Proceedings of The Thirteenth (2003) International Offshore and Polar Engineering Conference

Honolulu, Hawaii, USA, May 25 30, 2003

Copyright 2003 by The International Society of Offshore and Polar Engineers
ISBN 1 880653 -60 5 (Set); ISSN 1098 6189 (Set)

Fatigue Design of Offshore Floating Structures

Timo P. J. Mikkola, Ilari Silvola, Jyri-Pekka Arjava, Antti Ajosmki
PI-Rauma LTD
Pori, Finland
and

Timo Kukkanen
VTT
Espoo, Finland

scheduled based on the steel mill, yard and transport capacities and the
time for the necessary on-site installations. Structural design is done, to
large extent, concurrently with the fabrication. Offshore structures are
in practice prototypes designed specifically for certain field. Each
structure has to be analyzed and designed independently even though
similar details are often used. This is especially true for the spar and
truss spar structures, which are relatively new designs with short
history.

ABSTRACT
A novel fatigue analysis procedure is introduced for offshore floating
structures. It is based on separation of hydrodynamic load and
structural stress responses, effective fatigue load concept and response
interpolation. The actual fatigue analysis is simplified to a spreadsheet
calculation with a few directional fatigue effective load cases.
With the present approach fatigue damage estimates are calculated at
early stage of the project. Detailed information is produced on load and
stress responses of structural details for design development. A limited
effort is needed for updating the fatigue life estimates as the design and
stress and load analyses become more accurate.

The design process is always a challenge. The main dimensions might

be chosen relatively early but often the decisions, which are important
for the structural design, are delayed. Typically a large number of
major design changes tend to come up in the middle of the structural
design process or simply after the design is complete. Often the
significance of a certain design change for strength or fatigue becomes
clear only after costly and time consuming analyses.

KEY WORDS: Fatigue analysis; floating offshore structures

INTRODUCTION

A typical approach in structural design process is first to estimate the

extreme loading conditions and check the strength against them.
Fatigue is more complex to assess and consequently the fatigue
analysis is made afterwards. This has been acceptable approach as
fatigue analysis has typically resulted to local detail changes at the
most. With the present designs this is not anymore true. Fatigue has
become equally important design driver with strength. The nominal
stress levels, acceptable for strength, may need to be lowered due to
fatigue. Such changes may result in excessive costs, if done at a late
stage. It has become clear that fatigue design tools should be applied
already when first design checks are made.

Floating offshore structures are being utilized in remote offshore areas

where they need to keep position regardless of weather conditions. An
example is the spar design for which the number of installations is
about to reach 10 at the Gulf of Mexico. The service life of these
structures is typically at least 20 years. With high safety factor of 10 the
design life for safety critical details becomes 200 years. High safety
factors are necessary, as repairs would be very costly if at all possible.
Design of welded steel structures for offshore are affected by demands
for structural efficiency, faster production time and cost effectiveness,
which are not allowed to compromise the high safety requirement. In
todays offshore structures high strength steels are used, which mainly
helps in meeting the strength requirements with smaller scantlings and
less weight. The fatigue strength of welded joints is close to being
independent of steel strength. Similarly, optimization of designs
typically results in smaller scantlings. Both the high steel strength and
the design optimization will increase the nominal stress levels. As a
consequence the relative importance of fatigue as a design driver has
been increased.

The present paper introduces a fatigue analysis procedure developed for

truss spar structures. Such procedures become partly design specific but
certain features of the procedure are directly applicable for other
floating offshore structures as well.

FATIGUE ANALYSIS PROCEDURE

Fatigue analysis needs to include estimation of local cyclic stress
history at all potentially critical structural details. It is not dictated by a
single extreme loading condition, but the cumulative effect of all

An important milestone for any oilfield development is the time spent

up to the first oil. In order to minimize this time span projects are

186

is not sufficient against the outside pressure in vertical position. High

safety against fatigue is required especially at the connections of the
sections and for the nodal joints at the tubular truss section.

loading conditions. A large number of different sea state conditions

should be covered by the analysis.
Fatigue analysis of floating offshore structures is based on local
nominal or hot spot stress approach, see e.g. API RP-2A WSD (2002).
Fatigue strength is based on experimental fatigue test data and this part
is strictly defined by guidance documents such as UK HSE (1995) and
DNV RP-C203 (2000). Linear damage summation is used for the
variable amplitude load history.

Requirements for Design Tool for Fatigue Analysis

Optimal tool package for fatigue design analyses would include several
procedures with varying complexity. They should cover the needs
during the early design stage for feasibility assessment up to the final
fatigue analyses made for verifying the chosen design. Furthermore, the
analysis process should produce results in a condensed form applicable
for sensitivity studies. This is necessary in order to avoid complete reanalyses for the various design changes.

Fatigue loading on a floating offshore structure, such as the spar, is

caused by waves in practice. Main fatigue effects are caused by the 1st
order wave loads, but the higher order loads become significant also for
fatigue design. Varying wind causes an additional cyclic load
component, which must also be accounted in fatigue analysis. The only
practical option for load analysis for fatigue becomes a non-linear time
domain motion analysis. Stress analysis for fatigue must be done with
extremely fine FE models. Consequently extremely complex and time
consuming analysis process results if stress analysis is conducted in
time domain for a large number of sea states, see e.g. API RP-2A WSD
(2002).

A design oriented fatigue analysis procedure needs to be hierarchic

such that the different structural details can be treated with varying
accuracy levels. The most critical hot spots might need to be analyzed
using full time domain stress approach. For majority of details less
rigorous approaches are sufficient. Key items for such procedure are
separation of load and local stress responses and reduction of the
number of sea states analyzed.
For a fatigue analysis the hydrodynamic motion and load responses
must be transformed to relevant stress ranges compatible with the
fatigue strength data. The local stress response is completely
determined by the nominal stress level and the local structural details
and extremely refined FE models need to be applied. The
hydrodynamic load response varies along with the actual structure and,
although the fatigue analysis approaches may be the same for all
floating offshore structures, case specific procedures are needed.

WORKOVER /
COMPLETION RIG

50'-0"

55'-0"

CHAIN JACK
MWL

Motion and Global Load Response

188'-0"

220'-0"

HARD TANK

The spar structure is designed to have small response to waves,

particularly for heave. Good performance for roll and pitch is achieved
with large inertia compared to restoring moment. Large masses at the
ends of the structure together with deep shape compared to diameter
make spar structural design mostly a shear and bending problem. The
global loads are sensitive even to small changes in roll and pitch. The
structure has considerable slow frequency motions or, putting it another
way, the wave response is small.

STRAKE

493'-0"

MOORING FAIRLEAD

330'-0"

288'-0"

650'-0"

MOORING LINE

HEAVE PLATE
90'x90'
TRUSS
MIDSECTION

100'-0"

17'-0"

KEEL TANK

EXPORT SCR
WITH PULLTUBE

TOP TENSIONED
RISERS

NEPTUNE
CLASSIC SPAR

NANSEN/BOOMVANG
TRUSS SPAR

Fig. 1 Classic Spar (Neptune) and Truss Spar (Nansen).

The present paper introduces a fatigue design procedure developed for
the truss spar, see e.g. Bangs et. al. (2002). The truss spar is a
modification of classic spar, Fig. 1. The hull consists of a cylindrical
ring stiffened section called hard tank, a tubular space frame called
truss section and a ballast and flotation tank at the keel called soft tank,
see Fig. 1. The hard tank forms the buoyancy and has sufficient
strength against the outside pressure in vertical position. Soft tank
consists of a compartment for fixed ballast and a compartment for
flotation in horizontal position. The strength of the soft tank, if empty,

Fig. 2. Hydrodynamic model of a truss spar shown below the water

surface.
For long waves the hydrodynamic analysis can be conducted with
Morison elements. For all of the fatigue sea states this is not possible.
At least the hard tank needs to be modeled with diffracting panel
elements. It is customary to model all the large displacement structures,

187

of sea states with the interpolation approach. The analysis is conducted

only for a limited number of sea states and extrapolated for others.
Consequently the total number of sea states does not affect the analysis
complexity as long as it remains two-dimensional.

i.e. also soft tank and heave plates with diffracting panel elements.
Truss members and viscous drag are modeled with Morison elements.
A typical hydrodynamic model of a truss spar is shown in Fig. 2.
Wind, mooring lines and risers have little effect on force variation
directly, but they affect the overall motions, so they need to be included
in the hydrodynamic model.

Tp
Hs

2.555
0.000

Fatigue Load Description

Cyclic stress responses in the spar hull are caused by cyclic global
motion of the structure and cyclic local pressure loads. The fatigue
analysis is simplified if the fatigue loading of a structure can be
described by global load parameters independent on the local structural
design. The consequence is that the hydrodynamic load and structural
stress analyses can be conducted separately.
Three critical connections exist in the truss spar between the topside,
hard tank, truss section and soft tank. The fatigue loading at these
connections is well defined by the global cross sectional moment, shear
force and vertical force.

3.966
16.331
0.048
0.002
0.000

0.151
15.47
17.74
0.122
0.004
0.000

0.986
0.740
9.554
12.94
1.541
0.010
0.000

0.007
0.034
0.151
1.521
6.343
2.961
0.096
0.002
0.001

0.000
0.096
0.290
1.812
2.027
0.441
0.019
0.003
0.000

0.001
0.012
0.044
0.100
0.639
0.604
0.107
0.005
0.002
0.001
0.000

0.003
0.002
0.010
0.009
0.011
0.048
0.107
0.156
0.022
0.004
0.003
0.002
0.001
0.001
0.000

0.000
0.002
0.002
0.002
0.003
0.001
0.002
0.010
0.045
0.011
0.005
0.004
0.004
0.001
0.002
0.001
0.000
0.000

0.000
0.001
0.002
0.001
0.003
0.000
0.001
0.001
0.000
0.000
0.001
0.001
0.002
0.001
0.001
0.001
0.001
0.001
0.001
0.001

0.001
0.002
0.001
0.003
0.001
0.000
0.001
0.001
0.000
0.001
0.000
0.000
0.000
0.001
0.001
0.001
0.000

0.000
0.001
0.000
0.001
0.001
0.001
0.000
0.000
0.000

0.000

0.000
0.000
0.000

0.000
0.000
0.001
0.001
0.000
0.000
0.000
0.000
0.000

0.000
0.000

0.000

0.000

The truss section members are loaded primarily by loads due to the
global motions. The local pressure fluctuations at deep water are small
and pipe diameters small compared to hard or soft tanks. The truss
section member forces of different tubulars cannot be used as load
response parameters as they depend on the structural design. For the
truss section the global cross sectional bending moment, shear force
and vertical force are applied as the global load parameters similarly to
the connections. In addition, the total pressure loads on heave plates are
considered as global load parameters.

Fig. 3 Two dimensional scatter data for in-place fatigue sea states. Stars
denote the 9 sea states selected for numerical simulation.

LOAD PROCESSING FOR FATIGUE ANALYSIS

For a fatigue analysis the cyclic stress response is needed. In the
present procedure only the global load response need to be processed as
the stress responses are calculated only for the effective load cases.

Definition of the global load response becomes more difficult for the
hard tank section. Clearly, the loads dominating fatigue are the 1st order
wave loads and the other load effects are less significant. A frequency
domain spectral approach for fatigue is used for the hard tank section.
The contributions of the other than the 1st order wave loads are
controlled through the global cross section loads at both ends of the
hard tank section.

Sea State Fatigue Load Processing

The time domain simulations result in global load time histories, which
need to be processed for fatigue effective loads. An example of global
moment response at the cross section connecting the truss section and
the cylindrical hard tank section is shown in Fig. 4. Rainflow counting
procedure is applied for extracting load cycles from the time history,
which gives a moment load cycle histogram of Fig. 5. The presence of
low frequency response is clearly demonstrated by the power spectral
density plot, Fig. 6, for the same time history.

Also for the heave plates the fatigue analysis need to consider the local
cyclic pressure loads. Similarly to the hard tank a frequency domain
fatigue analysis is conducted combined with estimation of significance
of other fatigue load effects.

The accuracy of the time domain simulation depends on the time step
size and the simulation length. In addition the wave and wind load
representations need to be truly random. Testing with simulation time
step and length has shown that the standard deviation of the time series
response becomes stable quite easily with time steps already close to
one second. The time step size depends on wave frequency content but
under typical sea states result stability can be easily achieved.

Environmental load definition

Environmental loading for fatigue is typically described with a twoparameter site specific scatter data, Fig. 3. The scatter is specified
separately for different loading directions. The two main parameters are
sea state wave height Hs and peak period Tp. The wave spectrum type,
wind load and current are usually described as a function of sea state
wave height.

It would be tempting to use the Rayleigh model, when calculating the

fatigue effective loads, which would be clearly conservative, see Fig. 5.
It is based only on the process standard deviation for which numerical
stability and accuracy are easily achieved. Several tests with the fatigue
effective global loads for spar structures have given varying results for
the associated inherent conservatism. In some cases it is small enough
to justify the use of the Rayleigh model and sometimes not. The present
procedure applies Rainflow counting in order to reduce the systematic
inherent conservatism related to the use of the Rayleigh model.

The number of different sea states becomes easily quite large already
with the two-parameter description of the environmental loading.
Typically the analysts would like to reduce the number of sea states in
the scatter data and use so called condensed scatters. On the other hand
the complexity of the environmental loading description is bound to be
increased along with increasing weather data. A difficulty arises from
the condensing of the scatter data. It is not easy to control its effect on
the fatigue damage estimates.

The Rainflow counting result changes, when simulation time is

increased or time step size is reduced. The result always becomes

The present procedure practically solves the problem of large number

188

inaccurate for the low probability and high range load cycles, see Fig.
5. This is easily controlled by simulation length. The simulation length
must by sufficient such that the low probability large cycles do not
anymore affect the fatigue effective load result of Eq. (2). This limit is
typically reached already with relatively short 3 hour simulations.

The simulation time step size may affect the load cycle histogram
distribution shape. At present the same time step is used for all sea
states. A limited series of analyses are conducted at early stage of the
project to assure sufficient accuracy.

Effective Fatigue Load Concept

Fatigue damage is defined as number of load cycles divided by cycles
causing failure on given stress range level. The Palmgren-Miner linear
damage summation is applied for stress range histogram representing
variable amplitude loading. With S-N curve approach the resulting total
damage D at a certain hot spot yields equation

D = ni ( i ) / C
m

(1)

where the sum with index i is taken over the load histogram of ni cycles
and stress ranges

TIME00:15:00

00:00

00:30:00

00:45:00

01:00:00

01:15:00

01:30:00

01:45:00

02:00:00

02:15:00

02:30:00

02:45:00

03:00:
h

Fig. 4 Time history of global moment at spar hull cross section between
the truss and hard tank sections.

eff

Rainflow
Rayleigh

m
= ni ( i ) / N
i

1/ m

(2)

The fatigue effective loading is defined by a constant stress range

Probability

i . The m and C are the S-N curve parameters.

Fatigue equivalent loading is defined with this equation by writing it

for a fatigue equivalent constant amplitude loading and requiring that it
causes the same damage as the original variable amplitude loading.
This results in equation

eff

and number of cycles N. Either the stress range or number of

load cycles can be selected and the other parameter is then given by the
Eq. (2).

0.1

In the present approach linear relation is assumed between the global

load parameters and the stress ranges. Fatigue effective global load is
then defined with the same equation as the fatigue effective stress
range. The stress parameters in the Eq. (2) are simply replaced by the
global load parameter F.

0.01

Interpolation Procedure

0.001

Range M

Fig. 5 Load cycle histogram for the global moment response at spar
hull cross section between the truss and hard tank sections.

Due to the non-linear load response it is necessary to use non-linear

time domain simulation as a basis for the fatigue analysis. Motion
analysis for a single sea state is relatively small task but analyzing all
sea states from all loading directions is unpractical.

Spectral density S()

The solution is to conduct non-linear motion simulation for a selected

subset of the fatigue sea states and to use these results for estimating
the results for others. Calculation of the fatigue effective loads adds
relatively little for the non-linear time domain simulation of motions.
The fatigue effective loads for the other, non-simulated sea states need
then to be estimated.
The problem reduces to interpolation in two-parameter space defined
by the sea state wave height Hs and peak period Tp. A practical choice
is to use functions with a selected basic form and a polynomial
correction part. The basic form chosen for the load component was

F0 (H s , T p ) = f 0 H s T p

where
0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1.0

1.1

(3)

f 0 is the polynomial correction part and F0 is fatigue effective

global load, which in practice is moment or force in the cross section.

The same basic function was used for moments, shear forces and
normal forces. It is linear in both sea state wave height and peak period.
This was in good agreement with the simulation results for the wave
height dependency. For the peak period the simulation results showed a

1.2

Frequency (1/s)

Fig. 6 Power spectral density for the global moment response at spar
hull cross section between the truss and hard tank sections.

189

more complex dependency but apparently this behavior is difficult to

generalize. A linear first approximation will work well at least for sea
states close to the simulated sea state.

the sea states close to the boundaries of the scatter data were
avoided by the selection of sea states for simulation and the
selection of the basic functional forms. For load responses with
rapid increase for higher sea states the extrapolation towards
short periods and low wave heights sometimes produced
negative values. The values were set to zero for such cases. In
most cases no corrections were required.

The basic form for the number of load cycles was

N 0 (H s , T p ) = n0 / T p

where

n0 is

(4)

the polynomial correction part and N0 is the number of

load cycles corresponding to the fatigue effective load parameter. This

function fitted so well, that correction parts were close to constant in
most cases.
Simulation results

functions limited to 2nd order. The constant

f 0 = f oi Z i

f0

Number of Load Cycles

The correction part was developed using polynomials for both wave
height Hs and peak period Tp. Polynomial fitting in two dimensional
space becomes easily complex as accuracy need to be sustained at the
same time with reasonable functional forms and results at boundaries of
the interpolation area. These problems were solved using polynomial
became
(5)

where the sum is over the simulated sea states i,

function value for sea state i and
was written for

Zi

f 0i

is the correction

the polynomials. Similar equation

Tp (s)

Fig. 8 Load cycles for the fatigue effective moment load. Interpolation
results (black lines) are shown at boundaries of the scatter data.

n0 . The sea states were selected such that they formed

a relatively regular grid in the two dimensional (Hs,Tp)-space. For

several regular grid forms polynomials are readily available with the
property that their value is unity at one grid point and zero for others.
Such technique is frequently used in finite element programs. In the
present application a grid of nine sea states was used, which results in
complete 2nd order polynomials. In practice the polynomials use a local
co-ordinate system. The relation between the local co-ordinates and the
parameter Hs and Tp need to be solved by iteration.

FATIGUE ANALYSIS
Application of the present fatigue analysis procedure is illustrated with
the truss connection detail. This is a safety critical component as it
connects the truss section to the cylindrical hard tank section.

Load Response Processing

An example of the interpolated fatigue effective global bending

moment load, Fig. 7, and corresponding numbers of cycles, Fig. 8, is
given for the loads at truss connection. The interpolated result fits
exactly with the numerical simulation result for the 9 sea states.

Interpolated results

For fatigue analysis a set of sea states are selected for hydrodynamic
load simulation, see Fig. 3. For each simulated sea state the global load
responses, see e.g. Fig. 4, are run through the Rainflow counting
procedure, see e.g. Fig. 5, and processed for fatigue effective loads
using Eq. (2). The fatigue effective loads corresponding to the total
design life are calculated using again the Eq. (2) but this time the sum
is calculated over all sea states in the scatter data. The fatigue effective
load results for the non-simulated sea states are calculated using the
interpolation approach of Eq. (3) to Eq. (5).

Increasing Hs

Fatigue effective moment

Simulation results

The environmental loading probability is typically strongly directional,

Fig. 9. The load response shows much less directionality even at fatigue
damage level with S-N curve exponent 5 as shown in Fig. 9 for the
truss connection bending moment load. The small directionality in the
load response is caused by mooring and riser arrangements and the
directionality in the wind area. Selecting the maximum load response
loading direction for the simulations causes only a small additional
conservatism in fatigue life, see Fig. 9.

Stress Response Processing

Sea state peak period

Fig. 7 Fatigue effective moment load. Red lines show interpolated

results for the period ranges bounded by the scatter data.

For a complete fatigue analysis of the truss connection structure a large

number of structural details must be checked. The stresses are
calculated using a global FEA model, Fig. 10. The local stress response
is calculated using a refined 3D FEA model, Fig. 11, which models the
structural elements to a great accuracy. The relevant fatigue effective

The present interpolation approach gives correct results for the

sea states, which form the basis of the interpolation. Problems at

190

loading is described by the global fatigue effective bending moment

with associated shear force and vertical force load cases.

Load

NW

In fatigue analysis the maximum load response is combined with the

directional environmental loading data. The resulting directionality in
fatigue damage is relatively high, see Fig. 12, and it must be taken into
account. Consequently the stress analysis for the truss connection is
needed for four directional bending load cases and a single vertical
force load case. A single one-quarter FE model for the local detail
analysis is sufficient for covering all stress combinations at the
connection area.

Load response
NE

SW

SE
S

Fig. 9 Relative load probability and relative load response effect in

fatigue damage as function of environmental direction.
Fig. 11 Local detail model for fatigue analysis of the truss connection
to hard tank section.

East

Relative Damage

South-East
North

Sea state Tp (s)

Fig. 12 Relative damage as a function of environmental loading

direction.
Fatigue Strength Definition
The major guidance documents for fatigue strength applied for offshore
installations are UK HSE (1995) and DNV RP-C203 (2000), which
cover general welded joints as well as tubular joints, and API RP-2A
(2000), which concentrates on tubular joints. For the truss connection
analysis the UK HSE (1995) is applied.
The given basic fatigue strength data is very similar in all the
documents. The fatigue strength is defined with an S-N curve and a
local hot spot or geometric stress in all the documents. However, large
differences can be found, when analyzing specific details according to
the different documents.
Comparison of the UK HSE (1995) and DNV RP-C203 (2000)
thickness corrections shows very large effect in predicted fatigue life
for thick plates. For an 80 mm joint the fatigue life prediction differs by

Fig. 10 A global FE model of a truss spar.

191

a factor of 3.5 already due to thickness corrections, Fig. 13. The

comparison is given for S-N curve slope of 5 due to the high design life
of offshore structures, which results in the range of several hundred
millions of load cycles.

In the present procedure the combined damage DC is calculated by

equation

DC = D12 / n + D22 / n + 2 12 D11/ n D21/ n

Thickness penalty in fatigue life (exp. 5)

Penlty factor in life

n/2

(5)

where D1 and D2 are damages from the two different load cases, 12
their correlation and n the S-N curve exponent. The Eq. (5) is an
approximation, which becomes accurate for the Rayleigh model if the
zero crossing periods are the same. In the present application this
condition is almost met. A large difference in zero crossing periods
would be required for the Eq. (5) to become inaccurate.

16

UK HSE(95)
DNV D-curve

13

10

The present damages are calculated using the Rainflow count

procedure. This causes small inaccuracy in the Eq. (5) as the difference
between damage results based on the Rayleigh and Rainflow count
approaches are not necessarily the same for the two load cases.

1
0

20

40

60

80

100

The combined damage can be calculated accurately at time domain

level. Case studies have shown that Eq. (5) is very accurate for the
present application.

Thickness (mm)

Fig. 13 Thickness penalty in fatigue life.

Hierarchic Improvement of Fatigue Life Estimation

The other major difference in the documents is that the DNV RP-C 203
(2000) defines a single hot spot S-N curve applicable for all joints. The
UK HSE (1995), instead, requires using of the classified S-N curves
with hot spot stresses. This causes a difference in fatigue lives of a
factor of over 4 for example for the bracket toe details at the truss
connection.

In the course of the project several fatigue life estimates can be

calculated. The first fatigue life estimate can be based on the nominal
stresses caused by the fatigue effective global loads and estimated
stress concentration factors. This analysis is targeted to selecting the
detail dimensions such that nominal stress level is sufficiently low and
subsequent design changes are restricted to detail design category.

The hot spot stress is defined clearly theoretically but the definition
easily becomes unclear with the multitude of different real welded
joints. Accurate analysis of hot spot stress becomes easily a very
complex task, see e.g. Fricke (2001).

The accuracy of the fatigue life estimate can be improved by using the
global FE model, Fig. 10, results. At the final stage the fatigue life
estimation utilizes the local hot spot stresses. The stress analysis is
conducted in two phases using the global FEA model and the local
FEA model, Fig. 11. The local FEA model has FE mesh density
required for the hot spot stress analysis at various locations.

In the local detail FE models plate thickness sized elements are applied
to model the potentially fatigue critical areas. The hot spot stress is
taken directly at the model node at hot spot. The local stresses at a hot
spot typically vary as a function of load case. A conservative approach
is to calculate maximum principal stress range for each load case and
use these stresses in fatigue analysis regardless of principal stress
direction.

All potentially critical hot spots need to be covered by the analysis.

Majority of the locations are known in advance but additional hot spots
may arise based on the analysis results. The fatigue analysis results
typically cause several minor design changes. Both the additional hot
spots and design changes require revision of the local FE model.
Rerunning the global load analysis and global FEA are not needed.

Apparently the resulting hot spot stress is conservative but it is easy to

process as complexities associated with extrapolations from integration
point results are avoided. The amount of work associated in analyzing a
single hot spot must be minimized as a multitude of hot spots must be
covered. The conservative approach typically results in a small group
of hot spots with too low design life. For these details a more accurate
hot spot stress analysis is made using extrapolation techniques, see e.g.
Fricke (2001).

Similarly changes in the global load response affect only fatigue

effective loads. For a revised fatigue analysis due to changes in global
load response, rerunning of FE analyses is not necessary.

Possibilities for Further Accuracy Improvements

Very high conservatism is included in the fatigue analysis. Sometimes
this results in difficulties in meeting the required design life at certain
hot spots. The present approach offers various options for reducing the
conservatism from the analysis result. For example, the analysis can be
returned to sea state level using two-slope S-N curve. The total fatigue
damage is then calculated with the same interpolation approach.

Calculation of Total Damage

Fatigue damage is calculated separately for the bending and vertical
force load cases. At the truss connection detail the bending load
dominates while the vertical force load causes much lower stresses. The
damage from the non-dominating load case is very small yet it cannot
be neglected as the loads act simultaneously.

The stress responses can be calculated in time domain as the number of

hot spots with high damage is typically very limited. Again the analysis
is conducted for the limited set of sea states and interpolation approach
is applied when calculating the total damages.

For simultaneous load cases the damage sum is calculated with an

adaptation of the combined spectrum approach, see e.g. API RP 2SK
(1996). The standard deviation and zero crossing period for the sum of
two simultaneous processes can be easily calculated. Thus the Rayleigh
model is readily applicable for damage summation, which is the basis
of the combined spectrum approach.

The present procedure has been applied in several spar projects. In the
course of these projects various checks have been made for assuring the

192

accuracy of the fatigue damage estimate. Typically both the

intermediate results as well as the final fatigue life estimates have been
compared with results calculated with more rigorous analysis
procedures. The main conservatism in the present procedure appears to
be in the use of a single one-slope S-N curve. This conservatism largely
disappears if the slope is varied selectively based on sea state severity.

ACKNOWLEDGEMENTS
The authors gratefully acknowledge PI-Rauma for the challenging task
and support for preparing the present paper as well as the contributions
made by the several co-workers at PI-Rauma and VTT. TechnipCoflexip is acknowledged for the permission to publish the paper.
Technip Rauma Offshore, especially Mr. Joonas Miettinen is
acknowledged as the key originator for the present work. Special
acknowledgements are due to Technip Offshore Engineering Inc. for
the comparative analyses and comments in the course of the work.

Analysis Software
In the present work the AQWA and ASAS program packages of
Century Dynamics were used for hydrodynamic load and structural
stress analyses. Majority of the statistical processing of responses were
conducted with software developed by Kukkanen (1996). In addition a
large number of PI-Rauma in-house programs were used.

REFERENCES
API RP 2A-WSD (2000). Recommended Practice for Planning,
Designing and Constructing Offshore Platforms Working Stress
Design, 21st Ed., American Petroleum Industry, Washington D.C.
API RP 2SK (1996). Recommended Practice for Design and Analysis
of Stationkeeping Systems for Floating Structures, 2nd Ed.,
American Petroleum Institute, Washington D.C.
Bangs, AS, Miettinen, JA, Mikkola, TPJ, Silvola, I, and Beattie, SM
(2002). Design of the Truss Spars for the Nansen/Boomvang Field
Development, Proceedings of the Offshore Technology Conference,
OTC 14090, Houston, Texas.
DNV RP-C203 (2000). Fatigue Strength Analysis of Offshore Steel
Structures, Det Norske Veritas, Hvik, Norway.
Fricke, W (2001). Recommended Hot Spot Analysis Procedure for
Structural Details of FPSOs and Ships Based on Round-Robin FE
Analyses, Proc. 11th ISOPE, Stavanger.
Kukkanen, T (1996). Spectral fatigue analysis for ship structures,
uncertainties in fatigue actions, Helsinki University of Technology,
Helsinki.
UK HSE (1995). Offshore Installations, Guidance on Design,
Construction and Certification, 4th ed., Section 21, Steel.

In principal, the present procedure can use results from any similar
analysis programs. However, the models and analyses require quite
massive data transfer, which makes the in-house procedures essential
for effectiveness. The quality assurance becomes an important issue
due to the massive data transfer between programs.
The present procedure provides automatically intermediate results,
which offer themselves as natural checkpoints for the analysis process.

DISCUSSION AND CONCLUSIONS

An effective fatigue analysis procedure is introduced for design
analyses. The approach is especially suited for concurrent engineering
as it allows fatigue damage estimation based on results at early stages
of the project. The accuracy of this estimate can be hierarchically
increased as more results become available.
The developed procedure transforms the load responses from
hydrodynamic time domain simulation to equally damaging constant
amplitude fatigue effective loads with Rainflow counting for load
cycles and S-N curve approach for fatigue. The stress responses are
required only for few quasistatic load cases representing the fatigue
effective loads.
The unavoidable re-analyses become less laborious as the procedure
separates the hydrodynamic load and structural stress analysis. The
effect of changes in either load or stress responses can be incorporated
in the fatigue analysis very easily.
The directional environmental loading data is taken into account in the
fatigue analysis directly as it is defined without any condensing. The
number of sea states in the description does not affect the time required
for the analysis.
The procedure has been successfully applied in several truss spar
design projects, see e.g. Bangs et. al. (2002). The accuracy of the
present approach has been verified within these projects by
comparative fatigue analyses for various details with direct time
domain stress approaches.
The present approach can be relatively easily adapted for other offshore
applications. Even if the stress and load analyses cannot be separated,
the interpolation approach could be used. This alone would drastically
limit the number of sea states required for fatigue analysis without
affecting the accuracy.

193