Transcript ttf-philo

The “Design Wave Philosophy’’
*****
The “Design Wave Philosophy’’ *****
Calculation of the design wave
Wave forces on semi-submersible platforms
Wave forces and bending moments in FPSO-ships
Platform movements in large waves
Examples of heavy weather damage
What is a Rogue Wave ?
Why, where and when ?
Shall we design against Rogue and Freak Waves ?
What can a platform master do against Rogue and Freak Waves ?
Remote-sensing of sea conditions
Search And Rescue and emergency operations
Decision making in an emergency
Trondheim Tekniske Fagskole
The “Design Wave Philosophy’’
Trondheim Tekniske Fagskole
The “Design Wave Philosophy’’
• Standards,
class societies, rules and regulations
• Consequence-based design, safety factors, reliability
• Design wave vs. design sea state
• Statistics and extreme value theory
• Meaning of the 100-year (or 10000-year) wave
Trondheim Tekniske Fagskole
The “Design Wave Philosophy’’
Everything started in a way similar to the Oklahoma rush in the
Conquest of the West, but:
• Hurricane Anita in the Gulf of Mexico
• The design wave increased by 1 meter each year from return of
experience in the late 70’s North Sea
• The ”Alexander Kielland” accident occurred in 1980
Trondheim Tekniske Fagskole
The “Design Wave Philosophy’’
National requirements and shipping regulations from a large
amount of actors:
• National Agencies (Oljedirektoratet, HSE, ...)
• Classification Societies (DNV, API, Lloyds, BV, …, IACS)
• Standardisation bodies (ISO, Bnpé, DIN, …)
• Professional bodies (OGP)
Trondheim Tekniske Fagskole
The “Design Wave Philosophy’’
…ended up into a “philosophy for design”.
In the North Sea, design is determined by extreme
waves, and at the time (80’s), for fixed platforms with
quasi-static response, by the single largest wave that
would break the platform.
At that time, one would compute what happens with a
100-year wave and add a safety margin.
Trondheim Tekniske Fagskole
The “Design Wave Philosophy’’
A 100-year wave is the wave height that is exceeded in
average once every century over a large number of
centuries.
It is NOT exactly the same as having a 100-year
average interval between two exceedances, and NOT
AT ALL the same as being able to expect a duration of
the order of magnitude of 100 years before the next
after a given exceedance.
Trondheim Tekniske Fagskole
The “Design Wave Philosophy’’
• Standards, class societies, rules and regulations
• Consequence-based design, safety factors, reliability
(some points made by Markku Santala - Exxon)
• Design wave vs. design sea state
• Statistics and extreme value theory
• Meaning of the 100-year (or 10000-year) wave
Trondheim Tekniske Fagskole
Design Method Effectiveness
• Identification of controlling design conditions
– Failure to identify controlling conditions may impact project schedule or
lead to unacceptable performance
– Design practices that are over-conservative may not be cost effective
• For floating systems the maximum environment is not always
sufficient for design
– Maximum environment  maximum response
• Response-based methods provide an approach for identification of
controlling design conditions
– Implementation details key to effectiveness
Trondheim Tekniske Fagskole
Traditional procedures and limitations
• Fixed Platforms
– Response = f(Hmax) + secondary contributions (w s, v)
– Specifying the 100-year wave plus associated parameters leads to the 100-year
response approximately.
• Floaters
– Response=f(Hs, Tp,, ws, v, ) + secondary contributions
– Specifying the 100-year wave (or any other single parameter)
plus associated parameters DOES NOT necessarily lead to the 100-year response.
• Example limitations
– In central GoM where offset can be dominated by Loop Current in a VIV lock-in
condition.
– In western GoM responses can be dominated by wind plus associated conditions
Trondheim Tekniske Fagskole
Common “Patches”
•
Specify a set of 100-year cases and look for the dominant response.
Minimal specification might include:
–
–
–
–
100-year significant wave + associated wind and current
Range of associated spectral wave periods
100-year wind + associated wave and current
100-year current + associated wind and wave
•
Develop contours in Hs-Tp, Hs-ws, ws-v space to search for dominant
responses.
•
Multi-dimensional parameter contours —though theoretically possible—
are not necessarily practical or sufficient.
Trondheim Tekniske Fagskole
Response-Based Approach
• Methodology
– Determine limit state for critical systems
– Formulate response functions for each critical system element
• Realistic characterization effects of wind, wave, and current
• Computationally efficient
–
–
–
–
–
Develop long-term characterization of the environment
Simulate long-term response time history
Evaluate extreme response statistics
Identify environments that produced design response
Assess design for controlling environments
• Consideration
– Factors other than environmental conditions may have comparable
contribution
Trondheim Tekniske Fagskole
Traditional 100-yr Environments
(as per ISO regional annexes)
West Africa
GoM
central N. Sea
Hs
3.9 m
12.6 m
13.6 m
Tp, associated
15-17 s
14.6 s
15.5-19.4 s
ws, 1hr,10m
8 m/s*
46 m/s
35 m/s
* 3-second
gust is 30m/s. (due to West Africa squall conditions)
Trondheim Tekniske Fagskole
Why is the issue different for W. Africa?
• Response may be highly resonant near
its natural frequency.
• In the Gulf of Mexico, which is a semienclosed sea, there are no long period
waves to excite the heave resonance.
GOM – 100 YR Wave
• In environments like West Africa where
there are long period swells it may be
possible to excite this resonance.
• This comparison shows a heave
response more than 10 times greater in
a 1m, 25s swell than in the 100-year GoM
hurricane.
Trondheim Tekniske Fagskole
West Africa - Swell
Long-Term Characterization for Environment
•
45-Year Wave Hindcast
Assembling a long-term environmental
database can be problematic.
–
–
–
•
•
Wind and Waves - Hindcast data provided a 45year time history of continuous 6-hourly
“normal” winds and waves.
Squalls – Only one year of measured wind data
on the seasonal frequency and intensity.
Currents - A long-term synthetic time-series of
current based on a year of measurements.
For this region, squalls and currents have
little correlation to the swell dominated wave
environment.
Assembling long-term databases would be
more straight-forward in mature areas such as
the GoM or N. Sea but must still be done with
care.
Trondheim Tekniske Fagskole
45-Year Squall Distribution
Simulate Long-Term
Response Time History
Initialize & load
environmental database
Analyze next
seastate
Compute mean
forces & moments
Compute offset & resulting
mooring stiffness
Compute slow-drift,
wave-frequency and
wind-induced motions
at the keel
Compute min/max
stroke in seastate
Last seastate ?
no
yes
Archive results as input to
extreme value analysis
Trondheim Tekniske Fagskole
Extrapolation of Response to Extremes
Peak-Over-Threshold Analysis
•
With a 45-year sequence of responses,
extrapolation to a 100-year extreme is
straightforward.
•
If our response functions were perfect we could
use the results of the analysis directly. However,
the response model used was an approximation
and we can only use the analysis as a screening
tool to determine input conditions.
•
In past analyses in the GoM where we have used
extremely long synthetic time-series (500
years+), the 100-year response can simply be
picked out of the input database.
•
In this case we need to “back out” conditions
which lead to the 100-year response.
Trondheim Tekniske Fagskole
Determining the 100-Year Stroke Input Condition
•
To determine the environmental conditions which give rise to
the 100-year response we examine the conditions which
generated the largest peak responses.
•
None of the responses occurred in the region of the 100-year
Hs plus the “conservative” range on the associated T p. In fact
the 100-year response was more than 50% greater than the
response in the worst part of the 100-year Hs and associated
Tp range.
•
In this case the top ten responses were all caused by
conditions with long wave periods, modest wave heights and
negligible winds and currents.
•
The environmental conditions driving the 100-year stroke
response were backed out of the region of the top ten
responses using the response function.
• This result could have also been
determined by examining 100-year
Hs-Tp contours. And, for this case with a
known sharp resonance, a prudent design
team would explore this option in the
absence of having performed a response
analysis.
Trondheim Tekniske Fagskole
Design Cycle Considerations
•
The conditions determined by the response
analysis are dependent on the system
configuration.
•
In a subsequent design cycle where the DDCV
geometry and mass distribution was changed
the response analysis was re-run.
•
A case unrelated to swells emerged as the peak
case. A large tilt response to extreme wind
caused a large pull-down (right).
•
Here simply using Hs-Tp contours does not yield the critical response. Relying
contours requires examining other contour dimensions to ensure identification of
other conditions that may govern the extreme response.
Trondheim Tekniske Fagskole
Summary
•
Traditional methods based on SPJ experience are clearly dated and most of industry has made
some effort to move ahead with specifications of metocean conditions more appropriate for
floaters.
•
Specifying a limited set of cases (e.g. wind-dominated, wave dominated etc) in the absence of any
knowledge of the structure to be used is a first step but does not guarantee that the 100-year
response of every critical system element has been considered.
•
Judicious use of environmental contours and careful consideration of system resonance and
damping on various components of the system may lead to an acceptable range of design cases.
In cases where damping or VIV lock-in are an important part of the response it is not assured that
the contour approach will identify the critical cases.
•
Response-based analyses require designers and metocean specialists work together in a
collaborative (rather than sequential) mode to identify critical cases. Success requires :
–
–
–
–
the appropriate responses being screened,
a good input database,
good response models,
appropriate updates of response analysis as design matures.
Satisfying the above conditions is not easy and requires a non-trivial analysis and data gathering
effort.
Trondheim Tekniske Fagskole
The “Design Wave Philosophy’’
Main problems with the 100-year wave + safety factor approach:
• Failures occur for sub-extreme wave height combined with
other factors
• Actual level of safety is not known, not consistent over
different structures, and with sometimes costly
overconservativeness and sometimes dangerous
unconservativeness
Trondheim Tekniske Fagskole
The “Design Wave Philosophy’’
New “goal-based” approaches:
• Define target levels of reliability
• Probability of failure = Overall probability that simultaneously
“stochastic” action exceeds “stochastic” resistance
• Targets:
• 10-2 yearly: unmanned, no danger to environment
• 10-3 yearly: evacuatable, no danger to environment
• 10-4 yearly: manned, or danger to environment
10-4 yearly is similar to a 10000-year wave, it is also different.
Trondheim Tekniske Fagskole
The “Design Wave Philosophy’’
• Standards, class societies, rules and regulations
• Consequence-based design, safety factors, reliability
• Design wave vs. design sea state
• Statistics and extreme value theory
• Meaning of the 100-year (or 10000-year) wave
Trondheim Tekniske Fagskole
The “Design Wave Philosophy’’
For many kinds of structures, wave height is not the
only wave characteristic leading to failure.
Steepness, wavelength, wave groups, ringing,
springing, beam waves, etc. lead to consider one or
several sea states (durations of, say, 3 hours) as the
design conditions.
Trondheim Tekniske Fagskole
The “Design Wave Philosophy’’
Two ways to arrive to the “design wave”:
• Extrapolate the maximum waves measured in each
sea state
• Find the distribution of the largest Hs’s, and perform
convolution with the distribution of the ratio Hmax/Hs
The two methods should yield the same final value… if
assumptions are verified and database is sufficient.
Trondheim Tekniske Fagskole
The “Design Wave Philosophy’’
• Standards, class societies, rules and regulations
• Consequence-based design, safety factors, reliability
• Design wave vs. design sea state
• Statistics and extreme value theory
• Meaning of the 100-year (or 10000-year) wave
Trondheim Tekniske Fagskole
Statistics and extreme value theory
How can one extrapolate a few years of data to yearly
probabilities of occurrence of 10-4 ?
Extreme values theory is a very
Using measured or hindcast data
of a few decades, and the
“independent identically
distributed” assumption, it allows
to determine the likely distribution
of 10000 year extremes
Trondheim Tekniske Fagskole
powerful tool
Statistics and extreme value theory
Extreme values theory is a very
Trondheim Tekniske Fagskole
powerful tool
Statistics and extreme value theory
..., and not forgetting the “independent identically
distributed” assumption, ...
Trondheim Tekniske Fagskole
Statistics and extreme value theory
What does “independent identically distributed” mean ?
Independent, in practice, means that a single event should
not be counted more than once. Designers are very
concerned about independence, and tend to accept higher
uncertainties in order to ensure independence.
Often, they use POT (Peak Over Threshold) to retain only
one value per storm, and may even consider that 2 storms 3
days apart should be taken as a single one. In fact,
statisticians have shown that many kinds of slight
dependence do not spoil extreme value extrapolation.
Trondheim Tekniske Fagskole
Statistics and extreme value theory
What does “independent identically distributed” mean ?
Identically distributed means that events are of a single kind. A
typical case where it is not verified is locations where hurricanes
occur once in, say, 10 years. Extrapolation from the main bulk of
measurements is thus useless.
Identically distributed is very difficult to verify, so designers
have assumed it in many cases.
Hence the question whether rogue waves are “normal” extremes
or “ones from nowhere”, and its crucial importance.
Trondheim Tekniske Fagskole
The “Design Wave Philosophy’’
• Standards, class societies, rules and regulations
• Consequence-based design, safety factors, reliability
• Design wave vs. design sea state
• Statistics and extreme value theory
• Meaning of the 100-year (or 10000-year) wave
(Some points made by Sverre Haver - Statoil)
Trondheim Tekniske Fagskole
Jacket structure in the North Sea
Trondheim Tekniske Fagskole
Target Safety Level of Offshore Structures
By designing according to Norwegian Rules and Regulations, it is tacitly
assumed that the nominal annual probability of structural failure is
10-4 – 10-5 or lower.
 A structure should resist all wave events or wave induced load events
corresponding to an annual exceedance probability of 10-4 with a proper
margin (i.e. in worst case some local damage damage may be experienced).
 Quantity of concern regarding ultimate safety is therefore the very, very
upper tail of the annual distribution function of wave events and loads.
Trondheim Tekniske Fagskole
Target Safety Level of Offshore Structures
Regarding overload failures, industry aims to fulfill target by the following
design controls:
i) Ultimate Limit State (ULS)
Component based control ensuring that the 10-2 – annual probability loads
multiplied by a load factor are lower than a low percentile of the elastic
component capacity divided by a material factor.
ii) Accidental Limit State (ALS)
System based control ensuring that the 10-4 annual probability load is smaller
than the the system capacity.
Trondheim Tekniske Fagskole
Governing limit state (introducing the ugliness property)
ALS governs design
sc,ALS,2
Bad-behaving problem
Loadlevel
ULS governs
design
1.3*sc,ULS
sc,ALS,1
Well-behaving problem
sc,ULS
0
1
2
3
4
- log(annual exccedance probability)
Trondheim Tekniske Fagskole
5
If freak waves exist – what is the problem?
For ship and platforms, freak waves will
mainly represent a problem if their crest hits
a structural element which is not designed
for wave loads.
?
?
Trondheim Tekniske Fagskole