You are on page 1of 66

TECHNICAL REPORT

JOINT INDUSTRY PROJECT


GUIDELINE FOR OFFSHORE STRUCTURAL
RELIABILITY ANALYSIS
- APPLICATION TO TENSION LEG PLATFORMS

REPORT NO. 95-3197


REVISION NO. 02

DET NORSKE VERITAS

DET NORSKE VERITAS

TECHNICAL REPORT
Date of first issue:

Project No.:

5 September 1995

DET NORSKE VERITAS AS

22210120

Approved by:

Division Nordic Countries

Organisational unit:

istein Hagen
Principal Engineer

Structural Reliability and Marine


Technology

Client:

Client ref.:

Joint Industry Project

Rolf Skjong

Veritasveien
1,
N-1322 HVIK, Norway
Tel: +47 67
57
72
50
Fax: +47 67 57 74 74
Org. No: NO 945 748 931 MVA

Summary:

Characteristic features of TLPs are briefly described, and an overview is given of TLP response analysis
to environmental actions. Model uncertainties involved in the response analysis are discussed and
recommendations are given for use in reliability analysis. A set of limit states are discussed, that may be
considered in the Level III reliability analysis of TLP structural components, including:
tether foundations
stiffened shell structures in columns and pontoons
structural connections between deck girders, columns and pontoons,
air gap.
Experience from an example of reliability analysis an ultimate limit state for TLP tethers is summarised.
This ultimate limit state includes the combined effects of external pressure, axial tension, and bending
moment.
The modelling of the fatigue limit state for TLP tethers is discussed.
Recommendations for further work are included.

Report No.:

Subject Group:

95-3197

P12

Indexing terms

Report title:

Guideline for Offshore Structural


Analysis
- Application to Tension Leg Platforms

Reliability

tension_leg_platform,
tethers,
structural_reliability,
combined_loads

Work carried out by:

Jan Mathisen, Farrokh Nadim,


Oddrun Steinkjer, Inge Lotsberg

No distribution without permission from the


Client or responsible organisational unit

Work verified by:

Vigleik Hansen, istein Hagen


Date of this revision:

29 Nov. 1996

Rev. No.:

02

Limited
distribution
Det Norske Veritas
Number of pages:

64

Unrestricted distribution

DET NORSKE VERITAS AS, Head Office: Veritasvn. 1, N-1322 HVIK, Norway Org. No: NO 945 748 931 MVA

within

Guideline for Offshore Structural Reliability Analysis


- Application to Tension Leg Platforms
Report No. 95-3197, rev. 02

Table of Contents

Page No. i
Contents

Page

1. INTRODUCTION.........................................................................................................1
1.1 Objective
1
1.2 Definition of a TLP
1
1.3 Design Codes for TLPs
2
1.4 Arrangement of Report
2
2. RESPONSE TO ENVIRONMENTAL ACTIONS ..........................................................3
2.1 General Description
3
2.2 Classes of Response
4
2.3 Stillwater Loads
4
2.4 Water Level Loads
5
2.5 Mean Environmental Loads
5
2.6 Wave-Frequency Loads
5
2.7 Low-Frequency Loads
6
2.8 High-Frequency Loads
7
2.9 Other Effects
7
2.10 Summary of Response Components
8
3. MODEL UNCERTAINTY .............................................................................................9
3.1 General Remarks
9
3.2 Mean Environmental Loads
9
3.3 Wave-Frequency Loads
9
3.4 Low-Frequency Loads
10
3.5 High-Frequency Loads
11
3.6 Tether Response
11
3.7 Platform Structural Response
11
3.8 Capacity Formulations
11
3.9 Summary
11
4. DISCUSSION OF LIMIT STATES .............................................................................13
4.1 Component selection
13
4.2 Foundations
13
4.2.1 Tension pile anchors
13
4.2.1.1 Special considerations
15
4.2.2 Gravity and suction anchors
17
4.2.3 Cyclic shear strength of soil
19
4.3 Columns and Pontoons
19
4.3.1 OSCS bay failure
20
4.3.2 OSCS plate failure
21
4.3.3 Stiffened flat plates
21
4.3.4 Yielding
21

Guideline for Offshore Structural Reliability Analysis


- Application to Tension Leg Platforms
Report No. 95-3197, rev. 02

4.4 Main Structural Connections


4.5 Deck Girders
4.6 Air Gap

Page No. ii
Contents

22
22
22

5. SUMMARY OF APPLICATION EXAMPLES............................................................24


5.1 Principal Dimensions of Example Structure
24
5.2 Summary of Tether ULS example
24
5.2.1 Failure criterion
25
5.2.2 Analysis of input required by limit state function
26
5.2.3 Probabilistic formulation of the reliability problem
27
5.2.4 System response calculations
28
5.2.5 Stochastic variables
29
5.2.6 Coupling of system response calculations with reliability calculations
30
5.2.7 Initial reliability results
30
5.2.8 Consideration of the possibility of failure anywhere along the length of one
tether
32
5.2.9 Consideration of the possibility of failure in any of the TLP's tethers
32
5.2.10 Comparison of out-crossing results with simplified model
33
5.2.11 Sensitivity Analysis with Respect to Tether Ovality
34
5.2.12 Conclusions for tether ultimate limit state
35
5.3 Tether FLS Summary
35
5.3.1 Failure criterion
36
5.3.2 Analysis of Input required by limit state function
37
5.3.2.1 Material properties
37
5.3.2.2 Short and long term load-effects
38
5.3.2.3 Combination of axial and bending stresses
38
5.3.2.4 Load components
38
5.3.2.5 Cycle counting
39
5.3.2.6 Stress concentration factors
39
5.3.2.7 Model uncertainties
39
5.3.3 Probabilistic formulation of the reliability problem
40
5.3.4 Stochastic variables
40
5.3.5 System reliability
41
6. RECOMMENDATIONS FOR FURTHER WORK.......................................................42
6.1 A Method to Properly Manage the Effect of Low-Frequency Offset on WaveFrequency Response
42
6.2 Distributions and Dependencies
42
6.3 Outcrossing Analysis
43
6.4 Load Effects
43
6.5 Fatigue Limit State
43
6.6 Ovality
43
7. REFERENCES..........................................................................................................44
8. FIGURES ..................................................................................................................52

Guideline for Offshore Structural Reliability Analysis


- Application to Tension Leg Platforms

Page No. 1

Report No. 95-3197, rev. 02

1. INTRODUCTION
1.1 Objective
The objective of this section of the Guideline for Offshore Structural Reliability Analysis is to
provide the following information with respect to tension leg platforms:
an overview of the characteristics of that structure's response to environmental
actions,
detailed guidance on the reliability analysis of that structure with respect to several
important modes of failure,
examples of reliability analyses applied to selected failure modes for that structure
type.
The guidelines are intended for the application of Level III reliability analysis to the structure
type; i.e. in which the joint probability distribution of the uncertain parameters is used to
compute the probability of failure. This is usually a fairly demanding type of analysis, and is
primarily expected to be applied in the final design or verification of major load bearing
components of a structure. Hence, the guidelines to be prepared in this project will concentrate
on the requirements for these types of analysis, and will not make any attempt to embrace all
aspects of the design of an offshore structure. However, within these limitations, our aim is to
cover all aspects of this type of reliability analysis.

1.2 Definition of a TLP


A Tension Leg Platform (TLP) can be defined as a vertically moored (or tethered), buoyant and
compliant installation, where the excess buoyancy of the platform maintains tension in the tether
system. In general a TLP may be designed to serve a number of functional roles associated with
offshore oil and gas exploitation. TLPs are considered particularly suitable in medium to deep
water, implying water depth in the range from 300m to 1000m. Mercier et al. (1991) provide a
useful review of developments in and prospects for TLPs, and cite design studies for 2000m
water depth without finding any technical barriers at this depth.
The tethers are normally attached at the corners of the platform, such that heave, roll and pitch
motions are restrained, as indicated in Fig.1.1. An alternative is to attach the tethers centrally,
thus allowing roll and pitch, and restraining only heave motion. This alternative type is usually
called a heave resisted, or single leg TLP; c.f. White et al. (1988).
A TLP is usually considered compliant with respect to surge, sway and yaw motions. In deep
water a catenary mooring system may be used in addition to the tether system. The catenary
mooring system may either be used to actively position the TLP relative to wells, or to passively
reduce the total offset of the TLP in severe weather. Such a supplemental mooring is applied to
the Auger TLP.
In deep water the tethers are usually made nearly neutrally buoyant, due to weight
considerations. Bottom connections are made to a piled foundation, or to a gravity based
structure, or to suction anchors. The connection to the TLP is either by external attachment to
columns, or internally through the columns. Flex joints are required at both ends of the tethers to
reduce the applied bending moments in the tethers.

Guideline for Offshore Structural Reliability Analysis


- Application to Tension Leg Platforms

Page No. 2

Report No. 95-3197, rev. 02

1.3 Design Codes for TLPs


Existing design codes for TLPs naturally cover many more aspects of the design process than
are considered here. Code analysis procedures are simpler, but openings are provided for the
application of level III reliability analyses.
A recommended practice for TLP design is provided in API RP2T (1987). A working stress
design format is adopted in APIs recommended practice, but in section 4.2.1 it is recognised
that: A probabilistic analysis combined with reliability based design methods may provide a
more versatile method for developing an optimum design, but requires extensive analysis. API
has established a technical advisory committee, TAC93-20, to revise their recommended practice
for TLPs.
Design rules are also available from DNV (1989). In DNVs Rules for Fixed Offshore
Installations, the structural design may either be based on the allowable stress method or
according to the partial coefficient method, but in Ch.1, Sec.4, paragraph A104 it is also stated
that: Level 3 reliability analysis methods are mainly considered as applicable to unique, special
case design problems, for calibration of the level 1 methods and for conditions where limited
experience exists.

1.4 Arrangement of Report


The response of TLPs to environmental loads is described in chapter 2, together with methods
for computation of the resulting load effects. The model uncertainties associated with the
computation of these load-effects are discussed in chapter 3. Important limit states are described
in chapter 4, and the stochastic modelling of these failure modes are discussed . Chapter 5
provides a summary of a reliability analysis, made for an ultimate state of TLP tethers. The
details of this analysis are presented in a separate report by Mathisen et al. (1995).
Recommendations for further work are given in chapter 6, based mainly on experience gained in
developing this guideline. References are listed in chapter 7, and figures are grouped together in
chapter 8.
The present report is based on the general guidelines set out in the Guideline for Reliability
Analysis of Marine Structures - General, (Skjong et al., 1995). Companion applications are also
available for jacket (Sigurdsson et al., 1996) and jack-up structures (Croker, 1996).

Guideline for Offshore Structural Reliability Analysis


- Application to Tension Leg Platforms

Page No. 3

Report No. 95-3197, rev. 02

2. RESPONSE TO ENVIRONMENTAL ACTIONS


2.1 General Description
Under stillwater conditions; i.e. in the absence of wind, wave, current and tidal effects, the TLP
floats in a stationary, stillwater position. In these conditions, the dominant load effects are due to
weight, buoyancy, mooring pretension, and riser pretension. The centre of buoyancy and centre
of gravity (due to weight and pretension) are normally designed to be located on a vertical line
through the geometric centre of the hull.
Ordinary platform operations involve moving
equipment and supplies of appreciable weight, and lead to some redistribution of the total
pretension load between the tethers. Ballast control can be applied to keep these effects within
acceptable limits.
Water level changes due to tides and storm surge cause variations in the draught of the hull, since
the tethers and excess buoyancy prevent the platform rising and falling with the water level. The
change in draught results in a change in the buoyancy forces, and a corresponding change in the
tether tensions, unless compensated by ballast adjustment.
The response of TLPs to environmental actions may conveniently be treated as a quasistationary stochastic process, in common with most other offshore structures and ships. This
implies stationary response in the short term, when the wind wave and current conditions may be
considered stationary. While the long term, non-stationary response, is obtained by an
appropriately weighted combination of the short term, stationary response. Hence, the basic
problem involves response analysis under stationary environmental conditions. Each short term
environmental state may be characterised as a realisation of a stochastic vector, whose
components include the significant wave height, peak wave period, mean wind speed, current
speed, water level variation, and the directions of the environmental effects. The joint
distribution function of these environmental variables is required in a level III reliability analysis.
In a stationary environmental state, the environmental actions and system response may
conveniently be separated into mean, time-independent components, and time-varying
components. Since the TLP is compliant to horizontal environmental forces, the response to
mean environmental actions is most apparent as mean surge and sway offset of the platform, in
the direction of these environmental loads. A mean yaw angle may also be induced. The
relatively stiff tethers cause a corresponding set-down of the platform, as a second order effect,
which becomes apparent at large offsets (or large yaw angles). Mean environmental loads in a
vertical plane also cause heave, roll, and pitch platform motions, which are not easily apparent,
since they involve elastic strain of the tethers, of much smaller magnitude.
Time-varying, environmental loads yield corresponding, time-varying, platform motions, easily
apparent in the horizontal plane, as surge, sway and yaw, and less apparent in the vertical planes,
as heave, roll and pitch. The compliant characteristic in the horizontal plane leads to low natural
frequencies for surge, sway and yaw motions. The stiff characteristic in the vertical plane leads
to high natural frequencies for heave, roll and pitch motions. A TLP is normally designed so that
these natural frequencies are located on either side of the frequency band of encountered waves.
Hence, first order wave loads, at the encounter frequencies, do not tend to excite the natural
frequencies of rigid body platform motions. However, turbulent wind loads excite the compliant
modes, together with nonlinear, low-frequency, wave loads, referred to as slow drift forces.
Other categories of nonlinear, high-frequency, wave loads excite the stiff, vertical modes.
Common usage differentiates between two such forms of resonant response, in which:

Guideline for Offshore Structural Reliability Analysis


- Application to Tension Leg Platforms

Page No. 4

Report No. 95-3197, rev. 02

springing is of a relatively stationary nature, and


ringing is of a relatively transient nature.
The response of the TLP in the form of rigid body platform motions is emphasised in the
preceding paragraphs. This is natural, because these response modes are most easily apparent,
and because the hull and deck form a relatively stiff structure. This being the case, elastic
structural dynamics of the platform can usually be ignored, and quasistatic analysis can be
applied, provided the lowest platform natural frequency has been checked, and found to lie
outside the range of significant excitation frequencies. Furthermore, the elastic deformations of
the platform are relatively small, and can usually be neglected when analysing other parts of the
TLP system; viz. the tethers. Note however, that three, vertical, rigid body modes are
insufficient to completely explain tether tension measurements at four platform corners, and that
some of the unexplained variation is likely to be due to platform flexure.
Fairly good estimates of compliant platform motions can usually be obtained by treating the
tethers as inelastic strings, without local external loading. This model relies on the effectiveness
of the flex joints at the ends of the tethers in reducing bending effects. The tethers and risers
tend to contribute somewhat to the damping of the compliant platform motions, but the
corresponding local loads can be negligible. Elastic treatment of the tethers is necessary to
include the vertical modes of platform motion. This may be adequate in relatively shallow water.
In deeper water, it becomes necessary to allow for dynamic response in the structural analysis of
the tethers themselves, though tether dynamics may still have negligible effect on the tension
applied to the TLP hull. In very deep water, the exposed length of the tethers may be so large
that the local hydrodynamic loads applied to the tethers have a significant effect on the compliant
platform motions, and possibly on the vertical platform motions, too.

2.2 Classes of Response


In general the following Load Categories are defined when designing a TLP:
- Permanent Loads (P)
- Live Loads (L)
- Deformation Loads (D)
- Environmental Loads (E)
- Accidental Loads (A)
As stated above, this section mainly considers environmental loads and load-effects related to
TLPs, because the stochastic modelling of these load-effects is of prime concern in structural
reliability analysis. However other load categories also need to be addressed in this connection.
This is mainly related to permanent loads and deformation loads.

2.3 Stillwater Loads


Stillwater loads for a TLP will typically be selfweight of the structural components (topside,
hull, tethers, risers and foundations), hydrostatic pressure acting on hull, tethers, risers and
foundations, and soil reaction and friction forces on the foundations. All these loads are
categorised as Permanent Loads. Pretension in the tethers arises as a result of these loads, and
varies depending on actual design conditions and the TLP in question. Tether pretension values
in the range 25 - 35% relative to the total axial tether load budget may be given as a typical
figure. Mercier et al. (1991) give examples of total pretension relative to hull displacement
ranging from 16% to 27%. External hydrostatic pressure will influence design of columns,
pontoons, lower part of tethers and the foundations.

Guideline for Offshore Structural Reliability Analysis


- Application to Tension Leg Platforms

Page No. 5

Report No. 95-3197, rev. 02

The parameters governing stillwater loads are carefully controlled. Hence, it seems reasonable to
assume that the stillwater loads may be taken to be deterministic quantities, rather than stochastic
variables. Some support for this assumption is provided by Mercier et al.s (1991) report that
the pretensions in the tethers of the Jolliet TLP, as measured after installation, were within about
2% of each other.

2.4 Water Level Loads


Variations in hydrostatic pressure and buoyancy caused by changes in water level due to tides
and storm surge are to be regarded as environmental loads. Depending on the water level
variations and waterplane area of the TLP hull, these load effects may be 5 - 10% of the total
load budget for the tethers.
These loads are fairly simple functions of the water level. The water level variation due to tide
and storm surge can conveniently be treated as part of the vector of short term environmental
conditions. Hence the water level loads are handled as stochastic variables, which are time
independent within each short term environmental state. In the reliability analysis, these loads
may be grouped together with the mean environmental loads.

2.5 Mean Environmental Loads


Mean environmental loads are loads from current, mean wind and mean wave drift forces. These
forces result in a platform offset and an overturning moment acting on the hull, which introduce
axial loads in the tethers. The platform offset is accompanied by a set-down effect, which also
introduces loads in the tethers. The horizontal restoring characteristic of a TLP is nonlinear and
is mainly governed by tether and riser pretension, waterplane stiffness, and weight of tether/riser
system.
The current loads can be determined as drag forces, using appropriate drag coefficients. An
alternative approach may be to apply model testing. Mean wind loads can also be determined as
drag forces, using appropriate drag coefficients for this purpose. Wind model testing will
normally be required at some stage in the design process, due to the importance of wind loads
(both mean wind and dynamic wind loads). Mean wave drift forces are determined either by
diffraction/radiation theory computer programs, or by model testing.
The current speed, mean wind speed, and wave spectral parameters are components of the vector
of short term environmental conditions. Thus the mean loads resulting from these environmental
effects are stochastic variables, that vary between environmental states, but are time-independent
within each short term environmental state.

2.6 Wave-Frequency Loads


First order wave motions of the platform and direct wave loads on hull, tethers and risers will be
referred to as wave-frequency loads. The wave period interval associated with these loads will
typically be from 4 seconds up to 30 seconds. First order wave loads acting on the TLP hull may
be estimated based on diffraction/radiation theory, Morisons equation and/or other methods or
tests.
The wave-frequency loads are dependent on the short term wave conditions. They are also time
dependent; i.e. stochastic processes. Since the wave elevation process may be assumed
Gaussian, and the wave-frequency loads are obtained as a linear transformation of the wave

Guideline for Offshore Structural Reliability Analysis


- Application to Tension Leg Platforms

Page No. 6

Report No. 95-3197, rev. 02

elevation, the wave-frequency loads are also Gaussian, and may be characterised by a covariance
matrix, or by a corresponding set of power spectral densities and cross-spectral densities. This
type of information is usually required in the reliability analysis.
Rigid-body platform motions are first computed in the form of transfer functions (or response
amplitude operators) in regular, long-crested waves. The transfer functions may be combined
with wave spectra to compute the covariance matrix of the response. These computations are
linear, and require linearisation of the restoring forces that act on the platform. The linearised
restoring forces may be regarded as a function of the platform offset. In particular, the coupling
coefficients between the modes of motion increase with offset. Usual design practice is to repeat
the wave-frequency response analysis for linearisations at a few different offset locations, also
including the increased draught due to set-down. Appropriate sets of linearised response are then
applied in each of the standard design cases that are considered. In the reliability analysis, it may
be worthwhile to arrange interpolation of the linearised response, as a function of offset.
Alternatively, an iterative approach can be applied, by checking that the set of linearised transfer
functions applied in a reliability analysis is suitable at the offset conditions defined by the
resulting design point of the analysis.
When dynamic analysis of the tethers is required, then this can also take the form of a linearised
analysis. Transfer functions for the local tether response in terms of axial tension and bending
moments about two cross-sectional axes are obtained. The tether response is driven by the
transfer functions for the rigid body platform motions at the upper end, and by hydrodynamic
forces acting locally on the tethers. The linearised tether response can also usefully be treated as
a function of offset. Nonlinear, time domain analysis of the tether response is also a possibility,
but the increased amount of computational effort is usually undesirable in reliability analysis,
where the response has to be obtained in many environmental states. However, nonlinear tether
response calculations can be very useful to quantify the model uncertainty of the linearised
analysis.

2.7 Low-Frequency Loads


Low-frequency loads are caused by excitation due to wave drift forces and wind gust loading.
The magnitude of forcing for these two excitations are generally small, however in combination
with relatively low damping the responses may be severe. Responses will in this context mainly
be platform offset (surge, sway and yaw motions) and set-down due to offset. These slowly
varying motions will oscillate with the system eigenperiod of the relevant component. The range
of eigenperiods will typically be 1 minute - 3 minutes.
In irregular waves, 2nd order potential theory predicts a steady (see section 2.5: Mean
Environmental Loads) and a slowly varying horizontal force termed the wave drift force.
According to the theory this occurs at the difference frequencies of the wave energy. As
mentioned, the damping level is governing the response due to low frequency loads. Potential
damping is negligible at these wave periods. The same may be stated for viscous damping,
particularly if there is no current present. Wave drift damping is therefore an important damping
source which governs the slowly varying drift motions. Slow drift excitation forces have been
available from diffraction programs for two decades, at least in terms of the mean drift force in
regular waves, describing the main diagonal of the quadratic transfer function. This type of
information can be used to compute the slowly-varying drift force in irregular waves, by
application of Newmans (1974) approximation. In recent years, more advanced diffraction
programs have become available, able to compute the off-diagonal, bichromatic, and bidirectional parts of the drift-force quadratic transfer function. However, these computations are

Guideline for Offshore Structural Reliability Analysis


- Application to Tension Leg Platforms

Page No. 7

Report No. 95-3197, rev. 02

very time-consuming, and seem unlikely to be applied in reliability calculations, in the near
future.
The low-frequency loads are dependent on the short term wind and wave conditions. They are
also stochastic processes. Since the wind gusts are assumed Gaussian, the low-frequency wind
loads may also be taken as approximately Gaussian. However, the low-frequency wave loads are
not Gaussian (Nss, 1990). It is suggested that the combined low-frequency loads due to wind
and waves may be approximately treated as a Gaussian process, until the required distribution
theory has been more fully developed.
The tether response to the low-frequency platform motions can simply be determined by
quasistatic analysis. Corresponding low-frequency wind and wave moments also act on the
platform, but are not amplified by resonant platform motions in the same way that the horizontal
forces are. The tether response induced by the moments should be included, but can be treated in
a much simpler manner, provided it is confirmed that these components are relatively small.

2.8 High-Frequency Loads


Large volume offshore platforms like TLPs may experience significant resonant response to
wave induced loads at frequencies well above the wave encounter frequency. These responses
are referred to as springing and ringing (cf. section 2.1). Both of these phenomena induce heave,
roll and pitch motions of the TLP, which in turn induce axial oscillations in the tether system.
Generally, springing is important in connection with the fatigue design of tethers whereas ringing
is governing for the ultimate design of tethers and hull/topside structure.
Springing response may be determined by analytical methods via computation of Quadratic
Transfer Functions (QTF) for response to second-order wave forces, often referred to as sum
frequency forces. The springing response is heavily dependent on the level of damping; e.g.
potential, viscous, soil, structural and tether system damping. A total damping ratio less than 1%
of critical damping is typical in connection with the springing response. Fatigue in the tethers
may, therefore, be totally governed by springing with low values of damping in the dynamic
system. Nss (1994) provides some information on the probability distribution of springing
response, and correlation with wave-frequency response. Springing tends to be considered less
important for ultimate limit states, in severe waves, based on the characteristics of the quadratic
exciting forces. However, there is currently some discussion of the adequacy of 2nd order theory
for the computation of springing, cf. Mercier (1994).
At present there exists no commercial computer program for simulation of TLP ringing response.
As a consequence one has to rely on carefully performed model tests. Important factors
governing ringing response are; system eigenperiods (for heave, roll, and pitch), the profile and
kinematics of steep waves, interaction effects, and the line of action of ringing forces relative to
the rotational centre of the TLP. The damping level is considered to be of relatively minor
importance in connection with ringing phenomena. The response level due to ringing may be at
the same (and higher) level as the first order tether tensions due to platform heave, roll and pitch.
Ringing effects will not be included in this guideline, due to the difficulty of quantifying these
effects at present (cf. Mathisen et al., 1994a). Considerable effort is currently being expended to
improve this situation, and suitable models should hopefully be developed in the near future.

2.9 Other Effects


Mispositioning is said to occur if the tether foundations on the sea floor do not accurately mirror
the attachment points to the TLP. This is mainly related to installation procedures and the

Guideline for Offshore Structural Reliability Analysis


- Application to Tension Leg Platforms

Page No. 8

Report No. 95-3197, rev. 02

requirements for the foundations on the seabed. Mispositioning causes an uneven distribution of
tension between the tethers, which tends to increase with platform offset. Significant increases
in tension may be caused by mispositioning. Mispositioning needs to be taken into consideration
early in design. Hamilton (1993) provides an approach to the calculation of mispositioning
loads.

2.10 Summary of Response Components

Table 2.1 Summary of response components.


Component
Stillwater
Water level
Mean
environmental
Wavefrequency

Lowfrequency
Springing
Ringing
Mispositionin
g

Analysis Technique
Weight, buoyancy and foundation forces
- quasistatic
Buoyancy change
- quasistatic
Wind drag coefficients,
current drag coefficients,
wave drift from diffraction theory
-quasistatic
Diffraction theory for linear transfer functions
for rigid body response,
- may interpolate on offset
- quasistatic for hull response
- dynamic tether analysis
Wind spectrum and drag coefficients or
admittance function,
diffraction theory for wave drift,
- quasistatic
Diffraction theory

Distribution
Deterministic
Stochastic variable function of environmental
conditions
----- -----

Gaussian stochastic process


- & function of environment

Approx. Gaussian stochastic


process, or more precise
dstn. & function of environment

Guideline for Offshore Structural Reliability Analysis


- Application to Tension Leg Platforms

Page No. 9

Report No. 95-3197, rev. 02

3. MODEL UNCERTAINTY
3.1 General Remarks
Model uncertainty is a type of epistemic (knowledge) uncertainty, and defined in the General
Guidelines as: uncertainty due to imperfections and idealisations made in physical model
formulations for load and resistance as well as in choices of probability distribution types for
representation of uncertainties. Model uncertainties relevant for TLPs, and in particular for
TLP load and response analysis are addressed in this chapter. In most cases, it is convenient to
define a model uncertainty factor as a stochastic variable Ux associated with an intermediate
estimate of some response or load-effect x, such that the response value including model
uncertainty is the product Ux x.

3.2 Mean Environmental Loads


Marthinsen (1989) has compared computed mean surge of a TLP in an extreme environmental
state with model test results and obtained deviations of -16% excluding current effects, and -12%
with wind, waves and current. Marton and Mathisen (1993) have compared computed mean
surge of a TLP due to wind and waves with model test results in two environmental states, and
obtained deviations of +12% and -1%.
Banon et. al (1994) have applied a coefficient of variation of 20% to static offset, and 10% to the
tether tension induced by static offset and by mean moment, and Mathisen et al. (1994b) have
used 10% for TLP hull stresses due to mean offset, in two related reliability analyses.
The mean offset and related loads are largely dependent on aerodynamic drag coefficients,
hydrodynamic drag coefficients and mean wave drift force coefficients. The computed mean
force, and system response will closely reflect the accuracy of these coefficients under specific
environmental conditions. If the drag coefficients are based on model tests for the actual TLP
considered, and the wave drift forces also are from dedicated model tests or diffraction
calculations, then a coefficient of variation as low as 10% should be attainable. If the force
coefficients are based on generic data, then it would be prudent to increase the coefficient of
variation for the model uncertainty to about 20%. No bias should be included in this model
uncertainty unless strong evidence to the contrary is found for the TLP considered.

3.3 Wave-Frequency Loads


Eatock Taylor and Jefferys (1986) have compared calculated wave-frequency force coefficients
and transfer functions for motions of a TLP, supplied by 17 different organisations. This
comparison shows good agreement in the major trends in the transfer functions, but includes an
appreciable amount of variability between the individual results. For short wave periods, in
particular, the variability is large, and is related to the varying fineness of mesh applied in the
numerical modelling of the TLP hull by the different organisations. Jefferys (1987), and
Korsmeyer et al. (1988), and Newman and Lee (1992) provide some information on how the
mesh fineness affects the results.
In practice, the transfer functions are integrated with a wave spectrum to obtain the standard
deviation of the response in a sea state. It is more convenient to apply a model uncertainty factor
to such a response standard deviation, than to the underlying transfer function ordinates. In
principal, such a model uncertainty might be expected to vary with sea state: a relatively low
uncertainty would apply in mild weather, and a somewhat higher uncertainty would apply in
severe weather, when nonlinear effects become more important. In practice, insufficient

Guideline for Offshore Structural Reliability Analysis


- Application to Tension Leg Platforms

Page No. 10

Report No. 95-3197, rev. 02

information is available to specify the model uncertainty as a function of the sea state. The
model uncertainty is taken as a stochastic variable, independent of the variables describing the
sea state. However, it may be reasonable to adjust the magnitude of the uncertainty with respect
to the type of analysis; i.e. lower uncertainty for fatigue analysis, which is dominated by
moderate sea states, and higher uncertainty for overload analysis, which is dominated by extreme
sea states.
Nielsen (1992) gives a summary of the comparison of motion response for two floating
production platforms from 23 different institutions. A turret-positioned ship and a deep draft
floater are considered. The deep draft floater is somewhat similar to a TLP in this context. When
including all the results supplied, the average coefficient of variation (CoV) for the standard
deviation of the first order motions is found to be 16%. This is based on a severe sea state with a
significant wave height of 15.5 m. After a critical review of the results, the CoV is reduced to
9%.
Banon et al. (1994) and Mathisen et al. (1994b) describe results from a joint industry project to
calibrate load and resistance factored design equations for TLPs. In this project a CoV of 5%
was applied as model uncertainty to wave-frequency offset and hull stresses, while a CoV of
10% was applied to wave-frequency tether tension.
Provided an adequate discretisation is applied in the hydrodynamic response calculation, then it
is recommended that a CoV of 10% is applied as model uncertainty for the standard deviation of
wave-frequency response of the TLP. The mean value of the model uncertainty should be 1.0
unless specific information about bias is available; e.g. from comparison with full scale
measurements or model tests.

3.4 Low-Frequency Loads


Nielsen (1992) also gives a comparison of computed results for the standard deviation of the
low-frequency surge motion, reporting a coefficient of variation of 44% for the deep draft
floater, and that the large amount of scatter is primarily due to variation in the damping applied
to the low-frequency motion. Wind and current effects were not included in this comparison, so
that the damping was mainly slow-drift damping, for which computational models have only
recently been developed. In an extreme environmental state, with wind and current present,
additional aerodynamic damping and viscous damping should tend to reduce the sensitivity of the
results to the slow drift damping.
Marthinsen (1989) compares the computed standard deviation of combined wave-frequency and
low-frequency TLP surge in an extreme environmental state with model test results, and obtains
-1% without current, and -15% with wind, waves and current. The low-frequency surge is much
larger than the wave-frequency surge in this comparison. Marton and Mathisen compare the
computed standard deviation of low-frequency TLP surge in two environmental states and obtain
deviations of +18% and -8%.
Banon et al. (1994) have applied model uncertainties with 44% coefficient of variation to lowfrequency wind induced offset, 28% to low-frequency wave induced offset, and 21% to lowfrequency wave induced tension. Mathisen et al. (1994b) have applied a model uncertainty of
11% to low-frequency TLP hull TLP hull stresses.
It is recommended that a model uncertainty with coefficient of variation from 20% to 30% be
applied to load-effects related to low-frequency response of TLPs. Again, no bias should be
included unless specific information is available.

Guideline for Offshore Structural Reliability Analysis


- Application to Tension Leg Platforms

Page No. 11

Report No. 95-3197, rev. 02

3.5 High-Frequency Loads


Mercier and Niedzwecki (1994) have made a detailed comparison of computed second order
pressure on a truncated cylinder in regular waves with the results of model tests. This second
order pressure is closely related to the excitation force for high-frequency loads in TLPs. This
comparison shows fairly good agreement in long wave periods, but consistent under-prediction in
short wave periods, for the lowermost pressure sensors, by up to -70%. There is also
considerable uncertainty in the damping of the springing response. Limited information makes it
difficult to recommend a model uncertainty for springing loads on anything better than a purely
subjective basis. A coefficient of variation of 30% is suggested accordingly for computed
springing. If the springing response is based on model tests, then lower uncertainty may be
appropriate.
Ringing is a newer topic than springing, which seems more difficult to analyse, and for which
less information is available. The model uncertainty on this effect should be even larger.
Mathisen et al. (1994a) include some suggestions on how model uncertainty can be quantified for
ringing.

3.6 Tether Response


For most typical limit states, some structural analysis is required of the response to the applied
loads, often to obtain resulting stresses which can be inserted into the limit state formulation.
The structural analysis may be linear elastic or nonlinear plastic, and quasi-static or dynamic.
Linear, dynamic analysis is commonly applied to determine the local tether response. Quasistatic analysis may be adequate in shallow water. Such analysis is quite accurate, providing the
tether natural frequencies are outside the stable vortex shedding range, and the tethers are
sufficiently widely spaced to avoid interference. Banon et al. (1994) have applied a coefficient
of variation of 10% for the model uncertainty of the tether response analysis, and this value is
also recommended here.

3.7 Platform Structural Response


Linear, quasi-static analysis is commonly applied to determine the TLP hull stresses. Mathisen
et al. (1994) have applied a coefficient of variation of 5% for the model uncertainty of the hull
response analysis, and this value is also recommended here.

3.8 Capacity Formulations


The capacity formulations applied in the following are fairly complicated, and usually based to
some extent on comparison with empirical data. The model uncertainty in these capacity
formulations varies considerably, and should be determined individually, based on comparison
with empirical results. Comparison with detailed, nonlinear finite element analyses can
sometimes also be used to quantify the uncertainty of capacity formulations. It is important to
avoid any deliberate bias in the capacity formulation. Design codes often include some
intentional bias to ensure that the resulting designs are conservative. If a capacity formulation is
based on a design code, then any bias should be quantified, and removed in the reliability
analysis.

3.9 Summary
The model uncertainties suggested or recommended in the preceding paragraphs are summarised
in Table 3.1. A normal distribution is usually a reasonable choice. However, negative values for
model uncertainty factors are usually undesirable, and a log-normal distribution is an alternative
that can avoid negative values, if necessary, for large coefficients of variation. The mean value
of the model uncertainty factor should normally be set to unity. Bias can have a considerable

Guideline for Offshore Structural Reliability Analysis


- Application to Tension Leg Platforms

Page No. 12

Report No. 95-3197, rev. 02

effect on the reliability results, and should only be introduced where adequate information is
available to quantify the bias.
Table 3.1 Summary of model uncertainties
Load or response variable
Mean environmental loads
Wave-frequency loads
Low-frequency loads
Springing loads
Ringing loads
Tether response
Hull structural response
Capacity formulation

Coefficient of variation (%)


10
10
20 to 30
30
assess individually
10
5
assess individually

Guideline for Offshore Structural Reliability Analysis


- Application to Tension Leg Platforms

Page No. 13

Report No. 95-3197, rev. 02

4. DISCUSSION OF LIMIT STATES


4.1 Component selection
Structural reliability analysis is a potentially useful basis for design or verification of the
following critical components of TLP's:
(a) Tethers
(b) Foundations
(c) Stiffened shell structures in columns and pontoons
(d) Connections between columns and pontoons
(e) Deck girders
(f) Connections between deck girders and columns
The discussion of tether limit states will be covered in relation to the examples in chapter 5,
while other components from this list will be addressed in the following sections of this chapter.

4.2 Foundations
A tension leg platform (TLP) may be tethered to driven pile anchors (with or without template),
to gravity anchors, or to suction anchors penetrated into the seabed beneath each corner of the
platform.
The loads that act on the TLP foundation consist of permanent (sustained) tension, and
environmental forces due to waves, wind, currents, drift, and tides. The TLP also has two
additional loads as a result of setdown and mispositioning. Mispositioning loads occur when the
TLP is slightly offset from the desired location. Setdown results from the lateral displacement of
the hull, which tends to lower the mean hull position.
The resulting foundation loads can be classified into static permanent loads, variable mean loads,
and oscillatory loads. Only the pretension load (cf. 2.3) from the hull buoyancy is considered a
truly permanent, sustained load. Mean loads are relatively constant over a short term period of
the order of one or a few hours, but vary slowly over longer periods and from day to day. They
include the mean wind, mean current, mean wave drift, tides, as well as most of the contribution
from the mispositioning and setdown (cf. 2.4 & 2.5). These loads are treated as quasi-static
forces in the foundation design and can be up to 30% of the total foundation load (Clukey et al.,
1995). The oscillatory loads have periods of the order of seconds to minutes. These loads are
due mostly to wave forces, although there is some contribution from mispositioning and setdown
loads (cf. 2.6, 2.7 & 2.8).
When the combination of these loads is applied to a TLP during a design storm, the tethers
develop an inclination, which results in a lateral load on the foundation. Although the connection
between the tether and the foundation is hinged, the lateral force caused by the inclination results
in an overturning moment at mudline. This moment depends on the location of the hinge above
mudline.
The pull-out capacity of the anchors under these loading conditions is crucial for the successful
platform operation.
4.2.1 Tension pile anchors
Two loading conditions should be considered in the pile design:

Guideline for Offshore Structural Reliability Analysis


- Application to Tension Leg Platforms

Page No. 14

Report No. 95-3197, rev. 02

maximum (cyclic) tension during extreme wave loading


permanent (sustained) tension
The maximum tensile load is typically much larger than the permanent tension and is more likely
to govern the design. However, some clayey soils exhibit much lower shear strength under a
sustained static load than under a transient cyclic load. So, in principle, both loading conditions
should be considered with the respective relevant soil shear strengths at a clay site.
Deterministic studies (Doyle, 1994; and from Norwegian Geotechnical Institute files) show that
for the TLPs in operation today, the foundation safety factor for the permanent tension condition
is significantly higher than that for the extreme environmental condition. Considering further
that the uncertainty in the permanent tension is much lower than the uncertainty in environmental
loads, it is reasonable to evaluate the pile anchor reliability only under the extreme wave loads.
The forces acting on a pile anchor are schematically shown on Fig. 1. The limit state function for
the pile anchor may be formulated as:

g = Qu - L

(4.1)

where Qu is the pile pull-out capacity and L is the axial load at the pile top. Considering the
annual probability of pile anchor failure, the load variable L in the limit state represents the
annual extreme axial load at the pile top. The lateral load and overturning moment on the pile
top are small (the angle q in Fig.4.1 is less than 10 under the maximum tensile load during an
extreme storm event) and may be ignored in the limit state.
The axial tensile load on the pile is resisted by skin friction and submerged weight of the pile
(including the soil plug). Thus, the pull-out capacity of the pile may be written as:
n

Qu = f i Ai

+ W'

(4.2)

i =1

where n is the number of soil layers into which the pile has penetrated, fi is the unit shaft friction
in layer i, Ai is the outer contact area of pile shaft with soil in layer i, and W is the submerged
weight of pile and soil plug. Possible resistance due to suction at the pile tip is neglected.
The ultimate unit shaft friction in clays is typically estimated by one of the following methods:
f = a su
(4.3)
or

f = b s vo

(4.4)

f = l ( s vo + 2 su )

(4.5)

or
where su is the undrained soil shear strength, svo is the effective overburden stress at the point in
question, and a, b, and l are skin friction factors.
Correction factors to account for specific effects, for example pile length or cyclic loading, can
also be applied, but one should take caution to account for all effects as some of these are
inherently included in the empirical methods that are presently in use.
In sands, ultimate unit shaft friction is evaluated by the following equation:

Guideline for Offshore Structural Reliability Analysis


- Application to Tension Leg Platforms

Page No. 15

Report No. 95-3197, rev. 02

f = K s vo tan d f lim

(4.6)

where K is the coefficient of lateral soil stress, d is the soil-pile friction angle (often assumed to
be f - 5, where f is the effective soil friction angle), and flim is the limiting skin friction. In
Eq.4.6, the differences among the many design methods are found with respect to mainly two
factors:
Value of the earth pressure coefficient, and therefore effective stress, allowed in the
calculations of axial pile capacity for compression and tension loading.
Limiting side friction value, whether it should be applied and whether it should be made
dependent on relative density.
The simplified models used in current axial pile capacity calculation methods have been derived
predominantly from onshore load tests on small piles. These empirical methods have led to a
number of parameters specific to the estimation of ultimate capacity of driven tubular piles, for
example the factors a, b, and l for estimation of shaft friction in clay. The uncertainties in these
parameters (model uncertainty) as well as the uncertainties in soil properties and effective
overburden stress are discussed in Section 7.3 of the General Guideline for Offshore Structural
Reliability Analysis (Skjong et al., 1995).
4.2.1.1 Special considerations

Axial pile capacity in compression vs tension

The API RP2T (1987) recommends a bias factor, B, with respect to the pile capacities computed
on basis of API RP2A (1987 and later) for tension pile TLP applications. The value for B is
left to the designer. According to the Commentary of API RP2T, the bias factor is to account for
the following:

Uncertainty in understanding soil-pile behaviour under tensile loading


Lack of residual strength of the soil-pile system
Load redistribution capacities of the foundation
Relative difficulty of foundation installation
Relative integrity of soil samples
Relative character and reliability of load determination
Relative lack of ability to inspect and repair TLP foundation piles
Consequences of a foundation failure to the integrity of the overall structural system
(compared to a normal jacket)

Based on engineering judgement and available test data, Doyle (1994) comes up with a B value
of 1.5 (i.e. the axial pile capacity obtained by using the API RP2A method is divided by 1.5) for
the Auger TLP in the Gulf of Mexico (soft clay site).
It is apparent from the list above that the bias factor used in deterministic tension pile design is
not appropriate for reliability assessment of the foundation. Most of the effects covered by the
bias factor B can be explicitly accounted for in structural reliability assessment of foundation
components and in overall system reliability assessment.
As far as the soil strength is concerned, there is strong evidence that the ultimate skin friction in
clay is the same for piles under compression and piles under tension. However, the ultimate skin
friction in sand may be significantly lower for a tension pile. The 20th edition of API RP2A
(1993) suggests a value of 0.7 to 1 for K in Eq.4.6 for both compression and tension piles. Many

Guideline for Offshore Structural Reliability Analysis


- Application to Tension Leg Platforms

Page No. 16

Report No. 95-3197, rev. 02

researchers and practitioners believe this value is too high for a tension pile and consider a value
of 0.5 - 0.8 more reasonable for TLP applications. The difference between compression and
tension piles is taken into account in the next edition of API RP2A-LRFD, due to be released in
1996. The new draft recommendations read:
The unit shaft friction increases with increase in soil density and pile displacement ratio.
Some data suggests that the shaft capacity in compression may be higher than that in
tension due to change in Poissons ratio, differences in the total stress field (increasing in
compression, decreasing in tension), and changes in mean effective stress due to rotation
of principal stresses (De Nicola and Randolph, 1993).
For open-ended pipe piles driven unplugged, it is usually appropriate to assume K as 0.7
to 0.8 for compression loading. For tension loading, it is appropriate to assume K as 0.5
to 0.7. For full displacement piles (plugged or closed end), value of K under
compression loading may be assumed as 0.8 to 1.0. For tension loading, value of K for
full displacement piles may be assumed to be 0.7 to 0.8. For both low (unplugged) and
full (plugged) displacement piles in compression and tension loading, the lower value is
for loose to medium dense granular material and the high value is for dense to very dense
material. For further discussion, refer to Kraft (1991).

Cyclic loading effects

The loads induced by sea waves, wind and current on a TLP subject the foundation soil to
repetitive loading-unloading-reloading. The pore pressure generated by cyclic loading may
accumulate, leading to a reduction in the shear strength. On the other hand, rapid rate of loading
tends to increase the shear strength of clayey soils. The cyclic soil strength and deformation
characteristics are problem- and definition-dependent, and should be considered on a case by
case basis. The main factors governing the cyclic soil characteristics are:

Soil type
Composition of the storm-induced loads on the foundation
Foundation dimensions and drainage boundary conditions
Ratio of cyclic load amplitude to permanent and/or static load

When the sum of the permanent (sustained) load and the storm-induced static load on the
foundation is significantly larger than the cyclic load amplitude (one-way cyclic loading), then
the cyclic effects will be minor. This is the situation for tension pile anchors supporting a TLP:
the cyclic axial load amplitude at the pile top is always less than the mean tensile force exerted
by the tethers.
There are only a few published studies on the effects of cyclic loading on axial capacity of
tension piles. The studies performed by Goulois (1982) and Karlsrud and Nadim (1990) suggest
that the cyclic axial pile capacity in clay may be 85 - 110% of the static capacity for one-way
cyclic loading of a tension pile. A cyclic capacity greater than the reference static capacity
means that the strength increase due to rapid rate of loading is more dominant than the strength
degradation due to cyclic loading.

Soil layering

The soil strength properties of distinct soil layers can generally be assumed independent random
parameters in the pile capacity computations.

Guideline for Offshore Structural Reliability Analysis


- Application to Tension Leg Platforms

Page No. 17

Report No. 95-3197, rev. 02

Within a nominally uniform layer, however, one should be careful on how the sublayer
thicknesses and properties are chosen. Engineering properties of a nominally uniform soil layer
exhibit spatial variability in both horizontal and vertical directions. For example, North Sea soils
typically have a lateral scale of fluctuation in the order of 30 - 60m and a vertical scale of
fluctuation in the order of 3 - 5m (Keaveny et al., 1989). The scale of fluctuation in the vertical
direction is an important factor in reducing the dispersion of the axial pile capacity due to spatial
averaging. A practical approximation that is often used is to choose the sublayer thicknesses to
be of the same order as the vertical scale of fluctuation, and assume the sublayer strengths to be
independent random variables.
4.2.2 Gravity and suction anchors
Suction anchors with skirts penetrated into the seabed can be used to take vertical pull-out loads
and horizontal loads from TLPs and various mooring systems. Similar foundation systems can
also be used to take vertical pull-out loads at the windward legs of jackets and jack-up platforms.
The first application of this kind of anchor for a North Sea TLP is the Snorre TLP in soft clay in
more than 300 m water depth (Christophersen et al., 1992).
Suction anchors, like piles, are generally cylindrical in shape, but have larger diameters and
shallower penetration depths. A foundation unit may also consist of several cylinders attached
together. The term suction refers to two different factors: 1) the controlled reduction of the
pressure at the caisson head during the installation process, and 2) the passive reduction of
pressure occurring during uplift (reverse end-bearing).
The pull-out capacity of a suction anchor relies on generation of suction in the soil in the skirt
compartment(s) of the anchor and in the soil beneath the anchor under the maximum tension
loads. The TLP suction anchors installed to date (Snorre and Heidrun) are designed to resist the
permanent pretension loads by dead-weight and ballast (Clukey et al., 1995). Passive suction is
relied upon to resist the cyclic and quasi-static loads under the design storm.
A gravity anchor foundation may also be used to resist the loads exerted by the tethers. The
submerged dead-weight of a gravity anchor is greater than the maximum design tension load.
Since a gravity anchor does not rely on suction in soil to resist the tensile forces, it does not need
to be equipped with skirts. However, skirts may be used on a gravity anchor to improve its
stability when the tension loads are minimum and a large net compressive load due to the
submerged anchor weight acts on the soil beneath.
The most essential aspect of the foundation design of an anchor is to ensure that the foundation
soil has sufficient capacity to carry the large static and cyclic pull-out loads without excessive
displacements. A foundation failure may occur as large vertical, horizontal and rotational
displacements. The failure mode may be large cyclic displacements, large cyclically-induced
permanent displacements, or a combination of large permanent and cyclic displacements. The
limit state function for foundation capacity can be written in a format similar to that for a pile
anchor (Eq.4.1):
g = R-L
(4.7)
where L is the maximum tension load on the anchor and R is the foundation capacity.
For a gravity anchor (without skirts), R is simply the submerged weight of the anchor and L is
the vertical component of load. The procedure for estimating the foundation capacity for a
suction anchor is, however, more complicated. It is similar to the procedure for calculation of

Guideline for Offshore Structural Reliability Analysis


- Application to Tension Leg Platforms

Page No. 18

Report No. 95-3197, rev. 02

the foundation bearing capacity for a gravity platform under combined static and cyclic loads.
The most commonly used analysis method was proposed by Andersen and Lauritzsen (1988).
The principles and procedures used to calculate the bearing capacity of gravity platforms can
also be used to calculate the pull-out capacity for suction anchors. The only modification needed
to adapt the procedure to calculate the pull-out capacity for anchors is to use extension shear
strengths instead of compression shear strengths in the soil beneath the structure, and compression shear strengths instead of extension shear strengths in the soil outside the anchor
(Fig.4.2).
The pull-out capacity is calculated with limiting equilibrium analyses. Various types of failure
surfaces should be analysed and a search must be made to find the location of the surface that
gives the lowest bearing capacity. The critical type of failure surface will depend on several
factors such as the load inclination, the ratio between static and cyclic loads, the shear strength
profile, and the embedment depth to diameter ratio of the anchor.
The limiting equilibrium analyses are performed with the "cyclic shear strength" of the soil. The
cyclic shear strength can be determined from cyclic triaxial and direct simple shear (DSS)
laboratory tests consolidated to the in-situ vertical and horizontal effective stresses. The
laboratory tests are run to failure under conditions that simulate the stress conditions along the
potential failure surface as closely as possible (Fig.4.2). Details of the procedure to calculate
pull-out capacity are provided by Andersen et al. (1994).
To verify the procedure for calculating pull-out capacity of anchors, the Norwegian Geotechnical
Institute performed a series of one static and three cyclic field model tests for Saga Petroleum
when they decided to use concrete anchors for the TLP at the Snorre field (Dyvik et al., 1993).
Class A prediction (i.e. prediction made before the tests were carried out) of model anchor
capacities proved to be very accurate (Andersen et al., 1993). The predicted pull-out capacities
are summarised and compared to the measured pull-out capacities in Table 4.1. The comparison
shows that the predicted capacities agree very well with the measured capacities in the static test
and in all the three cyclic tests, independent of variations in cyclic load history, geometry and
load eccentricity.

Table 4.1

Comparison between predicted and measured pull-out capacities of suction


anchors in clay (Andersen et al., 1993)

Model

Test

Pull-out capacity (kN)

test No.

type

Predicted

Measured

Predicted/Measured

1
2
3
4

Static
Cyclic
Cyclic
Cyclic

138
118
105
92

137.7
112.9
99.5
90.5

1.00
1.05
1.06
1.02

The predicted critical failure surfaces for the static test (Model Test 1) and two of the cyclic tests
(Model Tests 2 and 3) were almost identical (Fig.4.3a). These tests had the same load eccentricity. Cyclic Model Test 4, with a greater load eccentricity than the other model tests, had a
different calculated critical failure surface, as shown in Fig.4.3b.

Guideline for Offshore Structural Reliability Analysis


- Application to Tension Leg Platforms

Page No. 19

Report No. 95-3197, rev. 02

These results show that the modelling uncertainty for the procedure outlined by Andersen et al.
(1992) for pull-out capacity of suction anchors in clay is relatively small. The key is an accurate
determination of cyclic soil strength under the applied loads.
4.2.3 Cyclic shear strength of soil
The cyclic shear strength of a soil element is defined as the sum of the average and cyclic shear
stresses that cause excessive permanent and/or cyclic shear strains after a given number of load
cycles or for a given cyclic load history:

f ,cy

= (t a + t cy ) f

(4.8)

where ta is the average shear stress, tcy is the cyclic shear stress amplitude (see Fig.4.2), and the
subscript f denotes failure.
The cyclic shear strength is not a unique soil property and depends on the loading history. Thus
the uncertainty in the cyclic shear strength is related to both the uncertainty in the soil
characteristics and the uncertainty in the load history during the design storm.
Keaveny et al. (1993) performed probabilistic stability analysis for a gravity suction anchor (gravity
anchor with skirts) for a TLP conceptual design at a stiff clay site in the North Sea. They used the
technique developed at NGI to estimate the uncertainty in the cyclic shear strength of the soil.
The uncertainty was estimated by Monte-Carlo simulation with Latin hypercube sampling of the
random variables that influence the cyclic strength. The study by Keaveny et al. showed that the
pull-out of the anchor was not the most critical mode of foundation failure because the tension
loads were resisted by the submerged weight of the anchor. The critical mode was foundation
sliding combined with partial lift-off of the anchor similar to that shown in Fig.4.2 and Fig.4.3a.
Using a second-order reliability method (SORM), the conditional reliability index given the 100year design storm was computed to b = 4.9 (Pf = 4.610-7). The low probability of nonperformance was consistent with the high material coefficients used in the deterministic design
of the anchors and pointed to conservatism in the anchor design for that TLP.

4.3 Columns and Pontoons


TLP columns and pontoons usually have a cylindrical or rectangular cross-section, and are
constructed as orthogonally stiffened shells. Local structure of these types needs to be designed
against buckling failure under compressive loads, and against yielding under tensile loads.
Buckling criteria are usually semi-empirical, calibrated against experimental data, and subject to
considerable uncertainty. A combination of stochastic load processes due to external pressure
and the force distribution in the platform structure is also required. These factors indicate that
structural reliability analysis is appropriate for the detailed analysis of such failure modes.
Such an analysis has been carried out as a joint industry project, with the objective of calibrating
simplified design equations for this type of structure, and reported by Mathisen et al. (1994b).
Three buckling failure modes were considered:
(i) Bay failure of orthogonally stiffened cylindrical shells (OSCS) - in which buckling of the
stringers together with the attached shell plate occurs, while the ring stiffeners remain stable,
(ii) Plate failure of orthogonally stiffened cylindrical shells - in which local buckling of the shell
plating occurs while both longitudinal and ring stiffeners remain stable,
(iii) Plate failure of stiffened flat plates - in which the panel itself buckles, while the stiffeners
remain stable.
A brief description of the limit states involved in these analyses will be given in the following.
The specific buckling criteria considered here may be taken as examples. The reliability analysis

Guideline for Offshore Structural Reliability Analysis


- Application to Tension Leg Platforms

Page No. 20

Report No. 95-3197, rev. 02

developed for these limit states can be adapted to buckling criteria for other failure modes, if
these are considered more relevant for a particular TLP.
4.3.1 OSCS bay failure
A buckling criterion referred to as the RCC (Rule Case Committee) criterion, and described by
Das et al. (1991) is applied. A very similar criterion is given in API Bulletin 2U (1987). The
corresponding limit state function may be written
2
2
r r

r r r
s1
k ( q , r ) s1 s2
s2
g(q , r , s ) = 1 -
(4.9)
r r
r r -
r r
r r +
rC 3 rC1 c1( q , r ) rC 2 c2 ( q , r ) rC 2 c2 ( q , r )
rC1 c1 ( q , r )
r
where vector q contains deterministic parameters describing the design of the particular
r
cylindrical shell, vector r contains realisations of time-independent stochastic variables used to
r
evaluate the capacity (model uncertainties, and material yield stress), and vector s contains

realisations of time-dependent, stochastic, stress processes acting on the structural component,


with s1 as the hoop stress, and s2 as the normal stress. c1 is the structural capacity for hoop
stress alone, c2 is the structural capacity for the normal stress alone, and k is an interaction
coefficient. rC1 , rC 2 , and rC 3 are capacity model uncertainty factors for hoop stress capacity,
normal stress capacity, and interaction term, respectively. Details of the capacity and interaction
terms are omitted here, but may be found in Das et al. (1991).
The limit state function in equation (4.9) requires a nonlinear combination of two stochastic
processes; viz. hoop stress s1 and normal stress s2 . The difficulties posed by this combination
may be overcome by application of the vector outcrossing algorithm, as described in the General
Guideline, section 3.6.2.2. To apply the outcrossing algorithm, it is necessary to include the
distributions of the time derivatives of the stochastic processes that are involved in the limit state
function.
Modelling of the time-independent stochastic variables with their respective distribution
functions is quite straightforward. Note that the model uncertainty in the hoop stress capacity
and in the normal stress capacity is quite high, with coefficients of variation of about 20%. The
software implementation of the capacity terms needs some care, to ensure that these terms are
continuous, and have continuous derivatives. These precautions help the reliability computations
to converge. Most of the work in the analysis involves the modelling of the load processes. The
basic load modelling is made conditional on the realisation of a short term stationary
environmental state, as for the tether ULS in the example report. Wave-frequency load-effects
are obtained via transfer functions, based on hydrodynamic analysis of the platform as a rigid
body, and FEM structural analysis of the platform response to the hydrodynamic loads. The FEM
analysis provides cross-sectional forces and bending moments. The normal stress at a particular
location is computed from the cross-sectional axial force and bending moments. The local hoop
stress is computed from the transfer function for external pressure and the cross-sectional
properties. The transfer functions for the hoop stress and normal stress are combined with the
wave spectrum to provide the covariance matrix for the stresses and their time derivatives. The
mean sectional forces are computed for three platform draughts, and fitted to an interpolation
function. A realisation of the low-frequency offset motion is used to compute the resulting
platform setdown, and instantaneous draught. The interpolation function is then applied to
obtain the corresponding sectional forces and local stresses due to mean and low-frequency
loads. The stress components and their time derivatives are modelled as a vector Gaussian
process, using these mean values and this covariance matrix. An overview of the computation is
given in Fig.4.4.

Guideline for Offshore Structural Reliability Analysis


- Application to Tension Leg Platforms

Page No. 21

Report No. 95-3197, rev. 02

On this basis, the outcrossing rate through the limit state function can be calculated for any
environmental state. The expectation of the outcrossing rate is taken with respect to the
distribution of the environmental variables (significant wave height, peak wave period, wind
speed, current speed, heading angles, etc.). The annual probability of failure conditional on the
r
time-independent variables r can then be computed. Finally, the marginal probability of failure
is obtained by integration with respect to the time-independent variables. This is the same type
of probabilistic formulation applied for the tether ULS in the TLP example report.
This reliability analysis was applied with two sets of environmental conditions, for the
Norwegian continental shelf, and for the Gulf of Mexico. The environmental model for the Gulf
of Mexico was only based on hurricane conditions, while the model for the Norwegian
continental shelf was based on all environmental conditions. Thus, for the Gulf of Mexico, the
probability of one or more hurricanes per year was included in the probabilistic formulation.
4.3.2 OSCS plate failure
The plate failure criterion for orthogonally stiffened cylindrical shells is based on API Bulletin
2U (1987), and the limit state equation may be written as in equation (4.9), but with different
formulae applied to compute the capacity terms c1 , c2 , the interaction term k, and the capacity
model uncertainty factors rC1 , rC 2 , rC 3 . The reliability analysis of this limit state is naturally
formulated in the same way as for bay failure.
4.3.3 Stiffened flat plates
A criterion developed by Chapman et al. (1991) was applied to buckling of stiffened flat plates.
The limit state function for this failure criterion may be written as

r r r
r r
2
2
2
2
g( q , r , s ) = rC s XY c 2 ( q , r ) - s X - k XY s X sY - s y

(4.10)

where the notation is similar to equation (4.9), with s X as the normal stress applied to the
shorter edge of the plate, sY is a function of the normal stress applied to the longer edge, s XY is
r r
a function of the applied shear stress, c( q , r ) is a function of the uniaxial stress capacity and the
applied pressure, and k XY is an interaction coefficient. Further details of the terms in this limit
state equation may be found in Chapman et al. (1991).
As with the previous two limit state functions, this one is also a nonlinear function of stochastic
processes, but now the number of stochastic load-effect processes has increased to four; viz.
stress on shorter side, stress on longer side, shear stress and pressure, which can be taken as the
r
components of vector s in this case. Again the vector outcrossing algorithm can handle the
combination of these effects in the limit state function. The computation of the load-effects, and
the probabilistic formulation is similar to the OSCS bay failure limit state described above.
4.3.4 Yielding
A useful criterion against yielding can be formulated in terms of the Von Mises equivalent stress,
with limit state function
2

g = sF - s X - sY + s X sY - 3 s XY

(4.11)

where sF is the yield stress of the material, s X , sY are normal stresses, and s XY is the shear
stress applied in the shell plating. This limit state is again a nonlinear combination of stochastic

Guideline for Offshore Structural Reliability Analysis


- Application to Tension Leg Platforms

Page No. 22

Report No. 95-3197, rev. 02

processes, which presents some difficulty and needs a vector outcrossing formulation. However,
the limit state surface can be viewed as an ellipsoid, and the outcrossing rate can be calculated
directly by numerical integration over this surface, as described by Madsen (1985). The vector
outcrossing algorithm which can compute the outcrossing rate using standard asymptotic
approximations from structural reliability methods is much more efficient than numerical
integration.
The computation of the local stresses and the probabilistic formulation for this limit state follows
the same procedure, as outlined for OSCS bay failure.

4.4 Main Structural Connections


Main structural connections in a TLP include:
connections between deck girders and the vertical columns, and
connections between the vertical columns and the pontoons.
The local structure in these connections may be quite complex, but it is amenable to FEM
analysis, which may be used to identify stress hot spots, that are potentially subject to fatigue
damage. The fatigue limit states for such hot spots correspond to those described in the Jacket
Application Guideline. The stress concentration factors and the load model will differ.
The dominant oscillatory loads are due to first order wave loads. These can be obtained through
conventional, hydrodynamic, diffraction analysis of the platform as a rigid body, combined with
FEM structural analysis of the elastic hull response. In addition, care should be taken to evaluate
any additional oscillatory loads due to springing or transverse vibration of the tethers. These
loads will certainly be present close to attachment points to the tethers, but may well be
negligible at some distance from these connections, dependent on how these loads are taken up
by the structure. Refer to section 5.3.2.4 for additional discussion of the oscillatory loads that act
on the tethers.

4.5 Deck Girders


The deck girders of a TLP also need to be designed against buckling and yielding. The same
limit states as discussed in sections 4.3.3 and 4.3.4 can be applied for this purpose. The main
difference is the absence of external water pressure. Other buckling criteria may also need to be
considered. The local effect of wind loads has to included in structure adjacent to the derrick
and flare boom, or other modules that are exposed to significant wind loads.

4.6 Air Gap


To avoid loads due to wave impacts against the underside of the TLP deck, it is usual to design
the air gap between the deck and the stillwater level with a very low probability of exceedance.
Although this is not a structural reliability problem, the methods of structural reliability are
useful to evaluate the combination of effects that can lead to wave impact under the deck. Banon
et al. (1994) have included the following limit state for air gap

g = zlevel - zcrest - z surge - ztide - z setdown

(4.12)

where zlevel is the stillwater air gap that has to be exceeded before a wave impact occurs, and the
other terms include the surface elevations due to the wave crest, storm surge, astronomical tide,
and platform setdown as a result of low-frequency offset. Various levels of sophistication can be
applied in the computation of the distributions of the surface elevation terms, and should be
reflected in the model uncertainties that have to be associated with each term. In fact, all the
surface elevation terms are stochastic processes, so there is some difficulty about the
combination of these terms, in general. In practice, the wave crest elevation is the dominant

Guideline for Offshore Structural Reliability Analysis


- Application to Tension Leg Platforms

Page No. 23

Report No. 95-3197, rev. 02

term, and most effort should be expended on the modelling of the annual extreme value
distribution for the wave crest elevation.

Guideline for Offshore Structural Reliability Analysis


- Application to Tension Leg Platforms

Page No. 24

Report No. 95-3197, rev. 02

5. SUMMARY OF APPLICATION EXAMPLES


Details of these examples are given in a separate report by Mathisen, Steinkjer, and Lotsberg
(1995), which will be referred to as the example report in this chapter. The limit states are
discussed here, and a summary of the results is included.

5.1 Principal Dimensions of Example Structure


A generic TLP has been defined with the main particulars given in Table 5.1. The platform can
be regarded as typical for North Sea conditions in medium deep waters (325m.).
Table 5.1 TLP Main Particulars
Displacement

200
000
tonnes
Total tether pretension
25 000 tonnes
Water depth
325 metres
Number of tethers
16
Vertical distance from MWL to top attachment point of 55.00 metres
tether
Centreline distance between columns
78.00 metres
Column outer diameter
28.50 metres
Draft
60.00 metres
Pontoon breadth
14.25 metres
Pontoon height
13.00 metres
Vertical centre of gravity, below MWL
10.00 metres
Vertical centre of buoyancy, below MWL
35.00 metres
Yaw radius of gyration
53.3 metres
The global coordinate system of the TLP with weather directions are given in Fig.5.1.

5.2 Summary of Tether ULS example


TLP tethers are specialised, highly-stressed components, of critical importance for the platform
operation and safety. Tether design involves both the combination of a number of stochastic
load components, and evaluation of a capacity which may usefully be treated as stochastic. This
makes tethers obvious candidates for structural reliability analysis. In the present example, an
ultimate limit state is considered which combines the effects of external pressure, axial tension,
and bending moment in a criterion for combined yielding and buckling of unstiffened cylinders.
Initial failure of a single tether is analysed. A brief survey of some previous related studies is
discussed below, before proceeding with the present example.
Stahl and Geyer (1985) developed a systems reliability analysis for TLP tether systems. Their
paper emphasises the modelling of system effects, based on Stahl and Geyer (1984), and does not
include much detail on the load or resistance variables. Only tension loads are considered in
their analysis, implying that all tethers at one corner of the TLP experience the same loads. On
this basis, a considerable systems effect is found; i.e. the reliability of a tether (or all the tethers
in one corner) is reduced because the same load is applied to all elements, and necessarily finds
the element with the lowest resistance. The limit state considered in the present report leads to

Guideline for Offshore Structural Reliability Analysis


- Application to Tension Leg Platforms

Page No. 25

Report No. 95-3197, rev. 02

the highest loads at the bottom of each tether, implying that the weakest element in a tether is not
necessarily subjected to the highest load, which may tend to reduce the systems effect.
Banon and Harding (1989) carried out a tether reliability analysis considering two limit states for
maximum and for minimum tether tension. Their paper addresses the combination of several
tension loads, and uses exceedance of the yield stress as the failure criterion. The load
components included correspond quite closely to those included in the present report. An
important difference is that the present report takes account of the effect of offset on the wavefrequency loads (cf. section 5.2.3), and dynamic analysis of the tethers is also applied. The point
crossing method applied by Banon and Harding is closely related to the vector out-crossing
technique applied here.
Harding and Banon (1989) and Banon et al. (1991) extended this analysis to include the effect of
underdeck wave impacts on the tether reliability. If the elevation of the deck above the stillwater
surface is inadequate, then underdeck wave impacts can occur, and induce significant loads in
the tethers, which will naturally affect the reliability considerably. It seems advisable to design a
reasonably generous clearance to ensure a very low probability of such wave impacts. If impacts
are to be tolerated, then more accurate analysis of the resulting underdeck loads is desirable.
Banon et al. (1994) also applied this reliability analysis, without deck impact, in a calibration of
a simplified design equation for the same tether limit state as studied in the present report; i.e.
now including the combined effect of axial tension, bending and external water pressure. The
vector out-crossing method was applied instead of the point crossing method.
Rooney et al. (1989) carried out a tether reliability analysis, using the load model described by
Marthinsen (1989), and including combination of axial stress due to tension and bending with
hoop stress due to external pressure in the limit state function. Turkstras rule was applied in the
load modelling, and the wave frequency stress was modelled with its extreme value distribution,
as the dominant stochastic process.
Lotsberg (1991) carried out a tether reliability analysis which also included the effect of
foundation-mispositioning on the tether tension, based on results obtained by Madsen et al.
(1987). A limit state for the combined effect of tension, bending and external pressure, taken
from the DNV Classification Note on Buckling Strength Analysis was applied, and Turkstras
rule was used in the modelling of the load effects. Both Rooney et al. and Lotsberg found the
randomness in the yield stress most important for the reliability result, followed by the
randomness in the wave-frequency load-effect, together with its model uncertainty. For other
offshore structures, the randomness in the wave-induced load tends to be relatively more
important. The difference in behaviour may well arise because of the relatively larger amount of
stillwater stresses present in a tether, due to pretension and external hydrostatic pressure. The
importance of the wave-frequency component in these analyses provides some justification for
the more detailed modelling of this component in the present analysis, including dynamic
response and the effect of offset on the wave-frequency response.
5.2.1 Failure criterion
A failure criterion has been developed by Moan et al. (1992) which is well-suited to analysis of
the tether body; viz. a cylindrical section subject to external pressure, axial tension, and bending
moments. The basic limit state function may be written as
n

s
s
p
g =1- a + b (5.1)
pc
s ac s bc
where s a is the applied axial stress, s ac is the axial stress capacity, which is set equal to the
yield stress, s b is the applied bending stress, s bc is the bending stress capacity, p is the
applied external pressure, pc is the buckling pressure, and n is an exponent, given by

Guideline for Offshore Structural Reliability Analysis


- Application to Tension Leg Platforms

Page No. 26

Report No. 95-3197, rev. 02

sa

n = na

sa

s ac

s b

s ac + s bc

1.7

+ nb

s b

s bc
sa

1.7

s b

s ac + s bc

1.7

(5.2)

na = 2 + 0.023 D t

(5.3)

nb = 1 + 300 t D

(5.4)

with as the diameter of the tether, and t as the wall thickness. The formulation is valid for
D / t in the range from 15 to 40. It is only intended for positive values of the three load-effects
involved. Further details of the underlying capacity variables are included in the example report.
The model uncertainty for this failure criterion has also been developed, based on comparisons
with experimental and nonlinear finite element results.
5.2.2 Analysis of input required by limit state function
The required material parameters are: yield stress, Youngs modulus, and Poissons ratio. The
ovality of the cylindrical section may also be included explicitly. However, the tethers are
initially assumed to be manufactured with high accuracy, and conform to the basis value of
0.15% ovality. The dimensions of the tether section are also required; viz. diameter and wall
thickness. Again, based on high accuracy manufacture, it should be appropriate to treat these
dimensions as deterministic values.
The required load-effects are external pressure, axial stress, and bending stress. The external
pressure is simply the hydrostatic pressure, including water level variation. Pressure variations
due to waves are usually negligible in the lower part of the tether, which is the most critical area
for a tether with constant cross-section. Hence the external pressure is treated as a timeindependent stochastic variable under short term, stationary environmental conditions. The axial
and bending stresses are time-dependent stochastic processes, involving several load-effects.
The failure criterion is formulated as a nonlinear combination of these load-effects. This leads to
some difficulty in determination of the extreme value of the combined load effect. This
difficulty is overcome by formulating the problem in terms of an out-crossing analysis, cf. Hagen
and Tvedt (1991). To comply with the requirements of this formulation, the time derivatives of
the stochastic processes involved in the failure criterion are also needed.
If the bending stress is a minor component, as may often be the case, then there is a possibility to
simplify the analysis. A combined normal stress ratio, can be obtained by linear combination of
the axial and bending stresses. The dependency of the exponent n on the bending stress can be
neglected in this case, and simplified to the value na . The extreme value distribution of the
normal stress ratio is obtainable for both the low-frequency and wave-frequency load-effects
separately, but not combined. However, combinations of these two load-effects can be based on
Turkstras principle.
The local stresses in the tether can be split into low-frequency (LF), wave-frequency (WF), and
high-frequency components (HF). High-frequency components are neglected in the present
analysis. High-frequency springing is primarily important for fatigue analysis, and expected to
be a minor component in the ultimate limit state. High-frequency ringing is neglected because
adequate response models that can easily be incorporated in the present example are not
available. Ringing should not be neglected in general. The low-frequency stress components
can conveniently be modelled as deterministic functions of the platform offset motion. Hence
the offset motion is an important element in this limit state. The wave-frequency stress

Guideline for Offshore Structural Reliability Analysis


- Application to Tension Leg Platforms

Page No. 27

Report No. 95-3197, rev. 02

components are computed as dynamic tether response to the wave-frequency platform motions.
The wave-frequency response of both platform and tethers is modified by the low-frequency
offset of the platform, and the analysis is formulated so that this effect can be included.
5.2.3 Probabilistic formulation of the reliability problem
Let us collect all time-independent stochastic variables related to the problem in the stochastic

vector R . The components of this vector will normally include all uncertainty related to the
capacity, and may also include model uncertainty related to the load-effects. Secondly, let us
collect all the stochastic variables required to describe a short-term, stationary environmental

state in the stochastic vector Y . These components will include the significant wave height, the
peak wave period, the mean wind speed, the current speed, the surface elevation due to tide and
storm surge, and the heading angle of the environmental effects relative to the platform.
Hydrodynamic analysis procedures are available to determine the short-term distribution

r r
FK Yr , Rr (k y , r )

of the low-frequency TLP offset motion K , conditional on realisations of the

r
r
Y
R
, and the time-independent vector . Similarly, hydrodynamic analysis
environmental state
r r
FS K ,Yr , Rr (s a k ,y , r )

procedures are also available to determine the short term distributions

r r
FS K ,Yr , Rr (s b k , y , r )
b

and

of the axial stress S a and bending stress S b acting at a chosen location in

a tether, conditional on the low-frequency offset K , the environmental state Y , and the time-

independent vector R . Fairly simple extensions of these results will also provide the short-term
distributions of the time derivatives of these stochastic processes; i.e. the low-frequency offset

& , the wave-frequency axial stress derivative S& a , and bending moment derivative S& b .
velocity K
This is the information required to apply the vector out-crossing algorithm to the limit state
function in equation (5.1), and obtain the short-term out-crossing rate

r r
n g (y , r )

, conditional on

r
r
Y
R
, and the time-independent vector . For completeness, information
the environmental state
on any stochastic dependency between these stochastic processes is also required. The expected
value of the outcrossing rate with respect to the environmental parameters must be computed by
integration before the expectation is taken with respect to the time independent parameters (c.f.
General Guideline section 3.6.2.2)

r r
r r
r
n g ( r ) = n g (y , r ) f Yr (y ) dy

(5.5)

Having obtained the out-crossing rate, the conditional first-passage probability of failure is
approximated by

Pf

r
R

(r ) = 1 - exp{-n g (r )l}

(5.6)

where l is the time duration considered, usually one year. The marginal probability of failure is
then obtained by the theorem of total probability

where

r
f Rr ( r )

Pf = Pf

r
R

(r )

r r
f Rr ( r ) dr

is the probability density of the time-independent variables.

(5.7)

Guideline for Offshore Structural Reliability Analysis


- Application to Tension Leg Platforms

Page No. 28

Report No. 95-3197, rev. 02

5.2.4 System response calculations


System response calculations are required to determine the short term distributions of the
platform offset motion, the axial stress and bending stress at the specific cross-section of the
tether that has been selected for analysis.
Fig.5.2 provides an overview of the offset calculation. The low-frequency platform motions in a
short term, stationary environmental state are excited by wind gust forces and second-order, lowfrequency wave forces. The low-frequency motion responses to these excitation forces have been
computed with the SWIM and MOTION programs (MIT, 1995), based on second-order
diffraction theory for an array of cylinders, by Emmerhoff and Sclavounos (1992). Wind load
excitation is also included. Surge, sway and yaw motions are computed in the time domain. The
largest mean LF yaw motion in any of the environmental states considered was less than 0.1
degree, and the largest LF yaw standard deviation was less than 0.4 degrees. The WF yaw
motion was even smaller. Hence the effect of yaw motion on the tether stresses was neglected.
The surge and sway motions were subsequently combined into a radial offset motion. Short term
statistics of the radial motion and its time derivative were computed from the time histories with
a small program written for this purpose (EXSWIM). These statistics comprise the mean value,
standard deviation, skewness and kurtosis. The distribution of the low-frequency motion is
based on these statistics. A normal distribution has been applied. The skewness was found to be
negligible. The kurtosis for the LF offset was found to be small in the mild environmental states,
and increased with significant wave height, to close to 2 in the most severe states. This indicates
considerable deviation from the normal distribution, which has zero kurtosis. It was intended to
use this information with an alternative distribution model for the offset motion, but this
alternative distribution was omitted due to lack of time.
Fig.5.3 provides an overview of the tether response calculation. The mean stresses and transfer
functions for dynamic tether response in regular waves were computed with the FRECOM
program (MCS, 1989). This provides a linear finite element analysis of the tether. The main
excitation is applied at the upper end, due to the corresponding transfer function for the TLP
motion at this point. Local hydrodynamic forces on the tether are computed by means of
Morisons equation. Correct phasing of the top point motions relative to the incoming wave
kinematics is obtained through the transfer function for the top point motion. The transfer
functions for axial and bending stresses at the lower end of the tether were combined with a
Jonswap wave spectrum, using the POSTRESP program (DNV SESAM, 1993), to provide the
covariance matrix for these two stresses and their time derivatives. These mean values and
covariance matrices for the stresses were computed for a number of short term, stationary,
environmental states, and mean offset positions of the tether top point. The mean offset of the
tether top point in the tether dynamic analysis will subsequently be taken to correspond to an
instantaneous value of the low-frequency platform motion, because the tether response to the
low-frequency motion can be considered to be quasi-static. Hence, more than one offset position
must be considered for the tether transfer functions, and also for the tether response in each
environmental state, not just the mean platform position in that state. The wave-frequency
stresses in a short term, stationary, environmental state, conditional on the offset, were modelled
as independent Gaussian processes, because the correlations were found to be negligible. If
more correlation had been found between the axial and bending stresses, then they and their time
derivatives could have been modelled as a Gaussian vector process.
The transfer functions for motions of the TLP platform as a rigid body in regular waves may be
computed with a diffraction program, such as WADAM (DNV SESAM, 1987). In principle,
these transfer functions should be computed as a function of the offset of the platform from the
mean stillwater position. The offset introduces coupling terms into the restoring matrix, which

Guideline for Offshore Structural Reliability Analysis


- Application to Tension Leg Platforms

Page No. 29

Report No. 95-3197, rev. 02

are essential to obtain the proper phase between surge and heave motions, with major
consequences for the resulting tether stresses. Unfortunately, transfer functions for platform
motions were only available at the mean stillwater position for this example, and resources were
not available to compute them as a function of offset. This led to some difficulties that are
described in the example report, and subsequently to a rough approximation for the effect of
offset on the tether response. The transfer functions for platform motions are normally computed
with respect to an origin in the stillwater plane, on a vertical through the centre of gravity. A
coordinate transformation is applied to obtain the transfer functions for motions at the tether top
connection to the platform, prior to use as input to the tether response computation.
5.2.5 Stochastic variables
Distributions for model uncertainties and some structural parameters are defined in Table 5.2.
The model uncertainty for the structural capacity variables has been established by Moan et al.
(1992). The model uncertainties for the response to environmental effects correspond closely to
the recommendations in Table 3.1. In the present example, the model uncertainty in the wavefrequency response also includes the uncertainty in the tether response.

Table 5.2 Time-independent stochastic variables. (Some of the deterministic


variables here could also quite well be treated as stochastic variables.)
Variable
Model unc.
collapse pressure
Model unc. axial collapse
stress
Model
unc.
bending
collapse stress
Model unc. mean offset
Model unc. LF offset
Model unc. WF response
Tether ovality
Young's modulus
Yield stress

Distribution
Normal

Parameters
Mean=0.95, CoV=0.13

Normal

Mean=1.20, CoV=0.08

Normal

Mean=1.00, CoV=0.07

Normal
Normal
Normal
Deterministic
Deterministic
Deterministic

Mean=1.0, CoV=0.10
Mean=1.0, CoV=0.20
Mean=1.0, CoV=0.15
0.25%
2.1e+8 kN/m2
3.58e+5 kN/m2

Distributions for the environmental variables are defined in Table 5.3, based on Bitner-Gregersen
and Havers (1989) model for Haltenbanken, on the Norwegian continental shelf.

Table 5.3 Environmental state variables.


Variable
Significant wave height
Peak wave period
10 min. wind speed
Current speed
Storm surge
Tide
Heading

Distribution
Weibull
Log-normal
Weibull
Weibull
Normal
Uniform
Uniform

Parameters
alpha=2.154, beta=1.273, gamma=0.763
Computed conditional on Hs
Computed conditional on Hs
Computed conditional on Hs
Computed conditional on Hs
(-0.75, 0.75)
(0, 45)

Guideline for Offshore Structural Reliability Analysis


- Application to Tension Leg Platforms

Page No. 30

Report No. 95-3197, rev. 02

Short term distributions for the stochastic processes involved in the system response are defined
in Table 5.4.

Table 5.4 Stochastic processes.


Variable
LF offset
LF offset velocity
WF axial force
WF bending moment
WF axial force velocity
WF bend.mom. velocity

Distribution
Normal
Normal
Normal
Normal
Normal
Normal

Parameters
Response surface
Response surface
Response surface
Response surface
Response surface
Response surface

5.2.6 Coupling of system response calculations with reliability calculations


An overview of the data flow in the reliability calculation is shown in Fig.5.4. This figure shows
that the system response computations (SWIM and FRECOM) are linked to the reliability
computations (PROBAN) by means of two response surfaces for offset and tether stresses. In the
reliability computations, an iteration is performed through the domain of the random variables, to
find the conditions most likely to lead to failure, known as the design point. During this iteration
process, the system response is required at each step; i.e. it should be available for any
environmental state. When the system response is easily computed, then the system response
computations can be linked directly to the reliability calculation, and carried out on demand, at
each iteration step. Such a direct link is impractical for complex system response, as in the
present example. In this case, the system response is computed for a series of environmental
states in advance, and interpolation between these results is applied during the reliability
analysis. Each system response variable (e.g. the mean offset) is a function of a vector of input
r
variables (e.g. the environmental state vector y ) and may be viewed as a surface in a
hyperspace. A response surface module developed by Mathisen (1993) is applied in the present
example. It is important to check that the response surface does not introduce any significant
inaccuracy into the computations, and this has been done in the example report.
5.2.7 Initial reliability results
The computed annual probability of failure is 1.5 10-4, with reliability index 3.61. This is fairly
close to a target index of 3.71, suggested in Table 2.7 of the general guideline, for serious
failures with reserve capacity. It indicates that the tethers of the generic TLP are almost
adequately dimensioned. However, it should be recalled that ringing effects have been neglected
in the present analysis, and these effects may well be significant in practice.
Some of the design point information from the reliability analysis is reproduced in Table 5.5.
The design point provides the values of the stochastic variables that the reliability analysis has
found most likely to lead to the specified event; i.e. tether failure due to overload. In general, the
design point should be examined carefully, both to check that the results seem reasonable, and
that the mathematical operations specified in the input file to PROBAN have been correctly
implemented. With the out-crossing analysis, information is now provided for two linearisation
points. The outer layer, for the time-independent variables comes first, and is followed by the
layer for the stochastic processes.

Guideline for Offshore Structural Reliability Analysis


- Application to Tension Leg Platforms

Page No. 31

Report No. 95-3197, rev. 02

In the outer layer, the dominant effect is provided by the low model uncertainty factor for the
axial stress capacity =1.02, compared to a mean value of 1.20. The model uncertainty for the
pressure capacity =0.85 is also low compared to a mean of 0.95. The model uncertainty for
wave-frequency load-effects =1.15 and is significant, while the model uncertainties for lowfrequency load-effects are relatively unimportant. The probabilities of the significant variables
are furthest from the median probability =0.5, with the axial stress capacity in the low end, and
the WF response model uncertainty in the high end. The failure event is attained by driving the
axial stress up, and these variables reflect that.
The inner layer applies to an auxiliary limit state function used with the vector out-crossing
algorithm, which includes both the basic failure criterion from equation (5.1), and the time
derivative of the failure criterion. This seems to have a slight effect on the design point, which is
not well understood yet. The most salient feature of the design point is a large significant wave
height of 20 m, which is needed to drive the loads up sufficiently to fail the tether. The other
environmental variables are within reasonable proximity of their mean values, conditional on this
large wave height. The TLP offset at 26.0 m is about one standard deviation above it's mean
value in the design point environmental state, while the WF axial force is right out in the tail of
its conditional distribution function for the environmental state. The components of the limit
state vector include the utilisation factors for the 3 load-effects, and show clearly that the design
point is dominated by the axial stress, with a utilisation of 0.82, with some contribution from the
pressure at a utilisation of 0.38, while the bending moment is insignificant. The table also shows
the percentage of the tether axial force contributed by the various components. The pretension is
just under half the total, and the WF tension is the dominant variable component.
Importance factors are often included to rank the significance of the various stochastic variables
in the computed reliability, and indicate how the reliability result would be affected if a
stochastic variable should be replaced by a corresponding deterministic value. The importance
factors are not so useful when many of the stochastic variables are not independent. This is the
case in the present analysis, where many of the various load effects are dependent on the same
environmental conditions. Hence, the importance factors are omitted here, but a fairly good
impression of the ranking of the stochastic variables is given by the deviation of the probability
for each variable from the median. The probabilities in Table 5.5 are related to the probability
distribution specified for each variable, and should be interpreted as conditional probabilities,
when the distribution parameters are computed from other stochastic variables.

Guideline for Offshore Structural Reliability Analysis


- Application to Tension Leg Platforms

Page No. 32

Report No. 95-3197, rev. 02

Table 5.5 Design point information for tether ULS from section 10.14 of example report.
Variable

Units

Value

Outer Layer of Reliability Analysis - time independent variables


Model uncertainty pressure capacity
0.848
Model uncertainty axial stress capacity
1.020
Model uncertainty bending stress capacity
1.000
Model uncertainty mean LF offset
1.034
Model uncertainty WF response
1.152
Model uncertainty LF std. deviation
1.020
Auxiliary std. normal variable for inner layer
-2.769
Inner Layer of reliability analysis - out-crossing algorithm
Significant wave height (m)
(m)
Peak wave period
(s)
Wind speed (1 hour)
(m/s)
Current speed
(m/s)
Tide and storm surge
(m)
Offset
(m)
Mean+LF tether axial force
(kN)
WF tether axial force
(kN)
Tether axial force from tide
(kN)
Tether axial force from moment of mean (kN)
environmental forces
Tether axial force from LF moment
(kN)
Sum tether axial force
(kN)
Tether axial stress
(kN/m2)
Tether bending stress
(kN/m2)
Hoop stress utilisation
Axial stress utilisation
Bending stress utilisation

20.0
24.7
45.4
0.423
0.855
26.0
13077 (48%)
9944 (36%)
1371
(5%)
2591
(9%)

Probability

0.203979
0.030222
0.499592
0.632814
0.845078
0.540584
0.002810

1.000000
0.586839
0.794112
0.528560
0.791520
0.997693

322
(1%)
27305 (100%)
299000
244
0.381
0.818
0.0006

5.2.8 Consideration of the possibility of failure anywhere along the length of one tether
The reliability analysis results above refer to the bottom of the tether. For the tether considered,
the effect of the increase in axial stress due to the tether weight, is less than the effect of reduced
pressure, when moving to the tether top. Hence, this tether is most likely to fail at the bottom. If
the variation in tether ovality and material parameters along the length is negligible, then the
probability of failure for the whole tether will be equal to the probability of failure at the bottom,
with the present limit state.
5.2.9 Consideration of the possibility of failure in any of the TLP's tethers
If all the tethers experience identical loads and have identical material properties, then the
probability of failure of any one of the tethers is the same as the probability of failure of a
specific tether. The effects of uncertain material properties are negligible in the present case.
However, the applied loads will vary. Tethers at the same corner will tend to experience very

Guideline for Offshore Structural Reliability Analysis


- Application to Tension Leg Platforms

Page No. 33

Report No. 95-3197, rev. 02

similar loads. Tethers at different corners will tend to experience quite different loads, at least at
any given time and heading angle for the environmental effects. It seems reasonable to assume
that the extreme value distributions for the tether loads at the 4 corners are independent. For an
axisymmetric system, this implies about a fourfold increase in the probability that any tether will
exceed a very extreme load level, compared to the probability for a single tether. The probability
of failure will increase somewhat less, since the same model uncertainty and uncertainty in the
capacity function is common to all the tethers. In practice, the system is unlikely to be
completely axisymmetric - the environmental loads are likely to be more severe from one
geographical direction. The corner which is subject the highest loads will have the tethers with
the highest probability of failure, and this will be very close to the probability of failure for any
one tether in the system.
The situation will be different in the fatigue limit state, when significant variation in the fatigue
resistance should be expected between the tethers.
5.2.10 Comparison of out-crossing results with simplified model
Since the out-crossing analysis is fairly new, some checking of the results is advisable. An
alternative formulation based on Turkstras hypothesis has also been tried out. This hypothesis
assumes that the extreme load-effect occurs simultaneously with an extreme value of one of the
stochastic processes that are involved. Appropriate distributions are applied for the other
stochastic variables, conditional on an extreme value of the selected variable. Two possible
combinations of variables have been considered, and combination B, with extreme wavefrequency load-effects, has been found to be more critical than combination A, with extreme
low-frequency load-effects. Results for combination B are reported in the following.
The basic comparison of the two approaches is initially carried out for a stationary environmental
state. The environmental state is specified by choosing a significant wave height, and fixing the
other environmental variables at their expected values, conditional on that significant wave
height. Diagonal seas are applied. Results for a comparison of the axial stress in the tether are
given in Table 5.6. Very good agreement is found. This also implies that there would be good
agreement in the probability of failure in a stationary environmental state, since the external
pressure varies relatively little, and the bending moment is insignificant.
Table 5.6 Probability of exceeding axial stress level 300000 kN/m2 during 3 hours in an
environmental state
Significant wave height (m)

Probability by outcrossing

Probability - combination B

20

0.969

0.968

19

0.888

0.884

18

0.663

0.657

17

0.298

0.293

16

0.0489

0.0477

15

0.00162

0.00157

14

7.7710-6

7.5310-6

13

5.0310-9

4.8910-9

12

3.0610-13

2.9710-13

Guideline for Offshore Structural Reliability Analysis


- Application to Tension Leg Platforms

Page No. 34

Report No. 95-3197, rev. 02

Next a comparison is made, including the full environmental climate, but retaining diagonal seas.
The results in Table 5.7 show considerable discrepancy between the two procedures, which is
surprising, considering the good agreement in individual environmental states. It seems likely
that the discrepancy arises through a weakness in the computation of the expectation of the outcrossing rate, with respect to the distribution of environmental states. Some additional
explanation and discussion is included in the example report. Little experience has been
available with the accuracy of the approximation applied for the expectation of the out-crossing
rate with respect to the distribution of the environmental states, prior to the present analysis.
Some comparisons made by Hagen (1992) for simple reliability problems have given acceptable
agreement. The present results indicate that it requires further study.
Table 5.7 Annual extreme value distribution for axial stress .
Axial stress level (kN/m2)

Probability by outcrossing

Probability - combination B

375000

1.9710-5

5.4510-6

350000

9.2910-5

2.2610-5

325000

4.8510-4

9.8810-5

300000

2.8210-3

4.6110-4

275000

1.8710-2

2.3610-3

250000

0.136

0.0140

225000

0.705

0.101

200000

0.9996

0.668

5.2.11 Sensitivity Analysis with Respect to Tether Ovality


The tether ovality of 0.25% initially specified implies high quality, expensive production
procedures for the tethers. Some interest has been expressed by one of the project sponsors in
the possibility of allowing larger ovality, to save costs in tether production. A sensitivity
analysis with respect to the ovality parameter has therefore been carried out.
In this case, the annual probability of failure has been computed, conditional on diagonal seas,
and using load combination B, rather than the out-crossing analysis applied in the base case. The
results are shown in Fig.5.5. A slight increase in the ovality, say from 0.25% to 0.5% only leads
to a small decrease in the reliability index, which could easily be compensated by adjusting
tether thickness and diameter. However, a larger increase in the ovality to 2.5% causes a severe
decrease in the reliability index, which would be more difficult to compensate.
The ovality parameter affects the reliability analysis through the buckling pressure only. Fig.5.6
shows how the buckling pressure varies with ovality. In addition, there is a relatively large
model uncertainty applied to the buckling pressure, with a CoV=13%. This model uncertainty
has a very marked effect on the reliability when the ovality is increased. If it is strongly
desirable to use pipes with ovality up to 2.5%, then it might well be worthwhile to do further
development work on the limit state function, with the intention of reducing the model
uncertainty of the buckling pressure.

Guideline for Offshore Structural Reliability Analysis


- Application to Tension Leg Platforms

Page No. 35

Report No. 95-3197, rev. 02

5.2.12 Conclusions for tether ultimate limit state


An example of an ultimate limit state for tether failure due to combined external water pressure,
axial force and bending moment has been presented and successfully analysed.
The system response model applied in the analysis allows the effect of the low-frequency offset
and set-down on the wave-frequency load-effects to be included. By combining these loadeffects in the reliability analysis, appropriate combinations of the low-frequency and wavefrequency effects can be made without assuming a fixed low-frequency offset, or computing a
simultaneous time history simulation of both effects.
When computing the tether response to wave-frequency motions of the platform at non-zero
offset, it is essential to take proper account of the strong coupling of the platform motions,
induced by the coupling terms in the restoring coefficient matrix.
When heading angle is modelled as a stochastic variable in a reliability analysis, then some extra
care is required compared to most other stochastic variables. This arises because the variable is
periodic or circular, and because maxima of load-effects may occur at more than one heading
angle. Confer section 5.5 of the example report for further details.
An outcrossing analysis was applied to accurately handle the nonlinear combination of loadeffects due to external water pressure, axial force and bending moment specified by the limit
state function. An alternative analysis of the load-combination was also developed, utilising the
strong correlation found between axial force and bending moment, as well as the relatively low
magnitude of bending stress. Close agreement was obtained between both formulations, for
probabilities computed conditional on an environmental state. Significant differences in
marginal probabilities were obtained when the effect of the distribution of environmental
variables was integrated out. Past experience provides a fair degree of confidence in the
calculation of marginal probabilities from the type of conditional result obtained from the
simplified method, whereas a fairly new method is applied to calculate marginal probabilities
from the outcrossing results. Hence:
The simplified approach, without outcrossing analysis, is adequate for application to this
tether limit state of the present TLP. It may also be applicable to a wide range of TLPs if the
bending stress is strongly correlated with the axial stress, or is insignificant in comparison
with the axial stress.
Some difficulty has been uncovered with the practical computation of the expectation of the
outcrossing rate with respect to the distribution of environmental states. The extent of this
difficulty, and approaches to resolve it, should be studied in more detail.
Sensitivity analysis with respect to tether ovality showed a manageable reduction in reliability
for a doubling in ovality from 0.25% to 0.5%, while a tenfold increase to 2.5% ovality resulted in
a very large reduction in reliability, which could not be compensated by a moderate increase in
tether wall thickness. This result is dependent on the fairly large uncertainty in the buckling
pressure. If there is further interest in allowing increased ovality, then it might be worthwhile to
check if this uncertainty could be reduced.

5.3 Tether FLS Summary


Fatigue analyses of the tethers is normally based on the Miner-Palmgren hypothesis for
cumulative damage and S-N data for structural fatigue capacities. Fracture mechanics analysis is
used to assess crack growth behaviour. A detailed discussion of the reliability analysis of fatigue
based on these two models is provided in the Jacket Application Guideline and Jacket Example
Report (Sigurdsson et al., 1996, and Sigurdsson and Cramer, 1996). The application of such a
fatigue limit state to TLP tethers is discussed in the TLP Example Report, and is summarised

Guideline for Offshore Structural Reliability Analysis


- Application to Tension Leg Platforms

Page No. 36

Report No. 95-3197, rev. 02

here. A brief survey of some previous studies of TLP tether fatigue is first included in the
following paragraphs.
Wirsching and Chen (1984, 1987) have applied reliability analysis of tether fatigue to the
calibration of a simplified fatigue design requirement for tethers. They have assumed the same
fatigue load is applied to all joints in a tether, and allowed for some difference in the fatigue
strength of the joints. Some discussion of the question of progressive collapse after an initial
tendon failure is also included.
Ximenes (1991), and Ximenes and Mansour (1991) have considered the effect of correlation
between the fatigue strengths of joints in the tethers in more detail. They have also quantified
the effect of time to repair after an initial tether failure, and considered the effect of inspection of
tether joints for developing fatigue cracks.
Jiao (1989) has included the effect of variation in the bending stress along the length of a tether
in 300 m water depth. In that case, the first bending mode has a considerable effect on the
combined stress, and tether joints near the middle of a tether experience a considerably higher
fatigue loading than joints near the ends. This significantly affects the system reliability.
Hovde (1995), and Hovde and Moan (1994), have studied correlation effects in some detail.
They have included the possibility of crack initiation at several sites along the weld seam of a
specific joint and found the number of possible crack sites to be an important parameter. All
failures are assumed to occur in tendons attached to the same TLP leg. This is conservative as
long as the legs are equally loaded, or the leg considered is more heavily loaded than the other
legs.
Winterstein et al. (1994) have included springing load effects in the fatigue analysis of a single
TLP joint using a second order Volterra series. They found a significant increase in the fatigue
damage due to the springing response, for the Snorre TLP in 320 m water depth.
Rhee and Stock (1991), and Rhee et al. (1993) have made a fracture mechanics fatigue
assessment of the threaded joint used in the Hutton TLP tethers, and compared deterministic and
probabilistic analyses.
All these references apply to the installed condition of the tethers, and this is the condition
discussed in the subsequent sections, too. However, a significant amount of fatigue damage may
arise during transportation and installation of the tethers, particularly when each tether is
assembled ashore and floated out in one piece. This source of fatigue damage can also be
incorporated in the following analysis, if necessary.
5.3.1 Failure criterion
The fatigue limit state is developed in more detail in the Jacket Application Report, and is only
briefly introduced here, to provide a basis for the discussion of items mainly related to TLPs.
The fatigue limit state function, based on the Miner-Palmgren hypothesis, may be expressed as

g = D-d

(5.8)

where D is the fatigue capacity and d is the accumulated fatigue damage defined by

d =
i

n( si ) ds
N ( si )
T u f S ( s)
ds
N ( s)

(5.9)

with n( si ) ds the number of applied stress cycles at stress range at level si , and N ( si ) the
number of stress cycles at that stress range required to cause failure. In the limit, when the
number of stress cycles is large, this may be expressed in terms of the long term probability

Guideline for Offshore Structural Reliability Analysis


- Application to Tension Leg Platforms

Page No. 37

Report No. 95-3197, rev. 02

density function for the stress ranges f S ( s) , the mean frequency of the stress cycles u , and the
time duration considered T . When the SN-curve is in the simple form

N ( s) = k s - m

(5.10)

with k and m experimentally determined parameters, then the damage function becomes

Tu
d=
k

f S ( s ) ds

Tu
=
E Sm
k

(5.11)

[ ]

[ ]

where E S m is the expected value of the m-th power of the stress ranges. If a bilinear form of
the SN-curve is applied, then the resulting damage expression is more complicated, and then
requires the expected value of the two alternative powers m1 and m2 of the stress ranges (cf.
example report section 6.2.1).
In a deterministic analysis, the fatigue capacity D might be set to unity, or to one over a safety
factor intended to allow for the uncertainty in the Miner-Palmgren hypothesis. In a reliability
analysis, the fatigue capacity is modelled as a random variable, to handle that uncertainty.
In a fracture mechanics fatigue analysis, the limit state function may be expressed as

g = aC - a N

(5.12)

where a C is a critical crack depth, at which the tether is considered to have failed, and a N is the
crack depth arising after N stress cycles. Further analysis of this limit state function (cf. Jacket
Application Guideline) shows that the crack depth arising is also due to the expected value of the
m-th power of the stress ranges. In this case, m is the exponent of the Paris-Erdogan crack
growth equation. The fracture mechanics approach is more complicated than the SN-curve
approach, but can be used to update the fatigue reliability predictions, with information from inservice inspection of the tethers for developing fatigue cracks.
5.3.2 Analysis of Input required by limit state function
5.3.2.1 Material properties
With the SN approach, the fatigue resistance of the tether joint material is modelled through the
parameters a , m of the SN-curve, as in equation (5.10). With the fracture mechanics approach,
the fatigue resistance of the tether joint material is modelled through the parameters C, m of the
Paris-Erdogan equation. In this case, the initial crack depth is also an important factor in the
fatigue resistance. The stochastic modelling of both these sets of parameters is discussed in the
Jacket Application Guideline. For combination of the probability of fatigue failure at various
locations in the tethers of one TLP, it is essential to consider the dependency of the fatigue
resistance parameters between different locations. How should this dependency be modelled,
and how can correlation coefficients (or other dependency parameters) be quantified? These
questions are addressed in the references quoted in the preceding section, with most detail in the
more recent references.

Guideline for Offshore Structural Reliability Analysis


- Application to Tension Leg Platforms

Page No. 38

Report No. 95-3197, rev. 02

5.3.2.2 Short and long term load-effects


The long term fatigue damage expressed by equation (5.11) is derived from short term models for
the stresses in stationary environmental states

d=

T
k

u (y ) E[(s(y ))
r

]f

r
Y

r
r
(y ) dy

(5.13)

where u (y ) is the frequency of the stress cycles in the short term environmental state denoted

by the environmental state vector y , with long term probability density f Yr (y ) , and the
expectation of the m-th power of the stress range is now also conditional on the environmental
state. When load-effects in different frequency bands are included, then it may be important to
include the frequency inside the integral, as in equation (5.13), rather than outside the integral, as
in equation (5.11). Note that this expression provides a deterministic result for the fatigue
damage, and that it only becomes stochastic after model uncertainties are introduced.
5.3.2.3 Combination of axial and bending stresses
In general, both tether tension and bending moment contribute to the normal stress at a specific
location in a tether joint, for which a fatigue analysis may be required. The axial stress induced
by the tether tension is normally the dominant oscillatory stress, so the bending component is
sometimes neglected to simplify the analysis. If the bending moment induces significant
oscillatory stresses, then the combined normal stress in an environmental state will vary, both
around a tether cross-section, and along the length of the tether. The dependency between
normal stresses at various locations also needs careful consideration in a systems analysis, which
combines the probability of fatigue failure at various locations. In an SN analysis, the axial
stress and bending stress are combined as the normal stress before calculating the fatigue
damage. In a fracture mechanics analysis, these two components may be required separately, to
compute the geometry function. The geometry function is used to take account of the local
geometry of the crack, when computing the stress-intensity factor from the applied, far-field
stress. If a welded tether joint is considered, then this is a joint in a thin-walled pipe, and there
will be little variation in the bending stress across the wall thickness. The effect of the bending
stress on the geometry function can be neglected in this case. If a threaded tether joint is
considered, then this may be relatively thick-walled, and it may be necessary to consider the
effect of the bending stress on the geometry function.
5.3.2.4 Load components
The following oscillatory effects may contribute to the axial and bending stresses at a location in
the tether:
(1) Wave-frequency motions of the tether top attachment point
(2) Springing (high-frequency) motions of the tether top attachment point
(3) External hydrodynamic drag along the length of the tether
(4) Dynamic response of the tether
(5) Transverse loads along the length of the tether due to vortex shedding effects
(6) Variation in platform buoyancy and weight forces due to tides, storm surge, and consumption
and replenishment of stores.
(7) Low-frequency motion of top tether attachment point
This list should be roughly in order of importance for fatigue analysis. For a TLP in shallow
water, such as the Hutton TLP in 148 m water depth, acceptable accuracy should be obtained by
only considering the wave-frequency motions (1). Winterstein et al. (1994) indicate that the
springing effect (2) must be considered for the Snorre TLP in 320 m water depth. Dynamic
analysis was applied for the test example considered in the present work in 325 m water depth,
but made little difference compared to quasi-static analysis. Jiao (1989) indicates that dynamic

Guideline for Offshore Structural Reliability Analysis


- Application to Tension Leg Platforms

Page No. 39

Report No. 95-3197, rev. 02

response is significant for his example TLP in 300 m water depth. Susceptibility to dynamic
response will be largely governed by the first mode of transverse vibrations, which is dependent
on the tether geometry and pretension. Jiao (1989) seems to have used a tether with a first
transverse natural period of 6.8 s, while the present example has a natural period of 5.1 s.
Lotsberg et al. (1988) considered a TLP tether with a natural period of 5.7 s in 310 m water
depth, and obtained twice the fatigue damage when the dynamic bending stress at the mid-length
of the tether was included. For a TLP in 1000 m water depth with first natural period at 26 s,
Hargroves and Jefferys (1989) indicate that dynamic response (3 & 4) is significant, and that
vortex shedding effects (5) can be beneficial, because the cross flow vibration tends to increase
the damping of the in-line vibrations. Daily changes in weight and buoyancy (6) have been
considered negligible in the TLP fatigue analyses reviewed here. Experience so far indicates that
the low-frequency motions (7) have negligible effect on fatigue in TLP tethers. However, the
low-frequency motions do have an effect on the fatigue of catenary mooring lines, so some effect
on unusual TLP tethers may be conceivable.
Tether mispositioning is sometimes included as a load component. More accurately, it may be
said to lead to a perturbation of the other load components. The effect of mispositioning should
be evaluated for a fatigue analysis.
5.3.2.5 Cycle counting
For a narrow banded Gaussian process the calculated fatigue damage in a short term,
environmental state may be found by a simple algorithm. When a stress process includes both
wave-frequency and high-frequency components, it becomes more wide-band, and the narrow
banded, Rayleigh stress range distribution assigns larger probabilities to larger stress ranges, so
that a conservative fatigue damage is predicted. Rainflow counting of the stress ranges from a
time series is considered to provide the most accurate analysis in such cases. However, this is a
computationally expensive technique, and more efficient, approximate algorithms are usually
employed for this purpose, as discussed in the TLP example report.
5.3.2.6 Stress concentration factors
The stress concentration factor at girth welds may be calculated by axisymmetric finite element
analysis. The fabrication tolerance should be accounted for in the analysis.
A stress concentration factor may be defined as the ratio of hot spot stress range over nominal
stress range. The hot spot stress is to be used together with the nominated S-N curve. Stress
concentrations in tethers (and pipelines) are due to eccentricities resulting from different sources.
These may be classified as concentricity(difference in pipe diameters), difference in thickness of
joined pipes, out of roundness, and centre eccentricity. The resulting eccentricity may
conservatively be evaluated by a direct summation of the contribution from the different sources.
Normally the eccentricity due to out of roundness is giving the largest contribution to the
resulting eccentricity. Additional details are included in the TLP example report.
Stress concentration factors also need to be considered for threaded joints.
5.3.2.7 Model uncertainties
Model uncertainties should be included as multiplicative factors on the following variables in the
fatigue limit state:
Computed wave-frequency stress
Computed high-frequency (springing) stress
Cycle counting algorithm for fatigue damage

Guideline for Offshore Structural Reliability Analysis


- Application to Tension Leg Platforms

Page No. 40

Report No. 95-3197, rev. 02

5.3.3 Probabilistic formulation of the reliability problem


Since the expectation of the m-th power of the stresses ranges is taken in equation (5.13), many
of the difficulties involved in handling the stochastic processes in the ultimate limit state are
avoided. The fatigue limit state at a single location only involves time-independent stochastic
variables, and can be computed fairly easily using a first order reliability method.
In order to apply separate model uncertainties to the wave-frequency and high frequency stress
components, these factors have to be applied before the cycle counting algorithm is used to
compute the expected value of the m-th power of the stress ranges, inside the integral in equation
(5.13). Hence, the short term response parameters used as input to the cycle counting algorithm
should be available to the reliability analysis for the full range of environmental states.

5.3.4 Stochastic variables


Table 5.8 gives a list of the stochastic variables involved in the fatigue limit state for a single
potential fatigue crack location, with an SN-curve type analysis.
Table 5.8 List of random variables for fatigue limit state
Random Variable
Uwf

Distribution
Normal, CoV=10%

Uhf

Normal, CoV=20%

Stress
concentration

Normal,. CoV= 5%
is suggested, but
should be further
evaluated for the
actual design.
Circle
distribution
before
installation
and as measured
after installation.
Normal, CoV=15%
Wirsching
&
Light
(1980),
Jiao (1989), Lie
& Fylling(1994)
m is fixed, for ln K General
see
General Guideline, Table
Guideline depending 7.10
on classification of
detail.
lognormal, median Wirsching (1984)
1.0 and CoV = 0.30.

Foundation
mispositioning
Ud

S-N data

Reference

Comment
Model
uncertainty
in
computed 1st order (wave
frequency) stress
Model
uncertainty
in
computed springing (high
frequency) stress
Median value to be evaluated
together with establishment of
requirements to fabrication
tolerances.

Model
uncertainty
for
inaccuracy in cycle counting
algorithm as compared to rain
flow counting

Guideline for Offshore Structural Reliability Analysis


- Application to Tension Leg Platforms

Page No. 41

Report No. 95-3197, rev. 02

5.3.5 System reliability


If the fatigue reliability analysis is extended from a single potential fatigue crack location to
include failure of a tether due to fatigue failure at an location, then this is referred to as a systems
analysis. This concept can be further extended to include the failure of the TLP, due to failure of
several tethers.
In a systems analysis, it is essential to apply a good model for the stochastic dependencies
between the various potential fatigue crack locations. For a chain under uniform tension, the
systems effect arises because the uniform tension finds the link with the extreme lowest strength.
If all the chain links should have exactly the same strength, then there would be no weakest link,
and no systems effect. The systems analysis is critically dependent on how the strengths of the
various links are related to each other. In a TLP, the fatigue resistance may vary around each
joint, and between joints. The applied fatigue load may vary around a joint, along the length of
the tether, and between tethers. The tether system is more complicated than the simple chain
example, because both load and resistance vary between potential fatigue crack locations.
The simplest approach to systems analysis of a single tether is to apply the highest fatigue load
anywhere in the tether to lowest fatigue resistance anywhere in the tether. This will yield a
conservative result for the probability of failure of the tether. The result may well be very
conservative. If a more accurate analysis is required, then the correlation models discussed by
Hovde (1995) should provide a useful basis.

Guideline for Offshore Structural Reliability Analysis


- Application to Tension Leg Platforms

Page No. 42

Report No. 95-3197, rev. 02

6. RECOMMENDATIONS FOR FURTHER WORK


6.1 A Method to Properly Manage the Effect of Low-Frequency Offset on
Wave-Frequency Response
Reliability analysis forces you to make arrangements to compute the system response and value
of the limit state function for any realisation of the input stochastic variables, or at least for quite
a range of values. At first this requirement can seem very arduous, but when you realise that it
may be enough to compute the system response for a limited set of input values, and then
interpolate between these values for the rest of the range, then this requirement is not quite so
troublesome. Then, when you actually do the reliability analysis, and examine the design point
results, you may realise that you have accomplished something very useful. Now it is no longer
necessary to initially assume some conservative value of an input variable, the reliability analysis
finds the right value of that variable for you.
Specifically, for TLP analysis, it is clear that the wave-frequency loads in the tethers are
dependent on the low-frequency platform offset for which they are calculated. In ordinary design
analyses, the tether loads are usually computed at zero offset, and at an extreme offset, and these
results are used to make some conservative choice about the tether design loads. In a reliability
analysis the tether loads should be computed for at least three offset values. This is enough to
arrange a second order polynomial as an interpolation function, which is very well-suited to the
combined effects of linear, horizontal, restoring forces and quadratic, vertical forces due to setdown. The resulting wave-frequency tether loads can then be interpolated for any value of lowfrequency offset, and the reliability analysis chooses this low-frequency offset for you.
This should be a real advantage in the tether analysis. Unfortunately, these results were not
accurately quantified in the example considered here, because of errors in the tether response
analysis.
The tether response calculations at non-zero offset should be corrected with the improved
understanding gained from this work, to obtain and evaluate more accurate results for this
effect.
The principle illustrated by this example is, of course, one of the main strengths of structural
reliability analysis; i.e. to help the analyst to make the right combination of several random
variables. Sometimes this principle gets buried behind the computational machinery.

6.2 Distributions and Dependencies


It is well recognised that non-Gaussian distributions are involved in TLP response; e.g. for lowfrequency offset excited by wave drift forces, or for springing response due to second order,
high-frequency wave forces.
More work is required to develop and obtain experience with alternatives to the Gaussian
distribution, especially for:
- low-frequency offset motion due to combined wind and wave loads, and
- springing response,
in a stationary environmental state.
Specifically, results for the kurtosis of the low-frequency offset were obtained in the system
response analysis for the TLP example, but there was insufficient time available to investigate
the effect of including the kurtosis in an alternative distribution model. This should be done.

Guideline for Offshore Structural Reliability Analysis


- Application to Tension Leg Platforms

Page No. 43

Report No. 95-3197, rev. 02

There are practical advantages in separating the response analysis due to low-frequency, wavefrequency, and high frequency effects. Hence, statistical distributions are primarily needed for
these response components separately. In the present work, these response processes have been
treated as independent in a stationary environmental state, while their dependencies on the
underlying environmental conditions have been included. The assumption of independence is
critical when the load-effects from the different frequency bands are combined.
More work is needed to investigate the independency assumptions, and develop suitable
models for any stochastic dependency that needs to be taken into account.

6.3 Outcrossing Analysis


Although the outcrossing analysis provides accurate and extremely useful results for combined
load-effects in a stationary environmental state, there appears to be some problem with the
algorithm used to compute the outcrossing rate expectation with respect to environmental states.
This problem should be investigated in more detail. An alternative algorithm for the
expectation of the outcrossing rate should be developed, if necessary.

6.4 Load Effects


Load-effects due to springing, ringing, and mispositioning have been excluded from the TLP
example analysed in the present project. More work on these load models for these effects is
needed, both to develop the models further, and to make use of them in reliability analysis.

6.5 Fatigue Limit State


A considerable amount of work has previously been published on the fatigue limit state for TLP
tethers. This work has largely concentrated on the systems effect arising when the possibility of
failure in many tether joints is taken into account, while fairly simple load models have been
applied. However, more sophisticated load models are needed to achieve accurate results at a
single joint, and to take proper account of the variation in fatigue load in a tether on the systems
effect.
A fatigue reliability analysis of a single tether joint should be developed, which includes
accurate modelling of wave-frequency loads, springing loads, and transverse tether dynamic
response.
This single joint analysis should be incorporated in a systems analysis for the probability of
fatigue failure anywhere in a single tether.

6.6 Ovality
Savings in the cost of tethers are possible if increased ovality of the tether cross-section can be
accepted. Some investigation of the effect of ovality on the ultimate limit state has been carried
out. However, ovality also affects the stress concentration factors at welded tether joints.
The sensitivity of the fatigue limit state to increased tether ovality should be investigated as
the next step in any further evaluation of the effect of tether ovality.

Guideline for Offshore Structural Reliability Analysis


- Application to Tension Leg Platforms

Page No. 44

Report No. 95-3197, rev. 02

7. REFERENCES
Andersen, K.H. and R. Lauritzsen (1988). "Bearing capacity for foundations with cyclic loads". J.
Geotech. Engrg., ASCE, 114(5), pp 540-555.
Andersen, K.H., R. Dyvik and K. Schrder (1992). "Pull-out capacity analyses of suction anchors
for tension leg platforms." Proceedings of the 6th International Conference on the Behaviour of
Offshore Structures, BOSS'92, Vol. II, Geotechnics, London, July.
Andersen, K.H., R. Dyvik, K. Schrder, O.E. Hansteen, S. Bysveen (1993). "Field tests of
anchors in clay II: Predictions and interpretation." ASCE, Journal of Geotechnical Engineering,
Vol. 119, No. 10.
API RP2A (1993). " Recommended practice for planning, designing and constructing fixed offshore
platforms." 20th edition, American Petroleum Institute, Washington D.C., USA.
API RP2T (1987). " Recommended practice for planning, designing and constructing tension leg
platforms." 1st edition, American Petroleum Institute, Washington D.C., USA.
API BUL 2U, (1987), Bulletin on Stability Design of Cylindrical Shells, American Petroleum
Institute, Washington DC.
Bai, Y., Igland, R., and Moan, T., (1993), Tube Collapse Under Combined Pressure, Tension
and Bending Loads, Int. J. of Offshore and Polar Engineering, Vol.3, No.2, pp.121-129.
Banon, H., Cornell, C.A., Harding, S.J., (1991), Probabilistic Combination of Forces in Tension
Leg Platform Tethers, Journal of Structural Engineering, Vol.117, No.5, pp.1532-1548, ASCE.
Banon, H., Harding, S.J., (1989), Methodology for Assessing Reliability of Tension Leg
Platform Tethers, Journal of Structural Engineering, Vol.115, No.9, pp.2243-2260, ASCE.
Banon, H., Toro, R.G., Jefferys, E.R., De, R.S., (1994), Development of Reliability-Based
Global Design Equations for TLPs, Proc. 13th OMAE, Vol.2, pp.335-344, Houston.
Bitner-Gregersen, E.M., and Haver, S., (1989), Joint Long Term Description of Environmental
Parameters for Structural Response Calculation, 2nd Int. Workshop on Wave Hindcasting and
Forecasting, Vancouver.
Bitner-Gregersen, E.M., and Haver, S., (1991), Joint Environmental Model for Reliability
Calculations, Det Norske Veritas Research , report no.90-2069, Hvik.
Bjerager, P., et al., (1988), Reliability Method for Marine Structures under Multiple
Environmental Load Processes, Proc. 5th BOSS, pp. 1239-1253, Trondheim.
Bothello, D.L.R., Finnigan, T.D., and Petrauskas, C., (1984), Model Test Evaluation of a
frequency Domain Procedure for Extreme Surge Response Prediction of Tension Leg Platforms,
Proc. OTC, paper no. 4658.

Guideline for Offshore Structural Reliability Analysis


- Application to Tension Leg Platforms

Page No. 45

Report No. 95-3197, rev. 02

Brekke, J.N., and Gardner, T.N., (1986), Analysis of Brief Tension Loss in TLP Tethers, Proc.
OMAE, Tokyo.
Burns, G.E., (1983), Calculating Viscous Drift of a tension Leg Platform, Proc. 2nd OMAE.
Chapman, J.C., Dowling, P.J., Chryssanthopoulos, M.K., Davidson, P.C., Bonello, M., Subbarao,
H., Yau, K.F., and Fieze, P.A., (1991), Design of Flat Stiffened Plating, CESLIC Report SP 9,
Chapman Dowling Associates, Engineering Structures Laboratories, Imperial College, London.
Chen, Y.N., Zimmer, R.A., de Oliveira, J.G., and Jan, H.Y., (1985), Buckling Strength of
Stiffened Cylinders: Model Experiments and Strength Formulations, Proc. 17th OTC, Houston.
Christophersen, H.P., S. Bysveen and O. Stve (1992). "Innovative foundation systems selected for
the Snorre field development", Proceedings of the Sixth International Conference on the Behaviour
of Offshore Structures, BOSS'92, Vol. II, Geotechnics, London, July.
Clukey, E.C., M.J. Morrison, J. Garnier and J.F. Cort (1995). "The response of suction caissons in
normally consolidated clays to cyclic TLP loading conditions." Offshore Technology Conference,
Houston, Texas, May, Paper OTC 7796.
Croker, A., Nadim, F., Lacasee, S., (1996), Guideline for Offshore Structural Reliability
Analysis - Application to Jack-Up Structures, Det Norske Veritas Classification, report no. 950072, Hvik.
Das, P.K., Faulkner, D., and Guedes da Silva, A., (1991), Limit State Formulations and
Modelling for Reliability-Based Analyses of Orthogonally Stiffened Cylindrical Shell Structural
Components, Report no. NAOE-91-26, Dept. Naval Arch. and Ocean Engng., Glasgow
University.
Davidsen, P.C., Chapman, J.C., Smith, C.S., and Dowling, P.J., (1992), The Design of Plate
Panels Subject to Biaxial Compression and Lateral Pressure, Trans RINA, Vol.134, Part A.
de Boom, W.C., Pinkster, J.A., Tan, S.G., (1983), Motion and Tether Force Prediction for a
Deepwater Tension Leg Platform, Proc. OTC, paper no. 4487/4437???.
De Nicola, A., Randolph, M.F., (1993), Tensile and Compressive Shaft Capacity of Piles in
Sand, JGE, ASCE, Vol. 119, No. 2, pp. 1952-1973.
Denise, J.P.F., and Heaf, N.J., (1979), A Comparison between Linear and Nonlinear Response
of a Proposed Tension Leg Production Platform, Proc. OTC.
DNV, (1987), Buckling Strength Analysis of Mobile Offshore Units, Classification Note no.
30.1, Det Norske Veritas, Hvik.
DNV, (1989), Rules for Classification of Fixed Offshore Installations, Part 3, Chapter 6, Special
designs; Tension Leg Platforms, Det Norske Veritas, Hvik.
DNV, (1992), Structural Reliability Analysis of Marine Structures, Class Note No.30.6, Det
Norske Veritas, Hvik.

Guideline for Offshore Structural Reliability Analysis


- Application to Tension Leg Platforms

Page No. 46

Report No. 95-3197, rev. 02

DNV SEASM (1987), WADAM - Wave Loading by Diffraction and Morison Theory, report
no.VT-86-3420, Hvik.
DNV SESAM, (1993), SESAM Users Manual POSTRESP, Interactive Post-Processor for
General Response Analysis, report no. 86-3315/Rev.5, Det Norske Veritas Sesam AS, Hvik.
Doyle, E.H. (1994). "Geotechnical considerations for foundation design of the Auger and Mars
TLPs." Proceedings of the 7th International Conference on the Behaviour of Offshore Structures,
BOSS'94, Cambridge, USA, Vol. I, pp 1-22.
Dyvik, R., K.H. Andersen, C. Madshus and T. Amundsen (1989). "Model tests of gravity
platforms: Description." J. Geotech. Engrg., ASCE, 115(11), 1532-1549.
Dyvik, R., K.H. Andersen, S.B. Hansen, H.P. Christophersen (1993). "Field tests of anchors in
clay I: Description." ASCE, Journal of Geotechnical Engineering, Vol. 119, No. 10.
Eatock Taylor, R., Jefferys, E.R., (1986), Variability of Hydrodynamic Load Predictions for a
Tension Leg Platform, Ocean Engng., Vol.13, No.5, pp.449-490.
Emmerhoff, O.J., Sclavounos, P.D., (1992), The Slow-Drift Simulation of Arrays of Vertical
Cylinders, Journal of Fluid Mechanics, Vol.242, pp.31-50.
Faltinsen, O.M., Van Hoof, R.W., Fylling, I.J., and Teigen, P.S., Theoretical and Experimental
Investigations of Tension Leg Platform Behavior, Proc. 3rd BOSS, MIT.
Faulkner, D., Birrell, N.D., and Stiansen, S.G., (1983), Development of a Reliability-Based
Code for the Structure of Tension Leg Platforms, Proc. 15th OTC, Houston.
Goulois, A.(1982). "Contribution to the study of tension piles under cyclic loading." PhD thesis,
Massachusetts Institute of Technology, Cambridge, USA.
Hagen, ., (1992), Structural Reliability Methods under Time Dependency, Det Norske Veritas,
report no.92-2013 (thesis), Hvik.
Hagen, ., and Tvedt, L., (1991), Vector Process Outcrossings as a Parallel System Sensitivity
Measure, J. Engng. Mech., ASCE Vol.117, No.10, pp.2201-2220.
Hamilton, J., (1993), Platform and Tether Structural Dynamics, Proc. Seminar on Tensioned
Buoyant Platforms, Bentham Press, London.
Harding, S.J., Banon, H., (1989), Reliability of TLP Tethers under Maximum and Minimum
Lifetime Loads, 21st Offshore Technology Conf., paper no. OTC5935, Hoston.
Hargroves, D., Jefferys, R., (1989), Analysis of Deepwater TLP Moorings, Proc. 8th Offshore
Mechanics and Artic Engineering Conf., The Hague.
Herfjord, A.K., Nielsen, F.G., (1991), Motion Response of Floating Units: Results from a
Comparative Study on Computer Programs, Proc. 10th OMAE, Stavanger.

Guideline for Offshore Structural Reliability Analysis


- Application to Tension Leg Platforms

Page No. 47

Report No. 95-3197, rev. 02

Hovde, G.O., (1995), Fatigue and Overload Reliability of Offshore Structural Systems,
Considering the Effect of Repair and Inspection, MTA report no.1995:104 (thesis), Division of
Marine Structures, Norwegian Inst. of Technology, Trondheim.
Hovde, G., and Moan, T., (1994), Fatigue Reliability of TLP Tether Systems, Proc. 13th
OMAE Conf. on Offshore Mechanics and Artic Engineering, Vol.2, pp.141-150, Houston.
Huijsmans, R.H.M., and Dekker, J.N., (1989), Combined Effect of Waves and Current on the
Motion Behavior and Tether Forces of a Deepwater Tension Leg Platform,, Proc. 21st OTC,
Houston.
Hunter, A.F., Zimmer, R.A., Wang, W.J., Bozeman, J.D., Adams, C.J., and Rager, B.L., (1990),
Designing the TLWP, Proc. 22nd OTC, Vol.3, pp.147-158, Houston.
Jefferys, E.R., (1987), Numerical problems of first order diffraction theory, 2nd Int. Workshop
on Water Waves and Floating Bodies, Report no. AM-87-06, University of Bristol.
Jefferys, E.R., and Patel, M.H., (1982), Dynamic Analysis Models of Tension Leg Platforms,
J. Energy Resources Technology, Vol.104.
Jiao, G., (1989), Reliability Analysis of Crack Growth under Random Loading Considering
Model Updating, MTA report no.1989:66 (thesis), Division of Marine Structures, Norwegian
Inst. of Technology, Trondheim.
Jiao, G., Moan, T., Marley, M., (1990), Reliability Analysis of TLP Tether Systems, Proc. 9th
OMAE, Vol.2, pp.65-71.
Kareem, A., and Li, Y., (1990), Response of Tension Leg Platforms to Wind, Waves and
Currents: a Frequency Domain Approach, Proc. 22nd OTC, Houston.
Karlsrud, K. and F. Nadim (1990). "Axial capacity of offshore piles in clay." Offshore
Technology Conference, Houston, Texas, May.
Keaveny, J.M., F. Nadim, and S. Lacasse (1989). "Autocorrelation functions for offshore
geotechnical data. " Proc., 5th International Conf. on Structural Safety and Reliability
(ICOSSAR'89), San Francisco, USA, 711 August.
Keaveny, J.M., F. Nadim and R. Rashedi (1993). "Probabilistic evaluation of cyclic capacity for
TLP foundation." Proceedings, 12th Int. Conf. on Offshore Mechanics and Arctic Eng., Glasgow
Scotland, June.
Korsmeyer, F.T., Lee, C.-H., Newman, J.N., and Sclavounos, P.D., (1988), The Analysis of
Wave Effects on Tension Leg Platforms, Proc. 7th OMAE, Vol.2, pp.1-14, Houston.
Kraft, L.M., (1991), Computing Axial Pile Capacity in Sands for Offshore Conditions, Marine
Geotechnology, Vol. 9, pp. 61-92.
Kung, C.J., and Wirsching, P.H., (1992), Fatigue and Fracture Reliability and Maintainability of
TLP Tendons, Proc. 11th OMAE, Vol.2, pp.15-21, Calgary.

Guideline for Offshore Structural Reliability Analysis


- Application to Tension Leg Platforms

Page No. 48

Report No. 95-3197, rev. 02

Larsen, C.M., White, C.N., Fylling, I.J., and Nordsve, N.T., (1984), Nonlinear Static and
Dynamic Behavior of Tension Leg Platform Tethers, Proc. 3rd OMAE, Vol.1, pp.41-50.
Lee, C.H., Newman, J.N., Kim, M.H., Yue, D.K.P., (1991), The Computation of Second-Order
Wave Loads, Proc. 10th OMAE, Vol.1A, pp.113-123.
Lie,H., Fylling,I., (1994), Evaluation of Methods for Fatigue Analysis of Offshore Mooring
Lines, 10th Offshore South East Asia Conf., Singapore.
Lotsberg, I., (1991), Probabilistic Design of the Tethers of a Tension Leg Platform, J. Offshore
Mechanics and Artic Engineering, Vol.113, pp.162-170.
Lotsberg, I., Dalane, J.I., Bakken, K., Gran, S., (1988), Fatigue Analysis of a Tension Leg
Platform, Det Norske Veritas, report no.88-2040, Hvik.
Madsen, H.O., (1985), Extreme Value Statistics for Nonlinear Stress Combination, Journal of
Engineering Mechanics, Vol.111, No.9, pp-1121-1129, ASCE.
Madsen, H.,O., Pedersen, C., Gran, S., (1987), Snorre TLP Tether Foundation Position Error
Sensitivity Study, A.S Veritas Research, report no.87-2024, Hvik.
Marthinsen, T., (1989), Hydrodynamics in TLP Design, Proc. 8th OMAE, Vol.1, pp.127-133,
The Hague.
Marton, I, Mathisen, J., (1993), Motions of a TLP in Wind and Waves, Det Norske Veritas
Research AS, report no.93-2018, Hvik.
Mathisen, J., (1993), A Polynomial Response Surface Module for Use in Structural Reliability
Computations, Det Norske Veritas Research report no.93-2030, Hvik.
Mathisen, J., Hansen, V., Steinkjer, O., (1995), Guideline for Offshore Structural Reliability
Analysis - Examples for Tension Leg Platforms, Det Norske Veritas Industry, report no. 953198, Hvik.
Mathisen, J., Nestegrd, A., Stansberg, C.T., Krokstad, J., (1994a), Preliminary Guidelines for
Determination of Ringing Effects in Design of Offshore Platforms, Det Norske Veritas Industry,
report no. 94-3536, Hvik.
Mathisen, J., Rashedi, R., Mrk, K., Zimmer, R., Skjong, R., (1994b), Reliability-Based LRFD
Code for TLP Hull Structure, Proc. 13th OMAE, Vol.2, pp.345-356, Houston.
MCS International, (1989), FRECOM-3D, Three Dimensional Frequency Domain Offshore
Analysis Program, Theory Manual, Version 2.1, Marine Computation Services, Galway.
Mercier, J.A., (1991), Reliability-Based Design Code for Floating Systems, 2nd SNAME
Offshore Symposium, Houston.
Mercier, J.A., Birrell, N.D., Chivvis, J.C., and Hunter, A.F., (1991),, Tension Leg Platforms Progress and Prospects, Trans. SNAME, Vol.99, pp.249-280.

Guideline for Offshore Structural Reliability Analysis


- Application to Tension Leg Platforms

Page No. 49

Report No. 95-3197, rev. 02

Mercier, J.A., Leverette, S.J., and Bilault, A.L., (1982), Evaluation of Hutton TLP Response to
Environmental Loads, Proc. 14th OTC, pp.585-598, Houston.
Mercier, R.S., Niedzwecki, J.M., (1994), Experimental Measurement of Second-Order
Diffraction by a Truncated Vertical Cylinder in Monochromatic Waves, 7th BOSS Int. Conf. on
the Behaviour of Offshore Structures, Vol.2, pp.265-287, Boston.
MIT, (1995), SWIM (Slow Wave Induced Motion), User Manual, Version 3.0, Massachusetts
Institute of Technology.
Moan, T., Estefan, S., and Svik, S., (1992), Limit State for Tendon and Production Riser
Bodies. Selection and Implementation of Limit States as Failure Functions and Design
Equations, SINTEF, Trondheim.
Natvig, B.J., and Vogel, H., (1991), Sum-Frequency Excitations in TLP Design, Proc. 1st Int.
Offshore and Polar Engng. Conf., Edinburgh.
Newman, J.N., (1974), Second-Order, Slowly-Varying Forces on Vessels in Irregular Waves,
Proc. Int. Symp. on the Dynamics of Marine Vehicles and Structures in Waves, Inst. of
Mechanical Engineers, London.
Newman, J.N., Lee, C.-H., (1992), Sensitivity of Wave Loads to the Discretization of Bodies,
Proc. BOSS 92, London.
Nielsen, F.G., (1992), FPS 2000 Project 1.1 Results from a Comparative Study of Computer
Programs - Summary Report, Norsk Hydro Research Centre, Doc.R-054095, Bergen.
Nss, A., (1989), Prediction of Extremes of Combined First-Order and Slow-Drift Motions of
Ofshore Structures, Applied Ocean research, Vol.11, No.2, pp.100-110.
Nss, A., (1990), Statistical analysis of nonlinear, second-order forces and motions of offshore
structures in short-crested random seas, Probabilistic Engng. Mechanics, Vol.5, No.4, pp.192203.
Nss, A., (1994), Statistics of Combined Linear and Quadratic Springing Response of a TLP in
Random Waves, J. Offshore Mechanics and Artic Engineering, Vol.116, No.3, pp.127-136.
Oortmerssen, G. van, (1991), Assessment of the Validity of Computer Models for the Prediction
of Dynamics of Floating Structures, Dynamics of Marine Vehicles and Structures in Waves,
pp.253-262, Elsevier, Amsterdam.
Paulling, J.R., (1977), Time Domain Simulation of Semisubmersible Platform Motion with
Application to the Tension Leg Platform, SNAME Spring Meeting.
Paulling, J.R., and Horton, E.E., (1971), Analysis of the Tension Leg Platform, Society of
Petroleum Engineers Journal, pp.285-294.
Rainey, R.C.T., (1978), The Dynamics of Tethered Platforms, Trans. RINA, Vol.120.

Guideline for Offshore Structural Reliability Analysis


- Application to Tension Leg Platforms

Page No. 50

Report No. 95-3197, rev. 02

Rhee, H.C., Stock, P.J., (1991), Fracture Mechanics Fatigue Life Analysis of Hutton TLP
Tether Threaded Joint, Proc. 10th OMAE Conf. on Offshore Mechanics and Artic Engineering,
Vol.3B, pp.519-528, Stavanger.
Rhee, H.C., Tallin, A., Stock, P.J., (1993), Comparison of Probabilistic and Deterministic
Fracture Mechanics Fatigue Life Assessment of Hutton TLP Tether Threaded Joint, Proc. 12th
OMAE Conf. on Offshore Mechanics and Artic Engineering, Vol.3B, pp.425-451, Glasgow.
Rodenbusch, G., (1986), Random Directional Wave Forces on Template Platforms, Proc. 18th
OTC, Vol.1, pp.147-156, paper no. 5098, Houston.
Rooney, P., Lereim, J., Madsen, H.O., (1989), Application of Probabilistic Methods for
Verification and Calibration of a TLP Tether System Design Based on Partial Coefficients, 8th
OMAE - Offshore Mechanics and Artic Engineering Conf., The Hague.
Salvesen, N, von Kerczek, C.H., Yue, D.K., and Stern, J., (1982), Computation of Nonlinear
Surge Motions of Tension Leg Platforms, Proc. 14th OTC, paper no. 4394.
Sigurdsson, G., Cramer, E., Lotsberg, I., Berge, B., (1996), Guideline for Offshore Structural
Reliability Analysis - Application to Jacket Platforms, Det Norske Veritas, report no. 95-3203,
Hvik.
Sigurdsson, G., Cramer, E., (1996), Guideline for Offshore Structural Reliability Analysis Examples for Jacket Platforms, Det Norske Veritas, report no. 95-3204, Hvik.
Skjong, R., Bitner-Gregersen, E., Cramer, E., Croker, A., Hagen, ., Korneliussen, G., Lacasse,
S., Lotsberg, I., Nadim, F., Ronold, K.O., (1995), Guideline for Reliability Analysis of Marine
Structures - General, Det Norske Veritas Research, report no.95-2018, Hvik.
Stahl, B., and Geyer, J.F., (1984), Fatigue Reliability of Parallel Member Systems, J. Structural
Engng, Vol.110, No.10, ASCE.
Stahl, B., and Geyer, J.F., (1985), Ultimate Strength Reliability of Tension Leg Platform
Systems, Proc. 17th Offshore Technology Conf., paper no. OTC4857, Houston.
Standing, R.G., (1991), The Verification and Validation of Numerical Models, with Examples
Taken from Wave Diffraction Theory, Wave Loading and Response, Dynamics of Marine
Vehicles and Structures in Waves, pp.263-278, Elsevier, Amsterdam.
Takagi, M., Arai, S-I., Takezawa, S., Tanaka, K., Takarada, N., (1985), A Comparison of
Methods for Calculating the Motions of a Semisubmersible, J. Ocean Engineering, Vol.12,
No.1.
Verley, R.L.P., and Moe, G., (1980), Hydrodynamic Damping of Offshore Structures in Waves
and Currents, Proc. OTC, paper no. 3798, Houston.
White, C.N., Erb, P.R., and Botros, F.R., (1988), The Single-Leg Tension Leg Platform: A
Cost-Effective Evolution of the TLP Concept, Proc. 20th OTC, pp.189-196, Houston.
White, C.N., Triantafyllou, M.S., and Erb, P.R., (1989), Response Cancellation as a Tool for
Single Leg TLP Optimization, Proc. 8th OMAE, pp. 159-166, The Hague.

Guideline for Offshore Structural Reliability Analysis


- Application to Tension Leg Platforms

Page No. 51

Report No. 95-3197, rev. 02

Winterstein, S.R., (1988), Nonlinear Vibration Models for Extremes and Fatigue, Engng.
Mechanics, ASCE, Vol.114, No.10, pp.1772-1790.
Winterstein, S.R., Ude, T.C., Marthinsen, T., (1994), Volterra Models of Ocean Structures;
Extreme and Fatigue Reliability, Journal of Engng. Mechanics, Vol.120, No.6, pp.1369-1385,
ASCE.
Wirsching, P.H., (1984), Fatigue Reliability of Offshore Structures, ASCE Journal of
Structural Engineering, Vol. 110, pp. 2340-2356.
Wirsching, P.H., and Chen, Y.N., (1984), A Preliminary Study of Fatigue Design Requirements
for Tension Leg Platforms, Proc. 4th Speciality Conf. on Probabilistic Methods and Structural
Reliability, pp.163-166, Berkeley.
Wirsching, P.H., and Chen, Y.N., (1987), Fatigue Design Criteria for TLP Tendons, J.
Structural Engng, ASCE, Vol.113, No.7, pp.1398-1414.
Wirsching, P. H. and Chen,Y. -N., (1988), Considerations of Probability-Based Fatigue Design
for Marine Structures, Marine Structures, Vol. 1, No. 1, pp 23-45.
Wirsching, P. H. and Light, M. C., (1980), Fatigue under Wide Band Random Stresses, Journal
of Structural Division, ASCE.
Ximenes, M.C.C., (1991), Fatigue Reliability and Inspection of TLP Tendon System, Marine
Technology, Vol.28, No.2, pp.99-110.
Ximenes, M.C.C, and Mansour, A.E., (1991), Fatigue System Reliability of TLP Tendons
Including Inspection Updating, Proc. 10th OMAE, Vol.2, pp.203-211.
Yosida, K., Ozaki, M., and Oka, N., (1984), Structural Response Analysis of Tension Leg
Platforms, J. Energy Resources Technology, Vol.106.
Zhao, R., Faltinsen, O.M., Krogstad, J.R., and Aanesland, V., (1988), Wave-Current Interaction
Effects on Large-Volume Structures, Proc. 5th BOSS, Trondheim.

Guideline for Offshore Structural Reliability Analysis


- Application to Tension Leg Platforms
Report No. 95-3197, rev. 02

8. FIGURES

Page No. 52

Guideline for Offshore Structural Reliability Analysis


- Application to Tension Leg Platforms

Page No. 53

Report No. 95-3197, rev. 02

offset

water
level
setdown

heave

Motion
Directions

Guideline for Offshore Structural Reliability Analysis


- Application to Tension Leg Platforms

Page No. 54

Report No. 95-3197, rev. 02

Fig.1.1 TLP motions:compliant in surge, sway and yaw, restrained in heave, roll and pitch.
Offset is the radial resultant of surge and sway. Set-down is a 2nd order result of offset, in the
vertical (negative heave) direction.

Guideline for Offshore Structural Reliability Analysis


- Application to Tension Leg Platforms
Report No. 95-3197, rev. 02

Fig.4.1 Forces acting on a tension pile anchor.

Page No. 55

Guideline for Offshore Structural Reliability Analysis


- Application to Tension Leg Platforms

Page No. 56

Report No. 95-3197, rev. 02

Fig.4.2 Simplified stress conditions for some elements along a potential failure surface in the soil.

Fig.4.3 Failure surfaces and observed displacements. a) Model Tests 1,


b) Model Test 4 (Anderson et al., 1993)

2, and 3,

Guideline for Offshore Structural Reliability Analysis


- Application to Tension Leg Platforms

Page No. 57

Report No. 95-3197, rev. 02

Fig.4.4 Overview of load combination calculation for bay failure limit state of orthogonally
stiffened cylindrical shells.

Guideline for Offshore Structural Reliability Analysis


- Application to Tension Leg Platforms

Page No. 58

Report No. 95-3197, rev. 02

weather 0
Z

weather 45
d
weather 90
d
Fig.5.1 TLP global coordinate system and definition of weather directions - section through
mean water level, seen from above.

Guideline for Offshore Structural Reliability Analysis


- Application to Tension Leg Platforms
Report No. 95-3197, rev. 02

Fig.5.2 Overview of TLP offset computation in a specific environmental state.

Page No. 59

Guideline for Offshore Structural Reliability Analysis


- Application to Tension Leg Platforms
Report No. 95-3197, rev. 02

Fig.5.3 Overview of tether response computation in a specific environmental state.

Page No. 60

Guideline for Offshore Structural Reliability Analysis


- Application to Tension Leg Platforms
Report No. 95-3197, rev. 02

Fig.5.4 Overview of tether ULS computation.

Page No. 61

Guideline for Offshore Structural Reliability Analysis


- Application to Tension Leg Platforms

Page No. 62

Report No. 95-3197, rev. 02

Fig.5.5 Sensitivity of tether reliability to tether ovaility.


conditional on diagonal seas.

Load combination B is applied,

Fig.5.6 Tether buckling pressure as a function of tether ovality.

You might also like