-
T--
.--
JC)
0
CL
(D
LL
ei
..
co
0
loool
ck:
Iq
MCI
w
w
x
w
p.
1:
1
0w
w 2
431
9
Cld
'-.
--1-
158 CHA'PTER
FIVE
ANALYSIS OF
PROBABILISTIC
STRUCTURAL RESPONSE
INTRODUCTION
5.1
description
The probabilistic
component of loading
is inadequate
to effectively
model all but
members. Extension of the probabilistic
methods
the simplest of structural
is, therefore,
required to, describe structural
response, such as stress
for
large
individual
deformation,
systems
comprising
of
a
number
of
or
a single
load components.
An extension
along
is
lines
these
developed
in
behave
an essentially
which
of structures
is not
frequency content of the excitation
of the
and consequently
structure
in this
Chapter
for
static
manner,
where the
the response
frequencies
to any natural
close
of the
analysis
is not
structure
is
This
for
dynamic
to
assumption
acceptable
many
magnification.
subject
fixed
discussed
The
structures
of
as
the
earlier.
generation
current
of
for
load
is
to
structures
which
all
restricted
member
components
procedure
may be expressed
in the
'Morison'
form subject
currents
and excluding
absence of uni-directional
in
loading
the
splash zone.
of
effects
The structure
volumes,
is
idealised
in the
to wave activity
the intermittency
into
as accurate
a model of the
set of nodes chosen so as to retain
behaviour
Linear
is
possible.
assumed where response
as
real structure
linear
loads.
the
as
a
combination
of
nodal
expressible
are
variables
discrete
pdf. s of short-term
Theoretical
loads
the
nodal
of
pdf.
variate
multi-variate
the
locations
response
which
can be expressed
Gaussian process
describing
of these
In this
nodes.
may be obtained
the water
itself.
in terms
particle
both
for
the correlations
which exist
account.
loads,
,
hence
the
and
non-linearity
nodal
and
motions,,
fully
motions
at
of response
loading
159
is
in Section
formulated
system,
5.3.
Extension
for
to
limitations.
However,
2.3.6),
Section
properties
the short-term
probabilistic
of response may
for which are
moments, expressions
of these
typical
compared with
including
5.2
are presented
from other
results
load or response
methods of
time-series
mathematical
in Section
methods to
prediction,
simulation.
5.2.1
distribution
distribution
Gaussian
process
locations
pdf.
of Wave Load
By analogy
force
idealisations
structural
of the probabilistic
of application
some results
as
properties
Long-term descriptions
of response- follow
of long-term
developed for the description
I
in
Section 5.5.
briefly
Finally,
force
to the univariate
made in relation
it is evident that the n-dimensional
wave
is a non-linear
involving
of each force
transformation
the particle
velocity
of the 2n-dimensional
and acceleration
at the
component.
X2i-1
where
+ X2i
2i-lIX2i-11
=uiK
Di
and
X
2i
i=1,
iK Ii
(5.2.1)
(5.2.2)
be
from
force
the 2n-dimensional
obtained
pdf.
may
The-n-dimensional
(Eq.
(1.53),
by
Appendix
one):
density
given
Gaussian
160
P(xl,
0.
11T
2n) -
exp{YrD)--e-t--FM]
2)n
(XIT
(XI,
X
,
where
...,
2n)
[M] - matrix of cross-covariances
for
cross-correlations
of
fM
which
are equal
(5.2.3)
to the
R12
Ri I
{X)}
[M]-l
Ri
2n
R21
R2n,
with
Rij
R2n
2n
,
(5.2.4)
= E{X ixiI
from
is
the,
the
E(Xi
Xi)
second
moment
corresponding
of
obtained
where
from
Section
2.3.5,
for
follows
example:
which
motions
particle
Rik
X4}
= E{Xj
60
= VrK-DI.K12 E{Ut
and
EfU1162)
(T) II
[Ru
T=o
IC12
force
The 2n-dimensional
pdf.
p(XI.
p(FI,
...,
2n
Appendix
1.3.2.9,
(see Section
transformation
may be obtained
... 2
using
the
One):
2n)
(5.2. S)
JDet [J]j
Fn+1 to F
the form
are
auxiliary
variables
of
2n
in
between
the
n),
which
are
necessary
mapping
F
=x 2j-I
. ..,
n+j
determinant
Det[J]
is
Jacobian,
the
the
and
spaces
of
two
probability
the
is unity.
for
the
system
above
which
force
where
components
; (j = 1,
n-dimensional
The required
by integrating
(5.2.3)
pdf.
p(F1,
F2
....
of force
Fn)=ff
thus
determined
from Eq.
variables:
00
p(F1,
_Co _C
n-fold
is
F2,
....
2n)
dF
n+i
dF
2n
(5.2.6)
161
Response of Linear
Quasi-static
S. 2.2
Structures
[A]
{F}
[A]
is
(5.2.7)
a transformation
matrix.
{Y)
from the stiffness
Nodal response variables
are obtained
equation
is
be
behaviour
the
to
structure
assumed
the
of
and
quasi-static
when
linear:
(5.2.8)
where
[K] is
{YI
the structural
[A]
[K]-'
=
{FI
[S]
=
The multi-variate
the multi-variate
stiffness
matrix,
hence:
{FI
(5.2.9)
pdf.
of nodal
pdf.
of force:
response,
1)
({F
p
Det
In order
(S. 2.10)
from
force
domain it
is
necessary
be
[S]
a square matrix.
must
thus
(5.2-9)
into
integrating
Eq.
them
and
the
out
variables
of
auxiliary
in the same way as the auxiliary
(5.2.10)
Eq.
forces
pdf.
multi-variate
in
the
previous
removed
were
Section.
form
indicated
is
lengthy since
the
above
extremely
of
approach
force
derive
of
the
pdf.
or response requires
n-dimensional
an n-fold
to
Furthermore to obtain the more practically
applicable
integration.
A direct
marginal
response
additional
(n-1)
unwanted
(n-1)
variable
YI:
distributions
fold
integration
response
variables.
(see Section
1.3.2.3,
must be carried
For example,
Appendix
One) an
the marginal
pdf..
of,
162 -
dY2,9. dY'
n
p({Yl)
P(Yj)
-00
(n- 1) fold
S. 3
(S. 2.11)
OF SYSTEMSOF TWOLOADCOMPONENTS
RESPONSE
and Marginal pdf. s of Response
Bi-variate
5.3.1
Consider the system sketched in Fig. 5.3.1 of two load components F1 and
F2 and assume response variables Y, and Y2 to be linearly
related to these
i. e:
loads according to Eq. (5.2.9),
'
Yl.
Sil
S12
Mor
Z-
4,
S22
S21
y2
S
where ij
{FI
the bi-variate
coefficients.
of the previous
ff
where
[M]
synunetric
expf-
vrD-e-e-tFM]
-00
I?
II.
I M-K
2
et
I dul
LM]
dU2
(S. 3.2)
matrix
be shown
can easily
is:
of response
pdf.
it
section
00
P(YIY2)
(5.3.1)
F2
influence
are constant
the procedure
Applying
[S]
.,
that
FI
of particle
kinematics
20RR
CrUl
2
Al
0
-R
1U2
Ill
UIU2
R.
.
UIU2
-R
.
UlU2
20
Cr
U2
UIU2
LRR.
UIC12
(ARG)= {Cil
UIU2
.0
UJU2
2+
2{C12
C24
C22
f1l
CII
ill
Ill
02
U2
2+
+
C33
C13
C34
U2
Ul
U2
2+
U2
C44
+
C14
f12 21
111 112 +
C23
ill
U2
C121
[M]
C.. are co-factors
of
13
(Fi
16, =
-K DI ullull)/K
Ii
62
112JU21)/K
(F2
J2
=
-K D2
Fl
and
F2
(S22
Yl
S12
Y2)/Det
[S]
(Sil
Y2
S21
Yl)/Det
[S]
follow
s
pdf.
The marginal
00
P(YI)
P(YI
Y2)
by integration,
dY2
for example:
(S. 3.3)
163
-
As a linear-combination
E(Fj
E[XIIXIIX3lX311
S12 E{Fj
by Eq.
(2.3.49)
+ E{XIIXIIX4}
of these kinematics
degree
has
load
of symmetry,
a
of
direction,
pdf.
it
or sign,
F)=
2j
p(F
p(-F
follows
XJ
S. 4, a solution
The particle
.
symmetrical
loads
+ E{X2
in Section
inducing
since
and:
+ E{X2X3lX31}
involving
Xi are solved later
The expectations
60
by Borgman
having been developed previously
(5.3.4)
kinematics
processes
and
the
equation,
retain
that
the bi-variate
in that:
-F 2j
li,
and
p(Fli,
In other
-F 2j
words,
)=
p(-Flig
that
2j)
of the bi-variate
part
pdf.
contained
within
any
domain
define
F,
F2
the
the complete
of
of
quadrants
and
two adjoining
densities
the
associated
with the remaining quadrants are
pdf. since
by
180*
this
through
probability
surface
rotating
about the
obtained
origin.
between load and response is linear,
as the transformation
is
in
bi-variate
degree
the
symmetry
present
of
response
same
the
by:
domain, as illustrated
Furthermore,
p(F lk ,F 2j)
P(Y
c;
li ,Y)= 2j
p(-F lk'
c=
F2.
t)
C=
const.,
P(-Y
ii
from Eq.
(5.2.10)
-Y 2j)
follows,
therefore,
that
164
IW
1
P(yl
0'
P(Yl)
00
Y2)' dy2;
[P(yl,
P(y2)
* P(Yl a 42)]
Y2)
dYl
0
(YI
domain
:0;
the
only
where
5.3.2
Method of Computation
pdf.
From Eqs.
(5.3. S)
Y2 4 ')
is considered.
Bi-variate
of Load
(5.3.3)
and (5.3.2)
it
follows
that
a 3-fold
of marginal probability
computation
required
for
its
integration
further,
required
cdf.
with a
integrations,
do not exist for'these
and solution
in
to
that
a
similar
manner
numerically
obtained
domain and described
briefly
in Chapter
of response,
'Closed-form'
must,
applied
is
integration
densities
for
variate
Using
solutions
therefore,
be
in the uni-
Three.
The computer programs developed for this procedure are indicated by name
in
Appendix
but
Six)
follow
(and
they
the
here
summarised
same
are
only
format
the
to
as
programs
referred
and
notation
and
earlier
general
documented
in
OSF2,
by
Reference
10.
Although
the
program
characterised
include
programs
time restrictions
fixed
Step_Fixed
of long-term conditions,
of this facility.
Range Numerical
density
ranges
of the response
run-
Integrations
is
found to be acceptable
9 Y2 4 8(YY2) are generally
with step
(where
domains
in
both
0.2a
a
the
represents
relevant
standard
of
widths
is
Rule'
tSimpson's
to yield
the marginal
applied
deviation)
when
densities
Velocity
numerical
components ul
integration.
achieved using
with Simpson's
step widths
of 0.2a
stability
over
step
a range of 5a in association
Rule summations.
Under certain circumstances the accuracies of the above numerical integra" as will be illustrated
later,
hence
inadequate,
is
and
attempts
tions
by
increase
the
to
accuracy
be
considering
shorter class widths
made
must
6rror
the
to'reduce
variables
of
accumulation.
ranges
expanded
over
165 densities
To minimise computational effort the bi-variate
are taken as
where the corresponding absolute
zero and the, computations curtailed
(5.3.1)
from
F2.
Eq.
forces
F1
and
of
exceed 10a of thoir
values
For
the above conditions computer run-times are
values.
marginal
CDC
65
ICL
7600
the
seconds
on
computer when using tho
approximately
(OPT
2).
of
operation
mode
=
optimised
Program Validation
To check the procedures the identity
Eq. (5.3.1), such that the bi-variate
loading
the latter
result,
uni-variate
solution.
A sketch
of a bi-variate
in terms
of the conditional
resulting
from application
matrix
distributions
and marginal
pdf.
of loading
distributions
is
of
5.3.2
and 5.3.3,
p(Fj
For the
in
the equivalent
made in Figs.
OS17.
of program
[S]
for
was applied
F2lF2)9
indicated
conditions
figures
these
2, in the extreme
curve
in the numerical
inaccuracies
by
caused
range,
integrations
departure
this
being
be explained
as will
later.
between the curves
The fit
load or response,
either
most
range
'linear'
5.3.5
a 2.0m.
1.0m.
diameter
the marginal
accumulation.
results
for
cdf.
To further
However,
slight
the marginal.
lowest
deviates
illustrate
system,
typical
of the
domain where, for
the
with
condition)
distribution.
and S. 3. S is
in the bi-variate
'inertial-dependent'
or
due to error
3 on Fig.
5.3.4
from computations
obtained
results
in Figs.
kurtosis
(the
in the extreme
this
see Fig.
condition
effect
curve
for
the
load
on the 2.0m.
S.
3.6,
1
Fig.
from the unishows a deviation
on
curve
diameter member,
distributions
2,
whilst
marginal
the
obtained
curve
when
cdf,
variate
follow.
diameter
S.
0m.
is
this
curve
exactly.
of
member,
other
follows,
be
behaviour
as
explained
noting that the cdf. of, the
may
This
kurtosis,
lowest
to,
the
assymtotes
extieme probability
with
of
variable
be
expected.
Consider,
for
as
might
example, the
variable,
the other
166
in Fig.
conditions
at a probability
from Fig.
directly
F2*
conditional
2.3.5)
are
in computing
truncated
in
contrast,
integrated
F1 = 8a.,
at
mass beyond
the
distribution
marginal
a greater
showing
also
indicated
the
marginal
limit,
at
possibly
spread
F2 and it
counterpart
will
of
F1,
in
larger
with
the
tails
similar
on Fig.
5.3.5
from
cdf,
curve
1,
which
and the
SaF2
it
follows
that
the
The
is
seen
the
is
force
the
marginal
to
contain
its
limit
curve
the
2,
its
than
distributions
conditional
of
the
'peaky'
more
limit
of
F21FO
p(Fj
because
domain
that
above
to
mass of
F2*
kurtosis,
univariate,
that
suggesting
,
be enhanced in the conditional
likely
F1 close
of
This
characteristics.
exhibit
about
cdf.
of
be expected
might
the
yield
are
probability
6a
approximately
limit.
this
that
BCF it is unlikely
any significant
F1
by
be
truncation
at even
omitted
will
to
they
whilst
of
2 on
From curve
represents
5.3.4
of F2 beyond 6a F2 the
the cdf.
determination
the
probability
F21F2).
p(Fj
probability
significant
this
(which
of kurtosis).
the value
given
distributions
densities,
In
2.3.5
at 8a F1 in Fig.
may be obtained
truncated
0.999965,
of approximately
Consequently,
6cr
where Force F, is
from Fig.
S. 3.5, '(or
Fig.
5.3.2
6a
F2
departure
occurs
characteristics
is
between
below
this
may
distributions.
be
it
in
deviations
that
to
can
concluded
avoid
the
comments
From
above
(or
F1
limits
F2.
Y1 and Y2) should
the
the
of
ranges
of
and
the extremes
be
deduced
to
cumulative
probabilities,
equivalent
which
may
correspond
from Fig.
objective
kurtosis,
force
be
or
response
must
provided without
extensions
increase in step widths if the accuracy of the numerical integrations
It
be
be
must
affected.
appreciated that with the
to
adversely
not
in rangesof
is
error
of run-time
of approximately
restrictions
distributions
6x
10-5%4
there will
11..
have confirmed
It
the
procedures.
computational
now remains to
of
the validity
to program OS17 which were introduced to
describe some modifications
Notwithstanding
the computations.
optimise
Integrations
to reduce
of Gaussian-Hermite
u,
bi-variate
where f(xljp
in the integration
3.5)
Using
this
technique
mn
(m)
(n)
of
the
C33
j-1
Ak
kl
f (X ijox
20
(5.3.6)
(ARG*)
,
r1
= exPl- -! -De-tt-FM]'
20
= (ARG) - Cil
(ARG*)
c33
u2k
-
ulj
u2Ij
[M]
2 Det
2
2k
Ai
.
by application
as:
2EAi
K K
jr)
l(CII
Il
12
IDet[S]1(2
Domains Using
run-times
2 Aet
--T-m
Y2)
P(yl
computer
may be expressed
pdf.
Over Velocity
(see Section
quadrature
variables
velocity
and
results
X2
2k
C33
(M)p
Ak
function
(n)
of Gaussian
coefficients
liermite
with
quadrature
xj#
the
2k
of order m, n,
respectively.
Some instability
illustrated
is
of solution
in curve
5.3.4
3 of Fig.
in using
experienced
for
the
situation
the method,
as
when m-n=
20,
distribution
However,
results.
oscillation
where
in
less
is
than
that
the
marked
present
uni-variate
much
effect
this
(see Section 3.5) and decays as the kurtosis
the
quadrature
of
application
decrease.
Use of this method is, therefore,
loading
components
the
of
dependence
drag
low
in the loading
input
for
of
conditions
acceptable
and with
reduced
m=n=
20 typical
to about 20 seconds,
marginal
run-time
only
requirements
for
for
Program OS17.
168
- .
Range Integration
Erograms
for
Response Variable
Y2
Fig.
where:
F2)/v/EiFI21
E{Fi
rF=
is
F21
E{F'l
and
Et-F221
Eq. (5.3.4).
with
given
(5.3.7)
Considerable
limited
have
over
mass
only
a
probability
pdf.
integrations.
in
the
covered
it
is
evident
to
the
4 in
is
integrations
excess
curves
of
force
to
a better
be expected.
be evenly
spread
these
(in
fit
to
This
procedure
over
the
entire
ranges
be'inaccurate.
will
effect
and correspond
10OMX2
yielding
a
"
'2
The behaviour
of these marginal
with
x,
in
errors
in
probabilities
deviating
still
this
cdf.
assymptotic
erroneous
considered
4A showing
demonstrate
accumulative
Normalising
range
figure,
fixed
the
over
on the
but
cumulative
unity.
but
5.3.2
of
a result
by the
distribution
might
Fig.
rF " 0.90.
yielding
probabilities
of
and 5.3.5
coefficient
distributions
the
of
conditions
integrations
that
and indicated
5.3.4
Figs.
correlation
in
5.3.7
Fig.
previously,
considered
Curves
from
these
the
s,
by dividing
in
uni-variate
in
significantly
accounts
numerical
high
the
probability
cases
force
the
for
at
force
the
cumulative
the
extreme
1.0)
excess
of
cdf.
than
the
extreme
the
error
region
produces
the
original
range,
by assuming
as
it
range.
in the
procedure is to assume that the percentage
A more logical
errors
distributions
integration
the
force
conditional
of
of
are of
numerical
It,
follows
therefore,
that the bulk of the
of
magnitude.
order
the same
in
densities
be
in
the
low
the
resulting
marginal
contained
will
error
force
of
range
This hypothesis
the upper
condition
probability.
where the conditional
masses are the largest.
by fitting
the marginal
may thus be investigated
cdf. at
I
obtained from the uni-variate
extreme to the probability
the remainder of the cdf. in reverse from this
and plotting
169
-
point
the originally
using
levels
successive'class
error
to the initial
of loading,
level
class
yields
such an approach
computed probability
extreme
cdf.
5.3.4.
cdf. s in
associated
constrains
the origin,
marginal
in Fig.
procedure
adjoining
a modified
distribution
the uni-variate
This
steps
which
In view
Fig.
5.3.5,
with
the
all
F-0.
Applying
follows
exactly
of the error
from the
resulting
discussed
truncation
the fit
as
adopted,
earlier,
range
force
In
level
be
5-6a,
the
a
at
region
of
made
should
case
the
uni-variate
cdf. again results.
with
agreement
exact
in the
in this
force
for
which an
to minimise
to the marginal
between
response
integrated
pdf.
over
p(Y1 Y21YI)
error
a floating
In these
range centred
and covering
only
in general
neither
OS14 and
in the integration
response
Y2 is
the
thereby
optimising
constant
mass,
probability
densities
Y1.
for
of
the
marginal
tion
Since
respectively
accumulation
response
is high.
variables
to replace
Yj nor Y2 will
step
be Gaussian
numerical
distributed,
integra-
the
Y21YO,
define
to
the
of
the
variance
p(YI
and
required
mean
of
values
Y1,
level
They
however,
be
Y2
are
unknown.
of
any
may,
at
Of
range
Gaussian values,
Section
by
1.3.3.2,
the
equivalent
see
approximated
Appendix
One:
E{y2lyll
I-- ry
and
a2
Y21YI
isthe
ry
where
'
21
YlAty231/Ety,
2- Cl
'2 ai2
ry
form of Eq.
(S. 3.8)
(S. 3.9)
(5.3.7).
170 -
These estimates
distributions
bi-variate
this technique
response
the
resulting
of
region of the probability
mass of
appears to locate the significant
p(y,
if
Y21YI)
covers
about
il6ay2ly,
centred
E(Y21Y11the
mean
at
an improvement
Although
producing
is not sufficiently
approach
sensitivity
of the numerical
limitations.
Consequently,
is
instructive
the multi-variate
Using
additional
inaccuracy
significant
for
analysis
must be expected
in
investigation
extension
integrations
of
of application
of
response.
ICL CDC 7600,
of at least
20 seconds.
at least
into
limitations
on the
5.2,
the
structural
mode of operation
Section
of error
given restrictions
on the
imposed by computational
integrations
here to demonstrate
the optimised
programs
the problem
the
Systems of Loads
Tri-variate
It
are high,
of the Impracticability
Illustration
range integration
suppress
investigations
S. 3.3
to
refined
when correlations
accumulation
fixed
on the
65 seconds whilst
With
loading
the tri-variate
computer
reference
to
in
to
those
the
applied
numerical
bi-variate
case, the number of computations of the argument in Eq. (S. 2.11)
factor
(20
least
81)
for
by
1600
OS14/
increased
be
=
of
at
x
a
would
for
bi-variate
the
those
system. Furthermore, a
required
OS15 over
integrations
of a similar
sensitivity
such run-times
of error
accumulation
bi-variate
loading,
the
tri-variateloading
were available,
experienced
these
problems
applications.
in view
in extension
might
of the increased
problems
from uni-variate
to
be expected
to be excessive
in
171 5.3.4
in
Systems Using
Bi-variate
using
included
Also
to 5.3.10.
5.3.8
Figs.
computed
cdf. s of response,
Some marginal
in
of Response from
Distributions
--Program OS18
Marginal
figures
the
from
fourth
the marginal
obtained
moments
statistical
the second and
Gaussian cdf. s of response.
kurtosis
the
equivalent
and
the
pdf-s,
the theoretical
addition,
from
to
presented
the
is
ation
is
This
Gaussian
from analysis
is
The hypothesis
in Figs.
tested
from the
cdf.
pierson-Holmes
found negligible
being
there
Six)
was written
jkppendix
Eq. (2.3.54),
-y over
a range
a range
In
Fig.
the
and
from
using
suggests
the
of
loading,
the
Gaussian
produce
drag/inertia
general
structuro
tests.
to 5.3.10
by plotting
the
moments of response,
deviation
the results
between
in the
Figures.
step
for
P,, (y)
81 steps
with
from
P-H distribution
integration
obtained
for
from
variable
dummy variable
5.3.8
a good fit
equivalent
is
P-H cdfs,
accumulation
in Section
departure
achieved
between
the deviations
The marginal
considerable
following
condition,
correlated
highly
5.3.4 and discussed
in
Fig.
4
to curve
which
exist
integrations
in the numerical
5.3.2.
the marginal
over
cdf.
in
resulting
as discussed
Fig.
However,
5.3.9
for
the procedure
in Section
distributions
probably
this
applied
5.3.2,
shows
that
loading,
of -5GI -
error
previously
for
from laboratory
to compute the
constant
system
5.3.8
loading
the transform-
of
of moments included
two sets
the
systems
the proposition
North
the Southern
that
apparent
to that
dominate
linear
through
estimates.
conditions.
effects
records
of strain
form
seen
the Gaussian
short-term
input
has supported
Tickell55
loading
for
pdf.
evidently
output.
immediately
may be of a similar
of response
multi-variate
since
in
force
the Pierson-Holmes
namely
from
is
The linear
nature
and hence non-linear.
to response for the general multi-variate
distribution
the
it
Figures
deviates
of response
drag dependent
from
in the
and are
'marginal'
computed
In
computed
included
S. 4, are
the numerically
with
agreement
show close
in Section
developed
the expressions
values
of
are estimates
172
normalisation
is
distribution
is
deviation,
Fig.
a result
densities
illustrates
5.3.10
the
bi-variate
predominantly
distribution
and exact
agreement
is
if
obtained
follows
the
It, therefore,
that
at the extreme.
inaccuracies
integrations
in
the
of numerical
fitted
in the marginal
concentrated
Finally,
the fit
improves
at low response
instability
4-fold
levels.
from the
of solutions
integration
numerical
the
for
P-11
The plot of the corresponding
condition.
inaccuracy
degree
from only a 2-fold
of
a
similar
shows
inertial
integration.
statistical
of the above analysis supporting
References 11 and 14.
Further
pdf. s.
these statements
results
are presented in
distribution
be
that
the
probability
of
therefore,
concluded
may,
described
by
is
Piersonthe
bi-variate
for
accurately
systems
response
it
follows
directly
Moreover,
that this statement
distribution.
Holmes
It
load
to
be
components.
systems
of
any
number
of
cover
generalised
may
at this stage that it has not been found possible
It should be explained
to prove
mathematically
that
the probability
structure
of a linear
following
distribution
P-H
the
variables
also
random
of
combination
However, as mentioned earlier,
the proposition
that distribution.
follows
from
by
the
experimental
measurementS55.
results
Is supported
fit
the
of response to the P-11
of
goodness
of
Further evidence
from the results
later
is presented
of mathematical
distribution
In view of the problems
simulation.
in
numerically
accumulation
series
appropriate
in
to make comparisons
insensitive
moments.
equivalence
cdf. s, it
is
of the probability
of the statistical
to such errors
between statistical
purposes
derived
associated
as discussed
with
considered
earlier.
moments of up to fourth
between the probability
order
error
to be more
structure
moments, which
Close
implies
distributions
time-
of response
are relatively
agreement
for
practical
producing
the
4-
173
-
from multi-variate
systems may
from knowledge of its second and fourth
Accepting
order
kinematics
of the particle
developed in the following
statistics
procedure,
moments.
statistical
response
at the nodal
sections,
This
many of the
avoids
A particular
nodal
expressed
in the form:
F1 + S2 F2
Si
Clearly,
since
distributions
symmetrical
E[Y
coefficients.
forces
all
k,
E{Y 21=
4)
E{Y.
of Y is
and have
all
given
i. e:
be zero,
ESi2
i=l
nnn
E{F
(5.4.2)
by:
n-1
=ESi4
study
odd integer
the fourth
Similarly,
then
=0k=
The second-moment
may be
+SnFn
....
Sn are flexibility
where SI,
taken
variable
response
2) +2EES.
i
i=1
j=i+i
Si
E{F
(5.4.3)
iF
moment of Y is:
E{Fj
41
i=l
n-1
nn
+6EES.
i=i j=i+i
E(F i3F
+4EESi3si
i=l i=l
jii
2S
n-1
n-3 n-2
+ 24 EEEEsisis
i=l j=i+l k=j+l
2E [p
2F
21
n-1
n
12 EE1:
i=l j=l k=j+l
jji
kii
S. 2S
S E{F
kijk
2FF
n
Y.=k+l
ED
F
Fk
kSt
iFi
1)
(5.4.4)
force
joint
hand side of
the
of
given on the right
moments
of
Evaluation
and (S. 4.4) may be achieved by expansion of the arguments in
Eqs- (5.4.3)
(5.2.2).
Eq.
The resulting
X
the
using
variables
expectatipns
terms of
determined
be
then
x
must
of
illustrated
known.
as
which are
f
5.2.1.
of X, Rip,
174
total
procedure it
this
Following
of 23 expectations
can be easily
are required
4}
E4
ES
E6
E{Xi
=H
2Xj2
Xj
3XI
{)C14
{X
X
=E
ii
4
- E[Xi
E7
E8
E9
I Y-j IXiIX 2)
E17 - E{X
ily 'i
k
41
Ix IXk
E18 - E{X
Xj
i1xi
j
E19 - E{X
Xk
XJO
iXj
E20 - E{XijX
Xk Xt I
IX
E21 - E{Xi XiXXi
Xk Xd
IXJIXklXkIXL}
E22 - E{XilXilXj
2}
X. 4 }
= E{Xi XJX i}
,
= E{Xi5jXijXjI
= E{Xi'41
E2 =
The first
E{Xie
.
(Eq.
klXkIXZIXLI)
El to E23
(5.4. S)
51 = 10SRi i4
(5.4.6)
moment is
bi-variate
the basic
result:
(5.4.7)
ii
of Price's
bi-variate
(1.60),
IX
= 3Rij
E3 = E{X iXiI=R
Other
Ix
One:
Appendix
El
E23 = E{Xi[XijX
of Expectations
Solution
application
moments require
Appendix One) for solution:
R-Ifi-
Xj2j
E4 = E{Xi2
for a
ElS W E{XiIX
iXj2Xk
E16 - E{XiIX
1Xj4Xk
Xi}
E10 = E{XijXi
1}
Ix
{Xjlxjlx
Ell -E
i i
,
lX
l}
E12 = E{Xi5lXilX j
j
5.4.1
expressions
= E{Xi
al
E{X
E2 =
i
E3 = E{Xi
El
shown that
as follows:
=4
E(X
ixi}
dR
ij
2R.. 2+R..
R..
ij
11 jj
+ E[X i2}
theorem
E(X
21
i
(5.4.8)
and similarly:
3
KX
Xj 1=
ES =
i
E6 = E{X i
--
44
xj2
22
.1
Uij
Rij
= 3Rii(4Rij
(5.4.9)
+ Rij
Rjj)
(5.4,10)
175 E[Xi4
E7
j 41 - 24R
4+
13
E(gi
R
ii
ii
the
using
= E{gj
gnj
921
9R
2+
(8Rij
following
property
gn- 1 E{gnIX,
92P ***'
(S. 4.11)
Rjj)
Rij
78 :
X29
...,
n-l
(S. 4.12)
where gi is
is
function
an arbitrary
the expected
of Xi and Efg
the values
of gn given
value
IXI X29
X
goes n-1
of X, to X
n-l'
Hence:
- E[XilXiIE[XjlXill
E8 - E[XiLXiLy-jl
mean value,
and the conditional
Appendix One, yielding:
R.
E8 =R
and using
E{IX
obtained
311
Eq. (1.34),
E8 =4R..
r2,'Tr
is
E{X IX
11
j
Appendix
One:
1/2 R..
13
11
(5.4.13)
Similarly:
E9 = EfXjXiLXj}
Applying
Price's
Appendix
One:
96
5/2 Ri.
Rij
=
42-7,1
theorem,
EIO = E(XjLXjLXj3)
using
iVf
(5.4.14)
Eq. (5.4.12)
Rij(3R
1/2
R.
11
and Eqs.
(1.58)
2
Rij
Rjj+
and (1.34),
(S. 4.1S)
11
Solution
Applying
Price's
theorem:
I.
R.
Ell
= E{XiLXiLXjLXjLl
E{IX
=4f
0
ix
dR
ij
+ EjXjjXjjI
E{Xj jXj I
(S. 4.16)
176
from
the mean-zero
and
= E{XjlXjl}
E{XilXill
E{jXij),
also
ax
{Sign(Xj)l
integral
of the
in Eq.
(5.4.16):
where 6(. ) is
= 26(Xi)
the dirac
delta
function.
Thus applying
Eq.
-0
R.
13.
11, = f E(sign(Xi)
dR
E{jXij)
E{jXjj)
sign(X
+
ij
0
(5.4.17)
1
11
from Eq. (1.34),
E{IX
Appendix One,
obtained
are
1
EfIx iXi
and
of X, and X
nature
Price's
Applying
symetrical
Price's
in the integral
to the argument
theorem
of
(5.4.17):
R.
E{sign(Xi)
4f
sign(X
S(X
E{6(Xi)
dR
ij
0
and
E{sign(Xi)}
also
f 6(Xi)
6(X
Appendix
Hence,
Xi=
/ (R
ii
ij
substituting
Ell
a result
Following
(5.4.18)
00
previously
0)
6(X
dX dX
i
P(xi
Xi=
Z P(X i=0,
where r-R
E{sign(X
=0
00
E{6(Xi)
+ E{sign(Xi))
One:
= 1/[27r(R
ii
R
ii
)1/2
(1
2) 1/2 ]
-r
Rii )1/2
(S. 4.20)
from
successively
[(4r 2+
obtained
2)
arc
(5.4.19)
Eq.
sin
by Borgman
6r(i
r+
to
-r
(S. 4.16)
2)1/2]
yields:
(S. 4.21)
60
177
2Ri3i R
J'{15
=
7r
E12 = E{XilXiLXjLXjL)
+ r(l
The remaining
if
therefore,
be reduced,
variables,
by substitution
of
Price's
theorem
One,
Appendix
covers
it
third
and fourth
proved
to
order
and using
differential
equations,
obtain
the
above
= E{ Xi2
Appendix
Xj
three
four
the
the
than
Price's
However,
and
1.3.3.3.
theorem
variables
result
of
solutions
to
solutions
to
attempts
the
(R
a,
=
where
ii
(R
R
ik
ii
R
jk
Rij
Rij
-
in
which
bi-variate
substitution.
E{X
Ix
ix
One:
(5.4.23)
Xi I=aixi+aix1
E{XklXi
ai=
any order.
expectations
2
j 10
E{Xi
E13 -
Eq. (1.55),
of
form
general
application
Section
to
reference
two
only
the
enable
(5.4.22)
and must,
involving
to
its
4))
2r
-
variables
expressions
in
to
r(I
+ 28r2
two
in
from
results
than
With
above.
that
(81
(5.4.12),
Eq.
using
be more difficult
which
problem
to
expectations
directly
this
to
possible,
apparent
multi-variate
apply
-r
more
as applied
is
2)1/2
involve
expectations
ar6 sin
2)
6r
+
/(Rij
R
jk)
Ri
j2)
Rjj
R )/(R
ik
ii
R
ii
-R ij
Hence:
E13 = E{Xj7_j
k'
-!
(5.4.24)
(E4)
+ai
= ai(ES)
similarly:
E14 = E{Xi4_. j
15a iR ii 2 Rij
E15 E-{XiLXiLXj2-Xkl
4a.
1R.
E16 z E{XiLXiLXj4-Xk}
r2ir
v,
4Rii
.
Rj j
(5.4.25)
+ aj (E6)
1/2 (3R
+ Rii
ij2
Rjj)
+ai
(E10)
(5.4.26)
z"'
[3a
iR
ii
6r 2+
1/2 (r 4+
1)
vr2ir
aR
1/2
r(15'+
lOr
2-r
4)]
second moment
178
21
E17 = E{XiLXiLXjLXjLXk
[R
(a
kk
lXjll
Price's
+ 2ai aj
(5.4.28)
2R.
.2R.
-1
j1
(12r
IT
(2r
3/2
ii
Expectation
which
Retaining
or
variance,
of Appendix
i=
is
mean of x 0
oc
are given
central
second
for
expressions
E{x
these
21XI,
0"
in Section
the conditional
the third
central
I=
1.3.3.2,
E{x
is
and the
(1
2+-2
oc
-r
2)1 /2
3) + (I
-r
(S. 4.30)
with
Eq.
(S. 4.29).
fourth
in Section
central
moment
Appendix
1.3.3.2,
moments.
Gaussian
variables
0n
to x
the
and
conditional
xn
01XIS ... '
2=
)21XII,
E{(x
cr
x
... '
oc
n"
0R
oc
resulting
conditional
second moment:
oc
Appendix
r(2r
2+
the conditional
One for
moment,
moments,
xn
having
second moment
Since
and fourth
the notation
the conditional
for
of the procedure
the third
One, by considering
r(l
2)1/2.
(5.4.29)
by comparison
an expression
from extension
follows
r+
13)
arc sin
(22r 2+ 8)]
11 is obtained
i
E18 requires
2+
3) arc sin
[6r
IT
lix
2+
3/2
3x3R
311
Ix
EfXilXiIX
theorem:
E[ IX
E(lXi3
is Ell
3 lxjlx
E{Xi
k21xix1
1)
31xjll
E[XilXiIX
where E{XilXilXjlXjll
and applying
E{X
] E[X.. IXilXjlXjll
jk)
.L
iRA+aiR
2 E(Xi3lXilX
- E[XiiXilXjlXjl
p(x
OIXII
moment must be zero, hence:
mean and
derivations.
is Gaussian
Xn)
... J,
in form
179
E{ (x 0 - ;-oc )'lxi,
E{x'Ixl,
0n
EX
x1x
....
n00
3i
....
ixi,
E(X
+ 3i2
0n
oc
3-
3x
2j+
oc
E{X21X1,
oc
j3
)Ixi,
x
. @. ,
0r
...,
0n
oc
3x
;2-
is
oc
x16o. j,
cc 0,
from above:
and-substitut*ng
E{xo'Ix'---'-'
From Eq.
EQXO -i
Returning
One,
Appendix
)41XII
...
oc
...,
O'Ixi.
Appendix
Ry
oc
4+
6a
oc
moment will
be:
2j
2+j
oc
oc
(5.4.32)
oc
= E{XilXilXjlXjl
in conjunction
Eq. (5.4.32)
central
E18:
to expectation
E18 = E{XiLXiLXjLXjLXk4i
and, using
oc
fourth
the
(S. 4.31)
xnI=
21
oc
9xnI=
the expectation
and expanding
Efx
xn'
(1.33),
I+i
"oc (30
=
(1.55)
Eqs.
with
One:
E18 =
13 ERkk
+ 6[R kk
+ 2a ia
+ai4
t 6a iji
+ai4
(ai
R
ik
(ai
E(IX
+aiR
R
+aiR
k
i3xi
3P
jk
+ai2
E{Xj3
'E{X Ixjjx
i
IXiIX
S.IX III
i
i
jk
)]2
[a
)]
i2
E{Xi3lXilXjlXjll
E[XilXiIX
+ 4a i3
3
EfXilXiIX
Ix 11
i
i
aj E{
Ix
i3
11]
IXiSlXj3l}
IX 11 + 4a. a
.3 E{
IX
3 X. 511
(S.
4.33)
180
Ix 11 is
where E(XijXijX
EfXislXiIX
E{jXi
3xi
Ell
Ix 11 is
i
i
311 is
by Eq. (5.4.29)
given
(5.4.30)
by Eq.
given
11
Ix
SIXiIX
is E12
E{X i
i
i
Price's
and applying
theorem:
2
2R.
E{Xil
JXi IX lix
jil
i
-2R.
1j1
11
7T
r(l
and
E(I Xi5xi
311
(4r 2+
2R.
5/ 2,
R. I)
11.
2)1/2
-r
17)
-r
(Sor2
[ISr
arc
3)
55)]
sin
(2r 2+
2)3/2
(S. 4.34)
r(4r
2+
3) +5 (1 -r
23)]
The remaining
therefore,
application
expectations,
a double
require
24r 2+
3/2
Tr
3(l
-
4+
r(8r
[3 arc sin
2)1/2.
(S. 4.3S)
four
and,
variables
of Eq. (5.4.12).
for conditional
coefficients
ai and ai have been used
statistics,
In the
on Xi and Xj*
of moments of Xk conditional
for the description
bi, bi and bk are required for description
of the
following coefficients
Xi,
XjA
Xk
Section
X.,
to
on
and
with
reference
conditional
of
moments
1.3.3.2, App8ndix One, are obtained as solution to:
So far,
[RI {b}
{R(,
=
)}
(5.4.36)
[R]
where
{bI
Hence, using
E19 = E{XiZj-k-h
abi'
E{X
where expectations
i2
I=
X'
iXk}+bi
Xi,
of variables
Xj* Xk;
One:
ixi2x
k}
xx
kil
'+ bk E{X
ixixk
2}
(S. 4.37)
181
E20
+bk
bi
ixk
2x
E{lXi3lX
=ai
21
xk
E{XilXiIX
bj
jkk
E(XilXi
IX 2x
(5.4.38)
{Xi3lXiIX
E
ai
k'
X+
21
3
Ix
xk
E{jXj
where
j,
E{Xilxilx
E{lXi3lX
= EJXilXilXk2
+ai
2} +a
E{XilXiIX
E{X Ix
i
ix
= ci E{lXi3lX k')
with ci and ck the solutions
E{lXillX
21
(S. 4.39)
3)
(S. 4.40)
kil
+ Ck E{XilXilXk3)
(S. S. 41)
to:
1
[R] [c) = {R
0)
(5.4.42)
[R]
is
the matrix of covariances
where
TT
{c) = (cij, c
k)
1= (Rij R
{R
jk)
0)
E{XilXiIX
31
which
is
1/2
.
11
r2-7,
.
21
in
obtained
.
11
iI=
(3R
ij2
derivation
the
16R.
E{Xi3lXilX
E10
4R.
E{ jXi3jX
Xis Xk
on the right
The expectations
of the form:
of variables
3/2
+ Rii
of
Rjj)
ElS,
(S. 4.43)
and applying
Price's
theorem:
R..
ij
(5.4.44)
-Tr
Yr2
The first
Ix Ix
i
k}+bi
i
jlXjl)Ck
E{XilXillX
21
31X0
(5.4.45)
182
Ix Ix
kaiEi
i
E{ IX-31X
on the right
the expectations
(S. 4.30).
The final
{X
IXiIX
1) +ai
i
31)
IX
3X
E{
E(X, IX, lx
in Eq. (SAAS),
expectation
Ix
JIXJIXk
21, is
and
E17.
E{XiLXiLXjLXjLXkLXkLXL'
E22
.=
E{lXi 3 Ix i Ix jlXklXkll
-bi
bk EfX i lxjlx
to
In
order
to
determine
+bi
klXkll
(5.4.46)
on the
expectations
right
of
expectations
conditional
1XiXiI.
Ef IX 31
k
former
be
the
that
expectation
shown
can
IIxixj1
I
{Xk
Xk
E
lix
jllXk3il
the
alternatively
It
the
expand
ix
L-{xjlxjllx
I Rkk - (ai
'
R
ik
hand
side
form
E[XklXkll
the
it
is
XiXiI
where 11
(a)
q_
42-7r
Q(a)
"0
fZ
a
j)'ERkk
or
+ aj Rjk
)].
[ (1
+ S12)
(1 - 2Q(fl))
(5.4.47)
2QZ(Q)]
= (a ixi+aix
necessary
(ai
-
R
ik
+aiR
jk
)] 1/2
L
a2l
2
(X) dX
functions
form
the
of
the
of
of Z(n) and Q(SI) prevent
presence
Clearly,
(5.4.46),
following
Eq.
by
further
of
substitution
expansion
any
far.
lines
Similar
the
the
(5.4.47)
of
method
adopted
so
along
Eq.
difficulties
EflXk3lXi
to be experienced
Xi }-
E23 = E{XiLXiLXjLXjLXkLXkLXLLX9.
Expansion of' this requires
,
have
a similar
which will
therefore,
in an expansion of
also
prevented.
Ll
the conditional
expectation,
direct
E(X
LNIN
solution
XiX0
is,
183
-.
were made to solve E22 and E23 by direct
application
of the
form
Price's
However, the shear
theorem.
of
general multi-variate
in
ity
the
results
practically
problem
unsurmountable
of
complex
I
from the formulation
difficulties,
as might be appreciated
of the problem
for E23 given below:
Attempts
Repeated
1.
application
of
D' (E23)
(const)
IIGR
Price's
theorem,
E{6(Xi)
6(X ) 6(X
i
(I. S9)
Eq.
Appendix
one yields
)i
(S. 4.48)
derivatives
)/IIDR
Dr'(.
product
of
a
represents
where
Xis Xjs X and XV
of variables
of cross-covariances
E{8(Xi)
where
right
hand side
30
30
The value
k)
1 :--
6 (V
of covariances
13 possible
combinations
li
3R?
ik
Dqt;
W
ik
DR
ij
Jk
DR
kt;
DR? W
DR
ik
it
Jk
DR
ii
DR
ii
DR DR
U
kk
of E23 must,
to the covariances
DR! ; DR2
ii
ik
DRlk
DR
it
30t
i
3Rij
3R
3Rjt. ? Rkt,* DRii
jk*
therefore,
of the right
DR
ik
pXk
and XV
the
may yield
(E23) as follows:
DR
ik;
3R
ik
Xj
derivatives
of partial
DR! ; DR!
it
it
DR
ij
integration
Xi,
of variables
DR
it
3qt
(5.4.49)
2vff21 f) -et-TM-]
from
of Eq. (5.4.48)
113R= 3R?
ii
six-fold
i)
[M] = matrix
At least
6(X
3Rjj;
satisfy
BRii
DR
ik
DR2
ik
3Rkj;
DR ;
it
Wt
BRit
BRjk;
BRjt;
3R
ii
DR
ii
each of the
hand side
Bqt
aRkk DR
U
DR
it
13 solutions
with
BRkt;
DR
ik
from
respect
184 'and
E23 by Numerical
E22
of
Solution
Integration
The, unsolved
describing
for
spaces.
probability
Xk'
hand'side
on the right
of Eq. (5.4.46),
E22, may be obtained from the tri-variate
space of Xi,
expectations
Xi and
example:
C
ix
3
Ixi
E{
jiXjlXklXkll
the other
Similarly,
following
expectations
00
{XilXiiX
E23 - ffff
{lXi31XJIXJIXklXkll
r" L ff
ix
ix
dX dX
Xj x)
kii
(5.4.50)
P(Xi
in an identical
manner:
dxi
P(X ixixkV
klxklxzlxtll
dx
_Co
the multi-variate
of the mean-zero
for
used for
notation
and Limitations
notation
the arguments
sea state
The
conditions.
X in the program
of expectations
Complexity
of variables
used herein
One.
on Structure
the computation
the expectations
with
statements,
dIk
in the
random variables
Appendix
and short-term
structure
covering
Gaussian
is obtained
Method of Computation
5.4.2
dxk
(5.4.51)
pdf.
above equations
d Xk
differs
comment
printed,
prevents
serious
confusion.
Numerical
integration
over
according
E22 and E23, respectively,
integration
by
step
constant
performed
of integration
is
the significant
using
the relevant
in form.
Gaussian
to Eqs.
range of this
Thereforej
probability
probability
spaces for
(5.4.50)
and (5.4.51), is
Rectangular
conditional
to centre
Rule summations.
probability
the computations
is
deviations
5
standard
of
a
range
of this conditional
over
performed
these moments being
mean value,
distribution
about its conditional
(1.57),
(I.
SS)
Appendix
One.
from
Eqs.
and
obtained
illustrates
the sensitivity
integrations
of the numerical
to
Table,
In
E18
is
included
in place of
the
chosen.
steps
of
the number.
derived
illustrate
tri-variate
the
numerically
expectation
since
E22 to
Table
5.4.1'
in the initial
investigation
the closed-form
solution
185
integration
and numerical
obtained
the results
in the Table
presented
of any application.
considered
to be generally
for
be Gaussian
set of
representative
follows
This
Although
required.
to a particular
correspond
member conditions,
will
was, therefore,
spaces
points,
resulting
of integration
for
definition
of the
variance
and
mean
conditional
between
for
differing
the variables
correlations
account
will
isation
of the grid
in
present
each application.
is
it
5.4.1
From Table
seen that
constant
integration
step
11
only
using
derived
expectations
which
results
steps, yields
by 35 steps, which
depart by less than 0.13% from the values obtained
represents
available
E18 derived
numerically
with
associated
1% of the values
the 11 step
small
sufficiently
computer
minimise
Computer Run-time
for
Values
solution..
found
were subsequently
obtained
purposes
therefore,
to be
to be
considered
its
to justify
of
The inaccuracy
analytically.
is,
integration
practical
'exact'
run-times.
Requirements
integration
with 3S steps in each cycle
numerical
when
above
is applied only systems of up to four nodes can be
As mentioned
of integration
in the 30 minute
analysed
11 steps
using
in the run-times
possible
to be the
at worst
within
integration
on the
available
run-times
Reduction
Liverpool.
University
of
the
at
of systems comprising
enables the analysis
of up to
to 11
12 nodes.
it
be
inferred
that much of the time
may
statements
the
above
From
is used in the numerical
integration
in
the
program
computer
consumption
with
associated
This
illustrated
be
may
combinations
fromthe
easily
solution
from Table
5.4.2
moments of response
shown that
of expectations
expectations
through
which
order
Eqs.
expectations
(5.4.3)
of force
and (5.4.4).
resulti
It
ng
'
can be
in the expan-,
E{F
E{F
FsI
E23
and
while
expectations
sion pf
p2FqFrI
pFqF rFor
latter.
fourfrom,
the
three
the
example,
expanding
and
oli'ly,
results
(S.
4.4)-in
in
Eq'.
X
terms
of
variables
using
expectations
variate
2, T a! 3. s
1, q
4:
for
2.1).
the
(S.
set
p
-Eq. .
.
force
186
FS)
E{F 12 : F2'
-E
+
4 XSIXSIX.
{X,
2XIIXIIX2
+ X14
X4
X3lX31XS
Fs
F'41
Xlh
XSIX51
2
X2
+
+ X14
X2
X31X31X5IX51X71X71
"
X2
X31X31X6
water
X4
+ 2XIIXIIX2
with
X4
X6
XO
XSIXSIXO
X8
X31X31X5IX5IX6
X2'X31X31X6
+
X2
XIIXIIXIIXIIXIIXSIXO
X6
+ X2
(5.4.52)
XO
X71X71
X2
X4
XSIXSIX71X71
+ X2
X4
X41
through
velocities
XIIXIIX4
XSIXSI
X3lX3lX6
2
X2
X4 X6)
+
+ XIIXIIX4
+
X71X71
XSIXSIXO
expectations
Variables
particle
X71X71
"
+ X2
underlined.
XS
2 X4
X2
+
+ XIIXIIX31X31X6
X51X51X71X71
XIIXIIX4
2XIIXIIX2lX31X$IX51XSI
X3lX3lX51XSI
IX71X71
+ XIIXIIX4
forming
X4 X6
X51X51
m E(XilXI-IX31X3IXSIXSIX71X11
+ XI-IXIIXIIXIIXI.
Arguments
X4
2
X4 X6 + X2 XSIX3lX6
+ 2XllXllX2
F2
SIXSI+
are related
and it
to the
following
E22
that,
the expansion of expectations
the above expressions
integrations
for
the numerical
in the manner of Eq. (5.4.46),
are required
Furthermore,
these
variables
only.
of
all expectations
combinations
E22 from Eq. (5.4.53)
with
associated
for
integration
numerical
5.4.4
In Section
expectation
approximations
for
may be solved
within
the cycles
of
E23.
expectations
in
be
for
solved
closed-form
the
may
which
obviating
need
considered
due
However,
integration.
to the considerable
rate of growth
numerical
be
increases,
illustrated
in
to
terms
computed
as
of
n
the
number
of
Table
5.4.2,
approximately
5.4.3
the upper
ceiling
35 nodes within
Validation
on system complexity
the existing
of the Theoretical
Response for
Simple
appears
run-times
to be
available.
of
Systems
(derived
theoretical
the
through the procedure of
moments
comparison of
Section 5.4.1 using Program OS19) for systems of 2-nodes with the moments,
obtained
bi-variate
domain,
pdf. s (resulting
using
the techniques
of Section
derived
5.3.4),
for
but
four-variate
the
the
three
expressions
all
on
check
a
and
provides
be
deduced
force,
from
Table
S.
4.2.
$uch a
may
as
of
expectations
both
by
illustrated
in
Figs.
3.8 to'
sets,
of
ai
moments
procedure
-S.
differences
theoretical
the
in
the
validity
of
moments
since
are
confirms
2%.
thin
about
wi
all cas es
187
Further,
confirmation
imersion
depth
of
the
same
at
Standard deviation
and kurtosis
figure
in
the
mpments
indicated
fs
are
pd.
marginal
Fig.
component forces'and
between the water
for
response
pair
to the correlation
attached
tion
at the other.
Fig.
5.4.3
shows a similar
respectively.
between
coefficient
the statistical
that
to Fig.
little
with
pair
between velocity
plot
by the correlation
5.4.1
for
different
member
deviation
(a
approach,
a value
weight
In both
conditions.
the
of
properties
largely
are governed
conditions
from the
values
relationships
these
spacing.
and crosses,
the correlation
cylinders
5.4.2
particle
change in horizontal
by circles
the
length
on two unit
is
the figure
included4n
illustrates
which
fourth
Also
with
5.4.1
force
of total
in the statistics
variations
in Fig.
is presented
standard
pdf.
are in
y=
kurtosis
the
values
of
corresponding
close agreement whilst
The major part
(ay =E {y4J/[E {y2132 ) are lower from the marginal pdf. s.
ay
be
be
in
to
deviation
the
of
a
result
of
can
shown
estimates
the
of
in
bi-variate
the
the
finite
range
applied
truncation
of
response
on
the
Clearly
of 8ay being
omission
the effect-of
point,
truncation
with
used herein
of high
corresponding
as discussed
response
small
levels
probability
in Section
above the
densities,
5.3.
is
estimation
the fourth
statistical
of Oy by consideration
of Fig.
moments obtained
and the true
SAA,
from histograms
moments defining
the pdf.
of the Pierson/
are compared.
8a, as applie4
cover the same range of the variable,
it
5.3
be
Section
the
that
and
can
seen
teappro.
of
ach
bi-,
varia.
the
OY
in
the
are
of
same
order
of
magnitude
observed
under-estimations
5.4-4.
from
Fig.
inferrid
those
The histograms,
in
as
188
Fig.
S. . S'indicates
length
below
of'a
obtained
deviations
for
indicates
S. 4.6
bi -variate
force
per unit
of immersion
in
the variation
loading
with
Figs.
4.1
-S.
in the
variation
and 5.4.3,
values
Again, the
are included.
by the upper
the
in the depth
member with*change
Fig.
level.
of. response
statistics
elevations
O..5m. diameter
water
still
of a and 0 for
the'variation
truncation
of the range
of response. considered.
has confirmed the validity
The above investigation
of those theoretical
in
5.4.1
in
derived
Section
for
OS19
the
and
programmed
expressions
fourth
and
second
moments of response,
The remaining
component'systems.
four
from
three
the
and
resulting
components;
the kurtosis
of total
members of identical
is
result
centroid
of the four
how this
two
in the fourth
moment,
in Eq.
expectations
that:
inertia
dominated
systems
of three
in space,
same location
To illustrate
for
of 3.0 results
a kurtosis
or four
for
are utilised
expressions,
force variate
by ensuring
(5.4.4)
which
equivalent
to that
which
would
at the
member group.
latter
requirement
is
satisfied,
consider
0.5m.
HI/3
= 9.3m,
F,
F4
then
as
subject
to the load, F5, on an identical
y=0.25(Fl
+ F2 + F3 + FO is equivalent
for
(100m/142.5m)
E{Fs2J
1.941
location
10s
N2
which,
x
at
=
member,
7.963.
0
an4 Fs =
with
to a sea state
Using
properties
of,,, Y are,
11 step numerical
E[Y2) -1.944
the
slight
values,
the,,,
aove
of.
from 100% correlation
deviations
nodes.
integration
x 105 N2 and
differences
resulting
between the particle
the corresponding
7.973,
within
0.15%
at the
189
Further
verification
typical
structures
of the procedures
in Section
by mathematical
time-series
to illustrate
Finally,
from their
results
5.6 by comparison
with
application
moments obtained
simulation.
the application
An alternative
solution
tion for
for
for
Ap2roximations
S. 4.4
of up to
in a and
idealised
for
as
approach
to
for
the
the other
expectations.
for XIXI,
approximations
'mean-square'
error in the statistical
from minimisation
resulting
Polynomial
sense,
of the
by Borgman 79 as
are given
follows:
i)
Linear
approximation
AR
XIXI
1/2
y 7r
(S. 4. S4)
xx
a result
the 'Morison'
in Section
applied,
previously
loading
3.2,
for
linearisation
of
mechanism.
Cubic approximation
Al-
XIXI
iii)
Quintic
(R
xx
1/2
in Fig.
offers
deviationswhilst
quintic"is
R
xx
i4
approximaipn
X+V
(5.4. SS)
3R 112)
xx
Uproximation
2
x1XI
1/2
X+
X3
and it
a reasonable
(S. 4. S6),
2R 172 - 60R
312
xx
xx
5.4.8
the cubic
X5
is
solution,
is very
evident
that
the
only
to about
linear
two standard.
three
and the.,
-'190
Re-stating7the,
E22
unsolved
E{XilXiIX
E23 -'E{XilXiIX
Using
expectations:
Ix Ix
i
i
klXklXtl
(S. 4. S7)
Ix Ix
i
i
klXkIXLIXZI'
(S. 4. S8)
from Eqs.
substitution
may be obtained
thqse-expectations
for
'Linear'-approximations
for
Substituiing
E22
(5.4.55)
(S. 4. S6),
to
approximations'
as follows:
XklXkl
8
R 1/2 ElXilXi
VT-f
7r kk
Ix Ix
j
JlXk
1/2
(E21)
Rkk
f
for
Substituting
'Cubic'
IX.
l
from Eq. (S. 4. SS):
X,
and
1/2
1/2
RU
for
approximation
for
Substituting
E22R
and, using
(S. 4.59)
XklXkl
E23 = - Rkk
iT
(E21)
(S. 4.60)
E22.
Xkxk
kk
to
1/2 (E21)
3 Xtj
(5.4.61)
KK
Eq. (5.4.12):
IX
X3x
E{XilXiIX iikz1111kjk
Ix
IX
bi E{
i3
EIX.
Ix
IX.
IX.
IX.
IX
IX.
E{X.,
XX
'1+
IX
IIX
bi
E{X
i
il
i 1
j
lxj
I
E{XilXiIX
xk4
+bk
i
defined
b's
the
are
where
E{XilXiIX
xk3
31
31
(5.4.62)
in Eq. (5.4.36):
Ix Ix 41 is
i k
i
E18
form,
the
are
the
expectations
of
remaining
same
and
(5.4.12)
(S.
Eqs.
4.31)'as
follows:
and
using
expanded
which may be
Xk'3)
F'{IXI'lXjlXjl,
Ur
kc
+ai3E
E{X13. IX
where
(a
RkkI
iR ik
are defined with
aa
derived
31)
51),
13X1
(5.4.63)
Appendix
One
Eq. (5.4.23)
for
previously
15X1
E{IX
+a13
(1.57),
of the r. h. s. of Eq.
E{IX
3 X1 31)
E{IX,
ai
+j3a 12ai
+ai
ll+"3akc2
IX
ai
i
Ix
jXjjI
JXijX
i5
i
t 3a iai2
2
akq
IX
(5.4.63)
have been
E18.
i
'Cubic/Linear'
for
approximation
from
Substituting
from
klXkl
E23
(5.4.54)
Eq.
and
X, IX, l from
for
(5.4.5S);
Eq.
1/2
E23
/2
Rkkl
iT
/2
(E21)
RjZ'
E[XilXilXjlXjIX
(S. 4.64)
is
h.
first
the
on
r.
s.
the
expectation
where
(S.
4.62).
form
Eq.
is
the
of
tion
of
for
Approximation
'Cubic'
(C)
for
Substituting
both X
klXkl
/2
RkkI
23
Ei
R
kXL
1/2
3R
31
second expecta-
E23
l
IX.
from Eq. (5.4.55):
X,
and
R 1/2 P21)
it
R 1/2
kk
E{XilXiIX
3R it 1/2
9.9.1/2
Ix
Ix
3)
kXi
3
Ix
JXijX
Ix
E{X
i
i
kXXI+
i
/2
3 RkkI
IX 3X
Ix
JXijX
E{X i
'i
jk
i
/2
9RkkI
R 1/2
I P,
(S. 4.6S)
form
(S.
Eq.
4.62)
the
two
are
of
the
expectations
of
and
middle
where
31
Ix
Ix
be expanded using expressions
for the
'can
E{Xilxilxi
i k3X1
in
Appendix
Five.
However,
this
summarised
as
moments,
conditional
expansion
Two additional
cubic
is
versions
approximations
approximation
some typical
seen that
for
results
from these
the cubic
improvement, in accuracy
approaches
approximation
for
including
approximation
are given
E22 offers
approximator
one including
in Table
the
for
5.4.3.
E23.
It
a considerable
and for
many terms
is
the" numerically
However,
computed values.
,
(i), and
in the table,
the error, As'
notably-(c),
for'ceitain'terms,
behaviour
for
is
but,
fortu'nately
Thezeason
this
not
clear
substantial.
-'
,
,
with lower numerical
it'appear. s. to apply to the expectations.
values and''.
final'moment
is
The. approximators.
the
for E23"
on
small.
hence their
effec.
in very
1-
to date.
ascertained
the'linear
the solution
some expectations
yields
closeagreement
are, evidently
Approximat6r.
deficient,
with
demonstrating
a higher
order
192
The polynomial expressions considered so far represent
I
in
'mean-square'
the
a statistical
error
which minimise
linear estimator A, is the solution to:
the estimators
sense. Tho
i
00
[E{(XIXI
rA
,V
-A
(XIXI -A
_a[f
IX)21]
IX)2
p(X) dX]
_00
)rielding:,
Ai
'E fIX31)/E[X2)
z,
and (1.34),
(1.33)
Eqs.
which, using
Eq. (5,4. S4).
in
[E(X
YAhence
4X4X2
E{X
E{X ii
No attempt
=0
(S. 4.66)
21
X2X
the best
procedure
)21]
Xk
A22
jxk
a similar
Following
defined
4x34xt2
4 X-A
kIXkI
IXk'I'
iix
E{X
A22
(X
approximaior
for
E23 may be
minimising:
(XklXklXtlXtl
A23
Xk
t)21
these
optimised
linear
estimates,
higher
This will
be the subject
order counterparts.
of
it
Is
for
that
the
anticipated
the
and
solutions
research
on-going
A23,
defining
A22
higher
the
and
and
corresponding
order
expectations
be
the
techniques
using
achieved
will
applied previously
coefficients,
indeed
or
in this
their
Chapter.
However,
from analogy
with
the expansion
of most
3
far,
E[XiJXiJX
XL31
so
namely
considered
expectation
complex
j1Xj1Xk
it may be deduced that the solutions
'the
Five,
in
Appendix
to
summarised
expectations
will
require
considerable
effort
both
in
in performing
the
expectations
the
and
expanding.
multialgebraically
from application
Price's
integratipns,
theorem.
resulting
of
stage
the presently
Although
the
approximations
are inadequate,
do not appear to deviate
too significantly
applied
moments of response
This behaviour,
from the true values.
resulting
found mbe
of four-node
typical
increases.
complexity
system
1Z-node system analysed
a,
from the increased
be expected,
As mentioned
in Section
such terms.
of the polynomial
approxima-
be
35
the
to
to
about
nodes
up
analysed against
of
systems
enables
integration
is
for
if
12
applied
solution
of
numerical
nodes
of
tions
limit
optimised
until
realised
developed
the full
However,
there
since
potential
is no means of judging
present
systems
in excess of
for
the
level
the existing
using
be
approximations
polynomial
S. S
for
growth results
of the form of E22 and E23,
in the number of
the inclusion
5.4.2,
as the
to Eqs.
with, reference
increasing
The error
to terms
attached
as might
Table 5.4.2,,
5.6.
0.1%, was
than
M. were obtained
of at worst
in Section
less
the errors
systems,
Errors
weighting
deviations
with
XIXI
of
have been
inaccuracy
for
approximations
12 components.
It has so far been demonstrated, in Section S. 3 and SA, that the shortof response may be described by the Pierson-Holmes pdf,
term distribution
has shown that the narrow-band 'type-21 cdf,
Tickellss
Eq. (2.3.54).
formulated in Section 2.3.6, for the distribution
of peak values shows a
fit
to prototype
reasonable
Southern North Sea.
Long-tem
obtained
that
those
distributions
the short-term
for
strain
zero-crossing
elevation,
surface
rates
according
may, therefore,
response
2.3.6.2
of response
to Eq.
in the
be
(2.3.67).
same as
This
is
again
for
(see
loading
Section
3.5).
single
member
made
a supposition
input
forming
data
to the long-term
the
analysis
wave climate
However,
discussed
in
in
Chapter
Four.
In
be
the
manner
many
processed
should
distribution
force,
is
the
of
response,
or
fatigue
range
analyses,
previously
required
description
rather,
is
se,
respon.
magnitude,
One possible.
if each positive
range results
by a negative
peak of equal
assumed to be followed
of response
yielding:
peak'of
194' -
YR=
Response range,
2Y
and
P (Y
(Y =YR /2)
R)=Pp
where Pp (Y) is
I
Although
tions
this
(5.5.1)
the cdf.
of positive
assumption
be likely
behaviour
of actual
it
L onguet-Higgins
for
accepted
the prediction
as developed by
of wave amplitude
of,
.'
be required
S. 6
to that
I
definitions
More rigorous
unlikely-that
inevitable
might
is
the
analogous
procedure
sinc e
distribution
from
height
the
wave
39
peak response.
their
evaluation.
In this
Section
some typical
The major
series
idealisations
structural
source
of comparison
simulation,
computed,
by other
Chapter
short-term
discrete
yielding
in this
under
derived
those
results
developed
the procedures
are applied
conditions
to
and the
methods.
from the results
of timeis simulated
at a
in time
the
particle
member forces
sets
and structural
responses are
into probability
can be formulated
which
The technique
moments of these variables.
developed for this procedure in Program OS23 (see Appendix Six) follows
79
by
Borgman
that described,
histograms
and statistical
t,
The structures
a fixed
considered
vertical
here are:
pile
these
two structures
has previously,
een described
of Reference
in Reference
15.
14 and tfie
In respect*of'
(ii).
19S
made by
the
the
(iii)
acknowledge
author
must
contribution
above,
and
his' colleague, Mr. Tickell,
who performed both the structural
deterministic
response computations presented here.
I:
Structure
Test
5.6.1,
Fixed
the determination
Consider
of total
and
Pile
or moment at the base of a
shear
fixed
cantilever
vertical
Vertical
analysis
inertia
drag
Equation,
the
terms in Morison's
and
of
strength
is a function
of depth of immersion below still-water-level,
as illustrated
,
in Section 3.
pdf. of load per unit
and hence the Pierson-Holmes
The relative
length
changes with
also
to yield
considered
Consequently,
a number of
a reasonable
to obtain
'Morison'
a realistic
components
must be
loading
to the actual
approximation
the member.
along
variation
Such'a
loading
of total
estimate
depth.
in
Fig.
shown
as
system,
simulation
M(t),
moment
techniques
quadrature
techniques
5.6.1,
In this
by Cheneo.
at the fixed
was recently
study
with
five
investigated
to
points
Q(t),
the shear,
were simulated
limit
using
using
computer
and
Gaussian
time,
as
fixed
end:
follows:
Quadrature
The Gaussian
n=
F()
F(aj)
abcissa coefficients
= weighting coefficients
number of Gauss points considered
function
the
of
polynomial
=
function
the
at specified
the
of
value
=,
Eq. (5.6.1)
Transforming
QM
is:
n
d = El Hj F(a i
j=
F()
where ai=
Hj
Formula"
F(t,
=f
to yield
the shear
n
z) dz =EHjF
(t,
point
aj
(5.6.2)
zi)i
,0
'd
f F(t,
0
M(t)
F(t,,
where
is
nd
z) z. dz =TH
J=
the force
F(t,
per un.itlength
-f. Z
at z'J:
(5.6.3)
196
distribution
to n force
idealised
along
components
as shown in Fig.
the pile
locations
at
z j*
Eqs.
S. 6.1
(5.6.2)
is
(S.
4.1),
form
Eq.
S
where
coefficients
the
as
are now
same
are of
di
(i
(-f
for
by
xHxz
xH
and
respectively,
weightings
replaced
2i
in
Table
6.1.
S.
take
the
values
given
which
and moment
the method of Section
Applying
5.4,
using
shear
into
response
result,
shear
and
which
of
upon
substitution
moments
Eq. (2.3.54),
using Program OS16, produce the Pierson-Holmes
cdf. s
in
Figs.
5.6.2
lines
5.6.3.
Points
the
and
solid
marked with
as
plotted
fourth
a triangle
lines
obtained
the Pierson-Holmes
represent
by the simulation.
moments. produced
Both sets
of results
signifies
deviations
simulation
in the latter
(2,000
The dotted
derived
histograms
the cumulative
represent
by Chen.
from simulation
cdf.
figures
on'the
the
ranges is probably
method caused in part by the finite
in the extreme
result
size
of inaccuracies
of the data
sets
points).
Test Structures
5.6.2
of the tests
The, followingzdescription
Reference
15.
5.6.
Conductor
and III
Il
from
'
tube
-2.1
Fig. 5.6.4a
properties
Modulus
Diameter
Cross-section
Moment of Inertia
0.001m.4
tF
197
The response
.
support.
replaced
Fig.
The. distributed
node 7.
by eq4valent
wave loading
loads
concentrated
5.6.2
Table
5.6.4b.
were bending
considered
Variables
gives
the
to be
was considered
at the positions
flexibility
at the
shown in
(S
coefficients
ij
a distributed
that
of lKN/m
load position.
particular
The conductor
Pierson-Moskowitz
with
spectrum
to a sea-state
subject
to a
equivalent
(111/3) of 9.3m.
wave height
a significant
2.0
CM
were
and C1) = 1.0 and the density
used
=
coefficients
3.
be
1000kg/m
taken to
The Morison
of water
corresponding
H1/3 of 9.3m. obtained
taken
arbitrarily
and as such,
approximately
to be Tz,
design
a realistic
for
period,
was
the spectrum
steepness
of
one-tenth.
Tower
5.6.2.2
The tower,
shown in Fig.
dimensions
only
analysis.
Encastre
5.6.5,
was treated
and no out-of-plane
support
actions
conditions
of the tower,
which
as a frame structure
were included
in two-
in the
for
be
1.95 x 1011 N/M2 and Fig. S. 6. S shows
to
taken
was
Young's modulus
The
individual
hydrodynamic
members.
equivalent
of
the properties
volumes
are quoted
derived from unit
5.6.3,
horizontal
together
forces
with
12 points
the flexibility
applied
at the
and these
values
coefficients
12 positions.
horizontal,
deflection
were
considered,
vAriables
of the
Three response
bottom
leg
force
the
a
of
main
and moment at load position,
at
deck, axial
between.
(see
Fig.
three,
in
positions.,
and
six
the
running
member
three,
1.0
Morison's roefficients,
were. taken to be C-2.0
and C
5.6.5).
IM
density
1000kg/m'..
the
load,
of
and
water
was
positions,,
at all
198
by a Piersonto a random sea represented
(111/3) of 6.9m.
Again,
wave height
significant
subject
the purposes
the problem
of comparison,
deterministic
'design'
corresponding
to the expected
height
wave with
was analysed
H-
14.2m.
and period
to a
subject
in a storm
TW9.3
of
secs,
12 hours
duration.
' Results
5.6.2.3
In Tables
moments of response
theoretical
5.4.1
and 5.4.2
expectations
also included
with
(using
approx
expectations
approximations
The approximation
in brackets.
method of Sections
the
using
for
of the polynomial
cubic
yields
in E23) are
second time
theoretical
by simulation
from 10,000
steps.
being
all
values,
result
with
data points
agreement
15%. although
with
even with
data
stable,
points the values are not completely
of
number
(Fe, Flo) and Vii,
deduced from the moments of force pairs
tower
1% of
within
results
values,
integration
numerical
of application
the results
the true
derived
considerably
the estimates
of the fourth
from the theoretical
values
As expected,
structure.
departure
greater
being
this
moment,
second
typical
of results
obtained
with
the
such a large
as may be
F12) of the
moment show
than
the
from finite
data
sets.
The deviation
theoretical
simulation
gives
procedure,
a measure of the
which
implies
by simulation
inaccuracy
likely
probable
obtained here.
by up to 15% from the results
between
the
response
of
properties
agreement
over-estimation
The general
resulting
bounds of inaccuracy'present
is
the
within
methods
further
verification
thereby-providing
loading,
procedures
developed
in this
Chapter.
from the
of the
of kurtosis
level
of
in the properties
of the probabilistic
of
provide
statistics
for
Although,
probabilistic
simulation,
This
true.
considered
method predicts
moments with
this
follows
of
of inaccuracy
of the
as being
generally
simulation
results
nature
be inferred
as might
than does
accuracy
to be independent
likely
therefore,
load
levels
and is,
sets
the approximate
greater
not be interpreted
should
result
here,
the applications
to those
of
of the number
from the similar
of
load.
the
is
the
approximate
of
probabilistic
accuracy
method
contrast
as mentioned in
to decrease with increase in system complexity,
In
likely
5.4.4.
Section
in design against
of extreme values of response are required
failure.
The most probable value of the extreme peak,
first'excursion
during a givIen period of exposure represents
(MpPV), of a random variable
For the P-H probability
of extreme peak values.
the mode of the'pdf.
Estimates
this
distribution
value
taken
off
of peak values,
3.4)
Fig.
to correspond
2.3.5,
at the probability
Considering
to that period of exposure.
corresponding
sea conditions,
hour period for the short-term
and assuming
here a 12-
level
force
peaks of
probabilities
and response
the number of
the relevant
of. non-exceedence
approximately'O.y
MPPV estimates
respectively...
.
are thus obtained
r
to equal
to the
using,
of the extremes
2.3,5
'Fig.
of the response
as summarised
in Tables
variables
200
of the extreme peak. load based on a deterministic
'Estimates,
height
during
the period
the
extreme
expected
representing
associated
with
in Tables
5.6.6
a steepness
of approximately
one-tenth
and 5.6.7.
Expected
values
than MPPV.estimates
extreme
wave,
of exposure
are also included
are slightly
skewness of extreme
greater
pdf. s.
it
is instructive
deviation
load,
be
The
or
of
response,
would
standard
distributed.
linearisation
kurtosis
be
the
and
with
would
underestimated
slightly
2.3.5
MPPV
Fig.
From
the
3.0.
then
response
would
at
constant
for the 12-hour period of exposure.
4.1cry.
to
approximately
correspond
in
deviation,
the
the linearisastandard
the
underestimation
omitting
between
26
therefore,
extreme
values
predict
of
response
and
tion would,
from
in
the
probabilistic
below
non-linear
those
method,
given
40%
Underestimates
5.6.7.
5.6.6
of this order may thus be
and
Tables
linear
from
the
test
analysis
of
spectral
a
structures.
expected
- -201 -
44
0
CA
cn
Ln
CA
0;
, .I
9
8
CA
Ch
0; C; C3
4)
P-4
u
u
.9u
Ch
9:
CA
ch
C!
9m
cn
Ol
Cl
Ln
C4
(14
0
Z
Ln
V-1
C4
Co
0
g
(4
V-4
Cd
10
r.
0
8 8
C; C;
4.'
tA
Ln
". 4
.0
ca
C;
r4
0
4
r-4
le
tn
M
"-4
g
0
0
41
0
., 4
r-
C;
P-4
0
C-)
0
C)
Pi
Ln
t-1)
9
rLn
g
r-4
<
8
co
00
in
IA
co
'Itt
C;
F-4
co
Ln
Ln
cn
,0
4
ce
10
fl%
t-
%D
Ln
Ln
cl
47%
V-4
L
IT
V-4
co
00
Ln
Ln
m
qlT
V-4
C;
0
. p4
41
10
r_
0
u
tA
+I
rn
I
4-4
0
0
b4
9
(D
uX
Cd 10
%.. 0
cd
CA
>
-, 4
Cl
I>
:3
:3
04
Cd
>
r-4
Cd
>
44
44
Nt
Co
W.
P4
4J
U)
im
44
0
1-%..
0
o
-H
P
C?
>.,
Cbt
44 X
.,. q e
-4 >
".,
r4
x
r4
.0X
0 %-o
14
z
P-4
C,
z:z
go
O.-NI 0--%
t-, )
C,4
w
%-.I
rn
C4
LQ
%-J
4-b
C13
to
I
LQ LQ
4J
0
.,. 4
I
pq
0
uR
4)
V-4 4J
w
Cd
"-4
$4
Cis
>
V-4 W
Cd :%
(A , 44
44
0
0k
4)
9.9m
to
4)
in
., 4
044
0
4j
cd
r4
>
0
13
-0
k
Cd
10
be
P
41
Ivi Cd
Dj
4,
0
tn
tL
0)
202
000
rq
Ln Ln Lr)
9-1Ln 0 Ln
m %0e 0 Ch
W-4
rq
lqr
kn
00
qe
rn
te)
Ln
r14
r4
01
u4
%-,0 nt
p4
94
N c% tn r-1
Ch
#-ll
P-4
al
%D "
88
00
CZCY&
0; Ler
0-4
I-A
M
0
im
94
%-#
.3
%0 r-1 00
9-4 cli
W-4
r4
f4 00000
tn
v--4 Co r
1--4 e4 rn
Co Ln et
W
9
Ch
17czar
93
cz
Ln
%D in
L)
IN
C/3
9.L.
94
(D V-1 e-i %D 0
9-4
Ln %D Ln 0
-4e %D0m
Ln 0
Ln 0
en 00
Ln
94
F-
In
-,
ll
11T
P.
t4 wl le Ln 0 4 tiP-4 P-4
Ln 00000
te qt
U)
CO
Ln
pi
Ln
4.4 4-b
0 r.
(U
0 r.
r, 0
0
ll
Ln
%-. 0
41 0
Co u
ti)
4)
41
0 -H gn
3$
3$
v> -ct Ln 0
c4 Ln 00000
P-4 P-4 I
en
nt
204
T A'B LE543
ITIVD T (I A Ir
DVCITT
IDDAU
TC
Response
Y=
force
Fi=
where
Member N=ber
Conditions:
1
2
3
4
1SOm; p-
ADDDnVTUAIrITflkTC
103 kg/m3;
A kfn
F1 + F2 + F3 + F4
on unit
length
member i
X(m)
Z(m)
D(m)
100.0
125.0
150.0
17S. 0
142. S
135.0
13S. 0
13S. 0
0.5
O. S
1.0
1.0
CM - 2.0;
CD m "C);
Numerical
Integration
(15 steps)
Method of Computation
in 0) 11
Ar?
Linear
Approximations
Cubic/Linear
Approximations
EXPECTATION
E23
E22
10.71
XIIXIIX3lX31XSIXSIX7lX71
X11XIIX2
X3jX3jX5jX5j
X11XIIX2
QX31X71X71
X11XIIX2
XSIX51X71X71
X3jX3jX,
XjjXjjX,
X31X3IX4
X11XIIX71X71
X3jX3jX4
X51XSIX7lX71
XSIXSIX6
jX,
7.197
7.181
1.760
-
5.788
1.818
-
7.173
1.745
-
1.780
6.139
-
2.420
X106
3.297
10.36
3.183
X108
X109
x1o"
X109
1.875
8.3S3
-
1.937
6.
S27
-
1.8S6
8.372
-
XJXJX31Q
2.297
1.3S8
2.393
X101*
X51XSIX6
X11XIIX71X71
1.812
1.467
1.823
X101,
XSIX5,1X6
X3jX3jX7jX7j
3.034
4.449
1.758
-
X106
X71X71X8
X11XIIX31X31
7.784
9.198
x1o,
X-JIXIIXEF
X11X1.1X51X51
-11.97
8.787
X7lX71XB
X11XIIX51X51
3.341
1.977
3.485
X101,
XIIXIIX3,
IX3 IXSIX51XBI
3.238
2.743
3.237,
X1010
n.,
,
IXIIXSIXSIX6lX7lX71
XI,
7.23S
7.468
7.210
xlO9
X1jX1jX4jX5jX5jX7jX7j
S. 009,
3.790
S. OS2
x1o,
p_4X2jX3jhjX5jX5jX7jX7j
T
,F
7.453
11.88
-14.63,
4.1913
(y4)
9.319
4.1097
11.99
-14.67
4.1881
X109
X109
X109
XIOS,
204
b4 44
..-1 44 ; -t
00
t2
Z0
e
Z0
00
N
IN
-0%-d
-0 IN
%-0
4-4
4-4 .4
4)
%D
le
Ln
hi
kn
P-4
f-4
%D
00
r-
Cb
%D
t.1
-4
r. 4
Ln
Ln
4.3
i-- -r--t
b4 =
0%
%0
en
%0
00
r_
c;
c;
Ln
10
%0
le
00
r_
--4
lit
clo
00
tA
en
00
00
%ID
r4
U)
En
co
p
c;
go
0
.4-J
r1
ri
>
%0
Kt
en
Ln
c;
t-4
1-4
En
(n
En
-rg
gn
U)
ul
Ln
Ln
Co
W-4
LA
8
m
tn
C,4
P-4
20S
V)
: r-t
r4
f-%
I4
Cd
Ad
%..0
z
C14
C3.
eq
C,
q
m
-8
00
i,Ln
F1.0
-*
on
C,
.4
%*
4"
'C'
C,
'C'
C11
C,
.4
.4
P4
0-4
C4
t-I
t,
0
Ln
%D
t')
--4
Ln co
w
ca
P.
C4
0
CD
cz
4
*
pq .4
%D C-1 tR:r
m
V
4J
r:
"
tf)
Ln
rr
Ln
qw
C31
40
.4
.4
0-4
a%
-t
ca
c3o
.4
10
eq
04
' *1 0 Co
00
as
%0
m
t*M
:H
11
ell
c3o
-4
"
orb
43 inCD
.4
.4
C'4 t-
o
r-
CA
4
'D
Ln
"D
P-4
C4
.4
1
v;
1
t4
m
C4
z
0
.9
4J
Ln
4-J
g%-o
a)
tn
14
r
1;
V-4
r-4
r
V-4
Ln Ln Ln Ln Ln 0
4)
go
-e
"d
a)
go
9.0
a co
-H (D
Ln
Cd
4)
>
40
44
Ln
4
"It
V-4
cli
-l
"4
r-4
C14
t-I
114,
m
m
'D
r-
(71
Ln
C,4
cr
rl:
P,4
Ln
%0
Ir-
00
(71
W. 0
0
. r4
4-)
. r4
tA
206
: r"%
.4C,
rJ1
o
%,
r,
14
z,
(D
93
C3
ca
cl,
fn
04
C4
04
M
P)
Ln
4
C4
00
r-Z
C4
r-.
-tt
"m
_4
;4
;;
Lu
C3,
-8
C3.
cb
i
1-1
.
C,
00
CA
14
C,
ko
"
t4
n
ccn
0,
4n
"C3.
.4
C3,4c3.4=
C3,
00
Ln
"
%o
P-4
Ln
"4
40
L4
C4
4N
4,
C,
cs
c,
c,
Ln
It?
co
C14
o4
c4
r-
rM
9
C
C4 .4;
Cq
"*CD
C)
CO
(71
M
L/I
Oo
Ln
%0
8
0
Ln
P-4
Cq v0 0
r., 00 IRT 00
CO
tn
00
C
r,
00
Ln
4m
C-4
Ln Ln 0
Cd
(n
Ln
u
rLo
0
u
P-4
9
0
IItIIII
=
+j
t4 -
to
la
40
%D
40
C21
C,
C,
t 4)
-4
00
C.
42
00
4
10
V-4
C3.
C,
C,
t-n C) -rm cn
0
Ln
C4
V-4
C
,
.
t*l
'D V, 00
a% 00 W)
0M
Ln 00 Oo
Ln M
LO
li
IIIII
C,
Ln
Ln ti)
Ln 00
L"
in
Lf,
Ln 0L n (4
u
-1
44
4)
Q
%D
bi
in
W)
to
in
to
r,
tel
co
00
C4
r-
C4
V)
;
;
lj 11;
t
14 rz 0 -
rl
t-
CA
t-,
r-4
t--.
r-
O-N
4
Ln Ln Ln (4
r4
C3
IT
IT
Ln Ln Ln
0
in
0
0.
'. 4
IN
tn
14
'o
r,
00
IM
0
V-4
. 4
r-4
IN
r-4
IZ
cd
.1
207
-
Ln
0
ca
Ln
t4)
0
r-
co Im %0 Ln
C71
LA rn N
%0
vt
m
-.0
%-D
-IT
4
m
r%0
m
F-4
. 1.
0
LI)
Ut)
bo
. rq
z
0
0
. r4
lid
0
4J
co
u
. rj
C/3
1--4
E-
Cd
4J
a
co
W4
r-
Ln
V)
00
tol
0
V-4
-r4
V)
W
in
.0
0
V-4
V-4
tn
r.
0
.4jrq
%-#
v
0
m
0
M
0
4"
0
r-. 4
F-4
V-4
V-4
C*4
0
ell
0
"I
t')
t4l
00
cn
r-4
m
V-4
%0
m
-4
%D
Go
%D
(n
0-4
te)
W)
00
F-4
%0
C14
CIO
0
F-4
V-4
Go
C4
Go
Cd
u
0
Lu
Cd
Ln
D r%.
ca
m
C4
"
00
%10
&A
M
Ln
0
tn
V)
%D
v-4
C V: 1
E-
+i
4-1
9
(U
-t
to
0
o
0
. r. 1
wi
F-4
1
0
W
0
to
0
w
Ln
C4
r-4
-,t
C4
Cd
c
-r4
e-%
Cd
%14
LLI
. rq
-4
00
r,
Rt 0
cn "
e4
40
wl
40
Ln
00
t-I
C14
0
m
%0 r-
C;
Itr
V-4
14
Ln
Cd
u
., 4
4
a
k
0
F-4
cl;
N
0
"
m t,
Ln 00
:r
Ln Ln C4 rl- ".4
P-4 (14 Ln %D 0 %0 r-
t C: t4 1; 0:
L4
tn
co m
:r
Ln o r-
tel
208
-
ul
rZ 0
LA
Co
Lo
En
4.3
cu
-. -#
rij
p4
9!
r-
ul
0
%D
Igt
(A
rA
00
-4
rz
(A
le
N
0
Co
0
fn
W
o
>
N
Q
10
0
z
Ln
EIi
C4
0
4J
Cd
(Yl
1-4
Cd
10
Co
00
ca
(D
.. o
%. -0
tl%
#-%
t
0
0
>. 1
Co
Co
c>
10
0
Z
., l
ELn
41
W
41
%t
41)
14
0
f.
'-I
>-
Co
Co
lit
ne
LA
Ln
%-. 0
r0
10
r0
%D
%.. f
209
-
z
0
F-4
CA
co
00 00
0
t4l
E-
6-4
t,.*)
Ln
qt
m
CA
t
U)
-4o
Ln
.4
t%
C)
rl
m
ul
te)
Ln
r4
t4
1%
4
.4
o
r14
to
.4
o
9-4
e4
r-
"t
"C4
4
eq
CY,
C71
o
00
00
t-I
Do
00
V-4
4
P-4
4
to
a% ,M
P4
o
-4
40
-q
x
CA
4--)
Ln
%D
%0
"4T
tn
8 "D En
LIrl
Ln Lf) 0
9
w
El)
;4
0
Cl.
U)
>
90
Co
Co
Co
C'4m
L
t4 14 1
ob
ob
-4
r-4
P-4
1-4
P-4
V-4
-4
at
m
t00
0
00
Ln
0
cz
M:r
W)
0
%D
co
00
qct
co
co
t-I
r-
00
r
co
tn
"t
ItT
44
cu
0
Ln
Co
Cl
M
": p
Ln
m
m
2 2 2 2 2 2 2 2 2 2 2 2
C-A
910
.ri
Cd
L4
r-.
C,
'D
P-4
Itt
IT
%D
m
-4
11
7-4
r-4
C4
C14
C4
1;
1:
%D
m
r-q
r-
4
r0
tIn
tn
CA
0%
ai
q.
tIn
tn
CA
Ln
q:r
t-I
Ln
--zr
m
P.
u
41
W
ca
in
4-J
"0
Cd
U)
-1
o
V-4
x
P.
'o
m
-4
Cd
r.L.
0
BC
P-I%
co
C*4
z
t,
0
r-4
rl
(n
0%
a%
40
40
14
o
C)
-6
o
.4
o
-'D
at
%D
r1l
1
0
4
co
0
V,4
u
. rq
4J
4)
14
0
4)
rl
co
0
4
C)
t-I
40
0
r-4
o
4
-4
x
x
00
Ln
Ln
1-4
(A
V)
4m
Ln
4
00
t-I
01
o
10
0
1"
0
'. 4
CPS
0
4
-,D
(76
co
N
I-
"T
RT
-4
r-4
U
I-. '
r-4
"
Itt
4
-4
x
-4
Ln
co
C4
C4
4
4 C14
C14
40
0
4c"
0
go
0
4
1
x
ko
"D
Ln "tR*. CA
m %D C4
c7i
C71
%0
P,
4
V-4
-4
-4
ell
4
It
.4
14
V3
o
%,
r-
co
C4
1;
14
(71
r-4
C4
210
-
ca
12!
U*i
P-4
Co
Co
(D
(n
Ln
k-
41
tz
rq
te
m
Ln
C!
00
12
!
P.4
%0
en
i. -N
c41
u3
cyb
al
10
%D
:z
u
04
Ln
u
4
0
W-
b-4
P-4
c.
lqt
Lli
ti
%D
0
Ln
%0
m
b-4
clo
ca
r-4
Cd
(4
ty%
Co
Co
W-4
0
. r4
4-1
u
4)
r-I
44
4)
P-1
Cd
t)
. ri
43
le
I ct
z
%_o
c r
c
le
CD
m
rq
rA
r4
Co
00
%-0
f.
%1-0
M-, %
%D le
r: c;
c4
t--%
Co
Ln
CYI
Ln
at
Co
00
r",
r-
%_o
%-0
pli
P-4
let
qe
Ft
%.. 0
oi
%. d
211
-
44
0
5
cn
9
Iti
0
cu
19
%0
>
ce
39 0
Ln
+-
%..d
P-1
11.1
C',
4.4
0W
z0
U)
t2
"I
ai
Co
Ln
tn b
ce r>
. b4
cu
P.1
Iti
r.
CI) 0
.0
=0
ja
0
p4
(A
ic
212
-
Iti
m
IA Cd
0Q
4)
4-b
4)
Im
co
0
t-
ri
qt
F-4
V)
P-4
4-3
\o
z
U)
Li
z0
>
C14
Q.
. 14
00
0
Sd
ce
00
m
%0
4-
U)
64
44
0p
U W.
4-)
V)
. r-l
r-4
tu
ba
g
r-%
,r -b
., -I
tn
. F.4 =)
.04
4)
tn
.0
0r
-4 o,
cl.
%-.0
0
C
Ln
Ln
C
Ln
b4 Z:
cd . 4
Ln
U!
Lf)
Ln
0
Cd
E-
in.,
in
ca
0
0
0
00
Ln
Ln
r-4
00
31:
0
(D
4-j
0
C14
-I--
Cd b
0 :1
$-4
4-4 U)
Iti
0 4)
924 U)
FIG.
Z--
5.3-1,
IDEALISED
SYSTEM OF TWO FORCE COMPONENTS
ON VERTICAL
CYLINDRICAL
MEMBERS SEPARATED
IN SPACE UNDER THE ACTION OF RANDOM WAVES.
DIRECTION OF
so
WAVE ADVANCE
S- Yl.L.
Di
UNIT
Z,
D2
P'
I'
F2
Z2
UNIT
SEA BED
X1
X2
214 -
CD: 1'0
C,z 103kg/m3
p. m spectrum
rF
T5T
JE
(Hij3x-9-3m)
7-
0-292
F2
-2-0
"0
.
.
14)
J4F
........
p (Fj
I)XI01 JI
2
Domin,
by
mmrsd
; omputctJm*
:0
llx
Mods
P(Fl F4
Mods -Z
P(Fj. FjjFj)
%bamm--.
0,
.;.
J.o
-2-0
0
( F2 1 cri )
1.0
n
3-0
.0
(FI. Fj I Fa )x 107
Fl +V4
Fj
ve
CONDITIONS
C 0-5m
D3 I-Om
Ole
150 m
Ca
2-0
Z, C 142-5m
ZI 2 135-OM
C-
1*0
ION GIM'
P. M spectrum
1-2
IE
XI: loom
Xjz 150M
9-3fn)
xO-292
F17E
'4.0
,I,!
In
Pe
14
".
f
1
IA4
D,
Zx
142-5!! s,,.
'O.'s i5ora
4.0
CM
,
2-0
4CD a "o
I,,
'C
o'103 kgIM3
p-M
spectrum (Hjjz
0-44OkN/M.
9*3m)
4A
3-2
7-.
At
11
-1
2-8
2-4
PIERSON - HOLMES
CDF--
-2-6
w
1.6
001,
1-2
.I
Ole
().4
1'
"V
Oc
M"Abtt.
'k Y
5.
4.
-E
N.-O
z
w 3-
_5
Ix
w
w
u 2-
I.
-0
1:.
ft
218
0
MARGINAL
FIG. 5.1.6.
CDFs
OF FORCE
DISTRIBUTION
JOINT
SYSTEM
COMPARED
OBTAINED
THE
UNIVARIATE
PIERSON_-', HOLMES
CDF.
I
GAUSSIAN
26
FROM
5CAL
12
CONDITIONS
24
22
2-Om
142.5 m
150 m
CD = l'O
CM = 2-0
20
103 kg/M3
P-M
spectrum
cr,
= 3-81 kN/M
(HI13 = 9-3m)
18
V
7,
16
LIJ 14
j
I-
z
12
ui
CL
ui 10
0
ULL.
so 60 70
to
so
ss
11.1 ss-1
98 99
OF
PROBABILITY
219
p (F, Fd x 101
FIG. 5.3.7
SKETCH
7
OF
OF
LOADING
ON A TWO
SYSTEM, AS
MEMBER
CONDITIONS.
DI : 2.0m. Zj: 12-5m. Xj: 100m
D,: 5,0m. *Zj-. 135-0m. Xi-. 100 m6
d -- 150M
Ct,? 2.0, CD: 1-0
BIVARIATE
INDICATED
WITH
0-85
IN FIG. 5.3.1.
OF
CORRELATION
BETWEEN
LOADS.
ez 103 kgIM3
p. m spectrum
5F
(H,,,: 9-3m)
E(FIF21
-E-tF,
085
FFT
2-1
EE
F2
Vo
1.0
*0
115'
%oh
z9
y0
-0
low2-0
-*
4-0
5.0
6.0
7-0
*le
VO
de
do,
00
F,
crF,
220
-
u
1w
.0
in
Ile
EE0
C)
0"
e_I
0
' e%
%
ui
%%
E
w
Aei
.10
A4
ty
(? 0
q.IV
lb
EEE
in 0
&6
; 1
0
*1
cr0
CL
LLI
C:
a
>F
C)'o
0
0
Ery
.01
"
EE
oo
my2"0
In
C3
41
0Z
Ula
0@ 00
HI
?Z
Cl.
En
0
E
j Q
U) 0
0
a_
W
w
a:
U0
z
0
do
E
HI
(NA) A
LV
ud'
. \17
".
13
$I
81
61,
41111
ca
09
Z. in
.?
CA
en
to
le
Q
.C)
tn
S
14
C'
m
ih
of
EEE
too
in A
ih
in
E
uu
0z
0>
3 -.eCL
3;
c
0
in
D
u
2-
cm
IL 4n
CID
ei
0
O
(NX) A
1*1
DISTRIBUTION
CUMULATIVE
FIGt5.3.9.
YaF,
FOR
+ F2
THE
OF
SYSTEM
FIG. 5.3-8.
01
12
11
z
Je
SKETCHED IN
GAUSSIAN
SCALE
'CONDITIONS
Zi x 142-5m.
5-Om
D2
1 om
:C
0-om
Z2 a 135-0m.
150 m
CD 3 1-0
CM x 2-0
d'
10
RESPONSE
10)
kglml
P -M spectrum
(HI/3 = 9-3m)
MOMENTS FROM MARGINAL
[Y4)
JY
E
2)
E
10-66
1-438
THEORETICAL MOMENTS
(FROM SECTION [541)
PDF
L
NORMALISED
MARGINAL CDF
1-426 10-79
ui
En
z
0
aU)
w
ix
C,
4
MARGINAL
CDF
(1
of
PROBABILITY
I"'
all
"40
214M 04"Mum
OF MON-EXCEEWNICIE
(019)
222
FIG. 5.3.10.
bF'RESPONSE
DISTRIBUTION.
CUMULATIVE
Y= Fl +0-5F2
SKETCHED
IN
FIG. 5.3.8.
I
IIIII
CONDITIONS
D1 = 5-OM
D2 = 5*Om
240
Cm = 2-0
P -M
(Hi/3
180
Z2
135 Om
CD :- 1*0
103 kgIM3
P
200
142-5m
= 30-om
150 m
220
Zj
spectrum
= 9.3m)
PH CDr
["*-FROM
MARGINAL
CDF.
i MOMENTS
160
MOMENTS
THEORETICAL
(FROM SECTION (5-4])
KURTOSIS
140
6-53x`1021-300x`106
v.
3-03
120
CDFs INVALID)
DUE TO ERRO
CCUMUL
A CCUMULATIO
100
50
60
40
20
50 10 70
91
93
99.8
PR,OBABILITY,
9-9
91193
OF
31.199 93.99119S."Ill
NON-EXCEEDENCE
33-ISS991
(019)
FIG. 5.41VARIATIONIN
THE STATISTICS OF TOTAL
FORCE Yo Fl# F*, (STANDARD DEVIATION C%flAND
KURTOSIS, EX. ) ON TWO UNIT LENGTH VERTICAL
MEMBERS
OF CYLINDRICAL
IN THE HORIZONTAL
SECTION WITH
CHANGE
BELOW.
CONDITIONS
da
150 m
Mw 9.3m
Cm a 2-0 Cd 1.0
ckla
D2 9 0-5m
1A
.x
ui
a
0
+
In
STANDARD DEVIATION
KURTOSIS
0
z
-00
cry
6
,cg!
I)
Ln
0
Nil
5
FOR 3 MELAM
!a
(a
4
I.0.
FOR 4. - ! JMWAS
100
.,
.V
HORIZONtAL, 'SPACING
5-OM SPACING, PY
, 'SPACING.
Py 1
'2-5rn
0011 SIIACING py
11
.
2.5m spocl"
py
;' 200
I
OF 'MEMB ERS -X
7-86
6-71
'-04
5-37
is,,,
6- 26
306
,: rl
224
5.4.2.
FIG.
VARIATION
OF PXRTICLE
KINEMATIC
WITH
CORRELATION
COEFFICIENTS,
CHANGE IN HORIZONTAL
SPACING.
0-8
IA
0-6-
A
+
0-4VEL/ACC
z
uW
%
0-2-
w+
0
u
z
0
01
ACCIACC
.w
cr - 0-2
0
u
VEL IVEL
-0-4
-OG
-(>8 -
loo.,
0
HORIZONTAL
300
200
SPACING
OF
MEMBERS
X (M)
Vz
in
l
De
3:
0
1-W 7
Z
> -i . 0
W 1,a
cc
LL
0
J
W
co
z
0
0
'm
w a
.4 w 0.
-J
0,
cc
n,
A.
Ix *
w x
>
0
z
I. -
ri
)1111
Xi
-4
.4
1
IV
<
>
u.
I..
cm
-Rj <
I.z
0
3:
w
jJ, O
cr
0
_ 40-0
X
I
w
> I...
Wit)
z
!iii
ix
o
z
cm
40
V.
sisoanx,
>
Ow
@w4
foo
NOIAVIA, W
UW313144303
q6
dW"Nvls
+
UJ
IE
a
0.6,
of
in
in
en c; 6
T
c% 6
I<
W1.
s <
ccLu
6
1
w0
1- w ;:
j u
0
in
do
V;,..
226
-
OF STATISTICAL MOMENTS
FIG. 5AAUNDERESTIMATION
DISTRIBUTION OBTAINED
OF THE PIERSONIHOLMES
AS
DESCRIBED
HISTOGRAMS
PROBABILITY
o
/
BE LOW.
C001
MH4
MH2
10
4567
KURTOSIS.
E{y2j
m
2442
In
p (y i
where
*2
A
y
Yi-
*1
42221
YMAX (N
MH2 =2E.
Ay
'
il"J1
presented
[expf-j(
*, Ij
and
random
M /M2
for
positive
values
of
AT,
+(J-l)
2=
Cy
distributed
Aj
*1MIN
'
y is
densities,
probability
3M2)/78]'
Id
i=l
Yi zF(i-l)A
yp
pNy
304
1)
y
The results
Zscretisailoni-
Nj
E
J'l
a Pierson/Holmes
and KURTOSIS of
Pierson/Holmes
as: -
27rcr
102
(M
02
E{y4j
;M=
form
discrete
a expressed.
defining
moments
statistical
y are: -
Theoretical
variable,
statistical
2=2EY4
(Y ) Ay
Yp
moments
graphically
Am-.110
2
NY, = 81
cr /20
N,; = 201
are
obtained
MH4
above
Y,
were
the
resulting
P(Yi)AY
determined
=,i-5cr
1 MIN
from
Using
the
following
histogram,
i, I fA,
t,
11
f..
III.,
t .I;!
w
o
w
w
44 0,
.44
cr.
u
in
6
at U.
0
ix
W
0 to
Io
wU
z
w
C)
ir
U.
W
Ix
zu
oPa
cc
W
>
(W)
Ak
11
.
9
C, j w
in.
19
IX 0
cc
uj
'.'A",
., , '.
1,
;I
. -0:.
EE0
Ps
wI.
C"
64
w
CL cc
-cy
JA
cr.
n
X
IE
0`7
Ao
-i
m0..
-.
hu
as U
-6
cr.
Iz
.
ww
W>
0< Z
U.
4
0
u
W '4(
CL
,
l<
- z
,
"
52<
'o
0w
U
ap
ow
1
Ul
-0
"
I
>
;7-
ic w
n 0
I-
.4
:10
In
id
z
0
0
cc
.2
44C
-Ili
_
a
,.
40
vo
pit
Cc
19
,%
,,
10
"!
W
'62
I
CC
-4W
cr
4
*0
.-
44
o'
ww
>
'0-
cc
z
W
z
uu
o
MZ
34 w
L6
mz
01,
LZU
L
C14
r)
N'SIN3NOdHOO
OD
30-UO: l
Nasklavoil
7"Ir 77
Tw:mpqm
or
io
w3ownN
Jo Hid
19.
4p a'
2 30
APPROXIMATION
TO XIXI
xlxl
Irx Crx
14Best
intic
f iqu
12
108-
6-
Best liner
f it.
42-
01234567
x
CrA
3.
2sl
FIG. 5.8-1.
WAVE
LOADING
ON SINGLE
ways propagation_
PILE.
Nr
r\
V
NJ
Instantaneous wave
load envelope
FM z)
Point
,,.,
IQ
d
zj
z
(t)
Sea bed.
mM
6-1
CONDITIONS.
IIIC
H1/3
4,8r"
50-Om
9-807mis2
0,5M
?4:
Co
2-0
f,
1-2
II
IV
232
4
%k%
LL
0
w
co
w
cc
w
L;j
U.
0
00
ID w
w
u
x
w
U
U 0:
a
z
z
2
U
u
Z
-i
U. U
FZ
2 'W"
'D
(10
W
u
z
w
0z
10
in U.
0
0
0<
CYa)
0
cr
CL
<
KL
W
M
0
w
In
D
'I
D
u
9-
jw
0
z
o
0
Ir
W
Z
oo
oo
Ir
W
z ww
:)
_3
-)
1
-4
Ln
4A
IA.
c2
lb
bb
NV3HS
61
03SICINVONVIS
'2
233
-
S.
,
0
w
LG
01
Z
w
1
0
1>
LL
0
w
u
z
w
00
co
%-4
W
w
u
x
w
Ft
Z
0
LL
0
Z
LL
0<
co
0
cc
IL
LL
in
4p.
Z4
W0
Ix
Ix
4
..:.
v)
31<
:)
k-
-1
44 4"
1 40
1)
10
iN3HOH
b
03SlC3NVCN, ViS
1
':
-2u-
195rn
ncl
170m
153rn
153.Om
V
ealn
141.7m
U2
1
4,
12 7.5 rn
-0.
.Z
110-5rn
24
102m
.z
93.5rn
85m
v
-SL
(D
76-5m
2-
68 m3 Aiel
M
--. 0.4
59-5rn
U7
34m
S.
42-5rn
v8
L@
Om
.2
om
x
(b)
NODE POSITIONS
LOAD POSITIONS
OF SIMPLE SYMMETRICAL
IN TWO DIMENSIONS.
25-Om
STRUCTURE
FRAME
'
25-Om
-1
r%
murizuritul
UWQP%
displacement.
+75m
V
piraction of,
wava advance
10-9/0-3
F11
C)
za-
0
00
0
0-7510-046/0-0031
0
Response Y3
Banding (nomant
0
0C0
0
65
6
C',
0
r,
9
0
C)
0
fl3
F12
7,
Fo
cy
CF
Still-water.
50M
+
E
9
in
F5
425m
7-0-7510-04610-0031
'o
0
6
GO
Response Y2
Axial force
8
9
0
0
0
F2
0
60
C)
,
E
0
in
CY
Om
-0.50/0-030/0-00087B
Outsidodiarnator(m)*/
structural
cross-saction
banding inertia
(m4)
(m2)
236
11 APTERSIX
CONCLUSIONS
A rational,
structures
offshore
for
calls
to the risk
approach
against
induced
fatigue
in the design
problem
and first
of the
properties
of
failure
excursion
of the probabilistic
an understanding
random loading
analysis
structural
response.
2.
A proper
treatment
is
Linearisation
of this
term permits
In the short-term,
crested
sea conditions
induced
loading
of several
linear
and applying
on short
members follows
are
analysis
hours,
assuming
long-
the wave
the Pierson-Holmes
pdf.
for
loads
of extreme
wave theory,
to the Gaussian
which degenerates
function
simplified
under-estimation
over a period
of
analysis
wave lengths.
to significant
the probabilistic
lattice
in relation
small
generally
for
essential
of the steel
structures
of the non-linear
pdf,
as the drag
effects
diminish.
Making narrow-band
justified
proposition
that
loading
derivative
is
is
statistically
shown to represent
which
S.
involve
reasonable
results
for
systems of multiple
diagram
this
of HI/3
load
time
the distribuyields
the Pierson-Holmes
model for
practicable
in realistic
assumption
which
approach,
of only
or response,
first
of peak load,
the further
of its
Furthermore,
cdf.
of loading,
independent
found to produce
of peak values.
the behaviour
by others,
previously
tion
prediction
for
assumptions
the
pdf.
is
the
applications
components.
description,
the bivariate
of loading
histogram,
are
237
to the moments of the surface
related
based on HI/3
derivation
spectrum
different
from
result
spectra
or both HI/3
Tz alone
alone,
of the departures
result
diagram
of the scatter
most classes
for
and, therefore,
and Tz as a
spectrum
surface
and the
been
has
From the results
it
presented
based on HI/3 only probably represents
in association
technique
the
with either
model.
assumed spectral
proposed that derivation
the most acceptable
Pierson-Moskowitz
or Jonswap spectral
procedure
is
climate
reducing
considerably
the computational
into
derived
peaks,
The long-term
extensively
on numerical
inherent
success for
instability
correct
values.
extreme
plots
loading
ibis
9.
each class
of
Of HI/3-
of wave loading,
between individual
values
probabilities
which
are
computer
Gaussian quadrature
run-time
techniques
minimisation
with
conditions
due to
relies
resources.
for
of wave loading
about
only
the
places
cdf. s which
at the disposal
describe
the short-term
of the design
the
engineer
accurate
problems, which
resources
integration
of the Pierson-Holmes
means for
information
basic
long-term
of
peak and
s
cdf.
computation
of
based on standardised
by hand has been developed,
for
conditions.
in the
of the cumulative
analysis
probabilistic
A technique
necessary
demand on computational
limited
of independence
found to follow
8.
of the extreme
abstracted
7.
z)
thereby
The only
with
the wave
of Iii/s
distributions
The probability
effort
associated
this
describing
values
descriptions.
long-term
Adopting
sea states
reduced
convolution
6. '
forms.
necessary
The importance
for
the complete
analysis.
description,
forming
input
to
238
-
in fatigue
importance
'one-year'
feasible
for
six
loading
of extreme
of the entire
A method of extrapolation
on the marginal
distribution
of the data
attached
to the extreme
of HI/3
limited
value
for
analysis
of recurrence
the tails
the design
of complete
of critical
'marginal',
for
effects
the structure
is
presented,
and
of the cdf. s
member loading
structures
response
response
variables
loading
from this
wave loading
near
the
free
application
have illustrated
distribution
'Morison'
pdf.
wave
The pdf.
for risk
The
surface
and
of
and linear.
to simple
the similarity
describing
of
assumed to be quasi-static
for
on many
of a complex
required
multi-variate
are a
pdf.
set of nodes.
variables
are of
where
and hence
and areas,
applies
two nodes,
'skew'
of individual
intermittency
Results
to most
are subject
frequency
properties
may be abstracted
procedure
based
of moments'
components,
wave
paper.
of individual,
'method
such variables
structure,
range of physically
where
which
to erroneously
descriptions
pre-requisite,
interconnected
be extrapolated
should
of the probabilistic
has been proposed,
'true'
to their
liable
on probability
probabilistic
load
conditions
in respect
The probabilistic
of Hi/3
of exposure
This
employed.
therefore,
extension
to either
pdf. s is
are,
data
to application
prior
long-periods
over
the presence
sea states
uncertainty
of particular
months to a synthetic
winter
load analysis.
fitting
12.
covering
to account
range,
equivalent
lower
analysis.
diagram
from a scatter
have significant
which'may
in the
load
on the cdf. s of
effect
10.
fluctuations
of seasonal
suppression
systems
comprising
of
239
-
13.
prevent
excessive
computer run-time
of the multi-variate
requirements
to systems
procedures
more
However,
load
two
than
assuming the short-term
points.
complex
be
Pierson-Holmes
form,
the
to
of
a proposition
response
pdf. of
data analysis
further
and laboratory
supported by prototype
performed
fourth
order
is
it
by Tickell5s,
fully
moments.
statistical
described
by its
Expressions
for
second and
these
moments
This
latter
integrations
expectations
variate
Various
obviate
in the fourth
accurate
for
development
the
solution
order
of more precise
four-
of certain
terms
from
moment of response.
35 nodes to be investigated.
use in complex
in part
results
these
approximations
the need for numerical
approximations
applied
for
required
for
of up to about
15.
on system complexity
restriction-
numerical
systems of up to 12
approach, peripitting
which
systems
However,
the presently
to be insufficiently
and a procedure
approximations
for
the
to typical
Some applications
of the probabilistic
procedures
illustrated
have
been
from
those
and
time
compared
with
structures
between
The
the results
agreement
shown
simulation.
of
series
has
different
approaches
provided
an indication
these
of the
developed
herein.
For systems
the
models
mathematical
of
validity
of up to
both
'accuracy'
and 'run-time'
in
14 above,
method would
advantages
'approximate'
over
appear
simulation.
probabilistic
simulation
for
to have
However,
method,
the estimation
and economic
of the moments of response provides the most reliable
for
12
than
systems
of
of
greater
analysis
nodes.
approach
240
-'
-
CHAPTERSEVEN
RECOMMENDATIONS
for
improvement
The outlined
procedure
incorporated
in the expression
be investigated.
should
This
for
will
permit
from systems
techniques
probabilistic
of the approximations
the fourth
moment of response
the
of
limited
at present
idealisations
components to structural
extension
12 load
to
3S
up to about
comprising
components.
developed herein to
procedures
of the probabilistic
be
performed with the aid of advice of a
should
structures
real
idealisation
nature on the optimal means of structural
structural
Application
2.
into
the limited
the method.
a description
This
should
of the response
as possible
from complex
areas
and volumes
nodal
load components,
into
by a study
be followed
is
variables
of
as
realistic
interest.
lumping
member
into
of the structure
sub-elements
each described
to retain
ways of
by single
by
can be handled
which
equivalent
values
of member diameter
objective
description
the total
contribution
more strictly
from the sub-element.
interest
3.
'
The importance
climate
of an accurate
on the prediction
to the response
description
of extreme
variables
of the long-term
values
or
of
wave
of wave load or
illustrated.
has
been
is
It
proposed that
response
structural
further work should be carried out on the problems of suppression
data
for
This
of
purposes
extrapolation.
will
variability
annuai
involve
the collection
form of scatter
diagrams
available.
Analysis
conju4ction,
with
typicality
of existing
of this
an analysis
of the climate
fitting
of
method
of HI/3
wave climate
and Tz,
data
(in
with
terms
measurements, in the
spanning
as many years
then be performed,
of wind records
recorded,
of the data
should
to ascertain
only)
as
in
the
of
to theoretical
241
I
distributions
probability
limited
the effects
of annual
minimised.,
variability'are
In this
such. that
extrapolation
be possible
will
from
directly
from the results
Items 1 to 3 above all follow
of
herein.
Each of these areas of further
the work described
in the proposals
of the University
research were included
of
in
Science
&
Engineering
Research
Council
the
-Liverpool,
funded Marine Technology Programme of the North West
Items 1 and 2 wore subsequently
Consortium of Universities.
1.1 which commenced in
approved, as part of Project
investigator
in
October 1978 with the author a joint
Professor
P. Holmes and
with his colleagues
association
The wave climate
Mr. R. G. Tickell.
programme was deferred
in view of the existence
of wave climate
analysis
programmes
being undertaken
at other establishments.
currently
4.
The probabilistic
described
analysis
herein
is
limited
to conditions
be
long-crested
to
assumed
and uni-directional.
are
waves
where
in addition
to the
on wave direction
the absence of information
sea state
short-term
and conservative
appropriate
of HI/3
parameters
and Tz it
to perform
for
In
may be considered
the probabilistic
analysis,
However,
the first
model-s for
Theoretical
energy
is dispersed
of this
behaviour
possible
discussed
sea state
scatter
this
situation
in the probabilistic
in predominant
to the next.
diagrams
could
If
where wave
Inclusion
wave direction.
analysis
is unlikely
to be
of the existing
techniques
describes
of directionality
extension
herein.
are available
as a straightforward
the variation
in direction
the variation
represents
wave direction
this
information
be formulated
for
then
a number of discrete
In this situation
the probabilistic
bands of wave direction.
herein
developed
be
'for
for
could
performed
response
each
analysis
to yield
the complete longconvolution
direction,
with a final
term characteristics.
4,
242, -
S.
As offshore
installations
will
investigated
herein.
for the
II
243
REFERENCES
1.
2.
DRAPER, L.
'Environmental
conditions'
Drilling
Proc. Symp. on Offshore
1970.
Architects,
Rigs,
The Royal
Inst.
WW4,1.
3.
BORGMAN,L. E.
for narrow-band
'Wave forces on piling
spectral
ASCE, 1965,91,
WW3,6S-90.
Waterways & Harbors Div.,
4.
BRETSCHNEIDER,C. L.
,
distribution
'Probability
Waterways & Harbors Div.,
S.
of wave force'
ASCE, 1967,93,
Naval
WW2, S.
6.
TUNG, C. C.
'Statistical
of wave force'
properties
Eng. Mechanics Div., ASCE, 1975, EM1,1-11.
7.
BATTJES, J. A.
'Long-term
wave height distribution
Isles'
British
Institute
National
of Oceanography,
at seven stations
Godalming,
around
1970, Report
on
Inst.
the
A44.
8.
SBDING, H.
loads of marine
'Calculation
of long-term. extreme loads and fatigue
structures'
in
Int. Symp. on 'The Dynamics of Marine Vehicles
and Structures
1974.
April
Mechanical Engineers,
Waves', Inst.
9.
10.
11,
12.
145-162.
244
13.
BURROWS,R.
'Probabilistic
model for the quasi-static
response of offshore
structures'
Dept. Civ. Eng., 1976, June.
MCE/2/76, Univ. Liverpool,
14.
BURROWS,R.
'Qua-s-i-static
response of offshore
structures
using
methods'
Modelling,
1977,1,
Applied Mathematical
September.
probabilistic
17.
BURROWS,R.
description
'Probabilistic
of the response of offshore
to random wave loading'
Proc. IAHR Symp. on 'Mechanics of Wave Induced Forces
1978, September.
Univ. Bristol,
BURROWS, R.
for the prediction
'A simple
technique
loading'
ICE,
& Theory,
Proc.
Pt. 2, Research
of
non-linear
1978,65,
structures
on Cylinders',
random
December,
wave
731-748.
18.
19.
BURROWS,R.
'A comparative
and probabilistic
study of simulation
structures'
wave loading on offshore
Dept. Civ. Eng., 1979.
MCE/l/79, Univ. Liverpool,
20.
BURROWS,R.
loading
'The choice of period when estimating
waves'
regular
Proc. Pt. 2, Research & Theory, ICE, 1979,67,
techniques
on cylinders
for
from
June.
21.
22.
BURROWS,R.
description
'Probabilistic
of wave load and structural
Computer Programs'
Dept. Civ. Eng., 1978, February.
Univ. Liverpool,
23.
24.
HOGBEN, N.
'Wave loads on structures',
BOSS 176, Behaviour of Offshore
1976,187-219.
'Technology,
WIEGEL, R. L.
Engineering,
'Oceanographical
1964.
Prentice-Hall,
Structures
Conf.,
response
Norwegian
Inst.
245
25.
MAWOTRA., A. K. and-PENZIEN, J.
'Non-deterministic
analysis
of offshore
structures'
ASCE, 1970,96,
EM6,985-1003.
Eng. Mechanics Div.,
26.
MILGRAM, J. H.
IWaves, and wave forces'
BOSS 176, Behaviour ofWfshore
Tech., 1976.
Structures
Conf.,
Norwegian
HOGBEN, N. et al
loading on offshore
'Estimation
of fluid
structures'
Proc. Pt. 2, Research & Theory, ICE, 1977,63,
September,
27.
Inst.
515-562.
BORGMAN,L. E.
28.
30.
31.
32.
-,
DET NORSKEVERITAS
'Rules for design, construction
and inspection
structures'
Oslo, 1974.
Det Norske Veritas,
SHOREPROTECTIONMANUAL
U. S. Army Coastal Eng. Research Center,
35.
36.
37.
38.
I.
of the data on
DEAN, R. G.
'Relative
validities
Waterways & Harbors
LAMB, H.
'Hydrodynamics'
Dover Publications,
New York,
,,
offshore
1973.
33.
34.
of fixed
1945.
STOKES, G. G.
'on the theory of oscillatory
waves'
Mathematical
papers, 1, Cambridge:
and physical
Press, 1880.
Cambridge Univ.
SKJELBREIA, L.
'Gravity
order approximation'
waves, Stokes' third
Calif.,
The Eng. Foundation Council on Wave Research,
Berkeley,
SKJELBREIA, L. and HENDRICKSON,J.
'Fifth
order gravity
wave theory'
Proc. 7th Conf. on Coastal Eng., Netherlands,
Berkeley,
Wave Research, The Eng. Foundation,
BORGMAN,L. E.
'Directional
spectra models for
Conf. Proc.
Offshore'Technology
PIERSON, W. J. Jr..
'Wind-generated
gravity
Advances in Geophysics,
design uses
1969, Paper 1069.
waves,
1955,2,93-178.
1959.
on
246
39.
40.
LONGUET-HIGGINS, M. S.
'The heights of sea waves,
Jnl. of Marine Research, 1952,11,
No. 3,245-
OCHI, M. K.
of extreme values'
'On prediction
Jnl.
of Ship Research,
1973, March,
29-37.
41.
42.
SAETRE, H. J.
in the Northern North Sea'
Ibn high wave conditions
Oceanographic Sciences (formerly
Nat. Inst.
Inst.
of Oceanography),
Godalming, Report No. 3,1974.
43.
44.
HOWARD,D.
Ph. D. Thesis,
45.
LIN, Y. K.
'Probability
McGraw-Hill,
Univ.
Liverpool,
Dept.
Civ.
Eng.,
1974.
dynamics,
theory of structural
New York, 1967.
46.
47.
48.
CIRIA
'Dynamics of marine structures:
the dynamic
rdethods of calculating
to wave and current
subject
action'
response of fixed structures
Report URB, 1977.
49.
AKAIKE, H.
'Note on higher order
Statistical
Ann. Inst.
50.
spectral
Mathematics,
NAGATA, Y.
higher
'Lag joint
probability,
higher order spectrum'
No. 2,78-94.
La Mer, 1970,8,
order
1966,18,
co-variance
No. 1,123-126.
function
and
YAMANOPCHI, Y.
in
247
52
53.
54.
V,ASSILOPOULOS, L. A.
'The application
theory of non-linear
of statistical
in
random seas'
performance
ship motion
No. 150,54-65.
ISP, 1967,14,
FOSTER, E. T.
'Statistical
prediction
tower structures'
Eng. Lab.,
Hydraulics
1967, October.
of wave-induced
Report
HEL 9-14,
responses
Univ.
to
systems
in deep-ocean
California,
Berkeley,
Civ.
Eng.,
loads
1974.
55.
TICKELL, R. G.
'Continuous
members'
random wave loading on structural
1977, SS, No. S, May, 209-221.
Engineer,
The Structural
56.
BENDAT, J. S.
for random responses:
functions
'Probability
prediction
of peaks,
failures'
damage and catastrophic
fatigue
Los Angeles, California,
1964.
Corporation,
Measurement Analysis
57.
MINER, M. A.
damage in fatigue'
'Cumulative
1945,12,
Jnl. Applied Mechanics,
lS9-164.
58.
GUMBEL, E. J.
'Statistics
of extremes'
Columbia Univ. Press, 19S8.
59.
60.
BORGMAN,L. E.
'Statistical
models for ocean waves and wave forces'
Academic Press, London,
In 'Advances in Hydrosciencel,
139-181.
61.
NEUMANN,G.
lZur charakteristik
Arch. fiir Meteorol,
352-377.
des seeganges'
Geophys. und Bioclamatologie,
1972,8,
1954, Serie
A7,
62.
DARBYSHIRE, J.
in
Atlantic
North
Ocean'
the
'An investigation
storm
waves
of
London, 1955, Ser. A, Vol. 230, S60-569.
Proc. Royal Society,
63.
BRETSCHNEIDER,C. L. et al
'Data for high wave conditions
observed by the OWS"Weather
I
in December'19591
Zeitschrift,
Band 15, Heft 6,1962.
Deutsche HydrogTaphische
64.
ABRAHAMSEN',E.
'Recent'developments
design'
Det, Norske Veritas,
in the practical,
1967, Pub. 60.
philosophy
of ship
Reporter"
structural*
248
65.
HASSELMANN,K. et al
'Measurements of wind-wave growth and swell decay during the Joint
(JONSWAP)l
North Sea Wave Project
Zeitschrift,
Reihe A (8*), No. 12,1973.
Deutsche Hydrographische
66.
EWING,, J. A.
from the Joint North Sea Wave Project
'Some results
to
of interest
engineers'
in Waves',
Int. Symp. on Dynamics of Marine Vehicles
and Structures
1974.
Inst.
Mech. Eng., April
67.
PHILLIPS, 0. M.
'The dynamics of the upper ocean'
.Cambridge Univ. Press, 1966.
68.,
69.
KRYLOV, V. I.
'Approximate
calculation
1962.
Macmillan,
70.
71.
72.
of integrals'
PEDERSEN, B.
'Predictions
with special
emphasis
wave conditions
of long-term
the North Sea'
Proc. Int. Conf. on 'Port & Ocean Engineering
under Arctic
1971, Vol. II.
Trondheim,
Conditions',
DRAPER, L. and DRIVER, J. S.
'Winter waves in the Northern North
recorded by m. v. Famital
Institute
National
of Oceanography,
Int.
Report A. 48.
JASPER, N. H.
distribution
'Statistical
patterns
induced ship stresses
and motions,
Trans. Soc. Naval Arch. and Marine
1971, August,
WEIBULL, W.
distribution
'A statistical
of wide applicability'
ini.
of Applied Mech., 19S1, Vol. 18,293-297.
74.
NOLTE, K. G.
'Statistical
extreme sea state'
methods for determining
Second Int'. Conf. on 'Port and Ocean Eng. under Arctic
1973.
Reykjavik,
Univ. Iceland,
76.
SAETRE, H. J.
in the Northern
on high wave conditions
Godalming,
Inst., Oceanographic Sciences,
CHOW, V. T.
'Handbook of, applied
1964.
McGraw-Hill,
hydrology'.,
NIO
73.
7S.
on
North Sea'
197,4, Report
Conditions'
No. 3. -
249
77.
OBSERVERS
HANDBOOK
-METEOROLOGICAL
HMWO, 3rd Ed., 1969.
78.
PAPOULIS, A.
'Probability,
random variables
1965.
McGraw-Hill,
79.
BORGMAN,L. E
'Ocean wave simulation
J. Waterways & Harbors
and stochastic
processes,
for engineering
design'
Div.,
ASCE, 1969,95,
WW4,557-583.
80.
CHENp J. J.
'Simulated
random wave loading on horizontal
and vertical
members
structure'
of an offshore
for Diploma of Maritime
Civil
Engineering,
Univ.
Dissertation
Dept. Civ. Eng., 1975, September.
Liverpool,
81,
and continuum
PRICE, R.
'A useful theorem for non-linear
IRE, PGIT, Vol. IT-4,1958.
having
82.
devices
mechanics'
Gaussian
inputs'
25P, I X, 'ONE
'APPEND
ANALYSIS
OF
RANDOM VARIABLES
INTRODUCTION
The purpose
of all
I.
to outline
used to describe
statistical-parameters
No attempt
is
Appendix
of this
continuous
concepts
and
random phenomena.
a comprehensive
or rigorous
account,
or parameters.
such concepts
broadly
be
can
phenomenon
Any physical
deterministic
In these
averages.
statistical
of the quantity
to say that
a certain
in terms
for
cases,
at any point
is
there
exactly
in terms
of probability
time
in time
it
will
it
DESCRIPTIVE
The properties
is possible
exceed a
of
frequency
the
time,
are
outlined
below,
or
which
is
random
varying
a more detailed
covered
may be described
variables
domains.
amplitude
Description
Piers 0147 .
and frequency
The time
account
being
in
the
ampl itude
domain
in
the
following
Section
given
concerns
by Bendat
in
1.2.1.1
Basic
statistical
and
probability
1.3.
I. 1, the following
parameters
of a random variable
parameters
are basic:
x(t)
as demonstrated
either
domains
Time Domain
1.2.1
or
the
quantities,
is unknown but
that
probability
varying
statements
value.
particular
theory
either
of the problem.
1.2
of known parameters
value
as being
classified
or random.
Deterministic
Fig.
the essential
in
251
x(t)
..
10
0.1
.
T
Sample Time History
I. 1
fig.
_-
of Random Variable
lim
T+c*
x=
dt
(I. 1)
liM
(t)
.12
T+-
Tx
dt
(1.2)
t=0
liM
Cr
XT
)2
(x (,t)
.1
T4-
dt
(1.3)
t=O
the STANDARDDEVIATION:
where ax is
X2_X2
These parameters
to fully
continuous
However,
define
Mn
with
Mi
to be basic
because they
probability
distributions
most theoretical
for
random variables,
example,
this
may be necessary
'The n-th
are considered
order
M2
*X 2
order
dt
used to describe
distribution.
distributions
probability
parameters,
moment is defined
(statistical)
1.
n
lim T
X (t)
T-to
and
higher
to consider
all
the Gaussian
are sufficient
and so it
as follows:
as:
(1.4)
2S2
The n-th
central
order
defined
moment, or n-th
the mean, is
(x _ j) n dt
4" lim Tf
cn
T-11- t=O
with MC2
function
dependence of values
The general
at one time
of x(t)
is described
later,
T units
(1.5)
Auto-correlation
1.2.1.2
time,
moment about
as:
T
IM
order
to those
by the auto-correlation
at another
function:
T
R
x
which
(T)
liM
T)
T-*- t=O
dt. -L
T
(1.6)
T)
Rx (0)
=Rx (T)
(1.7)
IRx (T) 1
=x2,
(1.8)
function
Cross-correlation
1.2.1.3
The inter-relationship
X2(t)
X(t
X(t)
can be described
of random variables
by the cross-correlation
x1(t)
and
function:
T
R
(T)
XIX2
lim ;FfxIM
= T_).
t=O
_
X2 (t
+ T)
dt
(1.9)
(T)
R
XIX2
(I.
=,-R
Xjxj
(T)
and if
(0)
RX,
IX2
0, then x1(t)
and X2(t)
are uncorrelated.
10)
253
1.2.1.4
function,
Auto-covariance
functions
and cross-covariance
functions
correlation
T
il
f
[X(t)
Cx (T) = lim
T-*- t=O
T
f
[XI(t)
(T)
UM
C
XIX2
T-',- t=O
[X(t
for the
an adjustment
+ T) - X] dt. T
ill
to the
are similar
but include
corresponding
mean as follows:
and
and
coefficient
correlation
Auto-covariance
function
cross-covariance
[X2(t
T)
(I. 11)
dt.; F
X2]
(1.12)
(T),
R
RX(T)
to
and
revert
XIX2
are mean zero (ij i2 0 0)-
These expressions
random variables
The
defined
is
coefficient
correlation
when the
respectively,
as:
(0)
XIX2
r=XIX2
Cr
XI
(1.13)
Cr
X2
Stationarity
1.2.1.5
A random variable,
described
and ergodicy
is
x(t),
are independent
above,
its
if
stationarr
statistical
time
of the
properties,
origin.
A stationary
from other
calculated
if
ergodic
its
statistical
calculated
those
is
random variable
samples of x(t)
taken
properties
I. 1, are equal
during
to
interval.
Random data
ergodic
and hence
finite
1.2.2
1.2.2.1
time
the
properties
from
correctly
estimated
stationary
representing
interval,
physical
of
a single
T,
is
Power s2ectral
density
a random, variable,
function
a
as
value
x(t),
random
time
are
phenomena
history
sufficiently
can be
if
record,
the
long.
Approach)
function
function
describes
(or
autospectral
the distribution
density
function)
,
of the mean square
of frequency.
value
'between, w and w+
stationary
observed
considered
phenomena
I
generally
of the portion
Aw is
given
by:
of x(t)
within
the-frequency
band
of
254
lim
X2, (t,
T4- 0
AW)
where w is
in radian
12
S (W) '= lim
AW x
x
AW+O
density
(W
*T
R (T)
Tr 0x
00
2
f
RX(T)
-f
e-
iwT
function
may be defined
as:
AW)
For a stationary
random process,
function,
Eq. (1.6),
correlation
(1.15)
density
the spectral
function
by a Fourier
are related
and the
transform:
dT
COS WT dT
is
of x(t)
(1.16)
obtained
relationship
T=0:
and setting
R (0)
and is
(1.14)
measure.
S (w)
x
AW) dt
W,
(w)
=2=fS
thus equal
dw
(1.17)
the spectral
density
function
or
'spectrum'.
Spectral
1.2.2.2
moments
respect
to the origin,
w=0,
is
OP
)n s (w)
mf
dw
(1.18)
2=Rx
Px
hence
mo =
and
1.2.2.3
Spectral
The spectral
(0).
bandwidth
bandwidth
c is used in connection
with
the probabilistic
x(t),
giving a measure of the number
to the number of crossings
or peaks of the process in relation
of lextremal
i,
level,
the
and is defined as (see Section 1.3.4.4):
mean amplitude
of
properties
[(MOM4
M2
2)/MOM4]1/2
255
PROBABILITY THEORY
1.3
1.3.1
Tx
X+ Ax
Consider
over
time
a continuous
density
Probability
The pdf.
describes
p(x),
some defined
within
Prob[x
This
Cx+
x(t)
function
(x,
range
= lim
AX-+O
density
random variable,
x+
x(t),
as sketched
1-2.
(pdf.
the probability
and
p(X)
.2
from 0 to T in Fig.
a time interval
1.3.1.1
varying
that
take a value
x(t) will
Ax) at any instant
of time.
T
(x
Ax] = lim
T4Tx]
lim
T-4-
function
(1.20)
that:
p (X) I 0
and
00
fp (x) dx
-00
Cumulative
1.3.1.2
The cdf.
equal
describes
to a given
distribution
probability
the probability
value
of x(t)
function
having
(cdf. )
a value
less
than or
x:
x
P(x)
This
= Prob[x(t)
distribution
ic x]
function
=f
p(a)
-CO
da
(1.21)
that:
256
0 IC P (x)
ICI
and
P
21
X)
(x)
1.3.1.3
dx
Statistical
The n-th
order
EfxnI
moments.
about
the origin
is defined
as:
xn p (x) dx
=f
(1,22)
-00
I
is the expected
Ef
where
value,
value
of its
or expectance,
operator
the expected
argument.
to the description
With reference
and represents
of the time
domain in Section
1.2.1:
Mn =E {xnl
(1.23)
and in particular,
i=
is:
(1.24)
MI = E[xI
*X2=RX
The n-th
order
=M2=E
(1.25)
E{ (x - j)nj
and again with
{X21
00
=f
(x _ j)n
M
= E{ (x - X)
cn
p (x) dx
to Section
reference
n
C2
Skewness,
and
Kurtosis,
E{(x
y=Mr
=M
as:
(1.26)
1.2.1:
(1.27)
and in particular:
2=M
j)21
3/2
31(MC2)
2
/(M
C4
c2)
257
By the linear
of, E(
nature
2xi
r'
in
21
+x
2x E{x)
the central
+ R2
j2
E{
X21 _
=
M
'a M2
C2
M12 = Cr
x2
(1.28)
M3
3M2M1
+ 2MI3
M
C4
M4
4M3Mi
2
6M2M1
+
(1.29)
3M14
(1.30)
etc,
Expected values of functions
1.3.1.4
of E{
k292(41
Properties
1.3.1.5
functions
gj(x)
of a random variable
and 92(x) and constants
we may write:
klE{gl(x))
-1
k2E{92(X)}
k, and k2,
exp{-
/27 cr2
x
as sketched in Fig. 1.3a.
distributed
if
its
density
(1.32)
is clear
in Section 1.2.1.1.
258
Xn
XP
: 94
Xe
p(
p(X)
(-FC-2cr,)(Tc-crx)! (I*crX)(FC#2crx)
(a) Gaussian
Mean: x=
Mode:
,
xe divides
P (x)
(b) Rayleigh
pdf.
= 0.5
co
fp
Xe
(x) dx
-CO
Fig.
If
x(t)
1.3
Gaussian
given
pdf.
the pdf.
(x)
and Rayleigh
by [dp(x)/dx]
=0
two and is
given
into
(n
1.3
0
Probability
Properties
Density
Functions
as:
n
1) (y for n even
x
for n odd
(1.33)
moments are:
1.3
(n
...
/ k
2.
kI
an
7T
x
E{ IxIn,
by:
dx
E{xn}
1.3.1.6
C4
E{x}
xp represents
Median:
Z-
.1zzzZZZZ
.60xFIIzvz
n
1) cr for n = 2k
x
for n = 2k +1
following
of a random variable
k=
integer
(1.34)
the Rayleigh
distribution
The Rayleigh
P(X)"=
and is
density
function
is
x
(12
exp{ -xx
-T
a
sketched, on Fig.
I. 3b.
given
by:
;k0
Its
moments are:
(1.35)
259 /12E,
2
MnE2k {xn
nan for
hence:
(1.36)
2k
kla
n odd
for
2k
n-
even
/1:
2
ma
and
2
(2
ol
The concepts
Papoulis
defined
in this
outlined
Malti-variate
by its
first
one-parameter
a.
Section
in more depth.
are covered
by
cdf.
for
distribution
The cumulative
defined
an unsymmetrical
78
1.3.2.1
n random variables
xj(t),
xn (t)
....
is
as:
P(xl,
This
fully
is
Random Variables
Multiple
1.3.2
distribution
being
distribution,
...,
distribution
random variables
Probfxl(t)
xn)=
is
described
to x
xj
Multi-variate
1.3.2.2
The corresponding
***'
as the n-th
order
(1.37)
4 xnl
xn(t)
multi-variate
cdf.
of
n'
pdf.
function
density
in the previous
case considered
uni-variate
C, x
Section
and is
as in the
by
obtained
differentiation:
an P(xi,
...,
n)
P(xl. '
xn)
P(xl,
xn
f P(Xli,
xn) =f
-' ... J,
-00
00
and
3xl,
...,
(1.38)
Bx
n
XI
with
.. -S,
Xrl
) dxi,
...,
dx
(1.39)
(1.40)
I. S. 2.3
Marginal
Variables
can be integrated
pdfls
out of the multi-variate
pdf.
as follows;
00
p (x Ip
n-1
P(xl.
-CO
xd
n
xn
(I 4 1)'
*
260
Hence the uni-variate
or marginal pdf. of a particular
can be obtained from the n-th order multi-variate
pdf,
CO
CIP
f P(Xlj
_Co
_Co
(n-1) fold
Conditional
1.3.2.4
describes
function
each having
of random variables
dx
(1.42)
the probability
a prescribed
are fixed.
set of variables
remaining
dX 2...
xn)
... J'
pdf. s
density
The conditional
random variable
for example:
of a sub-set
value
of the
Hence:
P(Xlv
xklxk+l'
... 0
P(Xls
is
Xk+l'
*`
density
the conditional
xn)
of xj,
Xd
... s
P(xk+l'
xn)
"
....
xk assuming
(1.43)
for
values
.... % xn
Expected
1.3.2.5
to the definition
By analogy
case,
variate
of a function
values
of the expected
the expected
...,
The conditional
xn is
of x k+l"
g(xj,
g(xi,
E{g(xl,
xnff
... q
x k)Ixk+l'
...,.
xn)
....
) dxl,
xn
... 1
...,
(1.44)
xn)
P(xl--. '
of function
value
obtained
in the uniis:
operator
values
expected
expected
variables
00
f
g(xl,
=
n)1
_fCo -co
n-fold
Conditional
1.3.2.6
value
of function
value
Co
E{9(xl,
of n-random
x0
...,
pdf.
the values
given
as:
00
9(Xli,
_Co _Co
... -1
xk)
k-fold
P(xl,
xklxk+l'
... 0
*`
dx
xn3*dxl
k
(1.45)
1.3.2.7
Independent
Random variables
IXIM
4 XI)A
xj,
random variables
...,
Ix (t)
... 0
n
If
this
xn are said
4xnI
are independent
for
if
any values
the events
of
is satisfied.
condition
P(xl)'xn)
=
... p
.
P(xl)
P(Xlb ... 9 xn)
P(xlj,
to be independent
-'
P(xn)
P(x
n)
(1.46)
dx
261
and for
functions
arbitrary
E{gl(xl),
gi(xi):
gn (x dl
...,
- E(gi(xi)).
...
1.3.2.8
Cross-correlation
The joint
moment E{xIX2)
X2 and, with
EIXIX21
=R
Similarly,
EQXJ
where
C12
C12
R12
il)
:
will
(X2
i2)1
R2
xI(t),
"'s
known, together'with
Yi '2 gi(x',
If
and
(1.48)
between
cross-correlation
-
il)
(X2
=C
(0)
with
x,
and x2.
reference
for
XIX2
to
(1.49)
between xj and
X2
and
xd
is
0-
yl(t),
which
the malti-variate
the set
of functions:
-'
CI 2
of random variables
theprocedure
k>n,
first
If
of xi
i=1,
xn)
is
Yk(t3
... *
density p(xl,
and
...,
to determine
(1.50)
density
the multi-variate
of
Yk*
Following
of the product
of random variables
to be solved
The problem
value
as the cross-covariance
Transformations
Consider
they are
1.2.1.3:
as the
be defined
R,
when
1.3.2.9
are independent if
is
1.2.1.4;
Section
the expected
moment E{(xI
joint
the
n(xn)l
= R12
be defined
R12 will
where
is
(0)
XIX2
(1.47)
and cross-covariance
to Section
reference
E{g
k<n,
from Papoulis78:
Yk+ I 2',xk+ V
variables
yn 0x
n'
is
singular
of yj
and it
is necessary
yn
must be defined,
for
example:
to
262
-.
density
The multi-variate,
of yl(t),....,
Yn( t) for a given set of values
by
is
the system of simultaneous
solving
obtained
Yn
equations
YI.,
If
Xn*
.,
this
real
solution
then:
P(Yll
P(Xli,
Yn) m IJ(xi,
-'
xn)
... '
-x-)i
....
n
where J(-)
ag I
J(Xjf
side
Eqs.
(1-50).
P(Y"
**'
1.3.3'. 1
Multi-variate
on the right
of xi
as functions
of yi,
hand
the solutions
of
they
...,
Yk) dyk+1-'
00
f,
f
p(yl,
%'
.,
_Co
_Co
(n-k) fold
. $. 0
dy
(1.52)
Random Variables
Gaussian
MultiRle
1.3.3
Xn
were required,
Co
yk)
agn
are expressed
variables
auxiliary
Bxn
agn
--Xl
the Determinant
of the equation
given by:
Det
... 0xn)=
Det represents
matrix
Fl
axj
If
(1.51)
i
x)=1-... p n
P(xl,
{x)
where
(c]
n/2
(21r)
Xn)
... '
= square matrix of cross-covariances,
tc, -i
(1.53)
Conditional
The conditional
c
nn
pdf. s and moments
pdf. s obtained
uni-variate
Gaussian pdf.
are also
corresponding
conditional
of x0jrom
... 1 c i, n
Cni
pdf.
xl
TI
fx,
= (xi,
c1l,
J. 3.3.2
r[
fc11
D
et
vl
expf-
Gaussian
and are,
For example,
xO, . *,,
xn is:
the conditional,
"
263
(X0
p(xO xi
It
can be shown
i27r:
n2a2
a
x0c = E{Xolxi,
mean, x0c is
given
by:
++anxn
aixi
to:
{Ro}
=
[R] is
in which
(1.54)
OC
xn
...,
expf-
the conditional
that
Od')
(I. S6)
the matrix
R
ij
of cross-correlations
xi
of variables,
to
(1.48)).
(see
Eq.
xn
(Rol
is
the vector
(Rol,
{al
is
the vector
(a,,
The conditional
2=
oc
X1,
Xd
...,
(I. S7)
+a nR on)
Efxo'lxi,.
Price's
dc
(a,
Rol
Roo
+
=
...
and conditional
1.3.3.3
)21X,
-i
2,
is:
moment, a0c
or second central
variance
Ef(xo
R
R02P ...
'
on)
adT
...,
I=
xn
"-I
Price's
theorem
82
theorem
states
xn
and
arbitrary
... J,
nn
ak E{. 11 gi (x
cr
oc
,+R
oc
joint
and
that
for
functions
moments
a set
gi(xj),
E
of Gaussian
""
M=l
N, are integers
E{ 11 gi
i=l
the co-variance
lying
between
(x
(I. sq)
1 and n inclusive
distinct;
gn (x ):
n
ykm
NkmT
11 (a crs)iLJLi
mm
M=j
where rm, sm; m
random variables
integers
k" are positive
mm
N
EC
k where c. - is
im
im m
i appears
in, (r
sm);
264
6(x )'dengtes
6-th
the
gi
i
-1 when-r m sm
0 when rm
This
may be expressed
domain by
function
an arbitrary
9(XIX2)
the theorem
(1.60)
from this
formula
by a recursion
k
r),
from
X2
moment E{xI
(1.60):
Eq.
DE[xl
X2
r,
kr
E[xI
=
a C12
if
in the bi-variate
2n g(XIX2)
a
= E{
n
,n
or bi-variate
developed
and
as:
an E{g(xlX2)1
n
D C12
Joint
for
that
who states
taken at xi;
of gi(x),
sm
expression
Papoulis7e
derivative
r-1,
X2
(1.61)
C12 =0
kX2. r}
Efxj
= E{x
kI
and:
(1.62)
Efxr)
Eq. (1.61)
integrating
Therefore,
and using
(1.62)
we obtain
the
formula:
required
E{xj
1.3.4
k-1
X2
r,
kr
Probabilistic
C12
k-I
f
E{xj
0
X2
r-1
Characteristics
d C12 + E{X
of Threshold
kI
E(xrj
(1.63)
Crossings
and Peak
Values
The probability
theory
in the
covered
concerned
with
the characteristics
variable,
x(t)
of Fig.
1.2,
be
the probability
will
what
for
what proportion
or,
x?
previous
of the basic
enabling
that
x(t)
solution
Sections
has been
continuous
to questions
of the
threshold
exceeds a certain
random
the magnitude
of x(t)
exceed x?
characteristics
continuous
random variables
are required,
namely:
additional
form;
level
265 Threshold
(1)
a certain
crosses
description
-
Peak distributions
(2)
The following
sections
a more detailed
For a continuous
in x(t)
or troughs)
x in a given
of time.
period
from
given
crossings
random variable,
of threshold
of the
crossing
00
E{N(X)l
(positive
of peaks
crests
X(t)
outlining
account
Threshold
1.3.4.1
threshold
exceed a certain
which
of the number
maxima or minima;
peaks;
or negative
rate
description
of the number of times that
-,
threshold
x in a given period of time.
crossings
=f
lkl
level
x, E{N(x)l,
is
the expected
given
by:
(1.64)
dc
X)
p(x,
x(t),
_Co
where p(x,
itself
i(t),
1) is
the bi-variate
a continuous
of threshold
The rate
and since,
for
and its
of x(t)
pdf.
first
time
derivative
random variable.
from below is
crossing
the slope
such crossings,
in some applications
required
must be positive:
co
Ef N+ (x) IfIp
(x,
1)
(1.65)
dk
1
EfN(x))
T
Threshold
1.3.4.2
if
is
crossings
p (X,
:k(t) will
the bi-variate
k)
stationary
for
following
may be expressed
12"
=1 27r
cxplCY C1.
Gaussian
process
a Gaussian
independent
be statistically
pdf.
random variables
a mean-zero
stationary
x(t)
its derivative
Eq. (1.53)
for
distribution,
of x(t)
and using
as:
(1.66)
and
Cr.2
E{N+(x)
exp
27r CY
x
2 Cly
X
211
(1.67)
,I.
therefore,
given by:
(1.68)
266
in terms
be'expressed
may
(1.17),
from,
Eq.
since
This
spectral
density
(1.69)
Ij:k
functions
0:
when
CO
Cr
x2=
and
of the relevant
(1.70)
Hence:
E{N+(O))
27r
(1.71)
MO
Peak distributions
1.3.4.3
A positive
history
of x(t)
corresponds
to the condition
0
fR
E{M(x)l
0, R) dR dx
p(x,
(1.73)
x -00
and the total
00 0
ENT)
peaks is,
number of positive
p(x,
=-ff
therefore:
0, k) dR dx
(1.74)
0 -CO
k, R) is the tri-variate
where p(x,
derivatives.
to E{MTj gives
The ratio
of E{M(x))
peaks that
that
peaks,
this
the cdf.
PW=1
pdf.
level
the proportion
x.
is obtained
EfM(x)l
pdf.
Using
to the probability
of x(t)
and its
first
of the total
the heuristic
distribution
two time
number of
45
assumption
of the
as:
(1.75)
is:
0
(x)
p pI
=fRP
EtMTJ
(x,
_00
0, R) dR
(1.76)
267
Peak distributions
I. 3.4A
With
the
Gaussian
and
a..
using
x(t)
and its
derivatives,
terms
5:
4
and
(11.70)
The expected
the
of
S(w) dw -
total
number
and the
of
of
ratio
peaks
of
terms
which
may be
using
Eqs.
of
(1.69)
time
unit
is;
below
from
to
the
number
_M2
(1.78)
axaxm
(12)1/2
in Eq. (1.19),
defined
bandwidth,
per
zero-crossings
X2
E(MT]
functions
in
m2
number
E{N+(O)l
(1
(1.53),
Eq.
tri-
(1.77)
(1.2
OL
the
is:
peaks
positive
density
1.3.4.2
M14
r Tr
the
of
two
process
M4
1x
X
first
spectral
(04
Section
can be expressed,
in
expressed
made in
pdf.
of
variances
Gaussian
a. mean-zero
as those
same requirements
variate
the
for
UMO
M4
M2
2)/MO
M41
is:
1/2
(1,79)
Wideband processes
If
is
(a = 0) the process
c=1,
described
peaks than
as being
and there
wideband
zero crossings.
with
crossing
threshold
and all
is
(a = 1) the process
c=0,
positive
described
slope
positive
leads
narrow
to a single
magnitudes.
may be sidplified
as being
banded.
positive
Under these
E{N
Furthermore,
for
distribution
is
to:
ppW-ax
(1.80)
(0)
the narrow
Rayleigh,
x
peak
conditions
E{N+_(x) 1
PW-1p
Each
exp (-
band case it
as described
X2
2)
r2
C
cy
x'
can be shown
in Section
45
1.3.1.6,
that
this
peak
where:
(1.81)
268
peak distributions
Approximate
1.3.4.5
for
non-Gaussian
narrow
band
processes
Eq.
(1.80)
narrow
on the distribution
makes no restrictions
band process
f I:k Ip (0, A) ft
(X, A) di/
P (X) =i-fIIIP
_00
-00
simplification
is non-Gaussian:
00
CO
of the underlying
11 cdf.
is independent of i(t)
since:
p (X, 1) =pWp
for
have:
we
which
E(N+(0)1
= p(0)
and
P (x)
termed
the
where p(O)
Co
111 P(A)
f
0
di
(1.82)
(1.83)
p (0)
band cdf.
of x(t)
evaluated
at x
269
APPENDIXTW0
CONSTRUCTION AND USE OF
Probability
randomly
papers
and hydrology
meteorology
for
PAPERS
the analysis
fields
theoretical
a particular
plotsfor
in the
phenomena especially
varying
PROBABILITY
of oceanography,
to produce
line
straight
76
distribution
probability
of
distribution
data plot
Having
be supposed that
'
to the theoretical
a reasonably
gives
the phenomenon is
distribution
distributed
Alternatively,
in terms
estimates
the theoretical
of the statistical
deviation,
it
structure
by inserting
fitting
the
methods
may be defined
pdf.
set,
may
according
The probability
paper.
described
be
then
graphically
the
may
phenomenon
of
by
by
line
or
best fit
eye
using statistical
straight
to the data points.
fit,
line
straight
approximately
of that
of the
typically
a technique
the
known as the
method of moments.
of the scales
Derivations
papers
for
Gaussian,
probability
II. 1
for
below
are given
discrete
the preparation
Rayleigh,
log-Gaussian,
distributions
required
data
of probability
on these
papers
is
outlined
in
The Gaussian
cdf.
of variable
x2
KX). d. -f
-L
r2ST
x is
given
by:
2SI dx
exp{ -
is
mean'*zero, with
x
,here,
unit
variance.
;..
270
-
be
the
paper
may
constructed using
probability
case
for
linear
by
scale
one co-ordinate
on
a
x
plotting
method
obtaining the values of x corresponding to the probability
by
interpolation,
for
the
axis
other
of-the
scale
required
In this
evaluation
Following
of x in the region
axis and
levels
from tho
of interest.
this
an implicit
II. 1 may be
obtained.
11.2
in
Fig.
is
shown
paper
11.3
11.2.
form the
In standardised
cdf.
of variable
x may be expressed
as:
exp(- Taking
logs
of both
{log
sides:
(1 _ p(X))-111/2
e
/2
(11.3)
distribution
is a straight
line
Rayleigh
the
Thus a plot of
{log
(1
p(X))-1}1/2
and
as the other.
co-ordinate
_
x as one
e
indicated
in
herein
are
preference
probabilities
paper used
indicated
in
Fig. 11.3.
function
logarithmic
as
the
IIA
on paper
For the
to values
GUMBELPROBABILITY PAPER
P(x)
where a
= exp{-
exp[-
of variable
x may be expressed
as:
(IIA)
ct(x - u)])
Tr
Ar-,
0.5772 is
the mode of the pdf; and
u=x-a
deviation
the
and
standard
ax
mean
are
and
x
of x.
with
of
271
(11.4):
of Eq.
logarithms
Taking
(11.5)
a(x - u) =-
in
line.
P(x)]
log
loge
will
straight
result
a
a plot
elthan values of the
rather
Again, for the paper used herein probabilities
(11.5)
in
Eq.
function
the
side
of
shown
are
used
as
right-hand
on
Note that for graphical
on
representation
of a variable
Fig. IIA.
Thus,
this
of X vs.
be known.
11.5
P(x)
expf-
function
distribution
for
C is
Taking
x , A
(11.6)
x<A
of x;
value
and
parameter;
a shape parameter.
of Eq. (11.6)
logarithms
log
limiting
x Is
B>0
and c>0
for
lower
is
A
a
where
B is a scale
of a variable
log
{cl
e
P(X-11
-
twice:
=C
log
(x - A) -C
e
log
(11.7)
eB
is
distribution
line
Weibull
a
straight
on paper with
the
A plot of
P(x))
the
A) as one co-ordinate axis and log log
as
loge (x
e
e{(I
S.
II.
in
Fig.
other axis as shown
Clearly,
data
of the value
should
only
of A although
be plotted
little
on Weibull
attention
of x are often
is often
method for
paper
after
in place
the determination
determination
to this
in the
(x - A) on
of A is given
of
272
'goodness-of-fit'
x omission
departure
to a 'straight
of parameter
line
is
A in the plots
line
concerned.
causes
distortion
considerable
and
form.
Criteria
11.6
Consider
variable
assigning
random population
of
a sample of n measurem,ents of a certain
,
If these data points
to magnitude
are ranked according
x.
the rank m-I
the cumulative
probability
Prob (x 4=xml
to the smallest
histogram
value
xj
then
may be developed
from this
data
since:
(11.8)
1)(x )=2
mn
of exceedance of 1.0 to
However,-this approximation assigns probability
the highest Value, xnI which cannot be plotted on probability paper.
Consequently, a number of modified plotting techniques have been
data
be
by
Gumbelse,
to
which
enable
all
points
proposed, as compared
is
that
The
recommended
applied
plotting
position
commonly
most
plotted.
by Gumbeland sometimes referred to as the 'Weibull plotting position',
where:
2+1
(xm) =
n+
The expression
those derived
probabilities
with an insignificant
from Eq. (11.8) when n is large.
yields
(11.9)
deviation
from
273
'FIG. (Ils T.)
, tit l i .1
tj
NORMAL)
(OR
GAUSSIAN
PAPER.
PROBABICITY
111'
t
iL
ff
"I
ui
c m
51
11 i
1.11
J;
if
11
1!
M il l i 4 I N
i,j T1
i hi
it
i Wl I III Ii
I LL[d
I.
it til
I tI
1111
U4
It R i-I fi
1 1: ,-TH
.
II
Il l
w iA
W
1
Fi4
iE
,
14 -
11
i i.
44 I k +
It l
T1
; 7F
:. -. L-
=T
.
:-R r
I
_.
...
f i ll
IIi
I ti ll tH
II -I
i I HH Ill
I
T-
4
Ill't
41
-1 . -N;
1111
li t
J IM
II I I
II
til lfiff
141
7
_tr_.
TT
I
--
yz
,j1 . .....
Z4 1
111711
t4
41
M"
7. i.
I
FE
_
4 1
T
T[
: it
1
I
1
i t
++
4r
f li t I
If I I
"IH
1
.
L.
i ii4
,
.
+
I
T-H
T_
7
7
.F
It.
H: t
it .tI. . r
!1
.
11ii
4.
11
:
m
M
Mr
ti ll
44+ 4+
ri 4
1A
ff
tr F.:
'71V`
JIT
rt .
'I IIFI
"T'_
rl*",
T7r T
fit
Mr
Pit! i ! ..
it
%
ILI
1- i .
+ ;4 4'
T,
*
1-
77
it
111T
IIt
274
FIG. Ol 2.
LOG-GA*USSIAN(OR
LOG'-NORMAL)
PROBABILITY
PAPER.
'momm
=_
--r--..
I
.1
--
ziz
-: --
_. - --
:r
-,
-.
H
r7
TT
T1
4-
TT
TT
IaIY,
JJI1
,
--
$.
27S
FIG. (Il, "3. )
RAYLEIGH
PROBABILITY
PAPER.
I ,
,
,,
I
T
x :. fl- 44+i
: .4
I
t
M
m
'
Ir
V
J
1 1111 i
A !
U l
1,
I II
H IM
H IM
ff
1111111 1
T I
:T
If
4
-- - -- ---- -1
H T . .. ... : :: -,
- -:: : - #1 , T... t
H1.1
.... ... ... ..... .....
1 i ll ml
44-
'I
W-!
-- -
M INH
TT
I .
TI
II
..
. I.
1111 LLJ 1 1
11 4
-.
jf
i4
- -- I.-
---
- -- I
U ::1 '4
J
m
T
h
ii
ILH
L L-1
i
A
4 4 -w-
--t+H+H+F4
i
1 4
44 44
HIM 11111f i li
111:1i
HIM
11
M
I1.1
H11
ill114 -
11if
4 -
+444
ji
F.
II V.
ti
1+4+
p
rt
V + 44H++ +H14"
+ -+4 ---H
-4
44
....... -
[IT
+ -i- !,-
44 -
f-
-44 +4-
4 4t 4T UTT
1
i t
11
RI fl i
T7TR,
O
-+4
4
T +:
1
OTT
T+?
+. I
V1.
1 11 11
T-
1E
1
iLl
.1
T11
I"
' 11
E
++-
ILiJ!
INIM-.4
Fiiiii 111
10
lffl
41
11t
- . itff' ,
ill
!I: :.T
i f
-LLU
x,
276'-
G.
El
GUMEEL PROBABILITY
.,
PAPER.
a:
Iii
A.
-J
LAJ
I.,
T
i
-Ft,F
u I-
#
T11
p
4
4
-.
LI
IIIIf ill 1.
.4-
F
I
!1
11
1
.
-,JL
TI
14
JI
1H
t-
I,
4.
+,t
t
i
HI
4i4,
1
I Ill
U:1] 11F1
zl
1 1H
T
ItI
HE
I.
277
LG
(U.,
G.
5.
).,
WEIBULL
-E
PROBA131LITY PAPER.
Li
J.
-1
1
4,
I-
-i-
.....
Jj
1.4
-1
J,- t
i 4--
HI
I T-
4 tl Ii
T-1-1
278
A P. PEN
D'I
XTHREE
WINTER MONTHS
WAVE CLIMATE
covering both summer and winter periods for another North Sea location
,
were available to aid this extension but this data cannot be mentioned
here because of restrictions
'explicitly
on publication
and will be
referred to as (S) and (W) data, respectively.
Simple
empirical
scatter
diagram
a summer six-month
values.
III.
data
set.
on the
badly,
as described
below,
and thus
method of prediction
an
was developed.
from a winter
period
of a
These propositions
the prediction
of Tz is
In these
for
PREDICTION
The performance
from
'simulation
based
of
empirical
data
winter
on. the'winter
The empirical
the
(W) data
methods
methods
was investigated
set
with
the
for
actual
that
summer
the
(S)
predictions
data.
the distribution
of T
z
of the marginal
from
to summer. However, a plot
winter
remains
that
distribution
of T from the (W) and (S) data sets demonstrated
z
distributions
the
the.
there was a
shapes
of
were, similar,
although
'in
lower
towards
the
summer. This was also
values of T
general shift
z
bk the mean values of Tz for the (W) and (S) data sets, the
reflected
being, #PProximately
0.67 seconds lower.
latter
unchanged
I%
279
HI/3-over
InvIew
111.2
data
show a similar
spread
the
at Famita,
conditions
at this
plots
demonstrating
of
agreement
finite
only
larger
the
sized
between
a 10% departure
mean. The (W) data
35% greater
a mean value
the existance
The cumulative
location.
of more severe
sea
distributions
conditional
reasonable
a larger
of HI/S
sets
data
data
sets
were plotted
sets.
distribubetween
the
the
conditional
of
agreement
measure
of
tions of HI/3 for (W) and (FW) and the apparent shift,
with little
between summer, (S), and winter,
change in shape, of the Tz distributions
12
for the
(W), the following
procedure was developed and employedloo
(FS).
diagram
for
Famita
the
of
summer
scatter
prediction
As a result
The basic
assumption
probability
densities
the ratio
ratio
sets
for
for
a value
the winter
data
P(HI/31TZ=t+0.5)
IT
j3
(11
Tzzt)
p
0,
p(H;
rt) (FS)
3
'F(H ITz
(S)
z
.,
=t+0.5)
TZ =t
sets
QW)
(W)
would
at
'
Gii.
1)
280
This
0.5
includes
expression
observed
secondt),
an attempt
between
the
Substituting
about
distribution
of
T
p(Hi/3,
Z)
T- J
z
)=
zF
p(HI/aIT
(of
this
(111.2)
into
expression
distributions
shape of marginal
(III.
Eq.
1) and assuming
the
that
to be
the same:,-
[p(T
for
and likewise
S)]
densities,
the Famita
(W)
we obtain:
Tz=t+0.5
p(HI/3,
Tz=t)
p(HI/s,
Ensuring
that
the class
[NO(HI/3
for
widths
in the
(S) p(HI/3,
each pair
(111.3)
Tz=t+0.5)
of data
furthermore,
number of occurrences
(111.3)
Eq.
same,
Tz=t)
p("1/3,
=
(FS)
sets,
(W)
(W) and (S)
to be the
diagrams
becomes;
Tz=t)]
[F(HI/3,
(FS)
Tz=t)]
(S)
[NO(HI/3,
Tz=t+0.5)]
[F(HI/3,
Tz=t+0.5)]
(FW)
(W)
(IIA)
in
is
diagram
Tz)
NO(Hi/j,
the scatter
the
number
of
occurrences
where
T) is the
present. in the class centred at (HI/3, T)zz and F(HI/3,
in the*scatter
diagram
fraction,
of the total
number of occurrences
in'the
present
Eq. (IIIA)
the
class
could
(FS), scatter
account
for
histogramslof,
centred
at
Z).
a. variation
However,
HI/3' -, Hi 'T7
/III..
'A
T,
CY,
Z
data
normalised
for
a modification
H*
(HI/3,
and variance
sets,
to
of the conditional
by replacing
HI/3
in
variable:
(11111.5)
1
",./
281
The functions.
Eq, (111.4)
thus modified
histograms
to yield
and (S)'data
Eq., (III.
4).
as the best
is
'
I/Tz=t
of Fig.
of 111/3 (class
centres)
values
of Ii* for
(FS)
with an identical
figure.
(S)
expression
for
variable,
of Eq. (IIIA)
to the application
the computations
this
III-1.
the
using
summer condition
are carried
S)CFW)
i/Tz=t+O.
(HI
I/Tz=t+O.
S)
(111.6)
as indicated
(a/Tz-t)(FS)
H*, as described
in the
in terms
of HI/3
out over
the central
range of the
l.
as indicated
in Fig.
III.
be appreciated
at this
stage
histograms,
should
z)FS
in the sample computation
at the required
by continuous
NO(H*, T
(fll'/Tz=t
It
of
that:
assumptions
that
were approximated
by eye,
demonstrated
Numbers of occurrences
were obtained
fits
function,
This
These histograms
the required
of the
to F(Ii*,
conditional
sets.
obtained
curves,
derive
deviation
that
in the performance
of the
no assumptions
above procedure,
distributions
conditional
them to similar
forms.
The proposition
made
found
distributions
to show reasonable
agreement
these
that
were
was
is
it
the
true
hence
that
to
same
suppose
appear
reasonable
would
and
for the summer data sets and the procedure outlined
above is one method
of achieving
this
The computations
of T
Z,
scatter
for
functions
which
diagrams
measure of agreement.
outlined
there
of
F(. ) apd N
in Fig.
III.
were a sufficient
(W),
1 were carried
out
for
all
number of occurrences
construction
classes
in the,
of the
282
-
0'. ,,
..
On completion
predicted
number required,
and'consequently
Only 0.4% of the
reduction in the number bf occurrences in each class.
(S)
in
data set were omitted from the graphical
the
occurrence
's
assumed that there were no significant
procedures and it was, therefore,
omissions in the predicted scatter diagram.,
diagram
The scatter
according
redistributed
summation for
direct
by 0.5seconds
Tz offset
derived
to the time
the required
data
of
of the latter
classes
intervals
'one-year'
wave climate
to enable
as presented
in Fig.
2. . l.
111.3
The marginal
scatter
diagram
in Fig.
III.
for
density
probability
2a.
at Famita
and for
From these
is
plots
shifted,
histograms
observed
The marginal
probability
Section
The 'comparison
lying
data,
data
histograms
the overall
whilst
behaviour
set
for
of HI/3
both
empirical
the actual
estimates
shape of
to that
corresponds
the
the Famita
'comparison
summer data
its
shown in Fig.
sample of which
although the 'comparison
empirical
relationship
modeli
is
2b.
distributions
The cumulative
conditional
of H1/3 were also plotted
of the summer data,
the winter data and for all three predictions
is
winter
modeP of
in Fig. III.
methods are presented
a histogram with ordinates
generally
model' yields
between those derived from the two empirical
lower
of the complete
density
the predicted
summer diagram are shown
it is apparent that the distribution
in
Tz
the winter
than
present
of
values
is similar.
This
the two distributions
from comparison
of Tz for
modeP yields
in the marginal
evident
111.3.
These plots
values
distributions,
demonstrated
between those
for
a
that
of the
no such general
in the conditional'distributions
of HI/3.
The
283
relationships
a distinct
between these
to those
similarity
As a result
'comparison'
conditional
developed
selection
in preference
the tests
carried
diagrams
predicted
from-the
it
of the simple
modification
in the TZ distribution
shift
empirical
well
structure
type
may be found
that
its
However,
study.
of the scatter
is possible
that
models accounting
to perform
the
to warrant
models in this
and it
showed
was concluded
sufficiently
to the empirical
distributions
a
for
the
well.
equally
It
"
""
0 lit,
'284
T
NUM118 OF' OCCURENCE1, IN I
FiG. M. 1, PR R!rgy 12N CWL
SECONDS-"
'SUMIAER'SATTIER DIAGRAM FOR T
1'I'',,
-0
.
,Zv7
5-
HMO
TZ 7-5 sEc
(w),
3-
HW
Cr
VQ
+
++
12kX* c*ro
Jbi di lould)
-,!
ON
1.
am Li
023
F1040.TZa TO SEC$)
(S)
TA
T, Z'7-0 SEC$-5.
HVSz Hs
4-
Cr x as
3--
2-
1 11-
01
++
+
M) (F W) DATA 4030-
HIS' 3-094 m
20-
am
11-2:
123
114
10
1
2
50-
Tz P7-5 SECS
(b)
NO
H
--.P.
+
+
222m
(c
(c)
o
-2
Tz 0 7-0 sEc S
Hjx 3-:
-094
NO,(H* Tz 7-0sS)
59-
(F S), ,.
40
Ik
30
cr
0.969
20 -
(d)
C)
I,.
L4)
'!"
IL
3.0.9
o
, .
1.5 2R
-1
3.3
ATz
mp"v
30"
t4
SEC.
WINTERDATA
(AND SUMMERPREDICTIONS
USINGEMPIRICAL METHODS)
SUMMER PREDICTION
20.
OF SECTION W, 3)
10-
IF
01234
10
.56789
HISMGRAMS
11
12
13
14
15
ll
TZ (SECS)
30-
WINTER
DATA
3:
77t j
JL
JL
Tt
01
,,
i.
,
t
I.
,
"
,
,
,
.
I.,.
t,,
2-4 '.
.6 lz
6. O'." ,
i,..,
1-2
"'
I
A-
-V!i
A'
Ap
S2
1
zb' I
0
OD
w
x
w
cr.
RK
IL
W
ul
8e
in
287
APPENDIXF0UR
SHI-SQUARE GOODNESS OF
PROBABILITY
In Section 4,4.2
DISTRIBUTION
the distribution
FIT
TEST ON THE
OF SIGNIFICANT
WAVE HEIGHT
wave height
of significant
for the
for the
Gumbel distribution
structure quite well and the theoretical
data was obtained by method of moments, using the mean and variance of
histoThe fit is shown on Fig. 4.7 and the probability
the data set.
The purpose of this Appendix is
grams are given in Tables 4.1 and 4.3.
the hypothesis
to test
of the Chi-square
application
involves
The procedure
chi-square
ting
the
47 which
function
distribution
of
Table
into
IM
with
two classes
group.
of
XC2,
the
discrepancy
the
data
equivalence
scatter
diagram
The general
between
and the
is
the
Gumbel
by investiga-
tested
are grouped
into
the
the
an approximate
with
Of Xc2.
test.
a statistic,
measures
The hypothesis
sampling
of HI/3 is Gumbel by he
goodness-of-fit
use of
density
probability
function.
density
the
distribution
observed
the distribution
that
four
class
width
of 0.6m.
in Table
classes
consolidated
contained
IM
in each
represents
frequency,
to fall
is
fi*
in Table IM
. Gumbel.
288
The sample, st4tistic,
frequenciesfor
and expected
observed
discrepancy
classes
between the
as:
F2
2=EI,
xC
Fi.
all
(IV. 1)
of XC2 is
distribution,
approximately
n degrees
with
to Xn2'
equal
of freedom.
the chi-square
case n is equal
In this
to'11,
the fact
interval
(b)
with
defined
is
associated
to define
the parameters
the small
probability
Xn=ll;
The probability
thus
for
will
fall
in this
of significance
is:
of the
the region
region
its
if
the hypothesis
likelihood
of Table
is unlikely
for
of rejection
that
observed
is valid.
is
strong
the hypothesis
samples
Conversely
if
the
to question
as a is reduced.
reason
decreasing
(IV. 1):
8S. 001
Xc2
These'valties
rejected.
level
of the test
of a it
values
to satisfy
the hypothesis,
set used
small
sample fails
of the data
(IV. 2)
a represents
test
class
a as the
region
the last
in the remaining
Defining
test,
the frequency
that
of the chi-square
47 :
26.76
'1'
a=0.005
do not
at the'a
as indicating-.
distribution
thqt,
I,
satisfy
4;
289
deviation Xc2 as large as 26.76 is due to chance only is O.S%.
Consequently, the observedvalue
of X2= 8S. 001 has a much lower
C
being
by
of
caused
chance only, this therefore,
casts serious
probability
doubts on the hypothesis which may be considered unacceptable.
However, the major contribution
to the deviation occurs within the
first
two class intervals,
as seen from Table IV. 1. If these values are,
omitted the total deviation becomes:
X
16.257'.,
Xn=9;
a=0.05
chi-square
distribution,
16.92
now falls
is
insensitive
to minor changes in the
or wave load are relatively
lower range of H1/3, such as those r' esulting from the Gumbel assumption
heights less than 1.2m.
for significant
290
-.
TAB,
LEIV
COMPUTATION OF
fF
HI/3
F
F.
(M)
1
0.3
96
164
28.19S
0.9
402
293
40. S49
'3
1.5
389
377
0.382
2.1
321
349
2.246
2.7
24S
268
1.974
3.3
161
181
2.210
3.9
132
117
1.923
4. S
70
71
0.014
5.1
46
44
0.091
10
S. 7
23
2S
0.160
11
6.3
12
17
1.471
12
6.9
11
2.286
10
0.500
3.000
1924
1924
,6
7. S
13
8.1
8.7
14
9.3
Totals
14
(X, 2)14
Fi
2)
(x
c
12
14
1:
85.001
fF2
(IF.
= 16.257
8S. 001
2pl
-
.-
APPENDIXFIVE
EXPANSION OF E(X iLXiLXJLXjLXk3- - h3
-j
for
An expression
E(X
i1xi
is
etc.,
Ix
XZ31 in terms
jlXjlXk3
required
of the cross-
for solution
R,
E23, and
of
expectation
ij
,
hence the fourth moment of response, by application
of the cubic
for XIXI as shown in Section 5.4.4.
approximation
covariances
Using Eqs.
1
and (5.4.31):
(5.4.12)
31
I Y, Ix Ix
31X
(V.
1)
EJX
Y-i
E{Xi I Xi I Y,
E{X
=
iIY.
k3x1
iI
i
k3
ixixk)I
i
z
iIY,
13CyIC2
3 Ix Ix Ix 31 +bi
31
lXk
[bi
E{jXj
E{Xilxillx
=
i
i
k
j3
"bi
E{ Xi3
" 3b
12bi
3 ix
" 3bi bi2 E{lXi
" 3b ibk2
" 3bi2bk
"bk3
31
i3
+ 3b i2bk
Ix
E{Xi
31X
IX
41
E{XilXil
+ 3b ibk2
i
k
i
611
Ix
E{XilXiIX i Ix
i k
Ix
31XiIX
k
lXjlXk 31 + 6bi bj bk E{lXi3
Ix Ix 51 +bi3
i k
i
E{lXijlX
E{Xi1xi
llx
E{lXislx i Ix i Ix k 31
lXj5lX
E{XilXil
Ix Ix 31
i
i k
Ix
ix
41
E[XilXiIX
E18
=
i
k
i
E{lXi3lX
all
the remaining
(Eq.
41
41
lXk
31
lXj3lXk5)
(V. 2)
Xj3
Ix
Appendix One;
(5.4.63)
for
solution,
using:
Eq. (5.4.12);
ii)
iii)
moments:
292
(X
El.
l5akc
kIXiXi
E(X
6642+4+6
1Xj
manner it
31X
E{XilXillXj3lX
19
=E
r=l
(V. 4)
kc
One
expectatiop
is
given
(V. S)
EX(r)
iI
bi]
31 (3cf,
kc
are
EX(4) = E{X
i
ii
by Eq.
given
IX JXk4j
= E{XilXiIX
(5.4.63).
[3C, 2b 0
2
is E18.
Ix
5jXijX
i
= E{XilXiIX
3b
Sa
+
11
i
[3CjkC2 ai(b
+ 6ai
i3
b i2
bk+
ai
2bibk2
3)1
+ Sai3bk
51X 11 [3crkc2
i
i
3+
(b
a
bi2bk+a12bibk2
6a
3)1
kc
kc2
the required
J)Ck31 [RrtC2
expectations
EX(S)
kc
are as follows:
EX(2)
EX(3)
kc
EX(l)'=
above
l5a
(V. 3)
Eq. (5.4.23).
with
E(jX
and the
+
kc3
and
from Eq. (1.55),
Appendix
ix Ix
E{X jXijX
k3X2.31
i
i
i
EX(r)
25
Eq. (5.4.63);
with
by:
where terms
l0a
+
kc
kc
45a
+
k C'
kc
l5a
Xi
2
where akc is defined
2=aiXi+aiX
kc
defined
a.
a
with
In this
4.
are E12.
311 [3crkc2(a
12bibjbk+
+ l0a ijk bb
ibi+
3Oai2 aibi,
3a
ibi2bi+
bk2+
12a iabi2bk
20a i3ajbk3,
293
3X
EX(7) zz E(jX
51)
:
4(a'
Crkc
2bbb+
12a
+
ijkijk
+ 10a
EX(8)
I'akc2 (a
+ 8a iaibibbk+
2abb2
l0a'
+
10a
+'2a
jjkik
3 Xi Xi. xi
E[XilXiIXJ
EX(11)
EJjX
13Xi
3a
30a a2bb2+
'[Xi3lXi'Xja'Xjll
EX(10)
bb2+
2
bj
b
aj
12a
20a ab
2)1
bibk
EX(9) -''t[Xi
b3+
[9a
I Xj II
311 [9a
kC4
10a
ia2bibk2+
2b2b+abb
kC4
[9a
2a 2b2b
b2b+
j, iiiik
(b
bk+
12
Sai bi bk2+
(b 2b+
C4 ikjjk.
(2bi
12a12bk3
2)1
Sa
Sa 2b3
ik
bb2+
Sa
bibk+
b
Sa
i2 k3)]
bibk2+
10ai aibk3
2)]
+'Sa ibibk
["C'kc 6bk3
733222
jXijX
EX(15)
(bi
+ 3ai bi
bk+
3a
ibibk
3)]
+ai3bk
+ab
Ix 11 [ai
i i
7 Ix
11
i
[ajs(b
i3+
3a
3aibi2bk+
i2bibk2
3)]
IX
73222
11[3ai
Ef
=
iXi
2
2a.
bbb+
+
(a
ibi+
2abb2
Sa
kijikijkijk
ai bibi+
4a
iaibibk
'ab3+a3bb
+2a
JX.
3
711
X
[3a (ai b+aibib+
EX(16) = E{
.2322
i
2bbb+
2+
2a.
Sa a2b
+
b'
ijkijkk,
2)]
4ai aibjbk
2a
a3b3+a.
3 b. ' b 2)1.,
294
EX(17)
E(X
Xjjx
Ix 1) [Sai(a
i
ibi+
3a
2
a. b bk+8a12aibibibk+
+6a.,
iabibi
10a i2ai2bibk2
Sa
+ Sa 13ai2bk3+ai2bibj2+a13bi2bk+
E {Xi3lXiIX
EX(18)
622.222
a"
Sa
EX(19)
llx
13X
2a2bbb+
36a.
+
ai3
+ 30a
13bi3+
i2b3+
9a
9a
10a
i2aibibi2+
b* bk
a.
Sa.iaj3bibk
ia2b12b+
12a
i3abbk
iaibibi7
Oa 2a3bb2+
1iijkiiikik
bi3+
3a
'2'b
+ai
i2b+a3bi2bk+
k3
511 (a
E{IX
(a
abb,
i2a3b
..
11[3a
2)1
l2ai
20a
2)]
a3bi2bk
3a3b3
i3ab2bk
21
i3ai2bibk
rI
r
41
295
APPENDIXVI
LIST
Listings
of all
'Probabilistic
programs
OF COMPUTER PROGRAMS
are contained
in a separate
document entitled
Description
Response:
Computer Programs'.
Notes
(1)
Programs in FORTRANlanguage
ICL 1906S computer
at the University
computations
cylindrical
on the
use either
of Liverpool
or on the
Regional
of Manchester
based on horizontal
'Morison'
wave loading
on
members.
FOR SINGLE MEMBERS
Program
Name
for
at the University
developed.
Published
Program Description
_
Information
OSH
Theory,
listings
&
results
in Ref. 10
OSF2
Theory,
listings
&
results
in Ref. 10
Theory,
listings
&
results
in Ref, 10
Results
in Ref.
12
Theory
results
outlined
in Ref.
&
12,
P-M (Hi/3)
spectrum;
'
algorithm).
OSF3
OSF4
As OSF2'but
OSF5
OSF6
OSF7
with
1 Deak
for'
296
Program
, Name
Program Description
Published
Information
OSF8
Results
in Ref.
12
OSF9
Results
i n Ref.
12
Oslo'
P-M (Hj/3/Tz)
OS16
Pierson-Holmes
pdf.
moments. inc. narrow
cdf.
OS20
OS20A
Statistics
inc.
force
,
OS20ER,
OS21
OS22
OS25
(b)
spectrum
wave/current
of short-term
densities
spectral
Theory summarised in
Ref. 17
in
in
in
in
range
spect.
JONSWAPspectrum.
wave climate
described
from
(1)
developed
Extreme distribution
long-term
based on programs
peak cdf.
(indept.
OSF3/5/6/8
period
peaks;
single
for
set of cards available
of exposure;
into
insertion
OSF2-OS10).
programs
OSSXX
(2)
developed
Extreme distribution
Based on
short-term
cdfs.
extreme
program OSF5 (indept.
peaks;
single
period
of exposure).
Theory outlined
Ref. 17
OS6XX
(3)
developed
from
Extreme distribution
Based on program
long-term
peak cdf.
OSF6 (indept.
peaks;
multiple
periods
of
exposure).
OS26
(4)
As OSF5X based
in OS25.
OSF3X
OSF5X
OSF6X
OSF8X
on wave climate
from
used
in
in
297
(c)
Program
Name
Program Desc#_ption
Published
Information
Theory
results
14
&
outlined
in Refs. 11 &
OS14.
OSIS
Short-term
joint
and marginal distributions of. response for 2 member systems.
2nd and 4th moments of -marginal
response, kurtosis
and correlation
(various
by
coeff.
spectra,
num. int.
Gauss-Hermite Quadrature,
OS15
floating
variable
range on response Y2)-
OS17
OR$
Theory
results
14
&
outlined
in Refs.. ll
Y2).
on response
OS23
Theory outlined
Ref. 79
(b)
OS24
from time-series
Response distributions
kinematics
simulation
of water particle
(short-term;
spectra ).
various
in
in
MISCELLANEOUS
Long-term
from cdf.
cdf. of individual
wave height
Tz).
of HI/3 (with 'associated'
in
':
&