Marit Irene Kvittem PHD

Doctoral theses at NTNU, 2014:337
Marit Irene Kvittem

Marit Irene Kvittem
Modelling and response analysis for
fatigue design of a semi-submersible
wind turbine
ISBN 978-82-326-0586-6 (printed version)

ISBN 978-82-326-0587-3 (electronic version)
ISSN 1503-8181
Department of Marine Technology

Norwegian University of Science and Technology
Faculty of Engineering Science and Technology
NTNU
Marit Irene Kvittem
Modelling and response analysis for

fatigue design of a semi-submersible
wind turbine
Thesis for the degree of Philosophiae Doctor
Trondheim, November 2014

NTNU
Thesis for the degree of Philosophiae Doctor

© Marit Irene Kvittem
ISBN 978-82-326-0586-6 (printed version)

ISBN 978-82-326-0587-3 (electronic version)
ISSN 1503-8181
Printed by Skipnes Kommunikasjon as

Abstract
Offshore wind is seen as part of the solution for reducing the world’s dependency on non-
renewable sources with CO2 intensive production processes. Developing wind turbines
that float will enable us to harness wind in deep waters, where the winds are strong and
less turbulent than over land. It also means that wind farms can be installed in sites where
there are fewer conflicts of interest than onshore.
Both wind turbines and floating platforms are well known technologies, but combining
them can introduce unexpected interaction effects. A floating wind turbine will experience
large loads from wind, waves and rotating blades. These loads will be influenced by
each other to a larger extent than for fixed wind turbines, and the interaction needs to be
considered in order to predict the loads accurately. By coupling well known aero- servo-
elastic and hydro-elastic codes, which was done as a part of this thesis, the interaction
effects could be accounted for.
The large, fluctuating loads experienced by a floating wind turbine will cause fatigue
damage to the structure, and could be an important design driver. Accurately predicting
the fatigue life requires simulating a large number of load cases, and it also requires long
simulation durations due to slowly varying loads and stochastic uncertainty. This thesis
focuses on methods to reduce the simulation time for fatigue life prediction.
Using a catenary moored, three-column semi-submersible wind turbine with the NREL
5MW turbine mounted on one of the columns as a case study, the following aspects of
FWT fatigue analysis were investigated:
• Frequency domain versus time domain analysis.
• Simulation length, number of seeds and bin sizes for selection of environmental
conditions.
• Misaligned wind and wave directions.
• Morison versus potential theory for hydrodynamic modelling.
A frequency domain method for finding the variance spectrum for tower base bending
moment in moderate environmental was developed and compared to results from coupled
time domain simulations. The results showed that applying a frequency domain method
including an approximation of dynamic excitation of the first flexible bending mode of
the tower and blades, could predict bending moment standard deviations with reasonable
accuracy.
The necessary simulation length can be up to 3-6 hours in extreme value analysis for
compliant offshore platforms, but a long term fatigue analysis showed that reasonable
accuracy can be obtained using shorter simulation lengths and fewer random seeds. It
was also indicated that assuming wind and waves to come from the same direction will
give a conservative fatigue estimate.
i
In all the case studies, fatigue was calculated for nominal stress. As the purpose was
to study the simulation parameters affecting fatigue life, and not the fatigue life itself,
it was assumed that the hot-spot stress amplitudes for the different load components are
proportional to the nominal stress.
Proper representation of wave loads could be obtained using Morison’s equation, provided
that the inertial coefficients were determined separately for each sea state.
ii
Acknowledgements
The research was performed with support from the Norwegian Research Council and in-
dustrial partners through the PhD and post doc programme Norwegian Research Centre
for Offshore Wind Technology (NOWITECH), at the Centre for Ships and Ocean Struc-
tures (CeSOS) at NTNU.
I would like to thank my supervisor, Professor Torgeir Moan, for giving me the opportu-
nity to work with offshore wind turbine research. I have appreciated our discussions, and
his perspectives have always provided valuable input to the work as it progressed.
Thanks should also be extended to Adjunct Associate Professor Zhen Gao, for sharing
his insights in the field of hydrodynamics and fatigue, and for always showing a genuine
interest in helping with big and small questions. Harald Ormberg and Elizabeth Passano
from Marintek deserve gratitude for their work with developing Simo-Riflex-AeroDyn,
and also Tore Holmås and Reza Taghipour for help with the Usfos+VpOne code in the
initial phases of the PhD period. Next, I would like to thank the administration at CeSOS
and Institute of Marine Technology for their work to keep everything sailing smoothly.
I had several fruitful discussions, joint projects and out-of-work experiences with my
fellow PhD candidates, so great thanks to Erin Bachynski (also for agreeing with me
that it was OK to have one more beer on Saturday night even if we were going to work
on Sunday morning), Chenyu Luan, Yihan Xing, Madjid Karimirad, Mahmoud Etemad-
dar, Zhiyu Jiang, Lin Li, Valentin Chaubaud, Morten Pedersen, Wenbin Dong and many
other colleagues at CeSOS and the department of Marine Technology at NTNU. A special
thanks to Zhenju Chuang, who sisterly shared the confined space called office with me for
large parts of my time at MTS (and persistently tried to teach me the basics of Chinese).
I would like to express my gratitude to my parents and sister for their encouragement, and
last, but not least to Jabus for being the highlight of my day, every day (and for help with
the aerodynamics).
iii
iv
Contents
Abstract i
Acknowledgements iii
Contents vii
1 Introduction 1
1.1 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Research context . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.3 Objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.4 Scope . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.5 Papers and declaration of authorship . . . . . . . . . . . . . . . . . . . . 4
1.5.1 Additional papers . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.6 Main contributions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2 Background 11
2.1 Floating wind turbine concepts . . . . . . . . . . . . . . . . . . . . . . . 11
2.1.1 Semi-submersible wind turbines . . . . . . . . . . . . . . . . . . 13
2.2 Design and analysis of floating wind turbines . . . . . . . . . . . . . . . 14
2.2.1 Design load cases . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.2.2 Fatigue limit state for floating wind turbines . . . . . . . . . . . . 16
2.2.3 Model tests . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
2.2.4 Software for dynamic response analysis of floating wind turbines 19
3 Modelling of floating wind turbines 23

3.1 Structural dynamics of FWTs . . . . . . . . . . . . . . . . . . . . . . . . 23
3.1.1 Equation of motion . . . . . . . . . . . . . . . . . . . . . . . . . 23
3.1.2 Non-linear FEM . . . . . . . . . . . . . . . . . . . . . . . . . . 24
3.1.3 Coupled analysis . . . . . . . . . . . . . . . . . . . . . . . . . . 24
3.1.4 Development of a new analysis tool . . . . . . . . . . . . . . . . 25
3.2 Load Modelling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
3.2.1 Hydrodynamic loads . . . . . . . . . . . . . . . . . . . . . . . . 28
3.2.2 Aerodynamic loads by BEM . . . . . . . . . . . . . . . . . . . . 29
3.2.3 Aerodynamic loads by GDW . . . . . . . . . . . . . . . . . . . . 30
v
3.2.4 Thrust force on pitch regulated wind turbines . . . . . . . . . . . 31
3.3 Stochastic modelling of environmental loads . . . . . . . . . . . . . . . . 32
3.3.1 Long term distribution of wind and waves . . . . . . . . . . . . . 32
3.3.2 Short term distribution of wind and waves . . . . . . . . . . . . . 33
3.4 Response to stochastic loads . . . . . . . . . . . . . . . . . . . . . . . . 35
3.4.1 Variance spectrum and frequency-response function . . . . . . . 35
3.4.2 Variance spectrum from timeseries . . . . . . . . . . . . . . . . . 36
3.5 Fatigue in steel . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
3.5.1 Fatigue damage calculation by S-N curves . . . . . . . . . . . . . 36
3.5.2 Cycle counting algorithms . . . . . . . . . . . . . . . . . . . . . 38
3.5.3 Stress calculation . . . . . . . . . . . . . . . . . . . . . . . . . . 38
4 Case studies 41
4.1 Semi-submersible used in the case study . . . . . . . . . . . . . . . . . . 41
4.2 Hydrodynamic modelling in fully coupled simulations (Paper 1 [57]) . . . 42
4.3 Time domain analysis procedures for fatigue assessment of tower and
braces (Paper 2 [59]) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
4.4 Effect of misaligned wind and wave directions on tower fatigue (Paper 3
[8]) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
4.5 Frequency versus time domain for fatigue analysis (Paper 4 [60]) . . . . . 51
5 Conclusions and recommendations for further work 57

5.1 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
5.2 Recommendations for further work . . . . . . . . . . . . . . . . . . . . . 59
References 60
A Appended papers 69
Paper 1: Effects of hydrodynamic modelling in fully coupled simulations of a
semi-submersible wind turbine . . . . . . . . . . . . . . . . . . . . . . . 71
Paper 2: Time domain analysis procedures for fatigue assessment of a semi-
submersible wind turbine . . . . . . . . . . . . . . . . . . . . . . . . . . 85
Paper 3: Wind-wave misalignment effects on floating wind turbines: motions
and tower fatigue load effects . . . . . . . . . . . . . . . . . . . . . . . . 119
Paper 4: Frequency versus time domain analysis for fatigue of a semi-submersible
wind turbine . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
B Abstracts of additional papers 153
C Specifications for a semi-submersible wind turbine with active ballast system157
D Analytical Morison force transfer functions 165
E Standards and guidelines on FLS design 171
vi
F Use of buoyancy compensating force in coupled Simo-Riflex models 175
vii
viii
Nomenclature
Abbreviations RAO response amplitude operator

BEM blade element momentum SCF stress concentration factor
COD co-directional SDOF single degree of freedom system
DLC design load case SSWT single semi-submersible wind tur-
bine
DLL dynamic linked library
TD time domain
DOF degree of freedom
FD frequency domain TLP tension leg platform
FE finite element ULS ultimate limit state
FLS fatigue limit state UNI uni-directional
FWT Floating wind turbine Greek characters
LWT land based wind turbine κ(t) retardation function
MBS multi body simulation λ wave length
Mor. Morison ω circular frequency
MSL mean sea level ρa air density
MUL multi-directional ρw seawater density
MWL mean water level σ axial stress
NSS normal sea state τ time lag or shear stress
NTM normal turbulence model Ω rotor speed

NWLR normal water level range Latin characters
NWP normal wind profile model a’ radial induction factor
PDF probability density function a axial induction factor
PF potential flow d cylinder diameter
QS quasi-static dD drag force
ix
j stress range bin no. j dF Horizontal Morison force per unit
length
dM pitch moment
Fmoor force vector from mooring
m slope parameter for S-N curve
Fturb force vector from turbine
ni number of stress cycles
r, ṙ, r̈ FEM global nodal displace- F
wave
wave force vector
ment/velocity/acceleration vector Fwind wind force vector
t time MH FEM global added mass matrix
u,u̇ wave particle velocity/acceleration H wave height
u wave particle velocity vector Hs significant wave height
x,ẋ,ẍ local body displacement/velocity/ K material parameter for S-N curve
acceleration
KS FEM global stiffness matrix
x, ẋ, ẍ body motion/velocity/acceleration
vector dL lift force
A added mass matrix M mass matrix
A outer area or cross section area of MS FEM global mass matrix
cylindrical cross section
Mx torsional moment
B linear damping matrix
My , Mz bending moment
C restoring matrix
Ni number of stress cycles to failure
Cd quadratic drag coefficient
Nbins number of stress range bins
Cm added mass coefficient
Ip polar moment of area
Cq quadratic drag coefficient
Iy ,Iz second moment of area
CS FEM global damping matrix
Nx axial force
CX (τ ) correlation function for process X
R EXT
FEM external forces
D accumulated fatigue damage
S stress range
D diameter cylindrical cross section
SX (ω) variance spectrum for process X
F force vector
Urel relative velocity
A∞ added mass matrix high frequency
limit Tp wave peak period
FD diffraction force vector U mean wind speed
F FK
Froude Krylov force vector Vy , Vz shear force
x
Chapter 1
Introduction
The current chapter gives an overview of the motivation, research context, objectives and
scope of the work, lists the papers that form the basis of the thesis and the contributions
to the research community. The remaining chapters are devoted to:
Chapter 2: More about the background of the investigations and a review of previ-
ous work in the field.
Chapter 3: Theory behind modelling and response analysis analysis for fatigue de-
sign of floating wind turbines.
Chapter 4: Results from the case studies performed as a part of the thesis.
Chapter 5: Conclusions from the case studies and suggestions for further work.
The thesis consists of a collection of 4 papers, which are appended in App. A.
1.1 Motivation
The wind resource is renewable and the process of harnessing wind energy involves very
small CO2 emissions compared to other sources of electrical power. This is why wind
energy is seen as an important contributor to reaching the European Union’s Renewables
Directive; at least 20% of the member countries’ total energy consumption should be cov-
ered by renewable sources of energy within the year 2020. Limitations in available wind
farm sites on land pushes the industry to go offshore, in fact, offshore wind is currently
one of the fastest growing maritime sectors [3].
In deep water (> 45m), it is expected that floating wind turbines (FWTs) will be less ex-
pensive than bottom supported turbines [79]. A floating substructure also means that wind
farms can be installed farther from shore, and thus eliminate problems with visual con-
tamination and noise. Another important advantage of turbines far from shore is stronger
and less turbulent wind. Floating wind turbines are comparable to other offshore wind
1
turbines when it comes to environmental impact, and they have a much smaller impact on
global warming than for example a natural gas power plant [103]. Among countries that
have most of their offshore wind resources in deep water (more than 100 meters) are the
US, Japan, Norway and Portugal [69]. However, FWTs is a young field of research, and
no commercial turbines or parks with more than one unit have been installed at the time
of writing.
Locating wind turbines in deep water areas with strong wind will necessarily impose
large, irregular loads from wind, waves and current on the turbine and its support struc-
ture. Due to difficult access, inspection and repair is expensive, and high reliability of the
FWT is required for it to be economically feasible. Also, since wind parks consist of sev-
eral units with the same design and fabrication, the turbines are vulnerable to “common
cause” failures. This is why precise prediction of extreme load effects and fatigue damage
is important. Numerical modelling and analysis as well as wave tank and prototype test-
ing can be used for this purpose, preferably all three together, in the concept development
phase.
Floating wind turbines have to be designed to fulfil criteria with respect to both function
and safety. Sufficient resistance of the material, connections and other components has
to be ensured, in addition to floater stability. This is covered by the Ultimate Limit State
(ULS) criterion, the Fatigue Limit State (FLS) criterion and the Accidental Limit State
(ALS) criterion. Functional criteria are covered by the Serviceability Limit State (SLS),
which involve requirements to for instance power production, corrosion protection and
nacelle accelerations. The focus in this thesis will be on motion and fatigue failure.
Fatigue Limit State (FLS) refers to failure during cyclic loading, and the fatigue life is
defined as the predicted time before failure due to fatigue. Wind turbines in offshore
environments will experience cyclic loading and dynamic resonance effects from wind,
waves and rotating machinery, and thus fatigue must be addressed. In order to give a
realistic estimate of the fatigue life, stress histories must be predicted for all relevant
environmental conditions that can be experienced by the FWT throughout its lifetime.
This is a challenging task, due to several causes:
• Simulation tools for FWTs are in the developing stage.
• There is limited knowledge of the load effects on FWTs and the implications for
fatigue.
• A large number of design load cases and environmental conditions have to be in-
cluded in the FLS analysis.
• The return periods typically used as basis for wind- and wave statistics are different.
• Time domain analysis are recommended due to non-linear effects and coupling be-
tween response to wind and wave loading, and time domain analysis can be very
time consuming.
Since FWTs are new to both the research community and the industry, so far only limited
2
FLS studies have been performed. This thesis focuses on the establishing and applying
numerical analysis tools in the FLS design of floating wind turbines, seeking efficient
methods to predict the fatigue life.
1.2 Research context
This research was performed as a part of the PhD and post doc programme Norwegian
Research Centre for Offshore Wind Technology (NOWITECH), work package 3 - Novel
substructures for offshore wind turbines. The work was carried out at Centre for Ships
and Ocean Structures (CeSOS) at NTNU in Trondheim.
1.3 Objectives
Response analysis for FLS design of FWTs has to cover many design load conditions
(DLC), and within each DLC many environmental conditions. Stress analysis for FLS
involves modelling non-linear and highly dynamic responses to both deterministic and
stochastic loads, and requires sophisticated analysis tools and long simulation time. The
author is not aware of any full FLS design of a FWT to this date (a review is given in
Chap. 2), which is presumably because it is such a comprehensive task. For this reason
it is important to assess the necessary level of refinement of the analyses, i.e. modelling
details, number of load cases and analysis methods. Also, since there are many limitations
to how much detail numerical models can include in simulation tools, it is important to
check the validity of different theories and compare them.
Two questions are put forward here, where RQ1 covers a broad aspect of topics and RQ2
is partly a sub-question of RQ1. The underlying objective of these questions is to look at
areas where simplifications of the response analysis can be made, without compromising
the quality of the fatigue predictions. The following research questions form the basis for
the thesis:
• RQ1: How should loads and structural response be modelled in order to properly
predict the fatigue life of a floating wind turbine?
• RQ2: Can simplifications in the analysis be made in order to make the FLS analysis
more efficient?
At the time this project started, there were no standards or guidelines for design of FWTs.
During the course of this project, a few guidelines have been published [1, 27], but these
reflect the lack of experience with mass production of FWT units and long term operation
and maintenance of FWT farms. The reason for this is simply that the development of
FWT is still in the research stage; only a couple of prototypes exist world wide, and no
commercial wind farms using floating foundations for turbines have been installed yet.
3
This thesis aims to contribute to design guidelines, by studying the global mechanisms
that cause fatigue and by trying to come closer to defining a set of requirements for FLS
design.
1.4 Scope
The research questions were investigated using a semi-submersible wind turbine as the
basis for case studies, and focus was kept on fatigue in the support structure. The detailed
studies described in the papers cover:
• Requirements for hydrodynamic modelling. Morison versus potential theory with
diffraction for calculation of wave forces on a SSWT.
• Effect of simulation length, stochastic sample and bin size for selection of environ-
mental conditions for response analysis for FLS design.
• Effect of misaligned wind and waves in fatigue analyses.
• Linearised frequency domain analysis versus non-linear time domain analysis for
fatigue damage calculation.
A part of estimating fatigue damage is developing a numerical model to perform the
response analyses with. Development of FWTs was at an early stage, and no computer
codes suitable for the purpose existed at the beginning of this project. Thus, developing a
numerical tool and building a model in this also became part of the scope of this project.
The title of the thesis is Modelling and response analysis for fatigue design of a semi-
submersible wind turbine since the main focus is on the global dynamic analyses. Thus,
no local stress concentration factors (SCF) have been used in the investigations.
A semi-submersible floater with a single wind turbine (SSWT) placed on one of the three
columns was used in the study. The columns are connected by braces and four catenary
mooring lines. The word “turbine" is used throughout the thesis to describe the tower and
the rotor-nacelle assembly.
1.5 Papers and declaration of authorship

The following papers [57, 59, 8, 60] form the scope of this thesis.
4
P1 Marit I. Kvittem and Erin E. Bachynski and Torgeir Moan: Effects
of hydrodynamic modelling in fully coupled simulations of a semi-
submersible wind turbine, Elsevier Energy Procedia, vol 24, p.351-362,
2012
Relevance to this thesis: This paper presents the first analyses per-
formed with Simo-Riflex-AeroDyn. It also compares the response of a
SSWT to Morison type wave forces and the response to potential theory
forces including diffraction.
My contribution: This paper is the result of a cooperation with Erin E.
Bachynski. I laid out the basic structure of the coupling between Riflex
and AeroDyn, did some contributions to the programming and testing
of the software and performed the wave-only analyses. The coupled
analysis was a collaborative effort.
P2 Marit I. Kvittem and Torgeir Moan: Time domain analysis procedures

for fatigue assessment of a semi-submersible wind turbine, under review
for Journal of Marine Structures, 2014
Relevance to this thesis: The paper describes a long-term fatigue anal-
ysis of a SSWT tower and braces, focusing on the effect of simulation
duration, number of realisations and bin size for selection of environ-
mental conditions for FLS design.
My contribution: I performed a new design of a SSWT based on the
WindFloat concept to obtain more detailed information than what was
available. All the analyses and the post processing of the results were
carried out by me.
P3 Erin E. Bachynski, Marit I. Kvittem, Chenyu Luan and Torgeir Moan:

Wind-wave misalignment effects on floating wind turbines: motions and
tower fatigue damage, to be published in Journal of Offshore Mechanics
and Arctic Engineering in November 2014
Relevance to this thesis: The paper compares the effect of misaligned
wind and waves on short term fatigue damage for four different FWT
concepts; a spar, a TLP and two SSWT concepts.
My contribution: The study was a cooperation between Erin E.
Bachynski, Chenyu Luan and myself. The paper was initiated and writ-
ten by E. E. Bachynski. C. Luan and I participated in determining the
analysis methodology and discussing the results. I performed the anal-
yses for one of the SSWT concepts.
5
P4 Marit I. Kvittem and Torgeir Moan: Frequency versus time domain for
fatigue analysis of a semi-submersible wind turbine, accepted for pub-
lication in Journal of Offshore Mechanics and Arctic Engineering 2014
Relevance to this thesis: The paper describes a linearised frequency
domain method to estimate tower base bending moment and fatigue
damage for a SSWT and compares the results to non-linear coupled
wind- and wave simulations.
My contribution: I developed a methodology to calculate bending mo-
ments from a frequency domain model of platform motions and to rep-
resent flexible dynamics of the turbine and performed the analysis.
1.5.1 Additional papers
The following papers [58, 61, 7, 77] are related to the work with FWT analysis, but are
not included as a part of this thesis. The reasons for not including the papers are that
they fall outside of the main topic of the thesis or that they are seen as preliminary studies
to the more complete papers 1-4. Since the papers still could be of interest to the FWT
research community, abstracts for the secondary papers are given in App. B.
P5 Marit I. Kvittem, Torgeir Moan, Zhen Gao and Chenyu Luan: Short-
term fatigue analysis of semi-submersible wind turbine tower, Proceed-
ings of the 30th International Conference on Ocean, Offshore and Arctic
Engineering, Rotterdam, The Netherlands, 2011
P6 Marit I. Kvittem and Torgeir Moan: Effect of mooring line modelling

on motions and structural fatigue damage for a semi-submersible wind
turbine, Proceedings of the twenty-second International Offshore and
Polar Engineering Conference, Rhodes, Greece, 2012
P7 Erin E. Bachynski, Mahmoud Etemaddar, Marit I. Kvittem, Chenyu

Luan and Torgeir Moan: Dynamic analysis of floating wind turbines
during pitch actuator fault, grid loss, and shutdown, Elsevier Energy
Procedia, vol 35, p.210-222, 2013
P8 Torgeir Moan, Zhen Gao, Madjid Karimirad, Erin E. Bachynski, Mah-

moud Etemaddar, Zhiyu Jiang, Marit I. Kvittem, Made Muliawan and
Yihan Xing: Recent developments of the design and analysis of floating
wind turbines, ISCOT: Developments in fixed & floating offshore struc-
tures, Busan, South Korea, The Royal Institution of Naval Architects,
2012.
6
1.6 Main contributions
The main contributions of this thesis to the floating wind turbine research community are
listed in this section. The relations between research questions, papers and contributions
are illustrated in Figure 1.1.
C1 A numerical tool for coupled analysis of floating wind turbines. Ex-

isting softwares Riflex and AeroDyn were coupled through a dynamic
linked library (DLL).
The DLL developed by PhD candidate Erin E. Bachynski and the author provided a cou-
pling between state-of-the-art simulation tools and the possibility of performing analysis
with a turbulent wind field, where turbine and platform response is fully coupled. The
performance of the tool was verified for a land based turbine and a spar wind turbine with
good agreement with other wind turbine analysis codes [84].
C2 Development of a simplified method to assess fatigue damage for the

tower base of a SSWT based on linear frequency-domain methods.
The linearised frequency domain analysis developed to calculate variance spectra for
tower base bending moment gave bending moments with reasonable accuracy for a lim-
ited selection of intermediate environmental conditions. A generalised degree of freedom
model applied to account for excitation of the first flexible mode of the turbine increased
the accuracy. Applying the bending moment model for calculation of fatigue gave in-
creased errors due to the exponential relationship between stress ranges and fatigue dam-
age in a Wöhler curve approach.
C3 Knowledge about the effect of diffraction forces for a SSWT with col-
umn diameters of 10 m and heave plates, and a proposed method to
numerically determine the added mass coefficients in the Morison type
forces.
It was found that for the SSWT studied in the thesis, the relative importance of viscous
forces and diffraction forces is comparable to what can be found for a fixed cylinder,
provided that the correct inertial coefficients in the Morison formulation are applied. For
a fixed cylinder, diffraction effects become important for diameter to wavelength ratios
above 0.14. Inertial coefficients had to be chosen based on the wave period for short
waves, and could not be assumed to be the same for every wave condition. Analytical
force transfer functions for including dynamic pressure in Morison models for a three
column semi-submersible were derived and applied with success.
7
C4 Investigations of required simulation lengths, number of realisations
and bin size for selection of environmental conditions for FLS design
of FWTs.
The accuracy of using shorter than 3 hour simulation length and fewer than 10 stochastic
seeds for response analysis for long term fatigue damage for a semi-submersible was
studied. Fatigue damage in the tower base was found to be less sensitive to simulation
length than in the main beams, pontoons and braces, but calculating damage based on
1-hour simulations gave less than 4% error compared to using the 3-hour simulations.
Reducing the number of random seeds also gave acceptable accuracy compared to using
10 random samples.
C5 Knowledge about the effect of applying wind and waves from different
directions on motions and in FLS design of different concepts of FWTs.
Wind and waves coming from the same direction gave higher fatigue damage than the
misaligned cases. Even though increased wave induced motions were observed for mis-
aligned cases, the fact that wind and wave forces acted around different axes gave a better
distribution of stress around the circumference of the tower and thus reduced the fatigue
damage. The concept comparison showed that a semi-submersible with large mass per-
formed best with respect to motions and fatigue damage compared to a semi-submersible
with less inertia, a TLP and a spar buoy.
8
RQ1:
Modelling
Hydrodynamic Linearised
Numerical tool
model model
P1 P2 P3 P4
C1 C3 C4 C5 C2
Simulation Reduce number

Frequency
length, number of environmental
domain
of seeds conditions
RQ2:
Efficient
analysis
Figure 1.1: Relations between research questions, papers and contributions.
9
10
Chapter 2
Background
The general background for the investigations is described in this chapter. A brief intro-
duction to the various types of foundations for floating wind turbines is given, followed
by a review of the state-of-the-art in the field of design and modelling of floating wind
turbines, with emphasis on fatigue design and numerical simulation tools.
2.1 Floating wind turbine concepts
Development of floating concepts is still at an early stage. There are currently three full
scale prototype floating wind turbines in operation; Hywind spar in Norway [69, 96],
WindFloat semi-submersible in Portugal [69, 93, 92] and a downwind semi-submersible
off the coast of Fukushima, Japan [83]. There are also a number of concepts installed in
smaller scale in wave basins and in full scale environments [69].
In principle, most proposed concepts combine known technology for onshore turbines
mounted on floating substructures known from offshore oil and gas technology. Some of
the suggested floating support structures are [99]:
• Ballast stabilized spar-buoy with catenary mooring with one turbine (Hywind)
• Single tension leg moored spar with one turbine (Sway)
• Multiple tension leg moored semi-submersible with one turbine (Blue H, Pelastar)
• Semi-submersible with catenary mooring and one or multiple turbines (WindFloat,
Winflo, HiPR Wind, WindSea)
While most proposed concepts use horizontal axis upwind turbines, some use less tradi-
tional turbine technology, like two-bladed rotors (Winflo) or vertical axis turbines (Verti-
wind). There are also concepts that combine wave energy converters with wind turbines
on floating platforms (Pelagic Power, Poseidon).
11
(a) Winflo (b) CSC
(c) Hywind (d) Pelastar
(e) HiPRWind (f) WindFloat
Figure 2.1: Floating wind turbine concepts. Images were downloaded from project web
pages [88, 21, 4, 96] or received from project owners (Principle Power and CSC braceless
semi-submersible from NTNU).
12
Due to the cost of a floating foundation, it is expected that floating platforms must support
turbines with an effect larger than what is normal for onshore and offshore bottom fixed
turbines today. Although the full scale prototypes do not carry turbines larger than 2.3
MW, it is normal to assume at least 5 MW for commercial installations. In 2009 Jonkman
et. al [49] developed data for a generic 5 MW horizontal axis upwind turbine (NREL 5
MW baseline turbine), which has been widely used in the research community for analysis
of offshore wind turbines. In the IEA Offshore Code Comparison Collaboration Phase IV
(OC3) [42], where numerical aero-hydro-servo-elastic codes were compared in analysis
of a spar wind turbine, a tower suited for floating wind turbines was developed. This tower
is 10 m shorter than the original NREL 5 MW turbine, and is assumed to be mounted 10
m above sea level, to maintain a hub height of 90 m.
2.1.1 Semi-submersible wind turbines
A semi-submersible platform achieves stability by water plane stiffness, but also to some
extent by ballast in the columns or pontoons. Compared to a spar buoy, that obtains sta-
bility by having a deep draught ballasted substructure, it has a significantly lower draught,
and is thus a more flexible concept when it comes to suitable water depths, installation
and transport. A spar WT either needs deep harbours and seaways to be towed from the
yard to the installation site, or a more sophisticated transportation- and installation method
than what exists today. The shallow draught of a semi-submersible makes it suitable for
intermediate water depths, and for shallow water assembly yards.
The case studies in the thesis were carried out for a three-column SSWT with the NREL
5MW on the OC3 Hywind tower [42] placed on one of the columns, a concept inspired by
the generic WindFloat [93] (see Figure 2.1). The station keeping system for the generic
WindFloat consists of four catenary mooring lines anchored to the seabed, where two of
the mooring lines are connected to the column with the turbine, and one mooring line is
connected to each of the other two columns.
The WindFloat platform has an active ballast system to keep the turbine upright (see
Figure 2.2) and thus maximise power output as the wind direction and intensity change.
This is done by pumping water ballast between the columns, and the system has a reaction
time of 20 minutes to significant changes in the mean wind speed and direction.
Another semi-submersible WT that has been frequently used in FWT studies, for instance
in IEA Offshore Code Comparison Collaboration Phase II (OC4) [91], is the DeepCwind
concept. Compared to WindFloat and the semi-submersible studied in this thesis, it also
has three main columns and catenary moorings, but the turbine is placed on an additional
central column, and the displacement is close to three times higher. The DeepCwind
concept was used in the study of misaligned wind and wave conditions in Paper 3 [8] ,
together with the SSWT used in Papers 1-4 [57, 59, 8, 60] and two other concepts.
13
(a) Unballasted platform (b) Ballasted platform
with turbine in operation. with turbine in operation.
Figure 2.2: Illustration of ballast system.
2.2 Design and analysis of floating wind turbines
2.2.1 Design load cases
Since the field of floating wind turbines is new, only design of prototypes has been per-
formed. This means that designs may not be optimised designs, since the prototypes are
built to gain experience. Also, since there were no guidelines or standards that applied
to FWTs until recently, there was no defined set of load cases to run through in a design
process.
As mentioned in Chap. 1, FWTs have to be designed after Ultimate Limit State (ULS),
Accidental Limit State (ALS), Fatigue Limit State (FLS) and functional requirements (of-
ten grouped under Serviceability Limit State). The largest load experienced under harsh
environmental condition is normally defined under ULS, whereas loads due to accidental
events, for instance ship collisions or explosions, fall under the definition of an ALS. FLS
covers accumulated crack growth under cyclic loading in normal environmental condi-
tions.
The IEC 61400-3 standard for bottom fixed offshore wind turbines specifies a minimum
set of design load conditions for ULS and FLS. These consist of the conditions power pro-
duction, power production plus fault, start up, normal shut down, emergency shut down,
parked, parked plus fault and transport assembly and repair. For each of these conditions,
a subset of load conditions with the wind and wave condition and directionality, currents,
water level, limit state and partial safety factor for each of the load conditions in the sub-
set. For the full table of load cases, refer to [40, Chap. 7]. The load cases valid for FLS
are shown in Table 2.1. All the requirements for bottom fixed wind turbines are not neces-
sarily relevant for FWT design, while additional requirements apply in the DNV Offshore
14
Standard [27], for example requirements with regards to stability, mooring and anchoring.
The compliant nature of an FWT means that the recommended values for parameters like
length of extreme operating gust, wind and wave return periods or wave spectrum can
not be transferred directly from bottom fixed design codes. Other design load cases than
those specified in the IEC standard [40] may be governing.
Table 2.1: Design load cases in IEC [40] relevant for fatigue limit state.
Name DLC Wind Wave Wind/wave Current Water

dir. level
Power 1.2 NTM NSS COD, No NWLR

production joint PDF MUL current or ≥ MSL
Power 2.4 NTM NSS COD, No NWLR

production Hs = UNI current or ≥ MSL
plus fault E[Hs |Vhub ]
Start up 3.1 NWP NSS(or NWH) COD, No NWLR

Hs = UNI current or ≥ MSL
E[Hs |Vhub ]
Normal 4.1 NWP NSS(or NWH) COD, No NWLR

shutdown Hs = UNI current or ≥ MSL
E[Hs |Vhub ]
Parked 6.4 NTM NSS COD, No NWLR

joint PDF MUL current or ≥ MSL
Parked 6.4 NTM NSS COD, No NWLR

plus fault joint PDF MUL current or ≥ MSL
Transport, 6.4 NTM NSS COD, No NWLR

assembly, joint PDF MUL current or ≥ MSL
maintenance
and repair
Jonkman [51] ran through the ULS DLCs for power production in the IEC 61400-3 stan-
dard for one FWT (barge) and one land based wind turbine (LWT), and compared the
results. No capacity evaluation was performed, but quantification of the ratio between
floating and land based response parameters showed that design loads were expected to be
higher for the barge FWT than for the LWT. Pitch motion and inertia loads were pointed
out as the driver for these differences. He also showed that the design driving load condi-
tions are not necessarily the same for FWTs and LWTs; i.e. the normal turbulence models
15
were worse than the extreme operating gusts for the barge, whereas it was vice versa for
the LWT turbine.
A more comprehensive concept comparison running through both ULS and FLS for the
normal operational conditions of the IEC-standard for six different FWT concepts was
performed by Robertson and Jonkman [90]. The conclusion was the same as in [51], that
increased pitch and roll motions increase the design loads.
Karimirad and Moan [54] compared tower forces for a spar FWT in the operational con-
dition with maximum thrust force, to idling conditions in a harsh environment, and found
that the forces can be 2.3 times higher for the harsh environment condition, due to large
wave induced response in harsh sea states. Jiang et. al. [44] compared turbine transient
loads due to blade pitch system fault, grid loss, and shutdown events for a spar and for a
land based turbine. The work indicated that the increase in tower and blade forces during
a fault event was smaller for a FWT than for a LWT.
2.2.2 Fatigue limit state for floating wind turbines
Fatigue damage is known to be a problem for bottom fixed offshore wind turbine substruc-
tures, and is also expected to be significant for floating wind turbines (FWTs). Adequate
fatigue strength should be ensured by design as well as by inspection and repair. Wind
parks consist of units with similar designs, and are thus vulnerable to “common cause"
failures, which means that the economic consequences of poor fatigue design are serious.
It is important, therefore, to make good fatigue estimates early in the design process.
How comprehensive a fatigue damage analysis of a structure is, depends on the number
of load cases, the nature of the structural response to varying loads, the level of stochastic
uncertainty in the load models, the complexity of the detailed geometry and the response
analysis tool.
Unlike structural failure during extreme load conditions, fatigue cracks can also grow
under benign environmental conditions, which means that every event that can possibly
cause crack- initiation and growth must be included in the analysis. For a traditional off-
shore structure, that does not have significant dynamic response to turbulent wind, envi-
ronmental conditions for fatigue analyses are taken from scatter tables of two parameters;
significant wave height and wave peak period. When wind is introduced as a third param-
eter, environmental conditions are found in a scatter block, which dramatically increases
the number of load cases. Other parameters that further increase the number of load cases
necessary for fatigue analysis are wind- and wave directions, current and operational state
of the turbine.
To make accurate predictions, fully coupled time domain analyses should be performed,
but this is a challenging task. Firstly, at the beginning of this project, the number of
available and suitable analysis tools was limited. Secondly, such analyses are quite time
consuming, and a fatigue analysis ends up altogether being an extremely time consuming
16
process. This is why it is important both to assess the validity of the theories available in
computer codes today and to investigate the impact of measures for reducing the compu-
tation time for fatigue analysis.
There has been a couple of FLS studies for FWTs so far. As mentioned in the previous
subsection, Robertson and Jonkman [90] found that floater motion gave increased fatigue
damage compared to the land based case. This effect was particularly visible for fatigue
damage in the tower base. Haid et. al [35] studied the effect of simulation length on
fatigue and ultimate loads for the OC3 spar buoy wind turbine, and concluded that the fa-
tigue damage in the tower, blades and mooring system was more sensitive to the treatment
of residuals in rainflow cycle counting than to simulation length. Recent studies of a spar
FWT by Jiang et al. [45] showed that the contribution to blade, tower and mooring line
fatigue damage in transient fault conditions can be minimal if the appropriate shutdown
procedure is used.
A structural analysis including fatigue of the WindFloat platform was presented in [5,
92]. An analysis covering 12 different sea states gave a fatigue life of 36 years for a
stress concentration factor (SCF) of 2.0 for the braces in the platform and a fatigue life
of 15 years for an SCF of 6.0 for the tower base. No SCF based on weld geometry
was determined in these analyses, but a sensitivity study calculating fatigue based on
different SCFs was performed. These calculations did not include the cyclic rotor loads
from turbulent wind and rotating blades, but the study showed that fatigue may be a
critical limit state for floating wind turbines.
The FLS studies that have been performed so far, do not contain the full number of en-
vironmental conditions that the IEC-standard specifies. In [90, 75, 35] only one set of
waves is combined with each wind speed, whereas the standard specifies that a bin sizes
of 2 m/s wind, 0.5 m for wave heights and 0.5 s for wave period shall be applied to the
joint probability distribution [40].
Frequency domain analysis for fatigue
Fatigue analysis for wind turbines is normally performed in the time domain. For a semi-
submersible WT, interaction between rotor and platform motion and non-linearities due to
viscous forces, catenary mooring lines and large deflections make time-domain analysis
necessary. However, the solution process of a non-linear time domain finite element sys-
tem with many degrees of freedom is very time consuming, and is not ideal for long term
fatigue analysis, which requires consideration of many environmental conditions. Fre-
quency domain methods are significantly faster and are also recommended in guidelines
[24] for fatigue analyses of oil platforms subjected to wave loads.
Frequency domain methods have also been applied in the fatigue analysis of bottom fixed
offshore WTs. Kühn [55] investigated simplified analysis methods for fatigue calculation
for offshore bottom fixed monopile turbines, where the wave response was performed by
linear spectral analysis. Van der Tempel [100] performed support structure response anal-
17
yses in the frequency domain with spectral analysis for both wind and wave response, and
got acceptable accuracy compared to time domain simulations and full scale measure-
ments. In these studies, the wave and wind loads were treated separately, which can be
a valid assumption for a fixed, but not necessarily for a floating wind turbine, since there
is a stronger interaction between wave- and wind induced responses. Both Kühn and Van
der Tempel emphasized the importance of including the aerodynamic damping due to an
operating turbine in simplified modelling.
A couple of studies have used frequency domain for FWT analysis. Bachynski [9, 6]
applied the procedure of Wayman et. al [102], added a wind response model, and studied
the applicability of frequency domain methods in the early concept development phase of
tension leg platform wind turbines. The conclusion of these studies was that the accuracy
was too low to use for concept development, since they gave a poor representation of
low-frequent response and since flexible modes of tendons, tower and blades were not
considered. The wave frequency motions and accelerations were captured quite well.
A comparison between frequency and time domain wave response analyses for a barge
type floating wind turbine was conducted by Philippe et al. [86], with good results for
surge, pitch and heave motions. Other studies applying frequency domain analysis for
floating wind turbines have been reported in [14, 67, 101].
The above mentioned studies for FWTs did, however, not consider fatigue. Although the
frequency domain method has limitations, in particular for FWTs, it can be a useful tool
in performing early stage design assessments. Such an approach can be easier to under-
stand than numerical models for time domain analysis, and makes isolating the effect of
different physical factors easier.
2.2.3 Model tests
A first step in design is to have an analysis tool that can be used to run all the necessary
load cases. Such a tool needs to be verified in order to be suitable for design, both against
scaled model tests and full scale measurements. However, since the hydrodynamic Froude
scaling, geometric similarity and the aerodynamic Reynolds scaling are difficult to match
in the same model, scaling up the results properly is difficult. However, efforts have been
made to get reasonable scale models [34, 19].
A 1:47 scale model test for the Hywind concept was performed in the Marintek ocean
basin. These model tests were first compared to Simo-Riflex coupled to a simplified thrust
force model by Nielsen et. al. in 2006 [81]. In 2007 Skaare et al. [95] compared these
model tests to a model with state-of-the-art aerodynamics by using Simo-Riflex coupled
to Hawc2, with good agreement.
In a paper by Goupee et al. [34], a concept comparison of three concepts of FWTs (spar,
TLP and semi-submersible) is performed with scaled model tests (1:50) in the Marin
ocean basin. The University of Maine DeepCwind program followed up these model tests
18
by using them for calibrating simulations in Fast [19, 97]. The semi-submersible used in
these studies was the DeepCwind platform.
2.2.4 Software for dynamic response analysis of floating wind tur-

bines
Numerical analysis tools that have been applied in FWT research so far has been devel-
oped from existing tools, either software for land based turbines or software for offshore
structures. Other tools have been developed from scratch, but do not contain as many
features, and still have a way to go on validation. The most widely used software is de-
scribed below. During the course of this thesis, a number of codes under development
have emerged; Table 2.2 shows the current status, as of December 2013, of a selection of
available computer codes. At the beginning of this project, none of the existing computer
codes were found sufficiently sophisticated to model the SSWT in this project, so a new
tool was developed. This will be further described in Sec. 3.1.4.
A code-to-code comparison for FWT software is currently beeing performed in the IEA
Task 30 - OC4 Phase 2 [89], for a semi-submersible wind turbine. Task 23 [50] did the
same for a spar FWT. Worth mentioning is also the Upwind project, where the capabilities
of the FWT softwares that were available in 2010 [18] were summarised.
Two computer codes for dynamic response analysis of FWTs that originate from wind
turbine software (Fast and Hawc2) and two that originate from offshore structures anal-
ysis software (Simo-Riflex and Usfos) are described below. The descriptions contain the
history of development and the status of capabilities of their released version as per De-
cember 2013.
Fast
Fast is an open source, publicly available aeroelastic simulation code for two- and three-
bladed horizontal axis turbines [43] developed at the National Renewable Energy Lab-
oratory (NREL). It can be coupled to the commercially available FE structural solver
Msc.Adams. The aerodynamic forces are calculated by the module AeroDyn, which pro-
vides a BEM with various corrections (see [78]) and a GDW option. The structural dy-
namics are described by a combination of multibody dynamics and modal superposition
of pre-calculated mode shapes with small deflections. The number of global degrees of
freedom (DOFs) for the entire system in Fast (when it is not coupled to ADAMS) is lim-
ited to 24 for a three-bladed turbine. This means, for instance, that the tower lacks the
axial and twist DOF, and the blades lack the twist DOF.
In 2004, Withee [104] developed a model for coupled analysis of hydrodynamic and aero-
dynamic loads using Fast and structural solver Msc.Adams and wave forces based on slen-
der body theory. The hydrodynamic module was later expanded in a cooperation between
19
Table 2.2: Selected computer codes for analysis of floating wind turbines and their current
features as of December 2013 [43, 66, 73, 78, 89]. See the Nomenclature for abbrevia-
tions.
Name Company Structural Aero. Hydro. Mooring Comment

model model model model
Simo- MARIN- FE BEM PF+ FE Currently no

Riflex TEK Mor. yaw bearing.
no nacelle drag
Fast NREL Modal BEM+ PF+ QS No dynamic

GDW Mor. wake correction
for BEM
Usfos+ Sway FEM BEM Mor. FE

VpOne
Fast+ NREL FEM BEM PF+ FE De-coupled

OrcaFlex Mor. aero- and
hydrodynamics
3DFloat IFE FEM BEM Mor. FE No dynamic

stall model
Hawc2 DTU MBS BEM Mor. MBS
20
Massachusetts Institute of Technology (MIT) and NREL to include large volume body
dynamics, by applying input generated by potential theory software Wamit. This module
was included as a Fast module and named HydroDyn. HydroDyn applies the wave forces
in the 6 platform DOFs in Fast. A quasi-static non-linear mooring line model, based on
look-up tables, was added later [52].
Simo-Riflex
Simo [72] is a simulation tool for floating structures, developed by Marintek and exten-
sively used in the offshore industry. Coupled to Riflex [73], a finite element solver for
slender marine structures, they form a software package for non-linear time domain sim-
ulation of offshore structures.
The first development of Simo-Riflex with a wind turbine module considered the tower
and blades as rigid [32]. Later the wind turbine model was incorporated with flexible ele-
ments in Riflex [85, 70], but there were no options for turbulent wind or external control.
At the time of writing Simo-Riflex has fully coupled aero-hydro-servo-elastic capabili-
ties, however with the limitations described in Table 2.2. The development performed in
relation to this thesis is more thoroughly described in Sec. 3.1.4.
Usfos - VpOne
Usfos [37] is a non-linear static and dynamic FE solver tailored for space frame structures.
With a beam element turbine model, coupled to an aerodynamic BEM solver, the software
package Usfos-VpOne provides coupled offshore wind turbine analyses. It has been used
in the development of the Sway concept [38], and Karimirad et. al. [53] did validation
efforts of the Usfos-VpOne for a floating spar-type TLP.
The initial plan was to work with VpOne for a semi-submersible. It was, however, found
that since a semi-submersible can fall under the definition of a large volume structure,
and since there can be hydrodynamic interaction between the members of the platform,
the slender body theory available in Usfos was unfit for the purpose.
Hawc2
Hawc2[66], developed by Risø, has the state-of-the-art capabilities in aerodynamics for

land based wind turbine design. The structural model consists of multibody dynamics
combined with a linear finite element formulation [18]. Early developments for FWTs
included a coupling to Simo-Riflex, but Hawc2 has later been developed to have FWT
simulation options as a stand-alone program. Wave forces assume slender body theory
and the mooring line model is a quasi-static non-linear force-displacement relationship,
however, a FE mooring line model is under development.
21
Earlier versions of Hawc2 for slender FWTs lacked the dynamic pressure formulations,
as can be seen in a code-to-code-comparison between Simo-Riflex and Hawc2 [54], but
this has been added later.
22
Chapter 3
Modelling of floating wind turbines
Structural analysis of FWTs for design in all limit states require dynamic structural analy-
sis, where aerodynamic and hydrodynamic loads and interaction with the controller need
to be accounted for. The theory behind the global response analysis methods applied in
the case studies is described in this chapter.
3.1 Structural dynamics of FWTs

A floating wind turbine in operation will be experience time varying environmental loads
and periodic load variations associated with rotating blades. These loads can cause sig-
nificant dynamic effects, and a static approach when determining the load effects is not
sufficient. Waves, wind and rotating blades cause cyclic loads on the FWT, and cyclic
loads can cause fatigue damage in the structure. Secondary effects of currents, i.e. vortex
induced vibrations or motions, also cause cyclic loading, but it has not been considered in
the case studies.
3.1.1 Equation of motion
The linear equation of motion for a submerged rigid body with 6 degrees of freedom and
external forces can be expressed in the frequency domain as in Eq. 3.1 [80]. F(ω) forces
can represent wave forces, mooring forces or other external forces.

M + A(ω) ẍ(ω) + B(ω)ẋ(ω) + Cx(ω) = F(ω) (3.1)
For non-linear systems, solution of the equation of motion must be performed by iteration
in the time domain. Equation 3.1 can be transformed to a non-linear time domain model
(Eq. 3.2) by using the Cummins equation, which introduces an impulse-response (mem-
ory) function into the equation of motion [80, chap.14]. This Duhamel integral is called
23
the retardation function, is denoted by κ(t) and accounts for frequency dependent added
mass and linear radiation damping in the time domain.
Zt
∞
(M + A ) ẍ(t) + κ(t − τ )ẋ(τ ) dτ + Cx(t) = F(t) (3.2)
0
3.1.2 Non-linear FEM
Finite element analysis (FEM) is a numerical method that, when applied in structural anal-
ysis, approximates the solution of the differential equations for displacement by dividing
the structure into smaller elements with a reduced number of displacement patterns.
Non-linear FEM refers to solution methods that take the effects of large displacements,
non-linear material behaviour or changing boundary conditions into account. These ef-
fects introduce a memory effect, and traditional structural mechanics principles like su-
perposition of loads and responses are no longer valid. Non-linear formulations of loads
also introduce non-linearities, but these are not addressed in this section.
In floating wind turbine modelling, the relevant non-linearities are related to geometric
structural non-linearities, i.e. the effects of large displacements, and quadratic load formu-
lations in the thrust and drag forces. The current section briefly addresses the geometric
non-linearities and the solution methods inherent in structural FEM solver Riflex.
The non-linear equation for dynamic equilibrium of a system of finite elements can be
expressed by global matrices that contain the mass, damping and stiffness properties of
the finite elements (Eq. 3.3) [73]. External forces can typically be weight, buoyancy,
forced displacements from attached bodies (x,ẋ and ẍ in Eq. 3.2), viscous drag and wave
acceleration terms from Eq. 3.6 and wind forces from AeroDyn.

MS + MH r̈(t) + CS ṙ(t) + KS r(t) = REXT r(t), ṙ(t) (3.3)
Geometric non-linearities can be particularly important for wind turbines with large de-
flection of the blades and in mooring lines. The non-linear dynamic solver applies implicit
integration for increments of time; a Newmark-β method, to find the dynamic equilibrium
at every timestep [73]. The parameters in this method control the accuracy, numerical sta-
bility and numerical damping of the integration method [17]. This displacement driven
method allows unlimited displacements, but assumes small strains.
3.1.3 Coupled analysis
If an FWT is regarded as separate components; the floating platform and the wind tur-
bine, response analysis can be performed with software and analysis methodologies that
24
are well known within the respective offshore- and wind industries. Handling the sub-
structure and the wind turbine separately is often done for bottom fixed offshore wind
turbines. However, when considering a floating platform, moorings and a flexible turbine
separately, important coupling effects can be ignored.
The submerged part of a floating platform is normally analysed by rigid body dynamics.
Moorings are modelled by flexible, slender FEM. The solvers are thus separate, and in-
dividual analysis of the floater and the moorings will not include the interaction between
the two. For coupled analysis, information must be passed between the two solvers at
every time step of the analysis.
For a floating wind turbine, coupling effects are important. Coupling effects in this context
are:
• Aerodynamic damping from the rotor on floater motion
• Platform motions influence the wind force
• Mean position of the platform influences wind force and mooring force
• Mooring line dynamics (inertia and damping) influences platform motion
In a coupled analysis, the rigid body and the flexible elements are connected at com-
mon nodes. The motion equilibriums for the flexible element system and the rigid body
are solved individually, but simultaneously, in time domain, exchanging external forces
and displacements at every timestep [74]. Iteration is used at every timestep to achieve
equilibrium for both equations.
The force vector (F(t) in Eq. 3.2) for a rigid platform with catenary mooring lines and a
turbine in a coupled model, contains non-linear restoring, inertia and damping from the
mooring lines, wave forces, wind forces and inertia and damping forces from the turbine
(Eq.3.4). All the forces are functions of platform motions, velocities and acceleration, in
addition to time. If one assumes 1st order wave excitation, wave forces are only a function
of time. As mentioned in the previous section, REXT in Eq. 3.3 contains forced displace-
ments, velocities and accelerations from the rigid platform in the coupled solution.
F(t) = Fmoor (t) + Fwave (t) + Fwind (t) + Fturb (t) (3.4)
3.1.4 Development of a new analysis tool
At the start of this project, none of the software listed in Sec. 2.2.4 could provide a suffi-
ciently sophisticated combination of structural model, hydrodynamics and aerodynamics
to model a semi-submersible wind turbine. The Fast [43] package was the most complete
option, but it did not include a dynamic mooring line model or the option of vertical Mori-
son members. The SSWT used in the studies throughout this thesis has catenary mooring
lines and heave-plates, which makes these features important.
25
Another option existed, namely Marintek’s Simo-Riflex windturbine code. At the time, it
was being extended from representing the tower and blades as rigid to flexible structures
using FEM, i.e. going from a Simo-model of the turbine to a Riflex model. However, the
in-house blade element momentum (BEM) model was not validated, external controllers
were not possible and only constant wind could be applied. Nevertheless, since coupled
analyses with Simo-Riflex had been successfully performed by the offshore industry for
analysis of floating structures for years, and since close cooperation with Marintek was
possible, it was decided to develop this tool further. The open source aerodynamic Fast-
module AeroDyn [62], developed by NREL, is a stand-alone code that had earlier been
coupled to structural analysis software Msc.Adams [63] and SimPack [36]. AeroDyn was
found suitable for coupling to Riflex, since it is a widely used and validated code that can
read turbulent wind fields generated by TurbSim [48].
In cooperation with Elizabeth Passano and Harald Ormberg from Marintek, PhD candi-
date Erin E. Bachynski and the author developed a dynamic linked library (DLL) that at
the beginning of every timestep passes nodal positions, orientations and velocities from
Riflex to AeroDyn and sends the aerodynamic nodal loads from AeroDyn back to Riflex.
Simo-Riflex-AeroDyn
The fully coupled aero-hydro-servo-elastic simulation tool Simo-Riflex linked to Aero-

Dyn comprises all the aerodynamic features of AeroDyn [78]. The module receives nodal
velocities, orientations and positions from Riflex through the DLL, reads the local wind
velocity from a turbulent wind field box generated by TurbSim (or a defined constant
wind speed), iterates to find the induction factors, calculates the element loads based on
BEM (see Sec. 3.2.2), and returns the forces as nodal blade loads back to Riflex through
the DLL. Figure 3.1 illustrates the role of the DLL in this procedure. Validation work for
Simo-Riflex-AeroDyn is presented in [84] and in Paper 1 [57].
AeroDyn and DLL features worth mentioning are:
• Tower shadow model
• Tip, root and hub loss (Prandtl)
• Skewed inflow correction (Pitt-Peters)
• Generalised dynamic wake (GDW) option (dynamic inflow)
• Dynamic stall (Beddoes-Leishman)
• Wind shear
In Riflex, the tower and blades are modelled as lines of beam elements between supern-
odes. The lower tower node is connected through a master-slave DOF to the rigid body in
Simo. There is also a master-slave dependency between the tower top and the shaft. The
shaft is modelled as two beam elements with high stiffness, and the connetion between
26
them is the rotational spring in which generator torque is applied. The same type of con-
nection is used for the pitch actuators in the nodes between the hub element nodes and
the first blade nodes. The hub elements are also modelled as beams with high stiffness,
and they are rigidly connected to the low speed shaft. Figure 3.1 shows the layout of the
Riflex turbine model.
The blade elements are defined by their structural properties in Riflex, including struc-
tural twists around the blade axis. Modelling of centres of gravity that are eccentric to
the structural axis is, however, not possible. Aerodynamic properties, i.e. aerodynamic
coefficients and aerodynamic twist, are specified in AeroDyn.
The masses and inertias of the hub and the nacelle are modelled by rigid bodies connected
by a master-slave connectivity to the low speed shaft and the tower top, respectively.
low speed shaft/ Foil elements

flexjoint/
high speed shaft
Riflex NACELLE
HUB
• velocity
• orientation
• position
• lift Simo body

• drag Riflex supernodes
• pitch moment
Riflex line
DLL
Rigid supernode
connection
AeroDyn tower
z
y x
MAIN BODY
Figure 3.1: Illustration of beam element model for the turbine in Riflex.
During the course of this project, Riflex was opened for the possibility of using external
pitch and torque control algorithms written in Java. The rotor speed and blade pitch angles
are received from Riflex and the controller feeds back the actuator torque and blade pitch
angle.
27
3.2 Load Modelling
3.2.1 Hydrodynamic loads
Hydrodynamic loads from surface waves on floating structures are normally calculated
based on one out of two theories: Morison’s formula or potential flow theory.
Morison’s formula is a semi-empirical formula assuming that the size of the structure is
so small that it does not affect the waves, thus it is best suited for slender structures. In
this thesis, “potential theory" refers to solution of the velocity potential for a linearised
boundary value problem where the wave forces are found by pressure integration around
a rigid body [30]. In the case studies, the potential theory forces were obtained using
panel method software Wadam [23]. The potential theory solution (see Eq. 3.5) contains
one force component from the undisturbed wave field (Froude-Krylov) and a diffraction
component that accounts for the flow disturbance due to the presence of the body . The
potential flow solution approaches the Morison solution for a cylinder when the wave-
length to diameter ratio is large and non-linear viscous effects are negligible.
Fwave = FF K + FD (3.5)
Figure 3.2 shows a theory validity diagram for a fixed cylinder with a diameder d, and
indicates which effects are important for various wave-height-to-diameter and diameter-
to-wave-length ratios.
2.5
wave breaking limit HL =0.1418

2.0
1.5
H (-)
viscous effects
MORISON EQ
d
1.0
diffraction effects
0.5
DIFFRACTION THEORY
0.0
0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40
d (-)
λ
Figure 3.2: Theory validity diagram for wave forces on a fixed cylinder [13].
The non-linear FEM-solver applied in the analyses (Riflex) calculates wave forces on
slender members by Morison’s formula (Eq. 3.6), whereas the floating rigid body (Simo)
28
was modelled with a combination of potential theory forces and the viscous drag term
from Morison’s formula. Equation 3.6 shows how the axial, lateral or transversal force,
dF acting on a strip of the member, in its local coordinate system, are expressed in the
simulations.
ρw
dF = ρw Au̇ + ρw ACm (u̇ − ẍ) + DCd |u − ẋ| (u − ẋ) (3.6)
2
Surface piecing floating bodies, i.e. the columns in a semi-submersible, will experience a
vertical force resultant due to the unbalanced dynamic pressure at the base of the columns.
This is not included in the Morison formulation, but has to be taken into account [16].
Paper 1 compares a model of a SSWT with pure Morison type forces to a model with
potential theory forces. In a pure Morison model, as it is modelled in Simo, there was
no way to account for the unbalanced dynamic pressure at the base of the columns in the
program itself. An analytical force transfer function was therefore derived and added to
the columns in the analysis. The derivation and verification of these transfer functions are
shown in Appendix D.
3.2.2 Aerodynamic loads by BEM
Simple momentum theory considers the change in momentum in the wind flow over an
actuator disc with area AD . By applying conservation of momentum, an expression for the
thrust force (T) and power (P) acting on the actuator disc can then be derived as a function
of the far field wind velocity (U∞ ). The theory described here is based on Burton et al.[15,
Chap. 3].
2
T = 2ρa AD U∞ a(1 − a)
3 (3.7)
P = 2ρa AD U∞ a(1 − a)
The blade element theory assumes that the forces on the blades can be calculated based
on two-dimensional aerodynamic properties. Thus the forces are also only calculated
for two dimensions. The forces on a 2D blade element depend on the effective wind
velocity experienced by the element (relative velocity Urel ). The relative velocity is the
combination of the tangential velocity (Ωr(1 + a0 )) due to the rotating blade and the
velocity at the rotor disc (axial velocity U (1 − a)). The angle between these two velocity
components determines the inflow angle (φ). The angle of attack (α) is defined as the
angle between Urel and the chord c, and the sum of the angle of attack and the blade pitch
angle (γ) is the inflow angle. Aerodynamic coefficients for each aerofoil type are based
on experiments and expressed as a function of the angle of attack. Figure 3.3 shows the
velocities and forces acting on an aerofoil. The lift (L) and drag (D) forces are defined
as normal and parallel to Urel , respectively. In addition, there is a pitching moment (M )
29
around the aerodynamic centre. The aerodynamic force per unit length on the rotor blades,
referred to the aerodynamic centre, are defined by Eq. 3.8.
L
Urel
U (1 − a) α
φ
γ D
rotation plane −Ωr(1 + a′ ) c
Figure 3.3: Definition of axes and forces on a 2D aerofoil.
1 2
dL = Cl ρa cUrel
2
1 2
dD = Cd ρa cUrel (3.8)
2
1 2
dM = Cm ρa AcUrel
2
The axial induction factor is found by equating the trust force on all blade elements to the
rate of change of axial momentum, and similarly for the tangential induction factor. The
equations are solved by iteration.
The underlying assumptions of BEM theory introduce a number of inaccuracies. Assum-
ing that the flow over the blades is always in equilibrium, disregards the dynamics of
the wake. Also, pressure gradients in the span-wise direction of the blade, which can be
important in a case with a highly loaded rotor (high tip speed ratios), are ignored. In nu-
merical BEM solvers, various corrections are normally applied, for instance for tip loss,
flow separation, dynamic stall and dynamic inflow.
3.2.3 Aerodynamic loads by GDW
In AeroDyn, a second option for aerodynamic load calculation is available, namely the
generalised dynamic wake model (GDW). GDW was developed from helicopter theory
and is based on the potential flow solution of Laplace’s equation [78]. As the name im-
plies, the method takes into account the dynamics of the wake, i.e. the transient condition
30
when the inflow angle changes. In GDW, tip loss, dynamic stall and dynamic- and skewed
wake are inherent.
GDW is, however, not valid for highly loaded rotors, i.e. at low wind speeds. And
neither GDW nor BEM accounts for the effect of large deflections on the blades on the
aerodynamic forces.
In connection with the work with Paper 2, tower bending moments for different methods
for calculating the aerodynamic loads was compared with wind-only simulations. BEM
without dynamic wake correction, BEM with dynamic wake correction and GDW were
compared. The differences were significant for the case where the 3P frequency interferes
with the first fore-aft bending mode of the tower, i.e. where aerodynamic damping is im-
portant. The GDW method and the BEM method with dynamic wake correction gave rise
to significantly higher damping than the BEM method without dynamic wake correction.
This work is summarised in [56].
3.2.4 Thrust force on pitch regulated wind turbines
Wind turbine control systems ensure optimised power output and mitigation of loads un-
der high wind conditions by changing the blade pitch angle (and thus the angle of attack of
the aerofoils), the generator torque or the yaw position [15]. The turbine used in this the-
sis is a variable speed generator with active pitch control (pitch to feather). Variable speed
control means that the rotational speed is controlled by varying the generator torque, to
maximise the power output below rated wind speed. The active pitch control ensures that
the aerodynamic torque is kept constant above rated wind speed. Since the blades are
pitched out of the wind above rated wind speed, the thrust force reduces for higher wind
speeds, and the maximum thrust force occurs at the rated wind speed.
The active pitch controller for the NREL 5MW uses proportional-integral (PI) control of
the generator torque in the above rated - below cut-off region. This is a strategy that uses
measurements and control system actions in a closed loop to obtain the desired torque
[71, 49]. The loop can be described as a dynamic SDOF system with constants that
determine the response characteristics of the controller. The control system constants
from the OC3 Hywind spar study were applied [42] for the NREL 5 MW used in the
investigations. This controller is tuned to avoid negative damping above rated wind speed
due to the pitch motion of the OC3 Hywind platform. This was done by setting the
control system natural frequency to 0.2 rad/s, which is outside of the wave frequency
range and below the natural pitch frequency of the spar platform. 0.2 rad/s is above the
natural pitch frequency of the SSWT (0.17 rad/s) in this study, but setting constants to
get below this will give too much variation in the power production. Even though the
controller natural pitch frequency could not be set a value below this limit, applying the
OC3 Hywind constants gave less pitch motion than the onshore controller. The negative
damping instability was not observed for this platform, probably due to the heave plates,
which provide viscous damping for pitch motion.
31
3.3 Stochastic modelling of environmental loads
For engineering applications, the wind speed and the wave elevation create the basis for
calculating environmental loads on an offshore structure. These have variations in long
and short term, but the engineering approach is to assume stationary conditions for a
certain period of time (long term), with small scale variations around the mean conditions
(short term).
3.3.1 Long term distribution of wind and waves
Wind and wave statistics are normally based on wind speed, wave height, wave period
and directional data recorded over several years at a specific location. The measured
parameter is assumed stationary for a certain return period, and the mean value over this
period forms the basis for long term statistics. For wind it is normally the 10 minute
mean wind at 10 m above the ground that is measured [98]. Sea states of significant wave
height (Hs ) and wave peak period (Tp ) can be assumed to be stationary for 30 minutes to
10 hours (normally 3 hours), and normal engineering practice is that the Hs and Tp data
from measurements are fitted to a 3-parameter Weibull distribution [22].
The mean wind speed normally varies as a function of vertical distance to the surface and
the terrain roughness. Profiles describing this are called wind shear profiles, and can be
found in standards, e.g. the IEC standard [39]. For offshore wind, the surface roughness
is low, and an exponential profile is frequently used.
Since environmental condition statistics require measurement records over sever years,
not many joint wind and wave statistics are available. Johannesen et al. [47] fitted a 3-
parameter Weibull distribution to wind and wave measurements from the northern North
Sea during the period 1973-1999. This distribution has been applied to several offshore
wind turbine studies, e.g. in [54, 105, 29].
In a study of combined wind and wave power units, Marina Platform [87], joint wind
and wave distributions for five sites in European waters were established [68], based on
hindcast data of 1-hour averaged sea states and wind. Figure 3.4 shows the long term
joint probability distribution of mean wind, wave height and wave period for the Buoy
Cabo Silleiro site, off the coast of Portugal. Since the WindFloat prototype is located in
Portugal, environmental conditions from this site were chosen for the studies in papers
2-4.
The scarcity of simultaneously measured wind and wave data also means that there is
even less information about the directional scatter of wind and waves available. The
joint wind and wave distributions in [68], where the wind and waves were assumed uni-
directional, were based on data that also contain direction [20]. However, a large number
of data points per direction is needed in order to make a proper probability distribution
that also covers direction. Figure 3.5 shows the occurrence of misaligned wind and wave
32
5.0e-03
20 2.8e-03
Hub wind speed (m/s)

1.4e-03
15 6.6e-04
3.0e-04
10
1.0e-04
5 2.3e-05
1.0e-06
5.0e-07
15
2 4 10 (s)
6
Wave height 8 5 period
(m) 10 Wave
Figure 3.4: Joint wind and wave distribution.
conditions in the Buoy Cabo Silleiro site, and illustrates that considering multi-directional
wind and waves in the design of FWTs may be necessary.
6000
5000
No. of observations
4000
3000
2000
1000
00 20 40 60 80 100 120 140 160 180

Wind/wave misalignment (deg)
Figure 3.5: Observations of relative direction between wind and waves in the hindcast
data for site 3 in [68]. The distribution includes observations in all directions.
3.3.2 Short term distribution of wind and waves
For simulations of response to stochastic loads, environmental conditions chosen from

long term statistics are assumed to be stationary for the assumed return period. The short
term random variations in wave elevation, or around the mean wind speed, are normally
in the scale that gives structural dynamic response. For wind, variations around the mean
wind speed is called turbulence, whereas the random wave elevation is often referred to
33
as irregular waves. The short term recordings of the random variations can be fitted to a
Gaussian distribution. The individual maxima of this process are thus Weibull distributed
[80].
The wind field can be described by three velocity components, where the longitudinal
component varies around the mean wind speed, and the vertical and the transverse com-
ponents are zero mean processes.
The short term variations are normally described in the form of spectra that describe
the amplitudes and frequency content of the variations. A frequently used spectrum for
the North Sea, for developing waves created by wind friction, is a modified Pierson-
Moscowitz spectrum, referred to as the Jonswap spectrum [80]. A typical Jonswap spec-
trum is shown in Figure 3.6. Swell spectra, e.g. the Torsethaugen spectrum (wind sea and
swell), can also be important to include in response analysis of FWTs, if relevant for the
particular site conditions [27].
3.0
Hs=3.0 m
2.5 Tp=10.0 s
2.0
Sζ (ω) (m2 s/rad)
1.5
1.0
0.5
0.00.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4

ω (rad/s)
Figure 3.6: Joswap spectrum for distribution of wave frequencies for a wind sea state with
Hs = 3 m and Tp = 10 s.
Spatial turbulence models that are frequently used in wind turbine design are the Mann
turbulence model and the Kaimal spectral and exponential coherence model. Figure 3.7
shows an example of a Kaimal spectrum for 11.2 m/s (rated wind speed for the NREL
5MW), as given in [40]. The coherence model describes the spatial correlation of the
longitudinal velocity component.
The Kaimal spectrum, which is recommended in the standard for offshore wind turbines
[40] is based on onshore wind measurements. There is limited offshore wind measure-
ment data available, but 14 month long recordings of wind speeds from a site outside of
Frøya in Norway [2], showed more energy in the low-frequency range than the Kaimal
spectrum and other land based spectra. This spectrum is probably more relevant for
34
60
Uw=11.2 m/s
SU (ω) ((m2 /(s rad)) 50
40
30
20
10
00.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4

ω (rad/s)
Figure 3.7: Kaimal spectrum for turbulent wind speed for a mean wind speed of 11.2 m/s.
FWTs, but it was not available in the turbulence generator for AeroDyn, thus the Kaimal
spectrum was used in the case studies.
The timeseries of wave elevation or wind speed is generated by IFFT in both Riflex [73]
and in TurbSim [48].
3.4 Response to stochastic loads
3.4.1 Variance spectrum and frequency-response function
Short term variations in stationary wind or wave conditions can be described as a linear
combination of different frequency components. A useful way of defining the frequency
content of a process X is by the variance spectrum, which is defined as the Fourier trans-
form of the correlation function [80].
Z∞
1
SX (ω) = CX (τ )e−iωt dτ (3.9)
2π
∞
For a linear relationship between a force (F (ω)) and a process (wind speed or wave eleva-
tion), the variance spectrum for the force (SF (ω)) can be expressed by the variance spec-
trum for the wind- or wave load through a complex transfer function (HXF ), as shown in
Eq. 3.10. Here, S(ω) denotes the one-sided variance spectrum.
SF (ω) = |HXF (ω)|2 SX (ω) (3.10)
35
3.4.2 Variance spectrum from timeseries
In this thesis, spectral estimates directly from timeseries were determined by Fourier
transform. Depending on the application, e.g. whether or not smoothing would intro-
duce errors, different procedures were chosen.
The Cooley-Tukey direct fast Fourier transform (FFT) algorithm provides an efficient
way to estimate the variance spectrum from long signals. However, the procedure re-
quires some tapering to give a good estimate of the variance spectrum. One method is
computing the autocorrelation function of the timeseries by inverse FFT, apply a taper
window function, e.g. a Hanning window, to the autocorrelation estimate and apply FFT
to the tapered autocorrelation function. This procedure is described in detail by Bendat
and Piersol[12, chap. 5].
3.5 Fatigue in steel

Fatigue damage in metals is fracture that occurs due to cyclic stress variations even if
the stress level is below the yield stress of the material. The cause of the fracture is
microscopic imperfections that cause high local stress and gradual crack growth. For
steel, imperfections and microscopic cracks are most likely to be found in welds, thus the
weld is the weak point from a fatigue damage point of view. The cracks grow until the
remaining cross section area does not have enough capacity, and the whole member or
joint fails.
A fatigue process is normally classified as high cycle or low cycle fatigue. High cycle
fatigue has more than 104 load fluctuations during the lifetime, but has a lower stress
level, whereas low cycle fatigue has a lower number of cycles, but has stresses into the
plastic deformation range [65]. This thesis only addresses high cycle fatigue, since stress
levels above the yield stress is defined as structural failure and since the number of cycles
during the life of an FWT is high.
Fatigue damage estimation is performed based on long term stress distributions and a
material response model. For the material response model, either fracture mechanics, that
considers the different phases of crack growth and the local stresses in the crack (e.g.
Paris’ law), or an S-N curve approach can be applied. The latter was used in the case
studies, and the method is described below.
3.5.1 Fatigue damage calculation by S-N curves
The most common engineering approach to high cycle fatigue in steel welds is to assume
that the mean of the varying stress is insignificant and that it is purely the stress variation
that causes cracks to grow. The method is based on laboratory tests, where specimens are
subjected to stress ranges of varying amplitude (S) until it reaches failure. The Wöhler
36
approach assumes that the number of cycles until failure (N) has a linear relationship
(with slope m) with the logarithm of the stress amplitude (S) (see Eq. 3.11).
N = KS −m (3.11)
This relationship is normally referred to as an S-N curve. S-N curves for offhore welded
steel structures can be found in [25]. In beneficial conditions, e.g. where the structure has
corrosion protection or for structural parts in air, a higher number of stress cycles with
small amplitudes can be allowed, and the S to log(N) relationship can be described with
a bi-linear curve, as shown in Figure 3.8.
Figure 3.8: Typical bi-linear S-N-curve with an illustration of cyclic stress [25].
From a global FE model, only nominal stresses in the base cross sections can be obtained.
However, S-N curves refer to hot spot stress, which means the concentrated stress around
a sudden change in geometry, e.g. a weld toe. In a full fatigue analysis, there are different
ways of accounting for stress concentrations. Either, local analyses are performed to
determine hot spot stress directly, or so called Stress Concentration Factors (SCF) can be
applied to the nominal stress. The use of SCFs is most suitable for simple joints, and
can involve different SCFs for different stress components. Since no weld geometry was
considered in this thesis, no evaluation of hot-spot stress was performed. Reference is,
however, made to Fredheim [31], whp performed a local hot spot stress analysis for a
semi-submersible windturbine.
The fatigue damage for a long term stress range distribution can be found using the Miner-
Palmgren hypothesis (Eq. 3.12), which assumes that each stress fluctuation contributes
with linearly accumulated partial damage. The accumulated damage is found by adding
the contributions from each stress level.
N
X bins
nj
D= (3.12)
j
Nj
where nj is the number of stress cycles in the time history and Nj is the number of cycles
to failure according to the S-N curve. Failure occurs when D = 1.
37
3.5.2 Cycle counting algorithms
The nominal stress histories of structural members can be achieved from frequency do-
main estimates, or directly from stresses from time domain FE analyses.
To be able to calculate the damage from stress histories using S-N curves, counting of
cycles and stress levels must be performed. Different methods for cycle counting are
available: Range counting, peak-valley counting or level crossing method, but the cycle
counting method that gives the best results for fatigue analyses for wide banded stress
histories is the Rainflow counting method [76]. The Rainflow counting algorithm applied
in papers 2-4 uses the Rychlik [94] adjustment. Rychlik introduced a toplevel-up cycle
counting method, that is equivalent to the original Rainflow counting mehtod. The stress
ranges obtained by cycle counting are paired in cycle pairs and sorted into bins based on
the stress range.
Frequency domain methods for fatigue estimation use the fact that the stress amplitudes
are directly related to the variance of the process. For a narrow banded Gaussian process,
stress amplitudes will be Rayleigh distributed and estimation of fatigue is straight forward
if a frequency domain solution exists. For a wide banded process, an estimate based on
the narrow band approximation is too conservative [11].
Proposed methods to overcome this use empirical formulas, based on frequency domain
solutions, with correction for bandwidth parameter (Dirlik [28] and Benasciutti and Tovo
[11]) or combining damage from high-frequency and low-frequency parts of a bimodal
process (Jiao and Moan [46]). The latter was also investigated for application to trimodal
processes by Gao and Moan [33]. Benasciutti and Tovo have also investigated methods
for wide banded, non-Gaussian processes [10].
Fatigue damage from frequency domain methods are fast to compute, however, the meth-
ods do not provide a realistic alternative to time domain simulation and Rainflow counting
unless a frequency domain solution for the stress is available. This is the reason that fre-
quency domain methods for estimating fatigue damage were not used in any of the case
studies. Nevertheless, they can be valuable in a frequency domain response analysis, e.g.
in the study in Paper 4.
3.5.3 Stress calculation
Based on the element forces from the FE solver, nominal axial stress and shear stress was
calculated according to Eq. 3.13 [41]. The expression for shear stress assumes St. Venant
torsion for a circular member. Figure 3.9 shows the definitions of directions and axis for
the tower base.
38
loc 12
Wind loc 18 loc 6
θ z
loc 0
Figure 3.9: Top view of tower base cross section.
Nx My Mz
σ= + r1 cos θ − r1 sin θ
A Iy Iz
(3.13)
Mx 2Vy 2Vz
τ= R1 − cos θ − sin θ
Ip A A
39
40
Chapter 4
Case studies
This chapter presents the main results from the investigations in the thesis. Detailed
descriptions of the background and methodology of the investigations can be found in the
appended papers. The investigations seek to answer the research questions proposed in
Chap. 1:
• RQ1: How should loads and structural response be modelled in order to properly
predict the fatigue life of a floating wind turbine?
• RQ2: Can simplifications in the analysis be made in order to make the FLS analysis
more efficient?
4.1 Semi-submersible used in the case study
All the case studies were performed for a three-column single semi-submersible wind tur-
bine with the NREL 5MW on the OC3 Hywind tower [42] placed on one of the columns,
a concept inspired by WindFloat [93]. Paper 3 investigates the effect of misaligned wind
and waves on four different FWT concepts, but since the author’s contribution was anal-
ysis for one of the semi-submersibles, this semi-submersible model is the only one de-
scribed in this chapter.
Two different variations of the SSWT were applied in the investigations, they are, how-
ever, quite similar. Paper 1 used a SSWT very similar to the generic WindFloat [93],
whereas more details than what was available were required for the study in Paper 2.
Papers 2-4 use a model that was developed to have similar characteristics as the generic
WindFloat: The outer dimensions and geometry were kept as in [93], but wall thicknesses
of the members, location of the ballast tanks and other material and geometric properties
were chosen to give a mass distribution close to the generic WindFloat. Also, to achieve
more of a catenary effect from the mooring lines than what the mooring system specified
in [93] provides, the lines were lengthened and the anchor positions were changed. The
41
mooring stiffness for small displacements of the platform was, however, found to be the
same as specified in [93]. The resulting specifications for the modified platform in Papers
2-4 can be found in Appendix C.
Paper 1 used a modelling method where the submerged part of the platform of the SSWT
is modelled as one rigid body and the mooring lines and wind turbine tower and blades
were represented by flexible beam elements. The investigations in Paper 2, also focused
the forces in the braces of the platform. Thus, a multibody model, where the three columns
were modelled as individual rigid bodies connected by FE flexible beam elements, was
developed. The use of multibody modelling was verified against a single body model
in Appendix C. The effect of hydrodynamic interaction between the columns was taken
into account by including the full effect of the presence of the other bodies, whereas the
effect of motion of the other bodies was only partly accounted for. This is more thor-
oughly described in Paper 2, and an assessment of the effect of hydrodynamic interaction
is described in Appendix C.
Paper 1 did not include WindFloat’s active ballast system, as simulations with an oper-
ating turbine were limited. It was, however, observed, that the mean tilt that was experi-
enced with no ballast system, gave less rotor loads compared to an upright position. By a
simple evaluation of the overturning moment from the maximum thrust force of the 5MW
turbine (~730 kN) and the pitch restoring stiffness, which results a mean pitch angle of
around 13 degrees, it is clear that such an angle will change the loads compared to an up-
right position. Thus the ballast system was modelled in the subsequent investigations. It
was modelled by redistributing the mass between the columns to balance the thrust force
and rotor torque.
4.2 Hydrodynamic modelling in fully coupled simulations

(Paper 1 [57])
The paper describes a comparative study of responses to wave forces calculated by Mori-
son’s equation or by potential theory with additional Morison drag. It is also the first
publication verifying and applying Simo-Riflex-AeroDyn, therefore some analyses with
combined wind and waves were performed to study the effect of Morison versus potential
theory on turbine power production.
The shortcoming of the Morison model is that it can not describe diffraction effects, hy-
drodynamic interaction between platform members or frequency dependent added mass
and damping. On the other hand, potential theory does not include wave kinematics above
the MWL. Nevertheless, for a SDOF system in regular waves with one amplitude, it is
possible to achieve identical responses with the two formulations.
The Morison model can be implemented with and without forces above the MWL and
with and without considering the wave particle velocities and accelerations at updated
42
member positions. Regular wave analyses were performed to investigate the different im-
plementations of Morison’s formula in Simo. The added mass coefficients in the Morison
inertia forces were calculated from the potential theory solution with frequency dependent
added mass.
As described in Chapter 3, pure Morison forces on surface piercing members lack the
force from the unbalanced dynamic pressure on horizontal surfaces. It was found during
the study, that including the dynamic pressure was not possible using traditional input
to Simo. Thus, an analytical force transfer function for the SSWT was derived and ap-
plied together with the Morison forces (modified Morison). The derivation of the transfer
functions is shown in Appendix D.
Figure 4.1: Platform RAOs for surge, heave and pitch motion for variants of Morison
and for potential and drag force. Inertial coefficients are calculated from potential theory
added mass.
Figure 4.1 shows response amplitude operators (RAOs) for surge, heave and pitch motions
using potential theory and four approaches to Morison’s equation. Results shown in Fig.
4.1 are based on time domain analysis with regular waves, and the response amplitudes
are normalised with the input wave amplitude. The response outside the wave frequency
was filtered out. The linear potential theory solution without quadratic drag is also shown.
Pure Morison with forces calculated up to MWL overestimated heave and pitch motion
compared to the potential theory and drag model, but good agreement was obtained by
including forces up to wave elevation. Correcting for dynamic pressure under the columns
(modified Morison) also gave a better fit, but this method can also be improved further by
including forces up to wave elevation. Calculating the forces at the instantaneous position
did not have a significant effect for these cases. Diffraction effects seemed to be important
in heave response for periods lower than 6.9 s.
Sensitivity analysis of horizontal (subscript h) and vertical (subscript v) inertial coeffi-
cients was performed. Figure 4.2 shows that for each wave period, the response in the
43
potential theory model can be matched by choosing the correct coefficient. When com-
paring Figure 4.2 to the inertial coefficients calculated from potential theory added mass
(listed in Paper 1), it can be observed that the correct coefficients can not always be de-
rived from the potential theory solution.
Figure 4.2: Platform RAOs for surge, heave and pitch motion for pure Morison and for
potential and drag force for different inertia coefficients
Applying the same added mass coefficients as in the regular wave analyses in Figure 4.1,
irregular wave analysis was performed. Irregular waves contain more than one frequency,
but the Morison inertial coefficients were determined based on the wave peak period. The
mean values of the motions in the irregular wave analysis (shown in Paper 1), were higher
for almost all the load cases for the Morison force analyses than for the simulations with
potential forces, but only negligible differences were found for the standard deviation.
Morison type forces will normally give a third order drift component [30, chap. 5], but
this effect comes from the drag term, which is the same for both analyses. However, the
Morison model includes inertia force terms above the mean waver line, and the structure is
asymmetric about the normal to the wave propagation axis, which due to different phasing
of the forces, causes a mean force.
Table 4.1: Mean value and standard deviation of electrical power output in kW for coupled
simulations for one wind dominated and one wave dominated environmental condition.
Results are shown for a land based and a floating turbine.
LWT FWT - Potential + Drag FWY - Morison

µ σ µ σ µ σ
Wind dominated 4798 339 4767 384 4734 424.6
Wave dominated 1896 732 1774 619 1774 617
Comparing one wind dominated and one wave dominated load case in coupled wind and
44
wave analyses (Table 4.1), showed that the hydrodynamic model can have an important
effect on the electrical power output of the wind turbine. For the wave dominated case, the
mean power output was lower, and the standard deviation of electrical power was higher
for the Morison model compared to the potential theory force model. Compared to an
onshore turbine, the mean power output was lower for the SSWT. This is partly due to the
mean pitch angle of the turbine (no ballast system was applied in these analyses) and due
to the motions of the platform that affect the relative velocity seen by the blades.
4.3 Time domain analysis procedures for fatigue assess-

ment of tower and braces (Paper 2 [59])
Long term fatigue analysis for nominal stress in the tower and platform members of a
three-column semi-submersible was performed by fully coupled time domain analyses in
Simo-Riflex-AeroDyn. The aim of the study was to investigate the necessary simulation
duration, number of random realisations (seeds) and bin sizes for the discretisation of the
joint wind and wave distribution.
A total of 2316 3-hour time domain simulations were carried out:
• 3-hour realisations of 155 different environmental conditions with aligned wind and
waves (10 seeds)
• 10 3-hour realisations of 42 environmental conditions with misaligned wind and
waves (10 seeds)
• Additional 3-hour realisation of 346 different environmental conditions (1 seed)
The environmental conditions and their respective probabilities of occurrence were se-
lected from a joint wind- and wave distribution generated from hindcast data from the
Buoy Cabo Silleiro, off the coast of Portugal [68].
The 20 year fatigue damage was calculated by Rainflow counting of nominal stress histo-
ries in the tower base, pontoon, main beam and brace based on 10 min to 3-hour samples,
averaged over 10 seeds. As can be seen in Figure 4.3, estimating the total 20 year fatigue
based on 10x1-hour simulations of each load case underestimated damage by less than
4% compared to 3-hour simulations. Using 10x10-minute simulations gave up to 10%
smaller damage. Fatigue damage for the tower appeared to be less affected by simulation
length than the platform members, pontoon, main beam and brace.
A full evaluation of the required number of seeds in a statistical sense would require a
large sample of 3-hour simulations. A sample of 10, which was used in this study, is not
statistically significant. The focus was kept on including as many different environmental
conditions as possible within the time frame of this work, and thus the number of samples
could not be increased. Therefore, a method introduced by Langley [64] to estimate the
root mean square (RMS) error of a limited number of program runs was applied.
45
0
Relative error of total damage [%]
−2
Tower base
−4
Pontoon
−6 Main beam
Brace
−8
−10
−12
600 1800 3600 7200 10800
Simulation length (s)
Figure 4.3: Relative difference to 3-hour simulations in 20 year total damage for axial
stress fatigue. All results are calculated as the mean value based on 10 seeds.
The resulting RMS error is shown in Figure 4.4. Similar conclusions to what has been
described above can be drawn from the RMS estimate; errors are small in general, but
larger for 10-minute simulations, and the tower base was less sensitive to the number of
seeds than the platform members.
Figure 4.4 also illustrates an important point; that the statistic content is the same in six
10-minute samples as in one 1-hour sample, and thus the expected error is the same.
Based on the results in Figs. 4.3 and 4.4, bin size investigations were performed with
1x3-hour simulations for 501 environmental conditions. The total fatigue was calculated
for increasing bin sizes. Table 4.2 shows that by increasing the bin size from 1 to 2 m/s,
the total fatigue damage for the tower was reduced by 9%. However, a bin size of 2 m/s,
starting at cut-in wind speed (i.e. load cases include 3, 5, 7,.. m/s), means that the most
important resonance case at 8 m/s is left out. If a bin size of 2 m/s was used, and the 8 m/s
cases were included, the deviation in the total damages would be 0.3% for the tower, 3.2%
for the pontoon, 1% for the main beam and 3.1% for the brace. This indicates that a bin
size of 2 m/s will give an acceptable damage estimate if 3P resonance cases are included.
The reduction of fatigue damage with increasing wave period was fairly linear, which can
explain that the increase of wave period bin size does not cause significant changes in the
damage estimates in Table 4.2. Choosing a bin size of 2 s gave an acceptable result in this
case, but keep in mind that this conclusion depends on the response characteristics of the
platform.
Assuming that the stresses increase linearly with pitch angle, the fatigue damage increase
46
14
10min-Tower base 1h-Tower base 3h-Tower base
12 10min-Pontoon 1h-Pontoon 3h-Pontoon
10min-Main beam 1h-Main beam 3h-Main beam
10min-Brace 1h-Brace 3h-Brace
10
RMS error (%)
0
1 2 3 4 5 6 7 8 9 10
No. of samples
Figure 4.4: Estimated RMS error for 20 year damage, as a function of number of samples.
Table 4.2: 20 year fatigue damage for varying bin size, relative to the smallest bin size.
The bin size variation in one parameter was combined with bin sizes of 1 m/s for wind
speed, 0.5 m for wave height and 0.5 s for wave period. Please note that the selection of
cases representing Hs bin sizes below 1.5 m is limited.
Bin size Tower base Pontoon Main beam Brace

Uhub Drel (%) Drel (%) Drel (%) Drel (%)
1.0 m/s 100 100 100 100
2.0 m/s 91 90 89 88
3.0 m/s 88 85 79 85
4.0 m/s 85 83 83 81
Hs
0.5 m 100 100 100 100
1.0 m 101 101 100 99
1.5 m 96 94 96 94
2.0 m 41 38 44 38
Tp
0.5 s 100 100 100 100
1.0 s 100 100 101 100
1.5 s 104 103 101 103
2.0 s 103 105 104 104
47
will be in the order of the m-parameter in the S-N curve (which is 3 and 5 in this study,
see [25]). This non-linear relationship between motions and damage may be the reason
why damage is more sensitive to the increase of bin size for wave heights than for wave
periods. However, more cases with variation in wave height should be included to make
a firm conclusion about sensitivity to wave height bin size.
An issue that emerged during the long term fatigue analysis, was the uncertainty in the
model for calculation of aerodynamic loads. For wind speeds of 7-9 m/s, the blade passing
frequency excited the first fore-aft bending mode of the turbine. At resonance, damping
is important, and the different options for calculation of loads in AeroDyn: BEM and
GDW, gave significantly different levels of aerodynamic damping. This emphasises the
importance of assessing aerodynamic models for lower wind speeds, and also the need
for a different tower design.
4.4 Effect of misaligned wind and wave directions on tower

fatigue (Paper 3 [8])
Four concepts for FWTs were subjected to a selection of environmental conditions with
misaligned wind- and waves, to investigate the effect on tower loads and motions com-
pared to aligned wind- and wave conditions: A spar platform, a TLP, one semi-submersible
with the tower on a central column (semi 2) and the semi-submersible described in the
beginning of this chapter (semi 1). All of them support the NREL 5MW turbine [49] on
an OC3 tower [42], and the control system gains were adapted to the individual concepts.
10 1-hour samples of 6 representative environmental conditions were run in Simo-Riflex-
AeroDyn with 0◦ wind direction combined with 0◦ , 30◦ , 60◦ and 90◦ wave directions (see
Figure 4.5), i.e. 24 different conditions. Due to the asymmetry of the semi-submersible
described in this thesis, additional cases with 90◦ wind and 0◦ -90◦ waves were run for this
concept. Fatigue damage was calculated for both nominal axial stress and shear stress by
Rainflow counting of the stress timeseries for various cross section locations and with S-
N curves based on DNV recommendations [25]. Nominal axial stress was found to cause
significantly more fatigue damage than nominal shear stress.
Figure 4.6 shows a typical pattern for how the expected (average of 10 seeds) fatigue
damage for semi 1 varied around the circumference of the tower base for different wave
directions when the wind direction was kept constant. The peaks of fatigue damage oc-
curred approximately in line with the wind direction for wave directions 0◦ , 30◦ and 90◦ .
60◦ waves is an unfavourable direction for this semi-submersible, since it causes large
wave excitation, and thus the location of the largest fatigue damage was shifted. The
same pattern was observed for semi 2. The spar and the TLP showed patterns where the
location of maximum fatigue damage followed the wave direction (see Paper 3).
The aligned wind and wave case gave the highest fatigue damage in Figure 4.6. This
was observed to be the case also for the other environmental conditions and for the other
48
Figure 4.5: Wind and wave directions applied in the analyses.
4.0 1e−5
3.5
βwave=0 ◦
βwave=30 ◦
3.0 βwave=60 ◦
Tower base DRFC (-)
2.5 βwave=90 ◦
2.0
1.5
1.0
0.5
0.00 50 100 150 200 250 300 350
θ (deg)
Figure 4.6: Tower base expected fatigue damage due to axial stress as a function of cross
section location (90◦ and 270◦ are on the wind propagation axis) for semi 1 for Hs =5 m,
Tp =14 s and Uw =10 m/s. The wind direction is 0◦ and the wave direction is βwave .
49
−4 Tower Base DRFC(σx)
x 10
2
Spar
EC1
0
0 30 60 EC2 90
−4 EC3
x 10
1.5 EC4
EC5
1 EC6
TLP
0.5
0
0 30 60 90
−4
x 10
1.5
Semi−Sub 1
0.5
0
0 30 60 90
−5
x 10
Semi−Sub 2
0
0 30 60 90
−5
x 10
βwind = 90 deg
Semi−Sub 1,
0
0 30 60 90
βwave, deg
Figure 4.7: Tower base fatigue damage due to axial stress (caused by tower bending).
Note that the vertical scale varies among different platforms. EC1-EC6 denote the envi-
ronmental conditions that are listed in Paper 3.
50
three concepts. Figure 4.7 compares the expected maximum short term fatigue damage
due to nominal axial stress as a function of the wave direction. The maximum value that
occurred around the circumference was selected for each load case. Figure 4.7 also shows
that wind coming from 90◦ gave lower fatigue damage than the base case 0◦ . However,
another case that could be critical, based on the observations in this study, is the case were
both wind and waves come from 60◦ .
Similarly to what was seen in the work with Paper 2 and Paper 4, blade passing frequency
resonance occurred and gave significant fatigue damage also for the TLP and semi 2.
This explains why the benign environmental conditions EC1 (for the TLP) and EC2 (for
the semi-submersibles) show high damage in Figure 4.7. The wind speed at which the
3P resonance effect occurred varied slightly between the concepts, as a consequence of
different eigenfrequencies for the first flexible mode of the system.
An important result reported in this study was the observed low-frequency damping of
wind response due to wave motions - most likely a second order effect. This effect caused
increased surge motions in misaligned waves for the spar and the semi-submersibles in
the intermediate load cases.
4.5 Frequency versus time domain for fatigue analysis

(Paper 4 [60])
Time domain analyses of FWT systems are very time consuming, but necessary to be
able to capture non-linear effects and coupled response to wind, waves and control sys-
tem. Linearised frequency domain analyses have, however, given reasonable results for
response analyses for offshore bottom fixed and floating wind turbines in earlier studies,
as described in Chap. 2.
A frequency domain (FD) method for calculating tower base bending moments due to
wind- and wave loads was developed and compared with time-domain (TD) analyses. A
model based on a rigid structure was considered. In addition, an approximate representa-
tion, based on the generalised degree of freedom method, of excitation of the first fore-aft
bending mode of the tower was investigated for the FD solution. Timeseries of bending
moment were found by inverse Fourier transform of the FD variance spectra for 10 seeds,
and the standard deviations of the tower base bending moment were compared to those
of the TD solutions. Nominal stress in the tower base was calculated from the bending
moment histories, and fatigue damage was calculated by an S-N approach.
First, wind and wave analyses were performed separately, then the timeseries of bend-
ing moment from the FD analyses were superimposed and compared to the coupled TD
analyses.
Table 4.3 shows an overview of the different analyses performed. For a more thorough ex-
planation, including equations for the different models, see Paper 4. The three variations
51
of the frequency domain method can be characterised as follows:
• FD1 assumes a rigid structure.
• FD2 includes excitation of the first fore-aft bending mode of the turbine (tower and
blades) due to the direct wind force and moment, and due to platform motion.
• FD3 includes excitation of the first fore-aft bending mode of the turbine due to the
direct wind force only.
Table 4.3: Analysis method overview.
Time/ Lin./ 3D/ Flexible turbine

freq. non-lin. 2D model
TD time non-lin. 3D full
FD1 freq. lin. 2D none
FD2 freq. lin. 2D excited by the direct wind force
(for cases including wind)
and wave induced motion
FD3 freq. lin. 2D excited by direct wind force
Figure 4.8 shows a comparison of the results for the wave-only frequency domain analysis
where two frequency domain methods are considered (FD1 and FD2). The differences
in standard deviation between the FD and TD estimates of tower base bending moment
(My), short term fatigue damage (Drf c ), platform pitch motions (Y5 ) and surge and pitch
accelerations(Ÿ1 , Ÿ5 ) are shown. The prediction of platform motions and accelerations
gave below 14% deviation compared to the TD simulation. It is inherent in the method
that platform motions and accelerations are the same for all the FD methods.
The assumption that the turbine is rigid (FD1) seemed to give lower tower base bending
moment amplitudes in the wave-frequency range than observed for TD simulations, even
if the frequency domain prediction of motions and accelerations were good. Figure 4.8
shows that including the effect of excitation of the first bending mode of the turbine (FD2)
gave higher bending moment estimates, and thus came closer to the TD results.
In the evaluation of DRF C (the damage calculated by Rainflow counting of stress histories
and S-N curves) for FD1-FD2 compared to TD simulations, FD2 provided the best result
for all the wave-only cases (Figure 4.8).
The variance spectra for tower base bending moments from the FD wave-only analysis
were compared to spectra from TD simulations (see Paper 4). When applying the flexible
dynamic amplification correction (FD2), the solution improved. This indicates that part of
the reason why the bending moment was underestimated by the rigid turbine assumption,
is dynamic amplification of the lowest fore-aft bending mode of the turbine. The spectra
also showed that the flexible model captures some of the the excitation of the first eigen-
mode of the turbine (at 2.73 rad/s), however, with too low amplitudes for wave periods
9-10 s. The 3ω for these periods are 1.9-2.1 rad/s, i.e. there will be non-linear excitation
52
20 Std(Y )
5
Std(Ÿ )
1
0 Std(Ÿ )
5
Std(My) FD1
Drfc FD1
−20 Std(My) FD2
Error (%)
Drfc FD2
−40
−60
−80 0.5 2.0 2.0 2.0 3.0 2.0 3.5 5.0 3.5 3.5
9 9 12 16 10 11 15 13 12 14
Hs (m)/Tp (s)
Figure 4.8: Wave-only cases. Errors in FD estimate compared to TD simulations (mean

value over 10 samples). Results are shown for two frequency domain methods: FD1 and
FD2.
in the region of the first eigenmode of the turbine, which cannot be captured by a linear
FD model.
Figure 4.9 shows how the wind-only FD analyses compared to the TD analyses. The
standard deviations of accelerations were not predicted very well, but the standard devi-
ations of bending moment were still close to the TD results. This could mean that the
bending moments directly from the rotor force contribute significantly more to the tower
base bending moments than the wind induced platform motions. Figure 4.9 clearly shows
how including the wind force excitation of the lowest eigenmode of the turbine improves
the solution (FD3). FD2, which contains the same assumptions as FD3, and in addition
includes excitation of the flexible turbine mode due wind induced motion, gave slightly
less accurate estimates than FD3. This may be because FD2 includes motion induced
dynamics, and as seen in Figure 4.9, wind induced motions were not very well predicted.
The low-frequency wind induced accelerations were underestimated by the FD model for
all wind speeds, which can explain the poor estimate of wind induced platform acceler-
ation in Figure 4.9. The low frequency response proved to be very sensitive to both the
level of linearised hydrodynamic damping and the level of aerodynamic damping (which
was derived for a fixed turbine), but the main cause of poor representation of the low
frequency content is presumably non-linear effects being lost in the linearisation of the
thrust force.
By looking at the relative error for the FD estimates compared to the TD fatigue damage,
the errors range between 24-59% for wind-only, 1-49% for wave-only, and 8-50% for the
combined wind- and wave cases, even if bending moment prediction errors were much
53
20
Std(Y5 )
Std(Ÿ1 )
0
Std(Ÿ5 )
Std(My) FD1
−20 Drfc FD1
Std(My) FD3
Error (%)
−40 Drfc FD3

Std(My) FD2
−60 Drfc FD2
−80
−100 5 8 10 12 13 14 17 20
Uw (m/s)
Figure 4.9: Wind-only cases. Errors in FD estimate compared to simulations (mean value
over 10 samples). Results for three frequency domain methods: FD1-FD3.
20
Std(Y )
5
0
Std(Ÿ )
1
Std(Ÿ )
5
−20 Std(My) FD1

Drfc FD1
Error (%)
−40
Std(My) FD3
Drfc FD3
−60
Std(My) FD2
Drfc FD2
−80
−100 5 5 8 8 10 12 12 13 14 14 17 20 20
0.5 2.0 2.0 2.0 3.0 2.0 3.5 2.0 5.0 3.5 3.5 3.5 5.0
9 9 12 16 10 11 15 16 13 12 14 12 13
Uw (m/s)/Hs (m)/Tp (s)
Figure 4.10: Combined wind and waves. Errors in superimposed FD estimates compared
to TD simulations (mean value over 10 samples). Results are shown for FD1-FD3.
54
lower. This is an expected effect since using an S-N approach expresses the damage as
proportional to the stress amplitude to the power of m, which in this case was 3 or 5.
For combined wind and waves, Figure 4.10 shows that the errors were in the same range
as for the wind- and wave-only simulations, and the trends were similar, i.e. no obvious
coupling effects were observed in these results, except for the aerodynamic damping of the
wave induced motions, caused by an operating rotor, which was included in the combined
FD analysis.
An important coupling effect is the low-frequency aerodynamic damping due to wave
motion. However, the low-frequency motions were already "damped" in the FD analysis,
since the low-frequency content of wind-induced response was lost in the linearisation of
the thrust force. This caused the FD predictions of the combined wind and wave response
to be fairly close to the coupled TD response, even though it was not derived from physics
assuming wave motions.
55
56
Chapter 5
Conclusions and recommendations for

further work
The current chapter summarises the contributions to the community of FWT researchers
that resulted from the investigations performed during the course of the PhD period. Sug-
gestions for further work, based on issues that were highlighted during the investigations,
are described in this chapter.
5.1 Conclusions
This thesis primarily deals with development of numerical methods and their application
to fatigue design in an efficient manner, using a semi-submersible as a case study. The
emphasis has been on global response analysis for use in fatigue design, thus determi-
nation of stress concentrations were kept out of the scope. Coupling existing structural
(Riflex) and aerodynamic (AeroDyn) codes, to enable fully coupled simulation of the
semi-submersible, was also part of the work with this thesis.
Comparing wave force models based on a simplified model based on Morison forces de-
rived by a potential theory panel model (Paper 1) showed that by tuning the Morison
inertia coefficients, the same response could be achieved. However, these inertia coeffi-
cients could not be directly derived from potential theory, and neither could they be set to
a constant for all wave periods. Diffraction effects were found to be important for wave
periods below 7 s. The study also highlighted the importance of including the force due
to hydrodynamic pressure under surface piercing columns in the Morison model. The
analyses of an operating wind turbine with the two different theories showed that pitch
motion has large influence on the mean and the variation of power production, thus it is
important to model pitch motion correctly.
The successive studies concerned with simulation procedures and selection of environ-
mental conditions for predicting fatigue damage, with the objective of possibly reducing
57
the computation time without significantly compromising the accuracy. Sensitivity analy-
sis of simulation length showed that calculating fatigue damage from 1-hour simulations
gave less than 4% error in the total fatigue life compared to 3-hour simulations. 10-minute
simulations, which are normal for LWTs, gave larger errors for the platform components,
but gave less than 3% error for the tower base. This means that 10-minute simulations
were not sufficient for fatigue analysis of components with stresses mainly induced by
wave loads, but that 10-minute simulations may give a reasonable result for components
that are more influenced by wind loads. However, the Kaimal wind spectrum, that was
applied here, does not contain the long period variations which could cause response that
is more sensitive to simulation length.
The RMS error compared to the average of 10 3-hour realisations was less than 1% for 7
seeds of 3-hour simulations and less than 2% for 9 seeds of 1-hour simulations. Fatigue
damage in the platform members was more sensitive to the number of realisations and
simulations duration than the damage in the tower base.
While some standards for offshore wind turbines (e.g. DNV-OS-J101) recommends bin
sizes to be 1 m/s for U , 0.5 m for Hs and 0.5 s for Tp , when selecting environmental
conditions for running a complete fatigue life assessment, the findings in Paper 2 suggest
that increasing the bin size for U to 2 m/s will give an acceptable fatigue damage estimate
if the wind speed representative of each bin is selected with care. It was found that the
fatigue life was fairly insensitive to the Tp bin size, but careful selection of the represen-
tative Tp , such that resonance frequencies are included, is also important here. The study
did not include enough cases for the smallest bin size of Hs to make a firm conclusion on
the proper bin size, but since damage does not vary linearly with wave height, choosing
the midpoint of the Hs bin is not expected to be a conservative approach. If the wave bin
size is to be increased from the recommended value, a different method of choosing the
bin representative Hs should be considered.
In the studies of misaligned wind and wave conditions, it was found that motions per-
pendicular to the wind increased compared to the co-directional case when waves came
from the side. This effect was expected since the rotor damping was no longer present in
the misaligned case. However, since the forces contributing to fatigue damage act around
different axes, the stresses get a more favourable distribution around the circumference
of the tower cross section, and thus aligned wind and waves gave the highest fatigue
damage. This conclusion was drawn for all the four different concepts studied in Paper
3: A spar, a TLP and two semi-submersibles. The direction of the co-directional wind
and waves is nevertheless important; applying the direction that gives the highest wave
induced motions is expected to be a conservative approach.
As for all the other case studies, nominal stress in a net cross section of the tower base
was assumed in the investigations of misaligned wind and wave conditions. But in reality
there will be hot spots at different locations around the tower base circumference. This
has to be taken into account when adopting the conclusions from Paper 3.
A new frequency domain method to estimate the variance spectrum of tower base bend-
58
ing moment due to wind and wave loading was established in Paper 4. In addition to
establishing a general frequency domain model by assuming the structure to be rigid, a
generalised coordinates model was developed to include the flexural response of the first
fore-aft bending mode of the turbine tower and blades. In the moderate sea states studied,
the bending moment calculated by the frequency domain method compared fairly well to
non-linear time domain simulations. Low-frequency motions induced by the wind were
not very well represented by the frequency domain model, which seemed to be caused by
the linearisation of the thrust force. But since the bending moments seemed to be dom-
inated by the moment directly from the thrust force, the inaccurate motion estimates did
not have a large impact in this study.
When it comes to calculating fatigue damage by the frequency domain method, larger
errors were observed, which is an expected effect due to the exponential relationship
between fatigue damage and stress range. Still, the method gave good predictions for
some environmental conditions, and can provide a rough estimate of fatigue damage in
early design phases.
A coupling effect that was observed in the studies in Papers 3 and 4 is the change in
the wind-frequency content of the thrust force when waves are present. This effect was
confirmed by isolating the thrust force (see Paper 3) from an analysis without waves
and re-applying it in an analysis with waves. The physical cause of this effect was not
identified, but is expected to be either a result of non-linear terms in the thrust force
formulation or an effect inherent in the method for calculating aerodynamic loads (BEM).
5.2 Recommendations for further work
The following suggestions for further work followed from the investigations in Papers
1-4:
• The conclusions listed above are only proven for a particular semi-submersible
wind turbine (and a spar, a TLP and a different SSWT in Paper 3). Further in-
vestigations regarding the application of the methods outlined in the papers to other
concepts should be made.
• Uncertainties relating to the level of aerodynamic damping at low to intermediate
wind speeds emerged in the studies in Paper 2. BEM and GDW gave very different
results in 3P resonance of the tower first fore-aft bending mode. Wake dynamics
could be important at these wind speeds, but the GDW method is unstable for highly
loaded rotors. These uncertainties should be assessed to enable proper modelling at
low to intermediate wind speeds.
• 3P resonance was observed in Papers 2-4, also for another semi-submersible con-
cept and for a TLP. The tower should be carefully adapted to each platform type,
such that the eigenfrequencies are outside of the 3P range.
59
• Long term variations in the wind can cause responses in compliant FWTs like the
semi-submersible. Most wind spectra applied for wind turbines today do not con-
tain energy at low frequencies. These low frequency variations are not important for
bottom fixed structures. For FWTs, however, long term wind variations could have
an effect, and thus this should be investigated, using wind spectra from offshore
measurements.
• The frequency domain analysis in Paper 4 applied a linearised thrust force by taking
the variance spectra of the timeseries of thrust force for a floating wind turbine. This
method gave less low-frequent response than the non-linear model, but a better
linearisation method could improve frequency domain analysis for floating wind
turbines.
60
References
[1] American Bureau of Shipping. Floating Offshore Wind Turbine Installations, 2013.
[2] O. J. Andersen and J. Løvseth. The Frøya database and maritime boundary layer
wind description. Marine structures, 19(2):173–192, 2006.
[3] A. Arapogianni and et. al. Deep water - the next stop for offshore wind energy.
Technical report, EWEA, 2013.
[4] Glosten Associates. Technology summary: Pelastar for offshore wind. http:
//www.pelastarwind.com/. Accessed Dec 11, 2013.
[5] A. Aubault, C. Cermelli, and D. Roddier. Windfloat: A floating foundation for
offshore wind turbines part III: Structural analysis. In Proceedings of the ASME
28th International Conference on Ocean, Offshore and Arctic Engineering, pages
213–220, Honolulu, USA, 2009. Paper no. OMAE2009-79232.
[6] E. E. Bachynski. Design and dynamic analysis of tension leg platform wind tur-
bines. PhD thesis, Norwegian University of Science and Technology, 2014. Thesis
no. 4-2014.
[7] E. E. Bachynski, M. Etemaddar, M. I. Kvittem, C. Luan, and T. Moan. Dynamic
analysis of floating wind turbines during pitch actuator fault, grid loss, and shut-
down. Energy Procedia, 35:210–222, 2013.
[8] E. E. Bachynski, M. I. Kvittem, C. Luan, and T. Moan. Wind-wave misalignment
effects on floating wind turbines: Motions and tower load effects. Journal of Off-
shore Mechanics and Arctic Engineering, 136(4):041902, 2014.
[9] E. E. Bachynski and T. Moan. Linear and nonlinear analysis of tension leg plat-
form wind turbines. In Proceedings of the 22nd International Ocean and Polar
Engineering Conference, Rhodes, Greece, 2012.
[10] D. Benasciutti and R. Tovo. Cycle distribution and fatigue damage assessment in
broad-band non-gaussian random processes. Probabilistic engineering mechanics,
20(2):115–127, 2005.
[11] D. Benasciutti and R. Tovo. Spectral methods for lifetime prediction under wide-
band stationary random processes. International Journal of fatigue, 27(8):867–
877, 2005.
61
[12] J. S. Bendat and A. G. Piersol. Random data: analysis and measurement proce-
dures. John Wiley & Sons, 2011.
[13] M. M. Bernitsas. Viscous forces on circular cylinders in separated flows. case
study: Cable strumming. Lecture Notes for NA 621, 1982.
[14] B. H. Bulder, M. Th. van Hees, A. Henderson, R. H. M. Huijsmans, J. T. G. Pierik,
E. J. B. Snijders, G. H. Wijnants, and M. J. Wolf. Study to feasibility of and
boundary conditions for floating offshore wind turbines. Technical report, ECN,
MARIN, Lagerway the Windmaster, TNO, TUD, 2002.
[15] T. Burton, N. Jenkins, D. Sharpe, and E. Bossanyi. Wind energy handbook. John
Wiley & Sons, 2011.
[16] N. Clauss, E. Lehmann, and C. Ostergaard. Offshore Structures - Conceptual de-
sign and Hydromechanics, volume I. Springer-Verlag, 1988.
[17] R. D. Cook, D. S. Malkus, M. E. Plesha, and R. J. Witt. Concepts and applications
of finite element analysis. Wiley, 2002.
[18] A. Cordle and et al. Project UpWind: State-of-the-art in design tools for floating
offshore wind turbines. Technical report, UpWind, 2010.
[19] A. J. Coulling, A. J Goupee, A. N. Robertson, J. M. Jonkman, and H. J. Dagher.
Validation of a fast semi-submersible floating wind turbine numerical model with
deepcwind test data. Journal of Renewable and Sustainable Energy, 5:023116,
2013.
[20] L. Cradden, D. Ingram, and T. Davey. Marine renewable integrated application
platform - site assessment. Technical Report Deliverable D2.1, University of Ed-
inburgh, UK, 2010.
[21] DCNS. DCNS home page. http://en.dcnsgroup.com/. Accessed: Feb
23, 2014.
[22] Det Norske Veritas. DNV-RP-C205, Environmental Conditions and Environmental
Loads, 2007.
[23] Det Norske Veritas. SESAM User Manual Wadam, 2008.
[24] Det Norske Veritas. DNV-RP-C103, Column-stabilised units, 2010.
[25] Det Norske Veritas. DNV-RP-C203, Fatigue Design of Offshore Steel Structures,
2010.
[26] Det Norske Veritas. DNV-OS-C103, Column-stabilised units, 2012.
[27] Det Norske Veritas. DNV-RP-J103, Design of Floating Wind Turbine Structures,
2013.
[28] T. Dirlik. Application of computers in fatigue analysis. PhD thesis, University of
Warwick, 1985.
62
[29] W. Dong, T. Moan, and Z. Gao. Long-term fatigue analysis of multi-planar tubular
joints for jacket-type offshore wind turbine in time domain. Engineering Struc-
tures, 33(6):2002–2014, 2011.
[30] O. M. Faltinsen. Sea Loads on Ships and Ocean Structures. Cambridge University
Press, 1990.
[31] Ø. Fredheim. Fatigue analysis of column-brace connection in a semi-submersible

wind turbine. Master’s thesis, Norwegian University of Science and Technology,
2012.
[32] I. Fylling, K. Mo, K. Merz, and N. Luxcey. Floating wind turbine-response analysis
with rigid-body model. In Proceedings of the European Offshore Wind Conference
and Exhibition, September, pages 14–16, 2009.
[33] Z. Gao and T. Moan. Frequency-domain fatigue analysis of wide-band stationary

gaussian processes using a trimodal spectral formulation. International Journal of
Fatigue, 30(10-11):1944–1955, October-November 2008.
[34] A. Goupee, B. Koo, K. Lambrakos, and R. Kimball. Model tests for three floating
wind turbine concepts. In Offshore Technology Conference, 2012.
[35] L. Haid, G. Stewart, J. Jonkman, A. Robertson, M. Lackner, and D. Matha.

Simulation-length requirements in the loads analysis of offshore floating wind tur-
bines. In Proceedings of the 32nd International Conference on Ocean, Offshore
and Arctic Engineering, Nantes, France, 2013. Paper no. OMAE2013-11397.
[36] S. Hauptmann, S. Mulski, M. Kühn, and L. Mauer. Advanced drive train mod-
eling in a virtual wind turbine using the multibody simulation code simpack. In
European Wind Energy Conference (EWEC). Milan, Italy: European Wind Energy
Association, 2007.
[37] T. Holmås. USFOS User’s Manual, version 8-1. 2004.
[38] T. Holmås and J. Amdahl. Design of floating wind turbines. In Computational

methods in marine engineering III, Trondheim, Norway, 2009.
[39] International Electrotechnical Commission. IEC61400-1, Wind turbines - Part 1:

Wind turbines, design requirements, 2005.
[40] International Electrotechnical Commission. IEC61400-3, Wind turbines - Part 3:

Design requirements for offshore wind turbines, 2009.
[41] F. Irgens. Mechanics of Materials (in Norwegian). Tapir, 2006.
[42] J. J. Jonkman. Definition of the floating system for phase IV of OC3. Technical
report, National Renewable Energy Laboratory, 2010.
[43] M. L. Buhl Jr. J. Jonkman. FAST user’s guide, 2005.
63
[44] Z. Jiang, M. Karimirad, and T. Moan. Dynamic response analysis of wind turbines
under blade pitch system fault, grid loss, and shutdown events. Wind Energy, 2013.
[45] Z. Jiang, T. Moan, and Z. Gao. A comparative study of shutdown procedures on
the dynamic responses of wind turbines. 2014. Under review.
[46] G. Jiao and T. Moan. Probabilistic analysis of fatigue due to gaussian load pro-
cesses. Probabilistic Engineering Mechanics, 5(2):76–83, June 1990.
[47] K. Johannessen, T. S. Meling, and S. Haver. Joint distribution for wind and waves
in the northern north sea. International Journal of Offshore and Polar Engineering,
12(1), 2002.
[48] B. J. Jonkman. TurbSim user’s guide, 2009.
[49] J. Jonkman, S. Butterfield, W. Musial, and G. Scott. Definition of a 5-MW ref-
erence wind turbine for offshore system development. Technical report, NREL,
2009. NREL/TP-500-38060.
[50] J. Jonkman and W. Musial. Offshore code comparison collaboration (OC3) for
IEA Wind Task 23 offshore wind technology and deployment. Technical Report
NREL/TP-5000-48191, National Renewable Energy Laboratory, 2010.
[51] J. M. Jonkman. Dynamics modeling and loads analysis of an offshore floating wind
turbine. PhD thesis, NREL, 2007.
[52] J. M. Jonkman. Dynamics of offshore floating wind turbines - model development
and verification. Wind Energy, 12(5):459–492, 2009.
[53] M. Karimirad, Q. Meissonnier, Z. Gao, and T. Moan. Hydroelastic code-to-code
comparison for a tension leg spar-type floating wind turbine. Marine Structures,
24(4):412–435, 2011.
[54] M. Karimirad and T. Moan. Wave-and wind-induced dynamic response of a spar-
type offshore wind turbine. Journal of waterway, port, coastal, and ocean engi-
neering, 138(1):9–20, 2011.
[55] M. J. Kühn. Dynamics and design optimisation of offshore wind energy conversion
systems. PhD thesis, Technisce Universiteit Delft, 2001.
[56] M. I. Kvittem. Comparison of rotor forces from Riflex-AeroDyn and RiflexBEM
for a semi-submersible wind turbine. Unpublished report, July 2013.
[57] M. I. Kvittem, E. E. Bachynski, and T. Moan. Effects of hydrodynamic modelling
in fully coupled simulations of a semi-submersible wind turbine. Energy Procedia,
24:351–362, 2012.
[58] M. I. Kvittem and T. Moan. Effect of mooring line modelling on motions and struc-
tural fatigue damage for a semisubmersible wind turbine. In The Twenty-second
International Offshore and Polar Engineering Conference. International Society of
Offshore and Polar Engineers, 2012.
64
[59] M. I. Kvittem and T. Moan. Time domain analysis procedures for fatigue assess-
ment of a semi-submersible wind turbine. Marine Structures, 2014.
[60] M. I. Kvittem and T. Moan. Frequency versus time domain fatigue analysis of a
semi-submersible wind turbine tower. Journal of Offshore Mechanics and Arctic
Engineering, 137(1):011901, 2015.
[61] M. I. Kvittem, T. Moan, Z. Gao, and C. Luan. Short-term fatigue analysis of semi-
submersible wind turbine tower. In Proceedings of the ASME 30th International
Conference on Ocean, Offshore and Arctic Engineering, pages 751–759. American
Society of Mechanical Engineers, 2011.
[62] D. Laino and A. C. Hansen. Aerodyn user’s guide, 2002.
[63] D. Laino and A. C. Hansen. User’s Guide to the Computer Software Routines
AeroDyn Interface for ADAMS, 2002.
[64] R. S. Langley. On the time domain simulation of second order wave forces and
induced responses. Applied ocean research, 8(3):134–143, 1986.
[65] P. K. Larsen. Design of steel structures (in Norwegian). Tapir, 1997.
[66] T. J. Larsen. How 2 HAWC2, the user’s manual, 2009.
[67] K. H. Lee. Responses of floating wind turbines to wind and wave excitation. PhD
thesis, Massachusetts Institute of Technology, 2005.
[68] L. Li, Z. Gao, and T. Moan. Joint environmental data at five european offshore sites
for design of combined wind and wave energy devices. In Proceedings of ASME
32th International Conference on Ocean Offshore and Arctic Engineering, Nantes,
France, 2013. Paper no. OMAE2013-10156.
[69] Maine International Consulting LLC. Floating offshore wind foundations: Industry
consortia and projects in the united states, europe and japan. Technical report,
2013.
[70] N. Luxcey, H. Ormberg, and E. Passano. Global analysis of a floating wind tur-
bine using and aero-hydro-elastic numerical model. Part 2: Benchmark study. In
Proceedings of 30th International Conference on Ocean Offshore and Arctic Engi-
neering, Rotterdam, The Netherlands, 2011. Paper no. OMAE2011-50088.
[71] J. F. Manwell, J. G. McGowan, and A. L. Rogers. Wind Energy Explained. Wiley,

2009.
[72] MARINTEK. SIMO Theory Manual, 2012.
[73] MARINTEK. RIFLEX Theory Manual, 2013.
[74] MARINTEK and Det Norske Veritas. Sesam User Manual, DeepC Theory.
65
[75] D. Matha. Model development and loads analysis of an offshore wind turbine on
a tension leg platform with a comparison to other floating turbine concepts: April
2009. Master’s thesis, University of Colorado, 2010.
[76] M. Matsuishi and T. Endo. Fatigue of metals subjected to varying stress. Japan
Society of Mechanical Engineers, Fukuoka, Japan, pages 37–40, 1968.
[77] T. Moan, Z. Gao, M. Karimirad, E. E. Bachynski, Etemaddar M., Z. Jiang, M. I.

Kvittem, M. Muliawan, and Y. Xing. Recent developments of the design and anal-
ysis of floating wind turbines. In ISCOT: Developments in fixed & floating offshore
structures, Busan, South Korea, The Royal Institution of Naval Architects, 2012.
[78] P. J. Moriarty and A. C. Hansen. AeroDyn theory manual, 2005.
[79] W. Musial, S. Butterfield, and B. Ram. Energy from offshore wind. In Offshore
Technology Conference, pages 1888–1898, 2006.
[80] A. Naess and T. Moan. Stochastic dynamics of marine structures. Cambridge

University Press, 2012.
[81] F. G. Nielsen, T. D. Hanson, B. Skaare, et al. Integrated dynamic analysis of

floating offshore wind turbines. In Proceedings of the ASME 25th International
Conference on Offshore Mechanics and Arctic Engineering, Hamburg, Germany,
2006. Paper no. OMAE2006-92291.
[82] NORSOK. N-003, Action and action effects, 2007.
[83] Windpower Offshore. Floating turbines - japan enters the stage.

http://www.windpoweroffshore.com/article/1211680/
floating-turbines---japan-enters-stage. Accessed: Oct
16, 2013.
[84] H. Ormberg and E. E. Bachynski. Global analysis of floating wind turbines: Code
development, model sensitivity and benchmark study. In Proceedings of the 22nd
International Ocean and Polar Engineering Conference, Rhodes, Greece, 2012.
[85] H. Ormberg, E. Passano, and N. Luxcey. Global analysis of a floating wind tur-
bine using an aero-hydro-elastic model: Part 1: Code development and case study.
In Proceedings of 30th International Conference on Ocean Offshore and Arctic
Engineering, Rotterdam, The Netherlands, 2011. Paper no. OMAE2011-50088.
[86] M. Philippe, A. Babarit, and P. Ferrant. Comparison of time and frequency domain
simulations of an offshore floating wind turbine. In Proceedings of the ASME 30th
International Conference on Ocean, Offshore, and Arctic Engineering, Rotterdam,
the Netherlands, 2011. Paper no. OMAE2011-49722.
[87] Marina Platform. Marina platform web page. http://www.

marina-platform.info/, 2014. Accessed September 19, 2014.
66
[88] The HiPRWind Project. Hiprwind web page. http://www.hyperwind.eu/,
2013. Accessed May 23, 2013.
[89] A. Robertson. OC4 phase II - 7th full committee meeting. Presentation, June 2013.
[90] A. Robertson and J. Jonkman. Loads analysis of several offshore floating wind
turbine concepts. In 21st International Offshore and Polar Engineering conference,
2011. Rep. no. NREL/CP-5000-50539.
[91] A. Robertson, J. Jonkman, Musial W., F. Vorpahl, and W. Popko. Offshore code
comparison collaboration, continuation: Phase II results of a floating semisub-
mersible wind system. Technical report, National Renewable Energy Laboratory,
2013.
[92] D. Roddier, C. Cermelli, A. Aubault, and A. Weinstein. Windfloat: A floating foun-
dation for offshore wind turbines. Journal of Renewable and Sustainable Energy,
2:033104, 2010.
[93] D. Roddier, A. Peiffer, A. Aubault, and J. Weinstein. A generic 5MW WindFloat
for numerical tool validation and comparison against a generic spar. In Proceedings
of ASME 30th International Conference on Ocean Offshore and Arctic Engineer-
ing, Rotterdam, The Netherlands, 2011. Paper no. OMAE2011-50278.
[94] I. Rychlik. A new definition of the rainflow cycle counting method. International
journal of fatigue, 9(2):119–121, 1987.
[95] B. Skaare, T. D. Hanson, F. G. Nielsen, R. Yttervik, A. M. Hansen, K. Thomsen,
and T. J. Larsen. Integrated dynamic analysis of floating offshore wind turbines. In
European Wind Energy Conference and Exhibition (EWEA), 2007.
[96] Statoil. Hywind. http://www.statoil.com/. Accessed: Jan 2, 2014.
[97] G. Stewart, M. Lackner, A. Robertson, J. Jonkman, and A. J. Goupee. Calibra-
tion and validation of a fast floating wind turbine model of the deepcwind scaled
tension-leg platform. In 22nd International Offshore and Polar Engineering con-
ference, 2012. Rep. no. NREL/CP-5000-54822.
[98] E. N. Strømmen. Theory of bridge aerodynamics. Springer, 2010.
[99] J. Twidell and G. Gaudiosi. Offshore Wind Power. Multi Science Publishing, 2009.
[100] J. Van Der Tempel. Design of support structures for offshore wind turbines. PhD
thesis, Technisce Universiteit Delft, 2006.
[101] W. J. M. Vijfhuizen. Design of a wind and wave power barge. Master’s thesis,
MS Dissertation, Department of Naval Architecture and Mechanical Engineering,
Universities of Glasgow and Strathclyde, Scotland, 2006.
[102] E. N. Wayman, P. D. Sclavounos, S. Butterfield, J. Jonkman, and W. Musial. Cou-
pled dynamic modeling of floating wind turbine systems. In Offshore Technology
Conference, volume 139. Houston, 2006.
67
[103] J. Weinzettel, M. Reenaas, C. Solli, and E. G. Hertwich. Life cycle assessment of
a floating offshore wind turbine. Renewable Energy, 34(3):742–747, 2009.
[104] J. E. Withee. Fully coupled dynamic analysis of a floating wind turbine system.
PhD thesis, Massachusetts Institute of Technology, 2004.
[105] Y. Xing, M. Karimirad, and T. Moan. Modelling and analysis of floating spar-type
wind turbine drivetrain. Wind Energy, 2013.
68
Appendix A
Appended papers
69
70
Paper 1
Effects of hydrodynamic modelling in fully coupled simu-

lations of a semi-submersible wind turbine
By Marit I. Kvittem, Erin E. Bachynski and Torgeir Moan

In Elsevier Energy Procedia, vol 24, p.351-362, 2012
71
72
Energy
Procedia
Energy Procedia 24 (2013) 351–362
www.elsevier.com/locate/procedia
DeepWind, 19-20 January 2012, Trondheim, Norway
Effects of hydrodynamic modelling in fully coupled

simulations of a semi-submersible wind turbine
Marit I. Kvittema,b , Erin E. Bachynskib , Torgeir Moanb

a NOWITECH, NTNU, NO-7491 Trondheim, Norway
b Centre for Ships and Ocean Structures, NTNU, NO-7491 Trondheim, Norway
Abstract
This work examines the dynamic response of a single semi-submersible wind turbine (SSWT) based on different
hydrodynamic theories. Comparisons of platform motions and structural responses in the wind turbine are shown for
simulations for a model with linear potential flow solution and quadratic drag and simulations with only Morison-type
forces. The SSWT modelled in this study is based on WindFloat and carries the NREL 5MW wind turbine and should
be considered a large volume structure. This implies that diffraction effects should be considered by using potential flow
theory and viscous effects by Morison’s equation.
A new coupled simulation code was developed by linking the SIMO and RIFLEX hydrodynamic, structural, and
control system computational tools, from MARINTEK, with the aerodynamic forces and wind field generation capabili-
ties of AeroDyn and TurbSim, from NREL. In contrast to other available simulation codes, this combination enabled the
implementation of these two different hydrodynamic theories and offered the possibility of finite element mooring line
models. Wave-only simulations were considered first, in order to tune and compare potential theory versus the inertia
term in Morison’s equation. Some limited coupled wave-wind simulations give an indication of the extent to which
hydrodynamic modelling affects the global response.
The SSWT case study showed that the Morison model with forces integrated up to wave elevation gave a good
representation of the motions compared to the potential flow model with quadratic drag forces. It also showed that mo-
tions are sensitive to choice of added mass coefficients, stretching and dynamic pressure under the columns. Combined
wind and wave simulations, using a non-optimized control approach, showed that pitch motions influence the power
production and blade bending moments.
c 2012 Published by Elsevier Ltd.

Keywords:
offshore wind energy, semi-submersible, integrated analysis, hydrodynamics
1. Introduction
The majority of commercial and academic software for analyzing floating wind turbines (FWTs) have
been developed from analysis tools for onshore wind turbines. This often means that the software includes
advanced aerodynamics and limited hydrodynamics. In many cases, the only hydrodynamics model is
slender element theory with Morison-type forces on the submerged part of the structure. A few analysis
programs for FWTs come from the offshore industry, often with advanced hydrodynamics but simplified
aerodynamics. Certain concepts, such as the spar buoy FWT, are slender enough to justify the use of
352 M. Kvittem et al. / Energy Procedia 24 (2013) 351–362
Morison’s equation [1] together with a simplified treatment of pressure that causes heave motion. For large
volume structures such as barges or semi-submersibles, however, diffraction effects may be significant. On
the other hand, applying Morison’s equation makes it possible to account for non-linear effects that come
from calculating the wave forces in the instantaneous position of the platform. These effects may also be
important. The consequences of applying different hydrodynamic theories have not yet been studied due to
limitations in analysis tools.
The Morison equation has been used as the hydrodynamic model for for semi-submersible wind turbine
(SSWT) analysis before, by for instance Phuc and Ishihara [2]. They conclude that the Morison model
compares well with model tests for a SSWT with slender elements in regular waves. This result cannot be
assumed to hold for WindFloat and most other semi-submersibles with large-diameter elements, and must
also be examined critically for irregular wave conditions. A code-to-code comparison for calculating mo-
tions of a 1/64 scale drill rig semi-submersible was conducted by the ITTC Ocean Engineering Committee
[3]. Their conclusion was that both potential forces and viscous forces should be included in the analysis,
and that the effect of wave height was large in heave.
In this work the effect of hydrodynamic load modelling for a single SSWT is investigated. The SSWT
design under consideration is very similar to the WindFloat concept [4]. Since a proper description of
hydrodynamic loads requires both potential theory and viscous drag, quadratic drag elements are included
in both models presented here. For simplicity, the drag coefficients are the same for both the potential and
the Morison model. Added mass coefficients were calculated based on the frequency dependent added mass
from the potential theory solution. After coefficients were found, regular and irregular wave analyses were
performed for both the Morison model and the potential theory model, and response characteristics were
investigated and compared. Limited analyses with an operating turbine and turbulent wind load were run to
study the effect of hydrodynamic modelling on power production.
2. Methodology
2.1. Potential Theory and Morison’s Equation

Two practical options for hydrodynamic load calculation in a global analysis are potential flow theory
and Morison’s equation. The first order potential flow theory applied here considers the solution of a lin-
earized boundary value problem for inviscid, incompressible flow about a rigid body. This approach, using a
panel method solution, accounts for Froude-Krylov forces and diffraction effects for large volume structures.
The resulting solution is frequency-dependent and linear with respect to wave amplitude.
Morison’s equation is a semi-empirical method for calculating wave loads on slender structures. For a
fixed cylindrical pile, Morison’s equation is equivalent to the potential flow solution when the wavelength to
diameter (λ/D) ratio is large and viscous effects are negligible [5]. Morison’s equation does not, however,
account for diffraction effects, which, as a rule of thumb for fixed cylinders, are important for wavelengths
shorter than five times the diameter [6]. Furthermore, the Morison formulation is extended to non-slender
members, including, for example, the heave plates of the semi-submersible in this study. Due to the quadratic
drag force and the formulation in terms of relative velocities and accelerations, Morison’s equation is solved
in the time domain with frequency-independent coefficients.
Although we are interested in a coupled multiple degree-of-freedom (DOF) system, let us consider the
time-domain equation of motion for a floating single DOF system in order to compare these hydrodynamic
models. According to pure potential flow theory, the single DOF system takes the form of Eq. 1 [7]:
Z ∞
(M + A∞ ) ẍ(t) + κ(t − τ) ẋ(τ)dτ + Cx(t) + K (x(t)) = F FK + F D (1)
−∞
where M is the dry mass, A∞ is the added mass for high frequencies, x(t) is the system displacement,
κ(t − τ) is retardation function accounting for frequency-dependent added mass and damping, C is the
hydrostatic restoring force, F FK is the Froude-Krylov force and F D is diffraction force. The first and second
time derivatives are expressed by ẋ and ẍ. In practice, we also include a quadratic damping term (Cq ) to
approximate viscous effects, as well as non-linear restoring forces (K(x(t))) from the mooring system. The
M. Kvittem et al. / Energy Procedia 24 (2013) 351–362 353
quadratic damping term depends on the difference between the water particle velocity (u) and the body
velocity. The resulting equation of motion for potential flow including viscous drag is then:
Z ∞
(M + A∞ ) ẍ(t) + κ(t − τ) ẋ(τ)dτ + Cx(t) + K (x(t)) = F FK + F D + Cq |u − ẋ| (u − ẋ) (2)
−∞
The same system, including nonlinear restoring forces, according to Morison-type wave loading is ex-
pressed by Eq. 3, where we have the same quadratic damping and no retardation function [7].
M ẍ(t) + Cx(t) + K (x(t)) = (ρw V + ma )a − ma ẍ + Cq |u − ẋ| (u − ẋ) (3)

In Eq. 3, we have introduced the density of water (ρw ), the volume of displaced water (V), a constant added
mass (ma ), and the water particle acceleration (a).
Examining these two formulations, we can see that it is possible to tune the Morison equation coeffi-
cients to obtain identical responses for a single DOF system in regular waves of constant amplitude. The
Morison model cannot, however, necessarily capture the equivalent hydrodynamic coupling effects for a
multiple DOF system as in the potential flow formulation. Furthermore, the frequency-dependence and
linear damping contributions are lost. On the other hand, the potential flow formulation does not consider
wave particle accelerations above the waterline, and requires solving the Cummins equation, which intro-
duces some computational cost. The applicability of these theories to the considered platform, which is
described in the following section, is considered in greater detail in Section 2.3.
2.2. Single Semi-submersible Wind Turbine

The model investigated in this study is a semi-submersible substructure very similar to WindFloat [4],
with the NREL 5MW turbine [8] (see Fig. 2.2) and OC3 tower [9]. The SSWT is modeled in SIMO/RIFLEX,
a tool for coupled analysis of moored floating structures. Mooring lines, tower and blades are modeled by
flexible RIFLEX elements, while the floating body is modeled as rigid in SIMO, with linear hydrostatic
stiffness and coupled frequency-dependent added mass and linear damping. The force and motion transfer
functions and retardation functions are calculated with Wadam [10] potential theory software. SIMO has
the option to attach elements with Morison force model to the body, which was done here to add quadratic
drag. In the Morison model, the added mass, damping and force transfer functions are set to zero, while
both acceleration terms (added mass) and quadratic damping terms from Morison’s equation are included.
In both models, the non-dimensional vertical drag coefficient, C D = 2Cq /ρA pro j , for the heave plate is
given in [4] as 7.5 and we assume a horizontal drag coefficient of 1.0 for the columns based on DNV-RP-
C205 [11]. As a first comparison of the models, Table 1 lists the natural periods of the system found by
decay analyses for potential theory and for Morison equation with Ca = 1.0 for both columns and heave
plates.
Mode Panel T (s) Morison T (s)

Surge 99.8 97.3
Heave 19.9 21.4
Pitch 39.9 40.7
Table 1: Damped natural periods, assuming Ca = 1.0 for

both horizontal and vertical Morison forces. C D = 7.5 for
Fig. 1: WindFloat (courtesy of Principle Power) heave plates and C D = 1.0 for columns.
Four different variations of the Morison model were studied: pure Morison forces as in Eq. 3, with
forces integrated up to mean water level (1) or up to wave elevation (2), pure Morison including the effect
of calculating forces at instantaneous position (3), and Morison with a correction for dynamic pressure
under the columns (4). Due to the surface piercing elements, this dynamic pressure correction gives a better
representation of the forces in the vertical plane. This correction in Morison model (4) is implemented via
an analytically derived force transfer function applied to the platform in SIMO.
2.3. Added Mass Coefficients

The dimensional added mass (ma ) for the Morison model is computed as in Eq. 4 for horizontal forces
on the columns and Eq. 5 for vertical forces on the heave plates. In Eqs. 4- 5, R is the radius of the column
or plate and L is the length of the column. Similar equations are found in [12] and [13] for cylinders and
caissons, respectively. The values of the coefficients Ca must be tuned for different wave frequencies.
mhor hor 2
a = ρw C a πRcol Lcol (4)
2π 3
mver ver
a = ρw C a R (5)
3 plate
Table 2 shows the range of wave cases considered. The wave heading is normal to the rotor plane (that
is, in the x direction) as displayed in Fig. 2.2, hence only surge, heave and pitch motions are studied. These
cases are applied in both regular and irregular waves, where for irregular waves the periods are peak periods
and heights are significant heights. Morison added mass coefficients are calculated based on frequency
dependent added mass from potential theory. Two of the cases are also run with wind loads for an operating
turbine (see Tab. 3).
Table 2: Load cases
Wave period (s) 3.7 4.0 5.0 5.5 6.0 6.9 10.0 15.0 20.0 21.0
Wave height (m) 1.0 1.0 1.0 1.0 1.5 2.0 3.0 6.0 9.0 9.0
These load cases represent a range of different conditions, where different hydrodynamic effects may
be of importance. The wave heights are selected to be appropriate for the given period. Figure 2 shows
where the load cases in Tab. 2 are located in a theory validity diagram based on the cylinder diameter. The
diagram is valid for a cylinder fixed to the ground [14], so it only gives an indication of which effects may
be important for the freely floating body with multiple components. The conclusions drawn from Fig. 2 are
that for wave periods 3.7 s to 6.0 s, diffraction may be important, while for periods 20 s and 21 s viscous
drag may be important. Both Morison and diffraction theory may be applied to the intermediate cases, where
inertial effects dominate.
2.4. Fully Coupled Analyses

In order to understand the impact of the different hydrodynamic theories on wind turbine simulation
results, it is also important to consider the full floating system. Table 3 summarizes the two chosen wind-
wave conditions: Condition 1, representing below-rated wind and gentle seas, and Condition 2, representing
a typical operational condition. The wind fields, generated in TurbSim, correspond to the normal turbulence
model (NTM) for class B of the IEC 61400-3 standard, with the power law applied for the shear profile
[15]. At lower wind speed, the characteristic turbulence intensity is somewhat higher. The wave conditions
were chosen to approximately correspond with the wind field characteristics according to typical North Sea
conditions given by Faltinsen, 1990 [5].
Simulations for identical wind and wave time series were carried out for a land-based wind turbine
(disregarding all wave input) and for two different hydrodynamic models: the potential flow theory with
Fig. 2: Regions of validity of potential flow theory and Morison’s equation (assuming fixed bodies) [14]
Table 3: Combined wind-wave simulation conditions
Condition 1 Condition 2
Mean Wind Speed (90 m) 8 m/s 16 m/s
Power Law Exponent [15] 0.14 0.14
Characteristic Turbulence Intensity (90 m) 19.16 % 16.30 %
Significant Wave Height 1.5 m 6.0 m
Peak Wave Period 6.0 s 15.0 s
Simulation Length (excluding transient) 1800 s 1800 s
additional quadratic drag, and the pure Morison formula with integration up to the free surface (Morison
model (2)), with coefficients as in Tab. 4. Wind and waves were applied in the x direction, i.e. normal to the
rotor plane as displayed in Fig. 2.2.
The same generator torque and blade pitch control system were applied to all of the models, with the
gains described in [8]. Floating systems generally require modifications to the control system in order
to avoid negative feedback in over-rated wind conditions [16, 17], when the thrust force at the nacelle
decreases for increasing relative wind speed. The horizontal velocity due to pitch at WindFloat’s nacelle in
these conditions is sufficiently low such that the destabilizing effect is small compared to the hydrodynamic
damping, but an improved control system will be included in future work. It is also important to note that
the aerodynamic model employed in these simulations does not account for dynamic wake effects, which
may have important consequences for floating wind turbines due to the sheared inflow, which is exacerbated
by the mean platform pitch. The dynamic wake option is available in the AeroDyn code for sufficiently
high wind speed, but the BEM option with dynamic stall was applied here for consistency at different wind
speeds.
3. Coupled Wind-Wave Simulation Tool
A new coupled simulation code (S-R-A) was developed by linking the SIMO [18] and RIFLEX [19]
hydrodynamic, structural, and control system computational tools, from MARINTEK, with the aerodynamic
forces and wind field generation capabilities of AeroDyn and TurbSim, from NREL [20]. The simulation
tool employs the finite element solver available in the combined SIMO/RIFLEX tool, passing position and
velocity information to the aerodynamic code via DLL at the first iteration of each time step. The DLL
returns lumped forces along the wind turbine blades. An external control system applies the generator torque
according to a look-up table and blade pitch commands via PI control as in the NREL 5MW definition [8].
3.1. Finite Element Model

In the finite element model, the wind turbine tower is modeled with axisymmetric beam elements, while
the blades consist of doubly symmetric cross sections. In contrast to the FAST model, the model includes
the torsional degree of freedom of the blades. The control system, which is also coupled to the finite element
program, applies appropriate torque directly to the low speed shaft and sets the angle of the rigid connection
between the hub and blade root. Additional details regarding the wind turbine module of SIMO/RIFLEX
(without AeroDyn and TurbSim) can be found in [21, 22].
3.2. Aerodynamic Model

The AeroDyn program provides both blade element momentum (BEM) and generalized dynamic wake
(GDW) models for the aerodynamic force calculation [23]. The results shown in this paper employ the BEM
method with the Beddoes dynamic stall model, but no dynamic wake effects.
3.3. Verification of Land-Based Wind Turbine Performance

Prior to using the S-R-A code for simulation of a floating offshore wind turbine, the global performance
of a land-based wind turbine was compared against available tools such as FAST and HAWC2 [24]. Simu-
lations of the NREL 5 MW wind turbine show good agreement regarding power production, rotor rotation,
blade loads and deflections and tower loads. Fig. 3, as an example, compares several performance indicators
for the FAST and S-R-A codes. The S-R-A results are shown for a fully flexible model and for a model with
exaggerated torsional stiffness for the blades. As shown, the control pitch required at higher wind speeds
decreases when the blades are flexible in torsion, in agreement with published results [25].
BEM
45
40 RotSpeed (rpm) (FAST)

BldPitch (deg) (FAST)
GenTorque (kN) (FAST)
35
RotSpeed (rpm)
BldPitch (deg)
30
GenTorque (kN)
RotSpeed (rpm) (TS)
25
BldPitch (deg) (TS)
GenTorque (kN) (TS)
20
15
10
0
5 10 15 20 25
Wind speed (m/s)
Fig. 3: Global performance indicators for the land-based NREL 5MW wind turbine. Rotor speed, blade pitch, and
generator torque are compared for FAST, fully flexible RIFLEX-AeroDyn, and RIFLEX-AeroDyn with exaggerated
torsional blade stiffness (indicated TS).
4. Results
4.1. Regular Wave Condition

First, added mass coefficients were calculated based on added mass from the potential theory solution,
see Tab. 4. These coefficients were used for the Morison inertia forces in further analyses, if not stated
otherwise. Figure 4 shows response amplitude operators (RAOs) using potential theory and four approaches
to Morison equation. Results shown in Fig. 4 are based on time domain analysis with regular waves,
normalized with the input wave amplitude. The response outside the wave frequency was filtered out, so
only linear wave excitation is included. The linear potential theory solution without quadratic drag is also
shown. The wave heights used for these analyses are listed in Table 2. When quadratic drag is included,
there is a quadratic relation between wave height and response, and this effect was not considered in this
paper since the quadratic drag coefficients were the same for the two models.
Table 4: Coefficients calculated from Eq. 5 based on A(ω) from potential theory solution.
T P (s) 3.7 4.0 5.0 5.5 6.0 6.9 10.0 15.0 20.0 21.0
Cahor (-) 0.51 0.76 0.65 1.05 1.35 1.18 1.07 1.06 1.03 1.03
Caver (-) 0.88 0.88 0.88 0.88 0.89 0.89 0.90 0.91 0.91 0.91
From Fig. 4 it is clear that pure Morison with forces calculated up to mean water level overestimated
heave and pitch motion compared to the potential theory and drag model, but by including forces up to
wave elevation we got a good agreement. Correcting for dynamic pressure under the columns also gave a
better fit, but this method can be improved by including forces up to wave elevation. For wave periods 20
and 21 seconds, surge motions for the modified Morison case were lower than the rest. This may be due to
surge-pitch coupling effects. Calculating the forces at the instantaneous position did not have a significant
effect for these cases.
Diffraction effects seemed to be important in heave response for periods lower than 6.9 s, which is in
agreement with the theory validity diagram in 2.
Fig. 4: Platform RAOs for surge, heave and pitch motion for variants of Morison and for potential and drag force.
Inertial coefficients as in Table 4.
In addition, regular wave analyses with different horizontal (subscript h) and vertical (subscript v) inertial
coefficients and integration of forces up to wave elevation were performed. The results are shown in Fig.
5, and it is clear that for each wave period the response in the potential theory model can be matched by
choosing the correct coefficient. And the correct coefficients are not necessarily derived from the potential
theory solution.
Fig. 5: Platform RAOs for surge, heave and pitch motion for pure Morison and for potential and drag force for different
inertia coefficients
4.2. Irregular Wave Condition

After determining the coefficients in Tab. 4, the potential theory model and the pure Morison model
with forces up to wave elevation were exposed to irregular wave conditions. Figure 6 shows the statistical
properties of the irregular wave response for both the Morison model and the potential and drag model.
Linear potential damping was equal to zero since it was significantly smaller than the drag damping. The
wave time series was the same for both models in each load case. These results were based on one single
time series per load case, where the simulation time is 30 minutes. This is not a sufficient statistical basis,
but gives an indication of trends because the purpose is to compare between the Morison model and the
potential plus drag model, and the wave time series are identical.
There was little difference in standard deviations in all degrees of freedom. The mean value of surge
motion was higher for the Morison model than for the potential model for long wave periods. This may
be due to the extra drift force caused by the inertia forces above mean water level acting on an asymmetric
structure.
Fig. 6: Statistical properties for response to irregular waves with no wind
4.3. Turbulent Wind and Irregular Waves

A brief investigation of responses in turbulent wind and irregular wave conditions is presented here.
One should note that these results correspond to limited stochastic analyses and that future work to reduce
statistical uncertainties is planned.
Statistical results for the Condition 1 and 2 simulations are shown in Tables 5 and 6, respectively. In
Condition 1, the wind forces were dominant, and the wind turbulence dominated the variability of all of
the performance parameters. As shown in Table 5, there was very little statistical difference in the power
production, turbine performance, blade loads, and floater motions when comparing simulations with the
potential and Morison theory hydrodynamics models. The power production of the floating platform was,
however, somewhat lower than that of the land-based tower. The reduction in power production of the
floating platform compared to the land-based tower was in part due to the mean platform pitch, which
decreases the efficiency of the energy harvesting, and also due to the pitch motion of the turbine, which
followed the wind and reduced the relative velocity seen by the blades. The reduced mean power led to a
corresponding decrease in power variability. The effect of the motions on the blade loads was also evident:
there was a clear increase in the out-of-plane bending moment due to gravity effects. In-plane loads, which
are dominated by gravity effects as the blades rotate, were not significantly affected by the floating platform
motions.
Table 5: Wind-wave Condition 1 simulation statistics
Land-Based WF - Potential + Drag WF - Morison

µ σ µ σ µ σ
Electrical Power (kW) 1896 732 1774 619 1774 617
Generator Torque (kNm) 20.47 5.73 19.59 5.06 19.59 5.04
Blade Pitch (deg) 0 0 0 0 0 0
Rotor Speed (rpm) 9.39 0.96 9.24 0.83 9.23 0.83
Blade Root Out-Of-Plane Bending Moment (kNm) 5812 1450 5967 1550 5970 1571
Blade Root In-Plane Bending Moment (kNm) 596 2564 561 2513 561 2510
Surge (m) n/a n/a 11.59 2.85 11.94 3.09
Heave (m) n/a n/a -0.01 0.03 0.04 0.03
Pitch (deg) n/a n/a 6.42 1.67 6.53 1.66
In Condition 2 (Table 6), the effects of the wave forces were more evident. As in Condition 1, the
electrical power output from the floating turbines decreased compared to the land-based turbine. In contrast
to Condition 1, however, the variability of the electrical power and generator torque increased for the floating
platform. The power and generator torque varied slightly more for the Morison model than for the potential
theory model, which can be attributed to the small increase in platform pitch motions.
Table 6: Wind-wave Condition 2 simulation statistics
Land-Based WF - Potential + Drag WF - Morison

µ σ µ σ µ σ
Electrical Power (kW) 4798 339 4767 384 4734 424.6
Generator Torque (kNm) 41.33 2.46 41.08 2.81 40.78 3.10
Blade Pitch (deg) 11.15 2.92 10.46 3.54 10.47 3.56
Rotor Speed (rpm) 12.10 0.25 12.09 0.27 12.09 0.30
Blade Root Out-Of-Plane Bending Moment (kNm) 5205 1645 5847 1850 5837 1900
Blade Root In-Plane Bending Moment (kNm) 1180 2621 1155 2524 1116 2510
Surge (m) n/a n/a 12.72 2.19 13.57 2.33
Heave (m) n/a n/a -0.01 0.61 0.06 0.64
Pitch (deg) n/a n/a 7.18 1.87 7.37 1.96
To further demonstrate the differences between the hydrodynamic models when applied to coupled sim-
ulations, power spectra results from Condition 2 are shown in Fig. 7. The top left panel shows the wind in
the global x direction measured at the hub for all three models as well as the wave elevation at the origin.
The platform pitch for the two hydrodynamic models is shown in the second panel. As shown, the Morison
model gave larger variation in platform pitch at both the wind and wave frequencies.
The low-speed shaft rotation speed (ω, shown in the third panel) is a complex result of the incoming
wind, generator torque and blade pitch control actions, inertial effects, and platform motions. The controller
is able to regulate the wind-driven variations (slower than 0.6 rad/s), but does not correct for variations in the
wave frequency range. Thus, the differences in platform wave-induced motion can be seen in the rotation
speed spectrum.
The blade 1 out-of-plane (OOP) bending moment is similarly difficult to dissect. The bending moment
showed strong variation related to the blade pitch angle (not shown, but consistent between all three models)
with large 1p variations. The platform pitch motion increased the amplitude of the 1p cycles, largely due to
gravitational loading.
40 20
X−dir Wind (Potential) Potential+Drag
Wave; Wind (m2s; m2/s)
X−dir Wind (Morison) Morison

30 15
X−dir Wind (Land−Based)
Pitch (deg s)
2
Wave
20 10
10 5
0 0
0 0.5 1 1.5 0 0.5 1 1.5
Frequency (rad/s) Frequency (rad/s)
−3 6
x 10 x 10
2.5 12
Blade 1 OOP ( (kN−m) s)
Potential+Drag
2
2 Morison 10
Land−based
8
ω (rpm2s)
1.5
6
1
4
0.5 2
0 0
0 0.5 1 1.5 0 0.5 1 1.5
Frequency (rad/s) Frequency (rad/s)
Fig. 7: Selected variance spectra: combined wind-wave Condition 2
5. Conclusions
A comparison of wave-induced response of a semi-submersible wind turbine using Morison equation

and potential theory with Morison drag was performed by means of the analysis software SIMO/RIFLEX
coupled to AeroDyn. First, added mass coefficients for the Morison equation were calculated based on the
potential theory added mass. Then variations of models with wave forces from Morison’s equation were run
in regular and irregular wave analyses and compared to the potential theory solution. A sensitivity study
on inertia coefficients was also performed. To study the effect of Morison versus potential theory on power
production, fully coupled analyses with wind and waves were performed.
The results from these studies showed that it is possible to obtain the same response amplitudes in
regular waves for periods above 7 seconds by choosing proper coefficients. For wave periods below 7
seconds, diffraction effects became important for heave motions. Surge motions were fairly insensitive to
the choice of added mass coefficients. Results also showed that inertia coefficients directly calculated from
the potential theory solution do not necessarily give the best agreement between Morison’s equation and
potential theory.
Pure Morison forces overestimated heave and pitch motion compared to the potential theory solution,
but by including forces above mean water level good agreement was achieved. Also, adjusting pure Morison
with dynamic pressure under the columns improved the results, though the method for including the dynamic
pressure has potential for improvement. Calculating the forces at the updated position of the platform did
not have a significant effect.
In the irregular wave analysis there were no significant differences between the two theories in the
standard deviation of the motions. Differences in the mean surge motion were observed, which may be
caused by inertial forces above the mean water level acting on an asymmetric structure.
Pitch motions decreased the relative wind velocity seen by the turbine blades and led to a decrease in
power efficiency compared to a land-based turbine. As a consequence of larger pitch motions predicted with
the Morison model than with the potential flow model, the predicted power production was more variable
than in the potential theory model.
Having analysis software capable of describing advanced hydrodynamics is important when studying
large volume structures, but for this particular semi-submersible the findings showed that slender body
theory by Morison is sufficient for the wave periods between 7 and 21 seconds. The coupled simulation tool
made it possible to study the impact of pitch motions on power production and blade bending moments,
which showed to be significant. A better control strategy is required to improve the power performance
and reduce the pitch motions of the platform. Large pitch motions influence extreme loads, fatigue life and
power production of the system, so having a good prediction of responses to wind and waves is crucial.
References
[1] M. Karimirad, Z. Gao, T. Moan, Dynamic motion analysis of catenary moored spar wind turbine in extreme environmental
condition, in: Proceedings of the European Offshore Wind Conference, Sweden, 2009.
[2] P. V. Phuc, T. Ishihara, A study on the dynamic response of a semi-submersible floating offshore wind system part 2: Numerical
simulation, in: Proceedings of the International Conferences of Wind Engineering 12, Cairns, Australia, 2007.
[3] M. Takagi, A comparison of methods for calculating the motion of a semi-submersible, Ocean Engineering 12 (1) (1985) 45–97.
[4] D. Roddier, A. Peiffer, A. Aubault, J. Weinstein, A generic 5MW WindFloat for numerical tool validation and comparison against
a generic spar, in: Proceedings OMAE 2011, 2011, paper no. OMAE2011-50278.
[5] O. M. Faltinsen, Sea Loads on Ships and Ocean Structures, Cambridge University Press, 1990.
[6] N. Barltrop, Floating Structures: A Guide for Design and Analysis, 1998.
[7] MARINTEK, SIMO Theory Manual (2009).
[8] J. Jonkman, S. Butterfield, W. Musial, G. Scott, Definition of a 5-MW reference wind turbine for offshore system development,
Tech. rep., NREL, NREL/TP-500-38060 (2009).
[9] J. Jonkman, Definition of the Floating System for Phase IV of OC3, Tech. rep., National Renewable Energy Laboratory (2010).
[10] Det Norske Veritas, SESAM User Manual Wadam (2008).
[11] Det Norske Veritas, Environmental conditions and environmental loads, Tech. Rep. DNV-RP-C205 (2007).
[12] N. Clauss, E. Lehmann, C. Ostergaard, Offshore Structures - Conceptual design and Hydromechanics, 1988.
[13] S. Chakrabarti, Added mass and damping of a TLP column model, in: Proceedings of the Offshore Technology Conference,
Houston, Texas, 1990, pp. 559–571.
[14] M. M. Bernitsas, Viscous forces on circular cylinders in separated flows. case study: Cable strumming., lecture Notes for NA 621
(1982).
[15] International Electrotechnical Commission, Wind turbines - Part 3: Design requirements for offshore wind turbines, Tech. Rep.
IEC61400-3 (2009).
[16] T. J. Larsen, T. D. Hanson, A method to avoid negative damped low frequent tower vibrations for a floating, pitch controlled wind
turbine, Journal of Physics: Conference Series, The Second Conference on The Science of Making Torque from Wind 75.
[17] J. M. Jonkman, Influence of control on the pitch damping of a floating wind turbine, in: Proceedings of the 2008 ASME Wind
Energy Symposium, no. NREL/CP-500-42589, 2008.
[18] MARINTEK, SIMO User’s Manual (2007).
[19] MARINTEK, RIFLEX User’s Manual (2009).
[20] National Renewable Energy Laboratory - National Wind Technology Center, NWTC design codes (nov 2011).
URL http://wind.nrel.gov/designcodes/
[21] H. Ormberg, E. Passano, N. Luxcey, Global analysis of a floating wind turbine using an aero-hydro-elastic model: Part 1: Code
development and case study, in: Proceedings of the 30th International Conference on Ocean, Offshore, and Arctic Engineering,
Rotterdam, the Netherlands, no. OMAE2011-50114, 2011.
[22] N. Luxcey, H. Ormberg, E. Passano, Global analysis of a floating wind turbine using and aero-hydro-elastic numerical model.
Part 2: Benchmark study, in: Proceedings of the 30th International Conference on Ocean, Offshore, and Arctic Engineering,
Rotterdam, the Netherlands, no. OMAE2011-50088, 2011.
[23] P. J. Moriarty, A. C. Hansen, AeroDyn theory manual, Tech. Rep. NREL/TP-500-36881 (2005).
[24] T. J. Larsen, How 2 HAWC2, the user’s manual, Tech. Rep. Risø-R-1597(ver. 3-9) (2009).
[25] J. Jonkman, W. Musial, Offshore code comparison collaboration (OC3) for IEA Wind Task 23 offshore wind technology and
deployment, Tech. Rep. NREL/TP-5000-48191, National Renewable Energy Laboratory (2010).
Paper 2
Time domain analysis procedures for fatigue assessment of

a semi-submersible wind turbine
By Marit I. Kvittem and Torgeir Moan

Accepted for publication in Journal of Marine Structures, 2014
85
86
Time domain analysis procedures for fatigue
assessment of a semi-submersible wind turbine
Marit I. Kvittem∗1,2 and Torgeir Moan2
1
NOWITECH, Norwegian University of Science and Technology, Trondheim, Norway
2
CeSOS, Norwegian University of Science and Technology, Trondheim, Norway
September 21, 2014
Abstract
Long term time domain analysis of the nominal stress for fatigue as-
sessment of the tower and platform members of a three-column semi-
submersible was performed by fully coupled time domain analyses in Simo-
Riflex-AeroDyn. By combining the nominal stress ranges with stress con-
centration factors, hot spot stresses for fatigue damage calculation can be
obtained. The aim of the study was to investigate the necessary simulation
duration, number of random realisations and bin sizes for the discretisa-
tion of the joint wind and wave distribution. A total of 2316 3-hour time
domain simulations, were performed.
In mild sea states with wind speeds between 7 and 9 m/s, the tower
and pontoon experienced high fatigue damage due to resonance in the
first bending frequency of the tower from the tower wake blade passing
frequency (3P).
Important fatigue effects seemed to be captured by 1 hour simulations,
and the sensitivity to number of random realisations was low when run-
ning simulations of more than one hour. Fatigue damage for the tower
base converged faster with simulation duration and number of random
realisations than it did for the platform members.
Bin sizes of 2 m/s for wind, 1 s for wave periods and 1 m for wave
heights seemed to give acceptable estimates of total fatigue damage. It is,
however, important that wind speeds that give coinciding 3P and tower
resonance are included and that wave periods that give the largest pitch
motion are included in the analysis.
Keywords: offshore wind, semi-submersible, integrated analysis, fatigue
1 Introduction
Fatigue damage is known to be a problem for bottom fixed offshore wind turbine
substructures, and is also expected to be significant to floating wind turbines
(FWTs). Adequate fatigue strength should be ensured by design. Wind parks
consist of units with similar designs, and are thus vulnerable to “common cause”
∗ [email protected]
1
failures, which means that the economic consequences of poor fatigue design are
serious. It is important, therefore, to make good fatigue estimates early in the
design process.
Whereas for a wave only case, environmental conditions are taken from a
scatter table (two variables: wave height and period), environmental conditions
for a combined wind and wave case have to be taken from a scatter block (three
variables: wind speed, wave height and period). This increases the number of
combinations of wind and sea states that needs to be included in the fatigue
damage assessment significantly compared to an onshore turbine or a tradi-
tional offshore structure. Wind- and wave directions and current are additional
parameters that further increase the number of load cases.
The equations of motion for a wind turbine on a compliant sub-structure
has many non-linear contributions: Catenary mooring line forces, viscous and
aerodynamic forces and large displacements that require the loads to be calcu-
lated at the updated position. Due to these non-linearities the wind and wave
loads on the structure cannot be treated separately, which means that all com-
binations of wind and wave loads must be analysed individually. Analysis of a
non-linear system must also be performed in the time-domain, which is much
more computationally time consuming than in the frequency domain.
Another issue is the discrepancy between guidelines for onshore wind tur-
bines and for floating platforms when it comes to simulation length require-
ments. Due to the long natural periods of a compliant floating platform, it is
often necessary to simulate from 3 to 6 hours to capture slowly varying response
to wave and wind loads. This is emphasized in the new offshore standard for
floating wind turbines from DNV [1]. Fixed wind turbines have higher natu-
ral frequencies, and the slowly varying response will be static, thus the normal
simulation time for wind turbines is ten minutes. It is also common practice
to assume ten minute stationary wind in wind statistics, whereas it is 1 to 6
hours for waves. Karimirad and Moan [2] found that a minimum of 3 hour
simulations were needed to capture extreme bending moments for a 5 MW spar
turbine, unless proper extrapolation was used. However, extreme values relate
to the ultimate limit state, and do not normally contribute to fatigue due to the
high return period.
In summary, all of these factors lead to a large number of environmental
conditions that need to be simulated in the time domain for one to 6 hours.
Also, to account for statistical uncertainty, a number of different realisations of
the wind and wave histories must be included in the fatigue assessment. This
requires unrealistic amounts of computing capacity and time in the design phase,
and is the motivation for studying the parameters that make the execution time
so long.
A recent paper by Haid et al. [3] studied the effect of simulation length on
fatigue and ultimate loads for the OC3 spar buoy wind turbine, and concluded
that the fatigue damage in the tower, blades and mooring system was more sen-
sitive to the treatment of residuals in rainflow cycle counting than to simulation
length. This work was done using the non-linear aero-hydro-servo-elastic tool
FAST.
Earlier work by the authors [4], analyses applying the simplified aerodynam-
ics model TDHmill in combination with Simo-Riflex indicated that 6-7 realisa-
tions of 1-hour wind and wave histories will give a fatigue estimate close to
the damage based on the average of 10 3-hour realisations. This was, however,
2
based on a limited number of environmental conditions.
The current study aims at assessing simulation requirements for fatigue dam-
age estimation, and the key questions are:
• How many realisations are needed to capture the effect of statistical un-
certainty?
• What simulation duration is necessary to capture the important effects of

slowly varying loads?
• What is the maximum bin size for the discretisation of the joint wind and
wave distribution?
Fatigue for a a three column, catenary moored semi-submersible with the

NREL 5MW [5] supported by the OC3 tower [6] (see Fig. 1) was examined. The
single semi-submersible wind turbine (SSWT) was inspired by WindFloat [7].
The simulation tool used was Simo-Riflex-AeroDyn from Marintek and CeSOS.
10 realisations of 197 3-hour environmental conditions were simulated, with a
CPU time approximately twice real time. The environmental conditions were
strategically chosen from a joint wind and wave distribution to identify special
expected and unexpected load effects.
These simulations serve as basis for comparisons of the fatigue damage for
varying simulation lengths (from 10 minutes to 3 hours) and for varying number
of realisations (1 to 10) of the same environmental conditions. The results were
also used to identify load cases and effects that contribute significantly to fatigue,
and to examine the effect of increasing the bin sizes for load case selection. The
effect of misaligned wind and waves was also examined.
2 Modelling of the semi-submersible wind tur-

bine and loads
2.1 Platform description
The WindFloat concept was chosen as inspiration for the floating platform in
this study, and the platform geometry given in [7] was kept, but since more de-
tailed specifications than what is available was necessary, a different design with
similar global characteristics was developed (Tab. 1). The new platform was
developed to have a mass distribution similar to WindFloat, but no extensive
design check was performed on the assumed wall thicknesses. Compared to the
OC4 DeepCWind concept, which is also a semi-submersible wind turbine con-
cept, the column plate thicknesses for this design are smaller, while the braces
and pontoons have thicker plates. The brace dimensions for the DeepCWind
platform were found sufficient in an initial design check [17] for the central North
Sea. Specifications for the turbine, the NREL 5MW reference turbine, can be
found in [5], and in [6] for the tower.
The control system constants from the OC3 Hywind spar study were applied
[6], with a constant torque strategy above rated wind speed. This controller
is tuned to avoid negative damping above rated wind speed due to the pitch
motion of the OC3 Hywind platform. This was done by setting the control
system natural frequency to 0.2 rad/s, which is outside of the wave frequency
3
(a) Structural model. Locations for
which fatigue results are presented
in Sec. 6 are marked with ’x’.
(-13.33, 23.00)
z
x (26.56, 0.00)
(-13.33, -23.00)
(b) Top view. The coordinate system z = 0 is in

the mean water line.
Figure 1: Semi-submersible wind turbine.
4
range and below the natural pitch frequency of the spar platform. 0.2 rad/s is
above the natural pitch frequency of the SSWT (0.17 rad/s) in this study, but
setting control constants to get below this will give too much variation in the
power production. Even though the controller natural pitch frequency could
not be set below this limit, applying the OC3 Hywind constants gave less pitch
motion than the land based controller. The negative damping instability was
not observed for this platform, probably due to the heave plates, which provide
viscous damping for pitch motion.
Table 1: Mass, damping and restoring data. All values are given for the platform
without the turbine (rotor-nacelle-assembly and tower), unless otherwise noted, and
are valid for zero thrust force ballast condition.
Displacement with WT 4810e6 kg

Mass with WT 4619e6 kg
Centre of gravity with WT (-0.331 m, 0.0 m, 1.489 m)
Mass 4019e6 kg
Centre of gravity (-4.300 m, 0.0 m, -7.857 m)
Radius of gyration Rxx 22.29 m
Radius of gyration Ryy 19.62 m
Radius of gyration Rzz 26.06 m
Radius of gyration Ryz 4.69 m
Hydrostatic stiffness Heave 2445 kN
Hydrostatic stiffness Roll 749815 kN m
Hydrostatic stiffness Pitch 7548571 kN m
2.1.1 Mooring
The platform is positioned by four mooring lines, two lines attached to column
1 and one line attached to each of the two other columns. In order to get
natural periods for surge, sway and yaw close to what is specified for the generic
WindFloat, a new mooring system was designed. A simple Riflex mooring line
model was used, together with the mooring line equations in Ref. [18], to achieve
the desired stiffness. The lines have 60 m of chain on top, 30 tonnes, 3.8 m3
clump weight between chain and polyester rope, 769.8 m of polyester rope in
the middle and 232.58 m of chain at the bottom.
2.1.2 Ballast system

The WindFloat platform has an active ballast system to keep the turbine upright
and thus maximise power output as the wind direction and intensity change.
This is done by pumping water ballast between the columns with a reaction time
of 20 minutes [7] to significant changes in the mean wind speed and direction.
The function of the ballast system is illustrated in Fig. 2 .
The initial ballast in the SSWT in these analyses was distributed to account
for the asymmetric loading of the platform, i.e. the weight of the turbine and
mooring system. During operation, the mean thrust force, tower drag and rotor
torque give the platform a constant tilt. To get an upright configuration of the
5
Table 2: Fairlead and anchor line positions.
x y z
Fairlead 1 31.56 0 -17
Fairlead 2 -15.78 -27.33 -17
Fairlead 3 -15.78 27.33 -17
Fairlead 4 31.56 0 -17
Anch 1 738.07 -706.51 -320
Anch 2 -721.76 -733.31 -320
Anch 3 -721.76 733.31 -320
Anch 4 738.07 706.51 -320
(a) Unballasted plat- (b) Ballasted platform

form with turbine in with turbine in opera-
operation. tion.
Figure 2: Illustration of ballast system.
6
platform in operation, a ballast system was modelled. Since the reaction time of
the ballast system is long compared to other the dynamic loads on the system, it
was modelled as constant ballast, and a new ballast distribution was calculated
for every mean wind speed. Thus, mass distribution and restoring was updated
for each environmental condition. The ballast distribution between column 1
and columns 2 and 3 (see numbering in Fig. 1a) was calculated from the thrust
force and tower drag for constant wind velocity with a power law wind shear.
The ballast distribution between column 2 and column 3 was calculated based
on the rotor torque.
Table 3: Mass distribution for different wind speeds. Mass and vertical position of
the centre of gravity (zG ) are given for the column, including heave plates and ballast.
Uhub Column 1 Column 2 Column 3

(m/s) Mass (kg) zG (m) Mass (kg) zG (m) Mass (kg) zG (m)
0.0 749,418 -10.358 1400,128 -7.918 1400,128 -7.918
8.0 658,582 -10.439 1441,151 -7.715 1449,940 -7.671
11.2 579,217 -10.384 1476,078 -7.539 1494,378 -7.446
25.0 659,277 -10.439 1435,887 -7.741 1454,509 -7.648
49.0 652,823 -10.440 1448,425 -7.678 1448,425 -7.678
Samples of masses and centre of gravity for different wind speeds are pre-
sented in Tab. 3. In the cases with wind coming in from the side, where the
rotor was rotated to be perpendicular to the wind, the same methodology to
balance the thrust force and torque was used.
2.1.3 Eigenperiods
Some important eigenperiods (damped), found by decay analyses of the no
thrust force ballast condition, are listed in Tab. 4. The tower first fore-aft
bending frequency is higher than the fixed NREL 5 MW tower frequencies, and
in this case that means that it will be within the range of the blade passing
frequencies for lower wind speeds. This would normally call for a redesign
of the tower. However, the analyses performed for this paper were also used
in a comparison study of fatigue for different FWT concepts, where the same
turbine and tower were used on different floating platforms [19]. In Ref. [19],
the resonance effect was seen also for a TLP and the semi-submersible used in
the OC4 study [20]. Thus, instead of optimising the design, it seems relevant
to keep this standard design and highlight the issue.
2.2 Coupled analysis code

The coupled analysis tool Simo-Riflex-AeroDyn was applied in this study. Simo-
Riflex from Marintek ([8], [9]) is suitable for time domain simulations of floating
rigid bodies (Simo) coupled to flexible beam elements (Riflex).
The structural model of the turbine tower, blades and shaft was made up
by beam elements. The rotor speed and the power output were controlled by a
generator torque controller a blade pitch controller, respectively. Aerodynamic
forces on the blades were calculated by AeroDyn [10], which reads a turbulent
7
Table 4: Damped natural periods for the platform ballasted for a zero wind condition.
Mode T (s)
Surge 107.0
Sway 124.8
Heave 19.9
Roll 35.6
Pitch 37.4
Yaw 68.5
Tower 1st 2.3
wind field and applies the aerodynamic nodal loads to the blade elements. This
method was verified by Ormberg and Bachynski [11].
Simo [12] solves the equation of motion in time domain for a single or multiple
rigid bodies based on a frequency domain solution of the hydrodynamic loads
by a panel method code (e.g. Wadam [13]). In addition, viscous forces can be
included by a drag formulation. The traditional modelling method is to assume
the complete platform as a rigid body. However, internal member forces of a
statistically indeterminate structure can not be extracted from such an analysis.
Thus, in order to get forces in the braces, the three columns of the platform
were modelled as individual Simo bodies with potential theory forces, and the
connecting braces were modelled as Riflex beam elements with Morison type
forces.
2.3 Hydrodynamic forces

2.3.1 Potential theory model for columns and heave plates
The dynamic equilibrium equation for the columns (and heave plates) include
mass forces, hydrostatic stiffness, gravity, buoyancy, external forces from Riflex,
retardation functions accounting for frequency dependent added mass and linear
damping, and wave excitation forces.
Wave force transfer functions according to the potential theory and the re-
tardation functions for the columns with heave plates were obtained using multi
body analysis in Wadam [13]. The heave plates were modelled by assuming a
flat plate at the bottom of the columns in the Wadam analysis.
Hydrodynamic interaction between the columns was taken into account by
applying the multi body analysis. However, some limitations on hydrodynamic
interaction are inherent in the coupled Simo-Riflex software. This means that
for one body, the effect of the presence of the other bodies was included in the
force transfer functions and retardation functions, but the effect of motion of
the other bodies was only accounted for in the force transfer functions, not in
the retardation functions. The wave forces from Wadam include Froude-Krylov
(FF K ) and diffraction (FD ) forces, as shown in Eq. 1.
FHY D = FF K + FD (1)
In addition to the wave forces according to linear potential theory, viscous
drag forces were included for both the columns and the heave plates. The drag
8
terms were calculated as described in the following section. The total wave
forces in 6 degrees of freedom then becomes a sum of Froude-Krylov, diffraction
and drag forces.
Difference frequency force transfer functions were not included since these
are computationally very expensive to find in a multibody panel model analysis.
A single body analysis in Wadam showed that the difference frequency forces
for this platform are most predominant in surge and sway. These motions
can contribute to fatigue, but in the current study excluding second order wave
forcing was a trade-off for being able to calculate fatigue damage in the platform
members.
2.3.2 Morison force model for beams, pontoons and braces

The wave forces on the slender beams, braces and mooring lines were calculated
by Morison’s formula (Eq. 2). The equation shows how the axial, lateral or
transversal force, dF acting on a strip of the member, in its local coordinate
system, are expressed in the simulations.
ρ
dF = ρAu̇ + ρACm (u̇ − r̈) + DCd |u − ṙ| (u − ṙ) (2)
2
A is the cross section area, D is the reference area normal to the flow di-
rection, i.e. the outer diameter for a circular cross section and the heave plate
area for heave motion of the heave plates, and ρ is the seawater density. u is
the wave particle velocity and r is the local member displacement, both in the
direction of the force. Cd is the non-dimensional drag term, which is direction
specific and Cm is the added mass coefficient, which is assumed to be zero for
the longitudinal direction.
Table 5 shows the Cm and Cd coefficients applied in the analysis. For the
mooring chains the non dimensional drag coefficients were taken from DNV-
OS-E301 [14]. A drag coefficient of 1.0 for the columns was chosen to get a
conservative estimate of the excitation forces on the columns. This value is
higher than recommended in DNV-RP-C205 [15] for fixed cylinders, and may
not give conservative damping values in surge and sway motions. Therefore,
a sensitivity analysis comparing fatigue damage in the tower and braces was
carried out with a drag coefficient of 0.7 for the columns, at relevant wave
periods. This analysis gave less than 1.5% difference in all cases, and provided
confidence that the results are not sensitive to the column drag coefficient. In
heave, pitch and roll, the heave plates contribute to most of the hydrodynamic
damping for this platform.
Vertical Morison drag terms for the heave plates and transversal/lateral drag
terms for the columns were calculated by the last term in Eq. 2. The heave
plate drag coefficient in Tab. 5 is based on model tests and numerical analysis
of the WindFloat concept [16].
2.4 Aerodynamic loads in low wind speeds

At low wind speeds, wake dynamics can be significant. AeroDyn has the option
of calculating the aerodynamic loads by simple BEM or by a generalized dynamic
wake (GDW) model. The BEM method in AeroDyn does not include a dynamic
wake correction, although it exists in other codes using BEM. In addition to
9
Table 5: Non-dimensional quadratic drag and added mass coefficients. For columns
and heave plates added masses are found by potential theory.
Component Transverse Cd (-) Longitudinal Cd (-) Cm (-)

Columns 1.0 0.0 -
Mooring chain 2.4 1.15 1.0
Mooring polyester rope 1.6 0.0 1.0
Main beam and pontoons 1.0 0.0 1.0
Braces 1.0 0.0 1.0
Heave plates 7.5 0.0 -
accounting for wake damping, GDW has the advantage of less computation
time than BEM. The validity of the GDW model, however, is limited to higher
wind speeds (unstable behaviour is reported for axial induction factors above
1/3, observed at 8 m/s and below for certain fixed turbines [10]). A floating
wind turbine will pitch and surge, which affects the relative wind speed, and
hence it will move in and out of the unstable range of GDW at wind speeds
close to the limit (8 m/s).
This lead to unphysical instabilities in the analyses with particular combi-
nations of waves and wind timeseries at 9 m/s. By changing to BEM, this effect
was avoided.
However, between 7 and 9 m/s tower resonance due to the blade passing
frequency (3P) load was observed, which led to high fatigue damage for these
cases. Damping determines the amplitude of resonant behaviour, and damping
may be lost or gained by including the dynamics of a wake. In a comparison
between a simulation with BEM and one with GDW, without the unstable be-
haviour, GDW gave significantly less fatigue for 9 m/s wind. Since this damping
is very important for the outcome of this study, and since the authors believe
that GDW gives a more physical representation of the aerodynamic loads at this
wind speed, GDW was used for the 9 m/s case. The few unstable analyses were
rerun with a different wind seed. For the 7 and 8 m/s case the rotor would be
too highly loaded, which would result in too many unstable analyses. Instead,
tower structural Rayleigh damping was increased from a stiffness proportional
coefficient 0.001 to 0.002, values which correspond to damping ratio of 0.3% and
0.6%, respectively, at the tower first fore-aft bending frequency. The original
OC3 tower specifies a structural damping ratio of 1.0%, which is constant for all
frequencies. Since only Rayleigh damping was available in the structural solver
used in this study, damping ratio was initially chosen to be 1.0% between the
first and second bending mode of the tower. Doubling the stiffness proportional
stiffness gives a damping ratio close to and below what was specified for the
OC3 tower, and is thus considered a conservative value.
10
3 Environmental data
3.1 Joint wind and wave distribution
In a study under the combined wind and wave power unit project, Marina
Platform, joint wind and wave distributions for five sites in European waters
were established [21], based on hindcast data of 1-hour averaged sea states and
wind. The WindFloat prototype is located in the Atlantic Sea, off the coast
of Portugal. Thus, the chosen environmental probability distribution of 1-hour
sea states applied in this analysis was taken from the Buoy Cabo Silleiro site,
off the coast of northern Portugal.
5.0e-03
20
2.8e-03
Hub wind speed (m/s)

1.4e-03
15
6.6e-04
3.0e-04
10
1.0e-04
2.3e-05
5
1.0e-06
5.0e-07
15
2 10 (s)
4 riod
6
Wave height (m 8
5 e pe
)
10 Wav
Figure 3: Joint wind and wave distribution.
3.2 Simulation length, wind and wave loads

Each environmental condition was simulated in Simo-Riflex-AeroDyn for 3 hours,
plus 350 seconds warm up and transient, with 10 different stochastic samples of
wind and wave timeseries. The 350 seconds of warm up and transient response
contains 50 seconds controller warm up and 150 seconds of constant wind, and
were removed from the results before post processing.
64-bit TurbSim [22] was used to simulate the 3-hour turbulent wind field.
Turbulence was generated applying the Kaimal spectrum and the IEC normal
turbulence model [23] with a reference turbulence intensity of 0.12. A power
law wind shear profile with exponent of 0.14 was used for the mean wind speed
component.
Since the power spectrum for wind measured over longer periods has small
variations in the 1-hour range [24], it seems to be valid to assume 1 hour sta-
tionary wind fields. More recent research suggests that this gap does not exist
[25], but the assumption has proven to give satisfactory models in the wind
industry [26]. For fatigue analysis the very low-frequency variation from 1-3
hour changes in the wind, will most likely not have any impact compared to the
high-frequency variations.
Irregular waves were generated in Simo with a 3-parameter Jonswap spec-
trum with a peak factor of 3.3, using Airy linear wave theory [12]. Long crested
11
waves were assumed. The misaligned wind and wave cases are described in Sec.
5.
4 Fatigue damage calculation

Axial stress σ in net cross sections of the different structural members was
calculated for 24 points around the circumference of the pipe cross sections. The
base metal cross sections of the members were used in the stress calculation and
it was assumed that hot spot stress amplitudes are proportional to the nominal
member stresses. Since it was the effects of simulation parameters that were
under investigation in this study, an SCF of 1.0 was applied (the actual hot spot
stress is proportional to this value). For a more detailed study with for multi
axial stress conditions, possible different SCFs for different stress components
need to be considered.
Nominal axial stress was calculated according to Eq. 4 [27]. Figure 4 shows
the definitions of directions and axis for the tower base.
loc 12
Wind loc 18 loc 6
θ z
loc 0
Figure 4: Top view of tower base cross section.
Nx My Mz
σ= + r1 cos θ − r1 sin θ (3)
A Iy Iz
Nx is the axial force, A is the nominal cross sectional area, My and Mz are
bending moments and Iy and Iz are the sectional moments of area.
Rainflow counting was performed for the stress timeseries and fatigue was
calculated by S-N-curves from DNV [28] and Miner’s sum. Residual half cycles
from the Rainflow counting were counted as half amplitudes.
Since fatigue damage occurs in welds rather than in the base material, S-N
curves for girth welds were used. Based on the joint geometry, S-N curve D for
σ was chosen, with categories ”air” for tower and ”cathodic protection” for the
other members. The offshore wind turbine standard DNV-OS-J101 [29] states
that the ”in seawater with cathodic protection” curve can be used for members
in the splash zone and water, assuming no repair and a guaranteed coating life
of 15 years. In this study the service life is 20 years, but since the aim of the
12
study was not to perform fatigue design, the 15 year coating lifetime rule is
disregarded.
To find the total damage over a service life of 20 years, damage estimates
were multiplied by the respective probability and a scaling factor to adjust for
the fact that the probability distribution yields for 1-hour sea states (T distr )
(see Eq. 4).
In traditional summations of fatigue damage, the probability is multiplied
by the bin volume to achieve a total probability of 1. When the environmental
conditions are not equally spaced, which was the fact in this study, it is not
PN lc
possible to find a bin volume, thus the scaling factor ptot / i=1 pi was intro-
duced instead. For the study of simulation length requirements and the number
of stochastic samples, this factor cancels out, since damage for the same envi-
ronmental conditions are compared, thus the factor will be a constant.
lc seeds
N N
ptot T distr 1 X X
Dtotal = N 20yr PN lc sim seeds
Dij pi (4)
i=1 pi T N i=1 j=1
In Eq. 4, Dtotal is the accumulated damage over 20 years for the included
environmental conditions, ptot is the total probability of all load cases within
the operation range of the turbine and with a probability above 10−4 , pi is
the point probability of environmental condition i, N 20yr is the number of 1-
hour sea states in 20 years, N seeds is the number of realisations of wind and
waves, N lc is the number of environmental conditions in the analysis, Dij is the
damage for a simulation for realisation number j of environmental condition i,
with duration T sim . T distr is the averaging period for the applied joint wind
and wave distribution, i.e. 1 hour.
Since this study focused on a turbine in operation, it was assumed that
it will be in operation the whole time during wind conditions between cut-in
and cut-out wind speeds. This is, of course, not a realistic assumption, but an
operation time factor will not affect the results in this study. For a full lifetime
fatigue assessment, downtime, along with survival, fault and installation, must
also be included in separate analyses.
5 Load case selection

The IEC standard for design requirements for offshore wind turbines [30] rec-
ommends the following bin size for joint wind and wave assessment:
• Mean wind speed: 2 m/s
• Significant wave height: 0.5 m

• Peak wave period: 0.5 s
The DNV standard for column stabilised units [31] specifies that fatigue
limit state analysis should cover events down to a probability level of 10−4 .
If the recommended bin sizes are used the total number of environmen-
tal conditions that have a marginal probability higher than 10−4 is 1539, not
accounting for the directionality of wind or waves. Each load case must be
simulated for at least 3-hours and with a certain number of different seeds for
13
random phase angles to capture worst representative wave loads. This requires
excessive simulation time and capacity.
The number of random seeds required by the IEC standard [30] is 6. But
since the structural response of a FWT to environmental loads is very different
from a fixed structure, more variability in the response statistics is a possi-
ble outcome of these analysis. Thus, with the intention of being able to see
convergence in the results, the number of seeds was increased to 10.
In this study, simulations were carried out in three phases: The first covered
the necessary simulation length and number of seeds, the second covered wind-
wave misalignment. After analysing the results from the first phase, additional
environmental conditions were added for the study of environmental condition
bin size in the third phase. A total of 2316 simulations were carried out:
• 3-hour realisations of 155 different environmental conditions with aligned

wind and waves (10 seeds)
• 3-hour realisations of 42 environmental conditions with misaligned wind
and waves (10 seeds)
• Additional 3-hour realisation of 346 different environmental conditions (1

seed, only used for studying the effect of bin size)
5.1 Phase 1 - Environmental conditions for simulation

lengths and number of seeds
The present analysis was based on the recommended bin sizes for waves from
the IEC standard, and the lower an upper turbine operation wind speed limits,
together with 1 m/s bins for mean wind speed. Applying the 10−4 probability
threshold together with turbine operation range limits, leaves 2201 load cases
that have a total accumulated probability of 81%. From these, 155 environmen-
tal conditions (see Fig. 5) were selected to reflect different load case groups:
1. Base case with large bin sizes

2. 5% most probable load cases
3. Special cases with critical frequencies

4. Variation in wave height
5. Variation in wave period
6. Variation in wind/wave misalignment
As a first step, a few load cases are selected as base cases, shown as “Base
case” in Fig. 5. The basic assumption in selecting these cases is that fatigue
damage increases with wave height (given wave period) and that important
dynamic effects are captured. A coarse grid of mean wind speeds and wave
periods was selected from the 2201 load cases with probability above 10−4 .
The 11 most probable environmental conditions were also chosen. The reason
for not including more is that one obtains many cases with low wind speed and
small waves and not much variation between the cases.
14
Load cases Base case
Special cases
Wave height
Wave period
5% most probable
6
5
Wave height (m)

4
3
2
1
0
20
15
)
10 iod (s
5 r
10 5 ve pe
Wind sp 15 Wa
eed (m/s) 20 0
Figure 5: The 155 environmental conditions analysed in this study. Wind speed
refers to hub wind.
To find how fatigue damage varies with wave height, some of the base cases
were run with varying wave height. The load cases were chosen such that
different wind speeds and peak periods were represented in the selection.
There are certain dynamic effects, which depend on wave period and wind
speed, that can cause large fatigue damage. To make sure that these were in-
cluded in the analyses, loads that might increase fatigue loading were identified:
• Wave length corresponding to when wave forces have 180 degree phase
difference on columns, in this case 7.2 s and 7.7 s
• Heave resonance at 19.9 s
• Tower wake 3P loads close to tower first mode fore-aft and side-to-side
bending at 2.3 s, which corresponds to mean wind speeds around 7 m/s
• Higher harmonics of drag forces (mainly 3ω) close to structural flexible
modes. Most relevant wave periods give 3ω in a range between rigid
body modes and the tower first bending mode, and other flexible natural
frequencies are expected to be higher, so only the shortest periods are
relevant to capture this phenomenon.
• Higher turbulence intensity at lower wind speeds
• Rated wind speed for maximum thrust force on rotor
Based on these criteria, combinations of wave peak periods 2.5, 6.5, 7.0, 7.5,
8., 16., 20 s and mean wind speeds 3., 5., 9., 11., 15., 23 m/s were chosen from
the 1539 load cases mentioned above. From these, combinations with the largest
wave height were chosen as the first step to screen load cases.
5.2 Phase 2 - Environmental conditions for wind-wave

misalignment
When wind and waves come from different directions, tower bending moments
can increase from the aligned case since the rotor no longer provides damping
15
for wave excitation. This has been documented through analyses for offshore
turbines with monopile foundations [32], and is also expected to be a problem
for floating wind turbines [33]. In the current study, some of the base cases were
run with misaligned wind and waves (0-90◦ ), both with head wind and wind
from the side with the rotor yawed 90◦ (see Fig. 6). Since a joint wind and
wave distribution with directional distribution did not exist for this site, and
since including directions will drastically increase the number of load cases, the
misaligned wind and wave cases were only performed to check the validity of
assuming unidirectional wind and long crested waves.
0◦ Wave
30◦
60◦ 90◦
0◦ Wind
Wind Wave
30◦
60◦ 90◦
(a) (b)
Figure 6: Top view illustration of wind and wave directions for the misaligned cases.
The turbine has a yaw mechanism to make the rotor plane normal to the
wind, and since the mass and restoring properties of the platform are not sym-
metric, wind coming from a different direction than the one assumed in this
study, causes different load responses. To get an impression of the effect of
wind direction, the misalignment cases were also run for wind from the side
(incoming wind angle of 90◦ ). When wind comes from the side, it will cause
the platform to yaw. Riflex can not yet model the yaw mechanism, but the
rotor should be kept normal to the wind, thus the platform was given an initial
yawed position, calculated based on the thrust force for a fixed turbine. This
was not done for yaw due to waves, since this yaw moment was observed to be
significantly lower than the yaw due to an eccentric thrust force.
5.3 Phase 3 - Environmental conditions for bin size eval-

uation
Since the results from the load cases listed above showed that determining fa-
tigue damage from only 1 realisation of a 3-hour environmental condition gave
reasonable estimates, one 3-hour realisation of evenly spread load cases were
performed to be able to say something about the discretisation of the environ-
mental conditions when calculating the total fatigue damage.
346 additional load cases were run, using a bin size of 1 m/s for wind speed,
1.5 m for wave height and 1 s for wave period. These load cases were only used
16
Base case
Load cases Special cases
Wave height
Wave period
6
5% most probable
Additional
5
Wave height (m)

4
0
20
15
0 10 (s)
5 iod
10 5 per
Wind sp 15 a ve
eed (m/s) 20 W
25 0
Figure 7: 346 additional environmental conditions, used for studying the effect of
bin size. Only one 3-hour realisation of the additional environmental conditions was
run.
to study how load case selection and bin sizes affect the total fatigue damage
estimate. The additional load cases, together with the original load cases, are
shown in Fig. 7.
17
6 Simulation results
Fatigue damage was calculated for both axial and shear stress components, but
shear stress fatigue damage was significantly lower than for axial stress. Thus,
the shear stress fatigue was excluded from the paper.
All results presented in this section refer to the cross section location with
the maximum 10-seed average damage. The points where fatigue was calculated
for the different members are shown in Fig. 1.
6.1 Load conditions that cause large fatigue damage

Wind speeds 7 m/s and 9 m/s caused high short term fatigue damage (Fig. 8)
compared to what one would expect for mild wind conditions. For the higher
wind speed cases (17 m/s and 20 m/s) fatigue damage is also high, but for these
cases wave height is higher, thus higher damage was expected.
Tp=6.0 s Tp=7.0 s Tp=8.0 s Tp=12.0 s Tp=16.0 s

Tp=6.5 s Tp=7.5 s Tp=10.0 s Tp=14.0 s
4.5e-04
4.1e-04
3.6e-04
3.2e-04
Damage (-)
2.7e-04
2.3e-04
1.8e-04
1.4e-04
9.0e-05
4.5e-05
0.0e+00
3 m/s 4 m/s 5 m/s 7 m/s 9 m/s 10 m/s 11 m/s 12 m/s 14 m/s 15 m/s 17 m/s 20 m/s
Umean
Figure 8: 3-hour fatigue damage in tower bottom for base case and special cases,
arranged by wind speed and wave period. Wave heights vary between 0.5-4.5 m for
below rated wind and 2.5-5.5 m for above rated wind.
Tower wake 3P loads at low wind speeds (approximately 2.67 rad/s for 7
m/s and 3.26 rad/s for 9 m/s) are close to the tower first bending mode (2.73
rad/s). This results in resonance in the tower at low wind speeds, which causes
high fatigue damage. The contribution from these load cases is large in the long
term fatigue analysis since this range of wind speeds occurs frequently.
This is normally not a problem for fixed foundations since the tower natural
frequency is lower. The natural frequency is lower both because there are no
rigid body modes and because towers are normally longer than for floating
platforms. For a floating platform, there is a risk that the first bending frequency
of the tower is within the three times rotational speed of the rotor in the below
rated operating state.
18
6.2 Simulation length and number of random realisations
The total 20 year fatigue damage was calculated by Eq. 4 for the environmental
conditions in Fig. 5. The number of samples varied from 1 to 10 and simulation
lengths were varied from 10 minutes to 3 hours. The 10 min to 2 hour simu-
lations were obtained by sampling stress histories from the steady state 3-hour
timeseries. The fatigue damages calculated based on 10 min to 2 hour durations
were scaled up for comparison with 3 hour damages.
Estimating the total 20 year fatigue based on 10x1-hour simulations of each
load case underestimated damage by less than 4% compared to 10x3-hour sim-
ulations (Fig. 9). Using 10x10-minute simulations underestimated the damage
by 10%. Fatigue damage for the tower appeared to be less affected by simulation
length than the platform members, pontoon, main beam and brace.
0
Relative error of total damage [%]
−2
Tower base
−4
Pontoon
−6 Main beam
Brace
−8
−10
−12
600 1800 3600 7200 10800
Simulation length (s)
Figure 9: Relative difference to 3-hour simulations in 20 year total damage (Eq. 4)

for axial stress fatigue. The average damage from 10 samples was used for all the
simulation lengths shown.
In essence, a 3-hour simulation contains the same statistical content as 3x1-

hour or 6x30-min simulations. I.e., using longer simulations reduces the number
of seeds required. It was found that the long term fatigue damage calculated
based on 10-minute simulations varied more with seed number, and gave larger
errors compared to the 10 seed average than for the longer simulation lengths
(Fig. 10). Another important observation is that the error for fatigue in the
tower base was very small.
A full evaluation of the required number of seeds in a statistical sense would
require a large sample of 3-hour simulations. A sample of 10, which was used in
this study, is not statistically significant. Moreover, the fact that the order of
the samples is random, contributes to the non-converging behaviour of the data
in Fig. 10. The focus was kept on including as many different environmental
conditions as possible within the time frame of this work, and thus the number of
samples could not be increased. Therefore, a method introduced by Langley [34]
to estimate the root mean square (RMS) error of a limited number of program
runs (Eq. 5) was applied.
19
10 min 30 min 1 hr 3 hr
3
2
1
0
−1
−2
−3
Tower base
Relative error of total damage (%)
−4
4
2
0
−2
−4
−6
−8 Pontoon
−10
4
2
0
−2
−4
−6
−8
−10 Main beam
5
0
−5
−10
−15
−20 Brace
1 2 3 4 5 6 7 8 9 10
No. of samples
Figure 10: 20 year damage as a function of the number of samples, calculated as the
cumulative average of N seeds , where N seeds varied from 1 to 10. The error refers to
the damage calculated from 10x3-hour simulations.
20
1 σD
δ=√ (5)
N seeds µD
µD and σD in this context are the mean value and the standard deviation,
respectively, of the 20 year fatigue damage.
The resulting RMS error is shown in Fig. 11. Similar conclusions to what
has been described earlier in this section can be drawn from the RMS estimate;
errors are small in general, but larger for 10-minute simulations, and the tower
base was less sensitive to the number of seeds than the platform members.
Figure 11 also illustrates an important point; that the statistic content is the
same in six 10-minute samples as in one 1-hour sample, and thus the expected
error is the same.
14
10min-Tower base 1h-Tower base 3h-Tower base
12 10min-Pontoon 1h-Pontoon 3h-Pontoon
10min-Main beam 1h-Main beam 3h-Main beam
10min-Brace 1h-Brace 3h-Brace
10
RMS error (%)
0
1 2 3 4 5 6 7 8 9 10
No. of samples
Figure 11: Estimated RMS error for 20 year damage, as a function of number of
samples.
6.3 Bin size and load case selection

The discretisation of long term environmental conditions (determining bin size)
must ensure that all important load effects are captured such that a realistic
and conservative estimate of the fatigue damage can be determined. Moreover,
a course grid of environmental conditions also means fewer simulations and thus
fewer random samples.
In this section, the influence of bin size on the 20 year fatigue damage is
investigated based on 1 realisation of the load cases labelled “additional” in
Fig. 7, as well as 10 seed average damage from the load cases in Fig. 5.
6.3.1 Bin size for wind speeds

The total fatigue was calculated for increasing bin sizes. Table 6 shows that
by increasing the bin size from 1 to 2 m/s, the total fatigue damage for the
21
tower was reduced by 9%. However, a bin size of 2 m/s, starting at cut-in wind
speed (i.e. load cases include 3, 5, 7,.. m/s), means that the most important
resonance case at 8 m/s is left out. If a bin size of 2 m/s was used, and the 8
m/s cases were included, the deviation in the total damages would be 0.3% for
the tower, 3.2% for the pontoon, 1% for the main beam and 3.1% for the brace.
This indicates that a bin size of 2 m/s will give an acceptable damage estimate
if 3P resonance cases are included.
Table 6: Variation of 20 year fatigue damage in tower base for varying bin size,
relative to the smallest bin size. The bin size is 1 m/s for wind speed, 0.5 m for wave
height and 0.5 s for wave period, unless otherwise specified. The damage is calculated
according to Eq. 6. Please note that the selection of Hs representing bin sizes below
1.5 m is limited.
Bin size Tower base Pontoon Main beam Brace No. of

Uhub Drel (%) Drel (%) Drel (%) Drel (%) env. cond.
1.0 m/s 100 100 100 100 478
2.0 m/s 91 90 89 88 219
3.0 m/s 88 85 79 85 152
4.0 m/s 85 83 83 81 120
Hs
0.5 m 100 100 100 100 478
1.0 m 101 101 100 99 234
1.5 m 96 94 96 94 410
2.0 m 41 38 44 38 101
Tp
0.5 s 100 100 100 100 478
1.0 s 100 100 101 100 442
1.5 s 104 103 101 103 161
2.0 s 103 105 104 104 245
The number of environmental conditions included in the calculation of the

total fatigue damage for different bin sizes is shown in the rightmost column
of Tab. 6. What can be observed, is that the total fatigue damage changes
more from 478 to around 200 environmental conditions when changing the wind
speed bin size, than when changing the wave height or wave period bin size.
The relatively low sensitivity to the number of seeds, which was described in
Sec. 6.2, is consistent with this. Thus, it seems that the total fatigue damage
was more sensitive to the bin size parameter than the number of environmental
conditions.
In the Hs section of Tab. 6, the number of environmental conditions is
smaller for bin size 1.0 m than for 1.5 m. The same can be observed for Tp .
This has to do with the distribution and the selection of simulated environmental
conditions (as shown in Fig. 7).
6.3.2 Bin size for wave periods

To be able to evaluate if the most important wave periods are included in the
total fatigue damage assessment, comparisons of damage for load cases with
22
varying wave period were carried out (“wave period” in Fig. 5). These com-
parisons indicated that pitch motion has a significant contribution to fatigue
damage in all members. Figures 12 and 13 show the highest damages for 10-12
s waves, which coincides with a peak in the pitch response amplitude operator.
The reduction of fatigue damage with increasing wave period in Figs. 12 and
13 is fairly linear, which can explain that the increase of wave period bin size
do not cause significant changes in the damage estimates in Tab. 6. Choosing a
bin size of 2 s gave an acceptable result in this case, but keep in mind that this
conclusion depends on the response characteristics of the platform.

4.2e-04
3.8e-04
3.3e-04
2.9e-04
Damage (-)
2.5e-04
2.1e-04
1.7e-04
1.3e-04
8.3e-05
4.2e-05
0.0e+00
4 m/s 7 m/s 10 m/s 12 m/s 14 m/s 20 m/s
2.0 m 4.5 m 5.0 m 5.0 m 4.0 m 5.5 m
Umean,Hs
Figure 12: 3-hour short term fatigue damage due to axial stress in the tower base for
varying wave period, averaged over 10 seeds. Maximum/minimum observed damage
in all realisations is shown with errorbars.
The selection of load cases called “additional” in Fig. 7 has a bin size of 1
s for wave periods, which means that the damage for bin size 0.5 s in Tab. 6
does not include all possible load cases with 0.5 s bin size. However, comparing
damage weighted with probability for the cases labelled “wave period” in Fig.
7 with a bin size of 1.0 s only gives 0.4-3.9% difference compared to a bin size
of 0.5 s.
6.3.3 Bin size for wave heights

Table 6 indicates that 1.0 m bin size for wave heights will give acceptable results.
The bin size for wave heights in the “additional” load cases (Fig. 7) is 1.5 m.
Calculating fatigue for the load cases that have a bin size of 0.5 m, 1.5 m bin
size gives 1.0-9.4% difference (17-32% for 4 m/s wind and 12 s waves).
One possible explanation for damage being more sensitive to the increase
of bin size for wave heights can be found in the wave induced pitch motion of
the platform. Assuming that only forces on the tower caused by pitch motion
causes tower fatigue, and that the stresses increase linearly with pitch angle, the
fatigue damage increase will be non-linear (in the order of the m-parameter in
the fatigue curve, which is 3 and 5 in this study, see [28]). Nonetheless, as can
be seen in Fig. 7, more cases with variation in wave height should be included
23
7.9e-07
7.1e-07
6.3e-07
5.5e-07
Damage (-)
4.7e-07
3.9e-07
3.2e-07
2.4e-07
1.6e-07
7.9e-08
0.0e+00
4 m/s 7 m/s 10 m/s 12 m/s 14 m/s 20 m/s
2.0 m 4.5 m 5.0 m 5.0 m 4.0 m 5.5 m
Umean,Hs
Figure 13: 3-hour short term fatigue damage due to axial stress in the brace for
different wind speeds, wave heights and periods, averaged over 10 seeds. Maxi-
mum/minimum observed damage in all realisations is shown with errorbars.
to make a firm conclusion about sensitivity to wave height bin size.
6.4 Total fatigue damage for other sites

Although this study was not an attempt to do full fatigue damage assessment,
and no explicit values for long term fatigue damage is presented, it should be
mentioned that the calculated damages indicate an unacceptably low fatigue life,
since installation, survival conditions, faults and parked condition also require
parts of the allowable fatigue damage, and since weld geometries most likely
will give stress concentration factors higher than 1.
In order to give an impression of how much the long term fatigue damage
can be reduced by applying environmental data from other sites, sensitivity was
performed by using the probability distributions from the other sites addressed
in [21]. The relative total 20-year fatigue damage (Eq. 4) compared to the
Portugal location was:
• 69-82% for site no. 1 (Atlantic, France)

• 89-94% for site no. 5 (Atlantic, Great Britain)
• 121-130% for site no. 14 (North Sea - Norway)

• 64-78% for site no. 15 (North Sea - Denmark)
The water depth in site no. 1, 5 and 15 is lower than assumed in this study,
and thus they are not relevant sites for this particular SSWT, considering that
the mooring design would have to be redesigned. The numbers do, however,
give an idea of the sensitivity of fatigue damage to wind- and wave distribution
at different sites.
24
6.5 Misaligned wind and wave conditions
Uni-directional wind and waves, with long crested waves, were assumed in the
above sections. Fatigue damage calculated based on misaligned wind and waves
indicated that this was a conservative assumption (Fig. 14). The reason for the
different conclusion in [33] and [32] will be further discussed in another paper
with misalignment as the main focus [19]. In short, the bending moments from
wind and waves act around different axes, and thus the contribution to fatigue
damage comes in different cross section locations. This is clearly seen for the
tower. Bottom fixed turbines get fatigue damage due to flexible modes, which do
not have very high hydrodynamic damping. For the SSWT large contributions
to fatigue damage come from the rigid body motions of the platform (roll and
pitch), and these motions have high drag damping from the heave plates. Wind
from other angles than 0◦ and 90◦ have not been simulated, and fatigue damage
may be higher with aligned wind and waves from other angles.
Figure 14a shows an incoming wind angle of 0◦ , and Fig. 14b shows the
90◦ case, both with the rotor facing the wind (see Fig. 6 for a definition of
the directions). In general, the 0◦ case displayed significantly higher short term
damage than the 90◦ case, which is most likely due to the wave excitation being
smaller when coming from the side. And even when the wave excitation come
from the front, the wind comes from the side, and the forces act around different
axes, causing less fatigue damage.
25
Θwave = 0 ◦ Θwave = 30 ◦ Θwave = 60 ◦ Θwave = 90 ◦
1.3e-04
1.2e-04
1.1e-04
9.4e-05
Damage (-)
8.0e-05
6.7e-05
5.4e-05
4.0e-05
2.7e-05
1.3e-05
0.0e+00
4 m/s 7 m/s 10 m/s 12 m/s 14 m/s 20 m/s
2m 4m 5m 5m 4m 6m
8s 12 s 14 s 12 s 10 s 14 s
Umean,Hs ,Tp
(a) Incoming wind angle 0◦ .
6.2e-05 Θwave = 0 ◦ Θwave = 30 ◦ Θwave = 60 ◦ Θwave = 90 ◦

5.6e-05
5.1e-05
4.5e-05
3.9e-05
Damage (-)
3.4e-05
2.8e-05
2.2e-05
1.7e-05
1.1e-05
5.6e-06
0.0e+00
4 m/s 7 m/s 10 m/s 12 m/s 14 m/s 20 m/s
2m 4m 5m 5m 4m 6m
8s 12 s 14 s 12 s 10 s 14 s
Umean,Hs ,Tp
(b) Incoming wind angle 90◦ .
Figure 14: 1-hour fatigue damage based on axial stress in tower base for misaligned
wind and waves, averaged over 10 seeds. Maximum/minimum observed damage in all
realisations is shown with errorbars. A definition of directions can be seen in Fig. 6.
26
7 Conclusions
This study deals with fatigue analysis for a wide range of environmental con-
ditions for a semi-submersible wind turbine with the purpose of studying the
effect of simulation length, the number of necessary realisations of wind and
wave loads, bin size and wind-wave misalignment.
The study showed that blade passing frequency resonance in the tower and
pitch motion of the platform were the most significant contributors to fatigue
damage in the tower and pontoon. Capturing resonant responses is important
when choosing load cases for fatigue analysis. Choosing load cases that do not
give high fatigue damage is also important, to get a realistic fatigue estimate
for the service life.
Based on 10x3-hour simulations of 155 environmental conditions, the sen-
sitivity to simulation length and number of realisations was investigated. By
calculating fatigue based on 10-minute samples, fatigue damage was underes-
timated by up to 10%. For 1-hour durations, the estimate was 4% below the
3-hour damage. The simulations showed that the overall trend for the fatigue
damage estimate decreased with number of random realisations. The RMS er-
ror compared to the average of 10 3-hour realisations was less than 1% for 7
seeds of 3-hour simulations and less than 2% for 9 seeds of 1-hour simulations.
Fatigue damage in the platform members was more sensitive to the number of
realisations and simulations duration than the damage in the tower base.
Based on the conclusions that only 1 3-hour realisation gave a satisfactory
fatigue damage estimate, 346 additional environmental conditions were analysed
to study the effect of bin size. A bin size of 2 m/s for wind speeds gave accept-
able damage estimates, provided that wind speeds for the important resonance
frequency at 8 m/s was included. For waves bin sizes of 1 s for Tp gave similar
estimates as 0.5 s, but with respect bin size for Hs , fatigue damage was more
sensitive to bin size.
Misaligned wind and waves seemed to give less fatigue than unidirectional
wind and waves, which means that it is conservative to assume unidirectionality.
However, which parts of the platform structure that experience the highest
fatigue damage, will depend on the wave direction.
Placing a turbine on a floating platforms alters the natural frequencies of the
turbine itself. For the platform in this study, the tower first fore-aft and side-
to-side bending frequencies are within the range of the blade passing frequency
for wind speeds 7-9 m/s. This lead to resonance in the tower for these wind
speeds. At resonance it is important to have a proper representation of the
aerodynamic and structural damping. This is however an issue in some state-
of-the-art aerodynamic models, since BEM without dynamic wake correction
does not include wake damping, and the generalised dynamic wake model in
AeroDyn is unstable for wind speeds around 8 m/s and below. This calls for
an improvement of aerodynamic codes applied in the analysis of floating wind
turbines.
Fatigue damages observed in this study indicate that fatigue damage will
be unacceptably high for the tower. Since resonant motion in the first bending
frequency of the tower contributes significantly to fatigue below rated wind
speed, towers should be designed for each platform type to have resonance
frequencies outside of the blade passing frequency range.
27
8 Acknowledgements
The authors wish to acknowledge the support from NOWITECH and the Nor-
wegian Research Council through Centre for Ships and Ocean Structures at
NTNU.
28
References
[1] Det Norske Veritas. DNV-RP-J103, Design of Floating Wind Turbine
Structures, 2013.
[2] M. Karimirad and T. Moan. Extreme dynamic structural response analysis
of catenary moored spar wind turbine in harsh environmental conditions.
Journal of offshore mechanics and Arctic engineering, 133(4), 2011.
[3] L. Haid, G. Stewart, J. Jonkman, A. Robertson, M. Lackner, and D. Matha.
Simulation-length requirements in the loads analysis of offshore floating
wind turbines. In Proceedings of the 32nd International Conference on
Ocean, Offshore and Arctic Engineering, Nantes, France, 2013.
[4] M. I. Kvittem, T. Moan, Z. Gao, and C. Luan. Short-term fatigue analysis
of semi-submersible wind turbine tower. In Proceedings of 30th Interna-
tional Conference on Ocean Offshore and Arctic Engineering, Rotterdam,
The Netherlands, 2011. Paper no. OMAE2011-50278.
[5] J. Jonkman, S. Butterfield, W. Musial, and G. Scott. Definition of a 5-MW

reference wind turbine for offshore system development. Technical report,
NREL, 2009. NREL/TP-500-38060.
[6] J. Jonkman. Definition of the floating system for phase IV of OC3. Tech-
nical report, National Renewable Energy Laboratory, 2010.
[7] D. Roddier, A. Peiffer, A. Aubault, and J. Weinstein. A generic 5MW
WindFloat for numerical tool validation and comparison against a generic
spar. In Proceedings of ASME 30th International Conference on Ocean Off-
shore and Arctic Engineering, Rotterdam, The Netherlands, 2011. Paper
no. OMAE2011-50278.
[8] MARINTEK. RIFLEX User’s Manual, 2013.

[9] MARINTEK. SIMO User’s Manual, 2013.
[10] D. Laino and A. C. Hansen. AeroDyn users guide, 2002.
[11] H. Ormberg and E. E. Bachynski. Global analysis of floating wind turbines:

Code development, model sensitivity and benchmark study. In Proceed-
ings of the 22nd International Ocean and Polar Engineering Conference,
Rhodes, Greece, 2012.
[12] MARINTEK. SIMO Theory Manual, 2012.
[13] Det Norske Veritas. SESAM User Manual Wadam, 2008.

[14] Det Norske Veritas. DNV-OS-E301, Offshore standard-position mooring,
2010.
[15] Det Norske Veritas. DNV-RP-C205, Environmental Conditions and Envi-
ronmental Loads, 2007.
[16] Principle Power. Wamit model for WindFloat 5 MW. Technical report,
Principle Power, 2010.
29
[17] C. Luan, Z. Gao, and T. Moan. Modelling and analysis of a semi-
submersible wind turbine with a central tower with emphasis on the brace
system. In Proceedings of the 32nd International Conference on Ocean,
Offshore and Arctic Engineering, Nantes, France, 2013.
[18] O. M. Faltinsen. Sea Loads on Ships and Ocean Structures. Cambridge
University Press, 1990.
[19] E. E. Bachynski, M. I. Kvittem, C. Luan, and T. Moan. Wind-wave mis-
alignment effects on floating wind turbines. To be published, 2014.
[20] A. Robertson, J. Jonkman, Musial W., F. Vorpahl, and W. Popko. Offshore
code comparison collaboration, continuation: Phase II results of a floating
semisubmersible wind system. Technical report, National Renewable En-
ergy Laboratory, 2013.
[21] L. Li, Z. Gao, and T. Moan. Joint environmental data at five european
offshore sites for design of combined wind and wave energy devices. In
Proceedings of ASME 32th International Conference on Ocean Offshore and
Arctic Engineering, Nantes, France, 2013. Paper no. OMAE2013-10156.
[22] B. J. Jonkman. TurbSim users guide, 2009.
[23] International Electrotechnical Commission. IEC61400-1, Wind turbines -

Part 1: Wind turbines, design requirements, 2005.
[24] I. Van der Hoven. Power spectrum of horizontal wind speed in the fre-
quency range from 0.0007 to 900 cycles per hour. Journal of Meteorology,
14(2):160–164, 1957.
[25] J. Apt. The spectrum of power from wind turbines. Journal of Power
Sources, 169(2):369–374, 2007.
[26] T. Burton, N. Jenkins, D. Sharpe, and E. Bossanyi. Wind energy handbook.
John Wiley & Sons, 2011.
[27] F. Irgens. Mechanics of Materials (in Norwegian). Tapir, 2006.

[28] Det Norske Veritas. DNV-RP-C203, Fatigue Design of Offshore Steel Struc-
tures, 2010.
[29] Det Norske Veritas. DNV-RP-J101, Design of Offshore Wind Turbine
Structures, 2010.
[30] International Electrotechnical Commission. IEC61400-3, Wind turbines -
Part 3: Design requirements for offshore wind turbines, 2009.
[31] Det Norske Veritas. DNV-RP-C103, Column-stabilised units, 2010.
[32] T. Fischer, W. Vries, P. Rainey, B. Schmidt, K. Argyriadis, and M. Kühn.

Offshore support structure optimization by means of integrated design and
controls. Wind Energy, 15(1):99–117, 2012.
30
[33] L. Barj, S. Stewart, G. Stewart, M. Lackner, J. Jonkman, A. Robertson,
L. Haid, and D. Matha. Impact of wind/wave misalignment in the loads
analysis of a floating wind turbine. In Poster presentation at AWEA Wind
Power, 2013.
[34] R. S. Langley. On the time domain simulation of second order wave forces
and induced responses. Applied ocean research, 8(3):134–143, 1986.
31
118
Paper 3
Wind-wave misalignment effects on floating wind turbines:

motions and tower fatigue load effects
By Erin E. Bachynski, Marit I. Kvittem, Chenyu Luan, Torgeir Moan

Accepted for publication in Journal of Offshore Mechanics and Arctic En-
gineering in November 2014
119
Is not included due to copyright
120
Paper 4
Frequency versus time domain analysis for fatigue of a

semi-submersible wind turbine
By Marit I. Kvittem and Torgeir Moan

Accepted for publication in Journal of Offshore Mechanics and Arctic En-
gineering in February 2015
137
Is not included due to copyright
138
Appendix B
Abstracts of additional papers
153
Short-term fatigue analysis of semi-submersible
wind turbine tower
Marit I. Kvittem∗1,2 , Torgeir Moan2 , Zhen Gao2 and Chenyu Luan3
1
2
3
NTNU, Norwegian University of Science and Technology, Trondheim, Norway
Published in Proceedings of 30th International Conference on Ocean, Offshore

and Arctic Engineering, 2011.
Abstract
Coupled time domain analyses of a semi-submersible wind turbine are
performed with the intention to study motions affecting fatigue damage
at the base of the tower. The software applied is SIMO/RIFLEX with the
extension TDHmill, which gives the wind thrust force and gyro moment
on the wind turbine as point loads in the tower top.
Short term environmental conditions are chosen from a joint wind and
wave distribution for a site in the Northern North Sea. Variance spec-
tra, mean value, standard deviation, kurtosis, skewness and Vanmarcke’s
bandwidth parameter are calculated for stresses at the base of the tower.
Damage is calculated for each short term condition by two methods; rain-
flow counting and narrow band approximation. The accuracy of narrow
band approximation estimates for fatigue are examined for the structure
in question.
Time domain simulations are carried out for different sea states and fa-
tigue damage is calculated for each case. Simulations show that turbulent
wind dominates the response at low wind speeds and the response spec-
tral density function tends to be very wide-banded. For wave dominated
response, spectra have lower bandwidth, and narrow banded approxima-
tion for fatigue damage gives estimates 20-50% above rainflow counted
damage.
1
Effect of mooring line modelling on motions and
structural fatigue damage for a semi-submersible
wind turbine
Marit I. Kvittem∗1,2 and Torgeir Moan2
1
2
Published in Proceedings of the Twenty-second International Offshore and Polar

Engineering Conference, 2012.
Abstract
Most current software applied for motion analysis of floating wind tur-
bines employ some kind of linearized mooring line model, ignoring mooring
line dynamics. The impact of these assumptions are well known for off-
shore oil platforms, but for a floating wind turbine the wind-to-wave load
ratio is higher, and the experiences from the offshore industry cannot be
applied directly. However, these experiences state that linearized models
can provide good prediction of fatigue damage applying frequency domain
analysis, which can significantly shorten analysis time.
In this paper a single semi-submersible wind turbine (SSWT) was in-
vestigated with a linearized mooring line model and compared with an ad-
vanced finite element mooring line model for three different mooring line
configurations. The parameters investigated were platform motion char-
acteristics and bending moment in the tower. These analyses were carried
out by a numerical tool obtained by integrating the hydrodynamic soft-
ware SIMO/RIFLEX and AeroDyn for aerodynamic loads on the blades
of the turbine.
The results from these analyses showed significant differences in the
standard deviations for sway and yaw motions and sideways bending mo-
ment in the tower between the linearized model and the full mooring line
model. The linearized model gave relatively good predictions of surge
motions of the platform and fore-aft bending moment in the tower. Wave
frequency response, which normally contributes most to fatigue, was not
particularly influenced by mooring line modellig. Nevertheless, an im-
proved model should include mooring line damping and coupling effects.
1
Dynamic analysis of floating wind turbines during
pitch actuator fault, grid loss, and shutdown
Erin E. Bachynski∗2 , Mahmoud Etemaddar2 , Marit I. Kvittem1,2 ,
Chenyu Luan1,2 and Torgeir Moan2
1
2
Published in Elsevier Energy Procedia, vol 35, p.210-222, 2013.
Abstract
Coupled non-linear aero-hydro-servo-elastic simulations of three types
of floating wind turbines (spar, semi- submersible, and tension leg plat-
form) are carried out for several fault cases over a range of environmental
conditions based on correlated wind and wave data from the North Sea.
Three particular fault scenarios are considered: 1) blade seize, where the
pitch actuator of one blade is blocked, 2) blade seize, recognized by the
controller and followed by shutdown (grid disconnection and aerodynamic
braking), and 3) grid loss followed by shutdown. The platform motions
and structural loads caused by fault events are compared to loads en-
countered during normal operation and during selected extreme weather
conditions. Although the global motions and mooring line loads tend to
be largest during storm conditions, selected platforms experience large
pitch or yaw motions due to blade seize and shutdown. Imbalance loads
due to blade seize can lead to particularly large loads on the blades and
tower, and the shutdown process can impose relatively large edgewise
blade loads.
1
Appendix C
Specifications for a semi-submersible

wind turbine with active ballast system
Platform specifications
The WindFloat concept was chosen as inspiration for the floating platform described in
this section, and the outer dimensions in [93] was kept, but since more detailed specifica-
tions than what is available was necessary, a different design with similar global charac-
teristics was developed.
Environment
kg
Depth 320 m. Seawater density 1025 m3
.
157
Platform geometry
Table C.1: Platform geometry.
Number of columns 3
Column diameter 10.0 m
Column center to center 46.0 m
Main beam and pontoon diameter 2.0 m
Braces diameter 1.5 m
Operating draught 17.0 m
Airgap 10.0 m
Heave plate edge length 15.0 m
Turbine tower foundation above MSL 10.0 m
Pontoon vertical position -15.0 m
Ballast tank bottom lower level -13.0 m
Steel
Table C.2: Properties for steel components.

kg
Density 7850 m 3
Column plate thickness 30 mm

Main beam and pontoon diameter plate thickness 30 mm
Braces plate thickness 25 mm
Heave plate effective thickness 50 mm
Turbine
Table C.3: Turbine data for NREL 5MW with OC3 tower.
Total mass 599,718 kg

Global centre of gravity (26.27 m,0.00 m, 64.13 m)
Hub height above MSL 90.0 m
Rated wind speed 11.2 ms
Mooring
The lines have 60 m of chain on top, 30 tonnes, 3.8 m3 clump weight between chain
and polyester rope, 769.8 m of polyester rope in the middle and 232.58 m of chain at the
bottom.
158
Table C.4: Fairlead and anchor line positions.
x y z
Fairlead 1 31.56 0 -17
Fairlead 2 -15.78 -27.33 -17
Fairlead 3 -15.78 27.33 -17
Fairlead 4 31.56 0 -17
Anch 1 738.07 -706.51 -320
Anch 2 -721.76 -733.31 -320
Anch 3 -721.76 733.31 -320
Anch 4 738.07 706.51 -320
Ballast system
The WindFloat platform has an active ballast system to keep the turbine upright and thus
maximise power output as the wind direction and intensity changes. This is done by
pumping water ballast between the columns, and the system reacts in 20 minutes to sig-
nificant permanent changes in the wind.
The ballast in the SSWT in these analyses was distributed to account for the asymmetric
loading of the platform, i.e. the weight of the turbine and mooring system. Other loads
that gives the platform a permanent tilt is the rotor thrust force and wind drag on the tower.
The rotor torque gives a sideways tilt. Since the reaction time of the ballast system is long
compared to other dynamic loads on the system, it was modelled as constant ballast.
A new ballast distribution is calculated for every wind speed. The ballast distribution
between column 1 and columns 2 and 3 is calculated from the thrust force and tower drag
for constant wind velocity with a power law wind shear. The ballast distribution between
column 2 and column 3 is calculated based on the rotor torque. Thus, mass distribution
and restoring is updated for each environmental condition. Samples of masses and centre
of gravity for different wind speed is presented in Table C.5.
Table C.5: Mass distribution for different wind speeds. Mass and centre of gravity is
given for the whole column.
Column 1 Column 2 Column 3

Wind speed ( ms ) Mass (kg) zCOG (m) Mass (kg) zCOG (m) Mass (kg) zCOG (m)
0.0 749,418 -10.358 1400,128 -7.918 1400,128 -7.918
8.0 658,582 -10.439 1441,151 -7.715 1449,940 -7.671
11.2 579,217 -10.384 1476,078 -7.539 1494,378 -7.446
25.0 659,277 -10.439 1435,887 -7.741 1454,509 -7.648
49.0 652,823 -10.440 1448,425 -7.678 1448,425 -7.678
159
Mass, damping and restoring
Table C.6: Mass, damping and restoring data. Inertia and restoring are given for platform
without turbine, ballasted for zero wind thrust.
Displacement 4810E6 kg
Mass 4619E6 kg
Centre of gravity (-0.331 m, 0.0 m, 1.489 m)
Mass without turbine and mooring 4019E6 kg
Centre of gravity without turbine and mooring (-4.300 m, 0.0 m, -7.857 m)
Radius of gyration Rxx 22.29 m
Radius of gyration Ryy 19.62 m
Radius of gyration Rzz 26.06 m
Radius of gyration Ryz 4.69 m
Hydrostatic stiffness Heave 2445E3 N
Hydrostatic stiffness Roll 749815E3 N m
Hydrostatic stiffness Pitch 7548571E3 N m
Viscous forces
The viscous forces on the slender elements of the platform are calculated by the Morison
formula. The coefficients applied are listed in Table C.7.
Table C.7: Non-dimensional quadratic drag coefficients.
Component Transverse Cd (-) Longitudinal Cd (-) Cm (-)

Columns 1.0 0.0 0.0
Mooring chain 2.4 1.15 1.0
Mooring polyester rope 1.6 0.0 1.0
Main beam and pontoons 1.0 0.0 1.0
Braces 1.0 0.0 1.0
Heave plates 7.5 0.0 0.0
Verification of multibody model

Traditional modelling of floater motion considers the complete platform as a rigid body
(Figure C.1a). However, internal member forces of a statistically indeterminate structure
can not be extracted from such an analysis. Thus, in order to get forces in the braces,
the three columns of the platform were modelled as individual Simo bodies with potential
theory forces, and the connecting braces were modelled as Riflex beam elements with
Morison type forces (Figure C.1b).
160
(a) Single body model
(b) Multi body model
Figure C.1: Illustration of the two different modelling strategies. The yellow parts are
modelled as rigid bodies and the grey parts are modelled as flexible beam elements.
161
Through decay simulations, it was confirmed that both the single body and the multibody
model gave the same natural periods for rigid body motion (Table C.8). Comparison was
also made for response in regular waves.
Hydrodynamic interaction between the columns can be taken into account by applying a
multi body analysis. However, some limitations on hydrodynamic interaction are inherent
in the coupled Simo-Riflex software. This means that for one body, the effect of the
presence of the other bodies was included in the force transfer functions and retardation
functions, but the effect of motion of the other bodies was only accounted for in the force
transfer functions, not in the retardation functions. Comparisons of added mass forces for
the multibody and the single body models are shown in Figures C.2 and C.3.
Natural periods
Table C.8: Natural periods.
Single body Multibody

Mode T (s) T (s)
Surge 107.0 107.0
Sway 124.8 124.8
Heave 19.9 19.9
Roll 35.6 35.6
Pitch 37.4 37.4
Yaw 68.5 68.5
Tower 1st 2.3 2.3
Added mass forces
Figures C.2 and C.3 show that the difference in added mass forces for three different
modelling strategies are similar for an acceleration of 1 m/s2 . The modelling strategies
are:
• One body with full hydrodynamic interaction.
• Three bodies with no interaction.
• Three bodies with interaction in the diagonal sub-matrices of the global added mass
matrix.
162
Figure C.2: Added mass forces due to 1 m/s2 accelerations for small wave periods.
Figure C.3: Added mass forces due to 1 m/s2 accelerations for wave period 9 s.
163
Response in regular waves
The response in regular waves was compared for the single body and the multibody model.
The braces were not included in the panel model for the single body model, but in the com-
parison, this was compensated for by adding extra restoring in the single body model and
by setting added mass of the braces in the multi body model to zero. Morison coefficients
for other parts were as given in Table C.7.
Table C.9: Load cases
Wave period (s) 3.7 4.0 5.0 5.5 6.0 6.9 10.0 15.0 20.0 21.0
Wave height (m) 1.0 1.0 1.0 1.0 1.5 2.0 3.0 6.0 9.0 9.0
Figure C.4 shows response amplitude operators obtained by running regular wave anal-
yses for different wave periods and heights, see Table C.9. The figure shows the mean
of the response peaks normalized by the wave height. There is good agreement between
the two models. Differences may be due to either interaction effects or that the choice of
added mass coefficients for the pontoons do not match the potential solution.
1.4
1.0 1.2
0.8
Heave (m/m)
1.0
Surge (m/m)
0.6 0.8
0.6
0.4
0.4
0.2 0.2
0.00 5 10 15 20 25 0.00 5 10 15 20 25
0.5 Wave Period (s)
MultiBody
0.4 SingleBody
Pitch (deg/m)
0.3
0.2
0.1
0.00 5 10 15 20 25
Wave Period (s)
Figure C.4: Response amplitude operators from regular wave analysis of single body
model and multi body model with flexible braces.
164
Appendix D
Analytical Morison force transfer

functions
Force transfer functions including vertical dynamic pres-

sure
A pure Morison type wave force model does not take into account the unbalanced vertical
component of the dynamic pressure under a surface piercing column. In this note, transfer
functions in heave, roll and pitch for Morison forces, including the effeect of dynamic
pressure, for a three-column semi-submersible with heave plates are derived. Vertical
forces do not contribute to surge, sway an yaw forces, thus they are not treated herein.
Reference to the theory behind the derivations can be found in [16], and the x-y plane
locations of the columns can be seen in Fig. D.1.
Table D.1: Nomenclature
g Gravitational acceleration
ω Wave frequency
t Time
k Wave number k = ω 2 /g
x,y,z Position in global coordinate system
β Wave heading angle
d Draught, positive
R Column radius
ρ Sea water density of mass
Ca Added mass coefficient
V Displaced volume of column and heave plate (small)
165
Velocity potential on complex form with unit wave amplitude:
g i(ωt+θ)
φ= ie (D.1)
ω
Where phase angle θ is:
θ = −kx cos β − ky sin β (D.2)
Particle accelerations in horizontal and vertical direction and dynamic pressure:

a1 = iω 2 ekz ei(ωt+θ)
a3 = −ω 2 ekz ei(ωt+θ)
(D.3)
pd = ρgekz ei(ωt+θ)
(-13.33, 23.00)
x (26.56, 0.00)
2R
(-13.33, -23.00)
Figure D.1: Top view
Vertical wave forces
For a unit wave in x = 0:
ζ(t) = eiωt (D.4)
Vertical inertial force, when assuming added mass can be calculated from the volume of
a half sphere and force due to the acceleration of the platform is ignored:
2π 3
F3inertia = (ρV + Ca ρ R )a3 (z = −d) (D.5)
3
166
which gives the following load amplitude:
∗ 2π 3 2 −kd
F3inertia = −(ρV + Ca ρ R )ω e (D.6)
3
For the semi-submersible we have surface piercing columns, and the dynamic pressure
acting on the bottom of these columns are not included in the Morison formulation. Hence
the vertical force due to dynamic pressure must also be included in the Morison model.
Vertical force due to dynamic pressure under columns can be found as follows:
F3dynpres = πR2 pd (D.7)
∗
F3dynpres = ρgπR2 e−kd (D.8)
Horizontal wave forces
The horizontal force is found by integrating the inertia term along the length of the column
(z direction).
Z0
F1inertia 2
= ρπR (1 + Ca ) a1 (z)dz (D.9)
−d
∗
F1inertia = iρgπR2 (1 + Ca )(1 − e−kd ) (D.10)
167
z
+
x z̄
d
Figure D.2: Wave forces, showing pressure and resulting force.
Heave, roll and pitch forces
Total surge, heave and pitch forces due to wave with heading β is taken as the sum of the
forces on each column. θ is a function of β and the position of each column.
 
n=3
X
inertia∗
F1∗ =  F1 eiθn  eiωt
1
 
Xn=3
∗ ∗
F3∗ =  (F3 inertia dynpres
+ F3 )e n  eiωt
iθ
1
(D.11)
 
Xn=3 n=3
X
∗ ∗ ∗
F5∗ =  F1inertia eiθn z̄ − (F3inertia + F3dynpres )eiθn xn  eiωt
1 1
168
And the point of attac for the resultant horizontal force on a column, z̄, is found by:
Z0
1 kz 1 1 −kd 1 −kd
z̄ = e zdz = − +e d+ e
A kA k k
−d
(D.12)
Z0
1
A= ekz dz = 1 − e−kd
k
−d
The forces obtained by Eq. D.11 are compared to a linear panel method solution in Figure
D.3. A comparison of the forces from Eq. D.11 without the dynamic pressure term is
shown in Fig. D.4. The importance of including the dynamic pressure in a pure Morison
force model is obvious, in particular for heave forces.
Figure D.3: Analytical Morison and potential theory transfer function excluding dynamic
pressure.
169
Figure D.4: Analytical Morison and potential theory transfer function including dynamic
pressure.
Contribution to excitation force due to dynamic pressure
Since this is not included for slender elements in Simo, a force transfer function is derived
analytically and applied to the Simo body to account for these forces. These transfer
functions are derived below. Transfer function for roll motion (important for 90 deg wave
heading) is also included.
n=3
X ∗
F3∗ = F3dynpres eiθn
1
n=3
X ∗
F4∗ = F3dynpres eiθn yn
1
(D.13)
n=3
X ∗
F5∗ = − F3dynpres eiθn xn
1
170
Appendix E
Standards and guidelines on FLS design
Since no standard or guidelines for FWTs existed at the time the work with this thesis
started, a review of existing guidelines that could form a basis for fatigue design of FWTs
was performed. Relevant guidelines were considered to be standards and guidelines for
floating offshore structures and bottom fixed offshore wind turbines. The current chapter
gives an overview of statements about fatigue design from relevant design codes.
IEC 61400-3: Wind turbines - design requirements for offshore wind turbines
• The joint probability distribution shall be extended to include wind and wave direc-
tions if necessary
• Minimum bin sizes: 2 m/s for wind speed, 0.5 m for Hs , 0.5 s for Tp and 30◦ for
wind and wave direction
• Design load cases are shown in Table 2.1
DNV-RP-F205: Global performance analysis of deepwater floating structures

• Frequency domain analysis is extensively used for fatigue analyses of floating units
• Coupled analysis (floater, mooring system, risers) needs to be performed to accu-
rately predict system response in fatigue load cases
DNV-OS-C103/DNV-RP-C103: Column stabilised units

• Site specific environmental data shall be undertaken in the analyses
• Local detailed FEM analysis of critical connections should be performed to find
appropriate SCFs
171
• If the semi-submersible classifies as a Y-unit in harsh environment, stochastic fa-
tigue analysis has to be performed, otherwise, simplified fatigue analysis as de-
scribed in [26] can be performed
• The stochastic analysis procedure is illustrated by Figure E.1
• For the stochastic analysis, a screening analysis using a simplified approach can be
performed to identify critical structural details
• The FLS response analysis should cover the range of probability levels from 10−1
to 10−4 , for exceedance of stress ranges in the platform lifetime
• Short crested waves with a cos4 spreading function should be applied in fatigue
analyses
• Fatigue analyses shall include directional probability of the environmental data
• Where non-linear effects may be considered insignificant, frequency domain anal-
ysis are sufficient
• For details not sufficiently covered by DNV-RP-C203 [25], local FE analysis must
be performed
Figure E.1: Stochastic method for frequency domain stress analysis for oil and gas plat-
forms [24].
NORSOK-N-003: Action and Action Effects

• Linear wave theory is relevant for fatigue analysis
• Difference frequency wave forces may be important for global motions
• Time domain methods are normally not required for fatigue analysis
• Simplified approaches to determine fatigue damage may be used if properly vali-
dated
• In addition to linear wave loads, sum-frequency wave loads, variable buoyancy,
slamming, vortex shedding and mechanical vibration can contribute to fatigue
172
• Stress ranges for wide-banded or non-Gaussian histories should be determined by
an appropriate cycle counting method, e.g. Rainflow counting
• A simplified method for establishing the relation between stress and wave height
for fatigue analysis may be used (see [82])
173
174
Appendix F
Use of buoyancy compensating force in

coupled Simo-Riflex models
This is a note prepared by the author on a modelling method for coupled Simo-Riflex,
developed by ph.d. candidate Chenyu Luan and the author.
Table F.1: Nomenclature
B = ρg∀ Buoyancy force

C(i, j) Restoring matrix term for indices (i,j)
COB Centre of buoyancy
COG Centre of gravity for rigid part of the floating body
G = mg Weight of the rigid part of the floating body
g Gravitational acceleration
m Mass of the rigid part of the floating body
W Weight of Riflex elements
xb ,yb ,zb Centre of buoyancy location in global coordinate system
xg ,yg ,zg Centre of gravity location in global coordinate system
ρ Sea water mass density
∀ Displaced volume of body
Buoyancy compensating force in Simo
Simo-Riflex is traditionally used for coupled analysis of platforms modelled as a rigid

Simo body with flexible Riflex elements for mooring lines. Restoring matrix and force
transfer functions are imported to Simo from pressure panel analyses in Wadam or Wamit.
A basic assumption in Simo is that the Simo floating body, without the weight of the Riflex
elements, is neutrally buoyant. In reality it is the weight of the floating body including the
175
flexible elements that is neutrally buoyant. In Simo the assumption is that the effects of
gravity and buoyancy in the equation of motion is taken into account through the restoring
matrix. This is not the case for Riflex, which thought the finite element method includes
gravity and buoyancy as nodal forces. This inconsistency is the reason why a so called
buoyancy compensating force, acting in the centre of buoyancy, must be specified in Simo.
This is a suitable way of modelling when the mooring line weight is applied symmetrically
and is much smaller than the platform weight. In the case of floating wind turbines, the
platform (the rigid part of the model) is not always symmetrically loaded with the weight
of the flexible elements (i.e. when the centre of buoyancy and the centre of gravity of the
platform, without Riflex elements, do not have the same x or y coordinates). Using the
method of the buoyancy compensating force will in this case create unphysical forces on
the platform. How Simo interprets the forces acting on the floating body with buoyancy
compensating force and asymmetrical load from riflex elements is illustrated in Fig. F.1.
As can be seen by Figure F.1, when modelling the traditional way in Simo, there will be a
moment acting on the body, that in reality is cancelled out by the gravity force. This will
give an incorrect static configuration and possibly problems with running the analysis. In
the dynamic analysis, when the body moves, this additional force will give an unphysical
restoring moment.
B=G+W
∆B=W
COB
COG COB
W G W
(a) Forces acting on the neutrally buoyant (b) Simo interpretation where B=G.
model.
Figure F.1: Discretisation of physical models with asymmetrical gravitational loads.

Forces refer to global directions (not following the body).
Proposed procedure for floating wind turbine models
The following proposed modelling method eliminates the problem of the unphysical Simo
body moment. It is also an attempt to have only "real" forces in the model, to make it more
intuitive and less prone to modelling errors. In the proposed method, an upward speci-
fied force at the buoyancy center of the hull to represent the buoyancy, and a downward
specified force to represent the gravity.
These specified forces will give restoring forces since their directions refer to global di-
rections. However, this effect is already included in the restoring matrix that is obtained
176
from Wadam or Wamit. In order to avoid to count the restoring effect twice, this part of
the restoring coefficients must be subtracted from the original restoring matrix, such that
the restoring coefficients only include the water line area stiffness.
This method also works for models where the weight of Riflex elements is symmetric
about the Simo body origin. Points 1) through 4b) serve as a general approach to mod-
elling of floating wind turbines in coupled Simo-Riflex. There are also other modelling
strategies, not mentioned here, that will give the same results.
1. Calculate mass, inertia and mass distribution for the floating wind turbine platform
with and without the turbine
2. Do hydrodynamic panel method analysis in Wadam or Wamit with input describing
the whole system of platform and turbine
3. Import output from Wadam/Wamit to Simo (Sima, DeepC or Simo inpmod)
4. In Simo sys-*.dat file, change the following to describe the platform without the
turbine and mooring lines:
(a) Mass and inertia
(b) Centre of gravity
(c) Add a specified force equal to the weight of the body (W) in the centre of
gravity OR use the GRAVITY FORCE INCLUDED option in Simo
(d) Add a specified force equal to the buoyancy (B) of the body in the centre of
buoyancy
(e) Subtract the part of the restoring terms (4,4), (5,5), (4,6) and (5,6) that are
caused by G and B, see Eq. F.1. The new restoring coefficients could also
be calculated in Wadam or Wamit by specifying the gravity centre and the
buoyancy centre at the same point.
C(4, 4)new = C(4, 4)W adam − ρg∀zb + mgzg
C(5, 5)new = C(5, 5)W adam − ρg∀zb + mgzg
(F.1)
C(4, 6)new = C(4, 6)W adam + ρg∀xb − mgxg
C(5, 6)new = C(5, 6)W adam + ρg∀yb − mgyg
177
178
Previous PhD theses published at the Departement of Marine Technology
(earlier: Faculty of Marine Technology)
NORWEGIAN UNIVERSITY OF SCIENCE AND TECHNOLOGY
Report Author Title

No.
Kavlie, Dag Optimization of Plane Elastic Grillages, 1967
Hansen, Hans R. Man-Machine Communication and Data-Storage

Methods in Ship Structural Design, 1971
Gisvold, Kaare M. A Method for non-linear mixed -integer

programming and its Application to Design
Problems, 1971
Lund, Sverre Tanker Frame Optimalization by means of SUMT-

Transformation and Behaviour Models, 1971
Vinje, Tor On Vibration of Spherical Shells Interacting with

Fluid, 1972
Lorentz, Jan D. Tank Arrangement for Crude Oil Carriers in

Accordance with the new Anti-Pollution
Regulations, 1975
Carlsen, Carl A. Computer-Aided Design of Tanker Structures, 1975
Larsen, Carl M. Static and Dynamic Analysis of Offshore Pipelines

during Installation, 1976
UR-79-01 Brigt Hatlestad, MK The finite element method used in a fatigue

evaluation of fixed offshore platforms. (Dr.Ing.
Thesis)
UR-79-02 Erik Pettersen, MK Analysis and design of cellular structures. (Dr.Ing.

Thesis)
UR-79-03 Sverre Valsgård, MK Finite difference and finite element methods applied
to nonlinear analysis of plated structures. (Dr.Ing.
Thesis)
UR-79-04 Nils T. Nordsve, MK Finite element collapse analysis of structural

members considering imperfections and stresses due
to fabrication. (Dr.Ing. Thesis)
UR-79-05 Ivar J. Fylling, MK Analysis of towline forces in ocean towing systems.

(Dr.Ing. Thesis)
UR-80-06 Nils Sandsmark, MM Analysis of Stationary and Transient Heat

Conduction by the Use of the Finite Element
Method. (Dr.Ing. Thesis)
UR-80-09 Sverre Haver, MK Analysis of uncertainties related to the stochastic

modeling of ocean waves. (Dr.Ing. Thesis)
UR-81-15 Odland, Jonas On the Strength of welded Ring stiffened cylindrical

Shells primarily subjected to axial Compression
UR-82-17 Engesvik, Knut Analysis of Uncertainties in the fatigue Capacity of

Welded Joints
UR-82-18 Rye, Henrik Ocean wave groups
UR-83-30 Eide, Oddvar Inge On Cumulative Fatigue Damage in Steel Welded

Joints
UR-83-33 Mo, Olav Stochastic Time Domain Analysis of Slender

Offshore Structures
UR-83-34 Amdahl, Jørgen Energy absorption in Ship-platform impacts
UR-84-37 Mørch, Morten Motions and mooring forces of semi submersibles as

determined by full-scale measurements and
theoretical analysis
UR-84-38 Soares, C. Guedes Probabilistic models for load effects in ship

structures
UR-84-39 Aarsnes, Jan V. Current forces on ships
UR-84-40 Czujko, Jerzy Collapse Analysis of Plates subjected to Biaxial

Compression and Lateral Load
UR-85-46 Alf G. Engseth, MK Finite element collapse analysis of tubular steel

offshore structures. (Dr.Ing. Thesis)
UR-86-47 Dengody Sheshappa, MP A Computer Design Model for Optimizing Fishing

Vessel Designs Based on Techno-Economic
Analysis. (Dr.Ing. Thesis)
UR-86-48 Vidar Aanesland, MH A Theoretical and Numerical Study of Ship Wave

Resistance. (Dr.Ing. Thesis)
UR-86-49 Heinz-Joachim Wessel, MK Fracture Mechanics Analysis of Crack Growth in

Plate Girders. (Dr.Ing. Thesis)
UR-86-50 Jon Taby, MK Ultimate and Post-ultimate Strength of Dented

Tubular Members. (Dr.Ing. Thesis)
UR-86-51 Walter Lian, MH A Numerical Study of Two-Dimensional Separated

Flow Past Bluff Bodies at Moderate KC-Numbers.
(Dr.Ing. Thesis)
UR-86-52 Bjørn Sortland, MH Force Measurements in Oscillating Flow on Ship

Sections and Circular Cylinders in a U-Tube Water
Tank. (Dr.Ing. Thesis)
UR-86-53 Kurt Strand, MM A System Dynamic Approach to One-dimensional

Fluid Flow. (Dr.Ing. Thesis)
UR-86-54 Arne Edvin Løken, MH Three Dimensional Second Order Hydrodynamic

Effects on Ocean Structures in Waves. (Dr.Ing.
Thesis)
UR-86-55 Sigurd Falch, MH A Numerical Study of Slamming of Two-

Dimensional Bodies. (Dr.Ing. Thesis)
UR-87-56 Arne Braathen, MH Application of a Vortex Tracking Method to the

Prediction of Roll Damping of a Two-Dimension
Floating Body. (Dr.Ing. Thesis)
UR-87-57 Bernt Leira, MK Gaussian Vector Processes for Reliability Analysis
involving Wave-Induced Load Effects. (Dr.Ing.
Thesis)
UR-87-58 Magnus Småvik, MM Thermal Load and Process Characteristics in a Two-

Stroke Diesel Engine with Thermal Barriers (in
Norwegian). (Dr.Ing. Thesis)
MTA-88- Bernt Arild Bremdal, MP An Investigation of Marine Installation Processes –

59 A Knowledge - Based Planning Approach. (Dr.Ing.
Thesis)
MTA-88- Xu Jun, MK Non-linear Dynamic Analysis of Space-framed

60 Offshore Structures. (Dr.Ing. Thesis)
MTA-89- Gang Miao, MH Hydrodynamic Forces and Dynamic Responses of

61 Circular Cylinders in Wave Zones. (Dr.Ing. Thesis)
MTA-89- Martin Greenhow, MH Linear and Non-Linear Studies of Waves and

62 Floating Bodies. Part I and Part II. (Dr.Techn.
Thesis)
MTA-89- Chang Li, MH Force Coefficients of Spheres and Cubes in

63 Oscillatory Flow with and without Current. (Dr.Ing.
Thesis
MTA-89- Hu Ying, MP A Study of Marketing and Design in Development

64 of Marine Transport Systems. (Dr.Ing. Thesis)
MTA-89- Arild Jæger, MH Seakeeping, Dynamic Stability and Performance of

65 a Wedge Shaped Planing Hull. (Dr.Ing. Thesis)
MTA-89- Chan Siu Hung, MM The dynamic characteristics of tilting-pad bearings

66
MTA-89- Kim Wikstrøm, MP Analysis av projekteringen for ett offshore projekt.
67 (Licenciat-avhandling)
MTA-89- Jiao Guoyang, MK Reliability Analysis of Crack Growth under

68 Random Loading, considering Model Updating.
(Dr.Ing. Thesis)
MTA-89- Arnt Olufsen, MK Uncertainty and Reliability Analysis of Fixed

69 Offshore Structures. (Dr.Ing. Thesis)
MTA-89- Wu Yu-Lin, MR System Reliability Analyses of Offshore Structures

70 using improved Truss and Beam Models. (Dr.Ing.
Thesis)
MTA-90- Jan Roger Hoff, MH Three-dimensional Green function of a vessel with

71 forward speed in waves. (Dr.Ing. Thesis)
MTA-90- Rong Zhao, MH Slow-Drift Motions of a Moored Two-Dimensional

72 Body in Irregular Waves. (Dr.Ing. Thesis)
MTA-90- Atle Minsaas, MP Economical Risk Analysis. (Dr.Ing. Thesis)

73
MTA-90- Knut-Aril Farnes, MK Long-term Statistics of Response in Non-linear
74 Marine Structures. (Dr.Ing. Thesis)
MTA-90- Torbjørn Sotberg, MK Application of Reliability Methods for Safety

75 Assessment of Submarine Pipelines. (Dr.Ing.
Thesis)
MTA-90- Zeuthen, Steffen, MP SEAMAID. A computational model of the design

76 process in a constraint-based logic programming
environment. An example from the offshore
domain. (Dr.Ing. Thesis)
MTA-91- Haagensen, Sven, MM Fuel Dependant Cyclic Variability in a Spark

77 Ignition Engine - An Optical Approach. (Dr.Ing.
Thesis)
MTA-91- Løland, Geir, MH Current forces on and flow through fish farms.
78 (Dr.Ing. Thesis)
MTA-91- Hoen, Christopher, MK System Identification of Structures Excited by

79 Stochastic Load Processes. (Dr.Ing. Thesis)
MTA-91- Haugen, Stein, MK Probabilistic Evaluation of Frequency of Collision

80 between Ships and Offshore Platforms. (Dr.Ing.
Thesis)
MTA-91- Sødahl, Nils, MK Methods for Design and Analysis of Flexible Risers.
81 (Dr.Ing. Thesis)
MTA-91- Ormberg, Harald, MK Non-linear Response Analysis of Floating Fish

82 Farm Systems. (Dr.Ing. Thesis)
MTA-91- Marley, Mark J., MK Time Variant Reliability under Fatigue Degradation.
83 (Dr.Ing. Thesis)
MTA-91- Krokstad, Jørgen R., MH Second-order Loads in Multidirectional Seas.

84 (Dr.Ing. Thesis)
MTA-91- Molteberg, Gunnar A., MM The Application of System Identification

85 Techniques to Performance Monitoring of Four
Stroke Turbocharged Diesel Engines. (Dr.Ing.
Thesis)
MTA-92- Mørch, Hans Jørgen Bjelke, MH Aspects of Hydrofoil Design: with Emphasis on
86 Hydrofoil Interaction in Calm Water. (Dr.Ing.
Thesis)
MTA-92- Chan Siu Hung, MM Nonlinear Analysis of Rotordynamic Instabilities in

87 Highspeed Turbomachinery. (Dr.Ing. Thesis)
MTA-92- Bessason, Bjarni, MK Assessment of Earthquake Loading and Response of

88 Seismically Isolated Bridges. (Dr.Ing. Thesis)
MTA-92- Langli, Geir, MP Improving Operational Safety through exploitation

89 of Design Knowledge - an investigation of offshore
platform safety. (Dr.Ing. Thesis)
MTA-92- Sævik, Svein, MK On Stresses and Fatigue in Flexible Pipes. (Dr.Ing.

90 Thesis)
MTA-92- Ask, Tor Ø., MM Ignition and Flame Growth in Lean Gas-Air
91 Mixtures. An Experimental Study with a Schlieren
System. (Dr.Ing. Thesis)
MTA-86- Hessen, Gunnar, MK Fracture Mechanics Analysis of Stiffened Tubular

92 Members. (Dr.Ing. Thesis)
MTA-93- Steinebach, Christian, MM Knowledge Based Systems for Diagnosis of
93 Rotating Machinery. (Dr.Ing. Thesis)
MTA-93- Dalane, Jan Inge, MK System Reliability in Design and Maintenance of

94 Fixed Offshore Structures. (Dr.Ing. Thesis)
MTA-93- Steen, Sverre, MH Cobblestone Effect on SES. (Dr.Ing. Thesis)

95
MTA-93- Karunakaran, Daniel, MK Nonlinear Dynamic Response and Reliability
96 Analysis of Drag-dominated Offshore Platforms.
(Dr.Ing. Thesis)
MTA-93- Hagen, Arnulf, MP The Framework of a Design Process Language.

97 (Dr.Ing. Thesis)
MTA-93- Nordrik, Rune, MM Investigation of Spark Ignition and Autoignition in

98 Methane and Air Using Computational Fluid
Dynamics and Chemical Reaction Kinetics. A
Numerical Study of Ignition Processes in Internal
Combustion Engines. (Dr.Ing. Thesis)
MTA-94- Passano, Elizabeth, MK Efficient Analysis of Nonlinear Slender Marine

99 Structures. (Dr.Ing. Thesis)
MTA-94- Kvålsvold, Jan, MH Hydroelastic Modelling of Wetdeck Slamming on

100 Multihull Vessels. (Dr.Ing. Thesis)
MTA-94- Bech, Sidsel M., MK Experimental and Numerical Determination of

102 Stiffness and Strength of GRP/PVC Sandwich
Structures. (Dr.Ing. Thesis)
MTA-95- Paulsen, Hallvard, MM A Study of Transient Jet and Spray using a

103 Schlieren Method and Digital Image Processing.
(Dr.Ing. Thesis)
MTA-95- Hovde, Geir Olav, MK Fatigue and Overload Reliability of Offshore

104 Structural Systems, Considering the Effect of
Inspection and Repair. (Dr.Ing. Thesis)
MTA-95- Wang, Xiaozhi, MK Reliability Analysis of Production Ships with

105 Emphasis on Load Combination and Ultimate
Strength. (Dr.Ing. Thesis)
MTA-95- Ulstein, Tore, MH Nonlinear Effects of a Flexible Stern Seal Bag on

106 Cobblestone Oscillations of an SES. (Dr.Ing.
Thesis)
MTA-95- Solaas, Frøydis, MH Analytical and Numerical Studies of Sloshing in

107 Tanks. (Dr.Ing. Thesis)
MTA-95- Hellan, Øyvind, MK Nonlinear Pushover and Cyclic Analyses in

108 Ultimate Limit State Design and Reassessment of
Tubular Steel Offshore Structures. (Dr.Ing. Thesis)
MTA-95- Hermundstad, Ole A., MK Theoretical and Experimental Hydroelastic Analysis

109 of High Speed Vessels. (Dr.Ing. Thesis)
MTA-96- Bratland, Anne K., MH Wave-Current Interaction Effects on Large-Volume

110 Bodies in Water of Finite Depth. (Dr.Ing. Thesis)
MTA-96- Herfjord, Kjell, MH A Study of Two-dimensional Separated Flow by a

111 Combination of the Finite Element Method and
Navier-Stokes Equations. (Dr.Ing. Thesis)
MTA-96- Æsøy, Vilmar, MM Hot Surface Assisted Compression Ignition in a

112 Direct Injection Natural Gas Engine. (Dr.Ing.
Thesis)
MTA-96- Eknes, Monika L., MK Escalation Scenarios Initiated by Gas Explosions on

113 Offshore Installations. (Dr.Ing. Thesis)
MTA-96- Erikstad, Stein O., MP A Decision Support Model for Preliminary Ship
114 Design. (Dr.Ing. Thesis)
MTA-96- Pedersen, Egil, MH A Nautical Study of Towed Marine Seismic

115 Streamer Cable Configurations. (Dr.Ing. Thesis)
MTA-97- Moksnes, Paul O., MM Modelling Two-Phase Thermo-Fluid Systems Using

116 Bond Graphs. (Dr.Ing. Thesis)
MTA-97- Halse, Karl H., MK On Vortex Shedding and Prediction of Vortex-

117 Induced Vibrations of Circular Cylinders. (Dr.Ing.
Thesis)
MTA-97- Igland, Ragnar T., MK Reliability Analysis of Pipelines during Laying,

118 considering Ultimate Strength under Combined
Loads. (Dr.Ing. Thesis)
MTA-97- Pedersen, Hans-P., MP Levendefiskteknologi for fiskefartøy. (Dr.Ing.

119 Thesis)
MTA-98- Vikestad, Kyrre, MK Multi-Frequency Response of a Cylinder Subjected

120 to Vortex Shedding and Support Motions. (Dr.Ing.
Thesis)
MTA-98- Azadi, Mohammad R. E., MK Analysis of Static and Dynamic Pile-Soil-Jacket

121 Behaviour. (Dr.Ing. Thesis)
MTA-98- Ulltang, Terje, MP A Communication Model for Product Information.

122 (Dr.Ing. Thesis)
MTA-98- Torbergsen, Erik, MM Impeller/Diffuser Interaction Forces in Centrifugal

123 Pumps. (Dr.Ing. Thesis)
MTA-98- Hansen, Edmond, MH A Discrete Element Model to Study Marginal Ice

124 Zone Dynamics and the Behaviour of Vessels
Moored in Broken Ice. (Dr.Ing. Thesis)
MTA-98- Videiro, Paulo M., MK Reliability Based Design of Marine Structures.

MTA-99- Mainçon, Philippe, MK Fatigue Reliability of Long Welds Application to

126 Titanium Risers. (Dr.Ing. Thesis)
MTA-99- Haugen, Elin M., MH Hydroelastic Analysis of Slamming on Stiffened

127 Plates with Application to Catamaran Wetdecks.
(Dr.Ing. Thesis)
MTA-99- Langhelle, Nina K., MK Experimental Validation and Calibration of

128 Nonlinear Finite Element Models for Use in Design
of Aluminium Structures Exposed to Fire. (Dr.Ing.
Thesis)
MTA-99- Berstad, Are J., MK Calculation of Fatigue Damage in Ship Structures.

MTA-99- Andersen, Trond M., MM Short Term Maintenance Planning. (Dr.Ing. Thesis)
130
MTA-99- Tveiten, Bård Wathne, MK Fatigue Assessment of Welded Aluminium Ship
131 Details. (Dr.Ing. Thesis)
MTA-99- Søreide, Fredrik, MP Applications of underwater technology in deep

132 water archaeology. Principles and practice. (Dr.Ing.
Thesis)
MTA-99- Tønnessen, Rune, MH A Finite Element Method Applied to Unsteady

133 Viscous Flow Around 2D Blunt Bodies With Sharp
Corners. (Dr.Ing. Thesis)
MTA-99- Elvekrok, Dag R., MP Engineering Integration in Field Development

134 Projects in the Norwegian Oil and Gas Industry. The
Supplier Management of Norne. (Dr.Ing. Thesis)
MTA-99- Fagerholt, Kjetil, MP Optimeringsbaserte Metoder for Ruteplanlegging

135 innen skipsfart. (Dr.Ing. Thesis)
MTA-99- Bysveen, Marie, MM Visualization in Two Directions on a Dynamic

136 Combustion Rig for Studies of Fuel Quality.
(Dr.Ing. Thesis)
MTA- Storteig, Eskild, MM Dynamic characteristics and leakage performance of

2000-137 liquid annular seals in centrifugal pumps. (Dr.Ing.
Thesis)
MTA- Sagli, Gro, MK Model uncertainty and simplified estimates of long

2000-138 term extremes of hull girder loads in ships. (Dr.Ing.
Thesis)
MTA- Tronstad, Harald, MK Nonlinear analysis and design of cable net structures
2000-139 like fishing gear based on the finite element method.
(Dr.Ing. Thesis)
MTA- Kroneberg, André, MP Innovation in shipping by using scenarios. (Dr.Ing.

2000-140 Thesis)
MTA- Haslum, Herbjørn Alf, MH Simplified methods applied to nonlinear motion of

2000-141 spar platforms. (Dr.Ing. Thesis)
MTA- Samdal, Ole Johan, MM Modelling of Degradation Mechanisms and Stressor

2001-142 Interaction on Static Mechanical Equipment
Residual Lifetime. (Dr.Ing. Thesis)
MTA- Baarholm, Rolf Jarle, MH Theoretical and experimental studies of wave

2001-143 impact underneath decks of offshore platforms.
(Dr.Ing. Thesis)
MTA- Wang, Lihua, MK Probabilistic Analysis of Nonlinear Wave-induced

2001-144 Loads on Ships. (Dr.Ing. Thesis)
MTA- Kristensen, Odd H. Holt, MK Ultimate Capacity of Aluminium Plates under

2001-145 Multiple Loads, Considering HAZ Properties.
(Dr.Ing. Thesis)
MTA- Greco, Marilena, MH A Two-Dimensional Study of Green-Water

2001-146 Loading. (Dr.Ing. Thesis)
MTA- Heggelund, Svein E., MK Calculation of Global Design Loads and Load
2001-147 Effects in Large High Speed Catamarans. (Dr.Ing.
Thesis)
MTA- Babalola, Olusegun T., MK Fatigue Strength of Titanium Risers – Defect

2001-148 Sensitivity. (Dr.Ing. Thesis)
MTA- Mohammed, Abuu K., MK Nonlinear Shell Finite Elements for Ultimate
2001-149 Strength and Collapse Analysis of Ship Structures.
(Dr.Ing. Thesis)
MTA- Holmedal, Lars E., MH Wave-current interactions in the vicinity of the sea
2002-150 bed. (Dr.Ing. Thesis)
MTA- Rognebakke, Olav F., MH Sloshing in rectangular tanks and interaction with
2002-151 ship motions. (Dr.Ing. Thesis)
MTA- Lader, Pål Furset, MH Geometry and Kinematics of Breaking Waves.

2002-152 (Dr.Ing. Thesis)
MTA- Yang, Qinzheng, MH Wash and wave resistance of ships in finite water
2002-153 depth. (Dr.Ing. Thesis)
MTA- Melhus, Øyvin, MM Utilization of VOC in Diesel Engines. Ignition and

2002-154 combustion of VOC released by crude oil tankers.
(Dr.Ing. Thesis)
MTA- Ronæss, Marit, MH Wave Induced Motions of Two Ships Advancing on

2002-155 Parallel Course. (Dr.Ing. Thesis)
MTA- Økland, Ole D., MK Numerical and experimental investigation of

2002-156 whipping in twin hull vessels exposed to severe wet
deck slamming. (Dr.Ing. Thesis)
MTA- Ge, Chunhua, MK Global Hydroelastic Response of Catamarans due to

2002-157 Wet Deck Slamming. (Dr.Ing. Thesis)
MTA- Byklum, Eirik, MK Nonlinear Shell Finite Elements for Ultimate

2002-158 Strength and Collapse Analysis of Ship Structures.
(Dr.Ing. Thesis)
IMT- Chen, Haibo, MK Probabilistic Evaluation of FPSO-Tanker Collision

2003-1 in Tandem Offloading Operation. (Dr.Ing. Thesis)
IMT- Skaugset, Kjetil Bjørn, MK On the Suppression of Vortex Induced Vibrations of

2003-2 Circular Cylinders by Radial Water Jets. (Dr.Ing.
Thesis)
IMT- Chezhian, Muthu Three-Dimensional Analysis of Slamming. (Dr.Ing.

2003-3 Thesis)
IMT- Buhaug, Øyvind Deposit Formation on Cylinder Liner Surfaces in

2003-4 Medium Speed Engines. (Dr.Ing. Thesis)
IMT- Tregde, Vidar Aspects of Ship Design: Optimization of Aft Hull

2003-5 with Inverse Geometry Design. (Dr.Ing. Thesis)
IMT- Wist, Hanne Therese Statistical Properties of Successive Ocean Wave

2003-6 Parameters. (Dr.Ing. Thesis)
IMT- Ransau, Samuel Numerical Methods for Flows with Evolving
2004-7 Interfaces. (Dr.Ing. Thesis)
IMT- Soma, Torkel Blue-Chip or Sub-Standard. A data interrogation

2004-8 approach of identity safety characteristics of
shipping organization. (Dr.Ing. Thesis)
IMT- Ersdal, Svein An experimental study of hydrodynamic forces on

2004-9 cylinders and cables in near axial flow. (Dr.Ing.
Thesis)
IMT- Brodtkorb, Per Andreas The Probability of Occurrence of Dangerous Wave

2005-10 Situations at Sea. (Dr.Ing. Thesis)
IMT- Yttervik, Rune Ocean current variability in relation to offshore

2005-11 engineering. (Dr.Ing. Thesis)
IMT- Fredheim, Arne Current Forces on Net-Structures. (Dr.Ing. Thesis)

2005-12
IMT- Heggernes, Kjetil Flow around marine structures. (Dr.Ing. Thesis
2005-13
IMT- Fouques, Sebastien Lagrangian Modelling of Ocean Surface Waves and
2005-14 Synthetic Aperture Radar Wave Measurements.
(Dr.Ing. Thesis)
IMT- Holm, Håvard Numerical calculation of viscous free surface flow

2006-15 around marine structures. (Dr.Ing. Thesis)
IMT- Bjørheim, Lars G. Failure Assessment of Long Through Thickness

2006-16 Fatigue Cracks in Ship Hulls. (Dr.Ing. Thesis)
IMT- Hansson, Lisbeth Safety Management for Prevention of Occupational

2006-17 Accidents. (Dr.Ing. Thesis)
IMT- Zhu, Xinying Application of the CIP Method to Strongly

2006-18 Nonlinear Wave-Body Interaction Problems.
(Dr.Ing. Thesis)
IMT- Reite, Karl Johan Modelling and Control of Trawl Systems. (Dr.Ing.
2006-19 Thesis)
IMT- Smogeli, Øyvind Notland Control of Marine Propellers. From Normal to

2006-20 Extreme Conditions. (Dr.Ing. Thesis)
IMT- Storhaug, Gaute Experimental Investigation of Wave Induced

2007-21 Vibrations and Their Effect on the Fatigue Loading
of Ships. (Dr.Ing. Thesis)
IMT- Sun, Hui A Boundary Element Method Applied to Strongly

2007-22 Nonlinear Wave-Body Interaction Problems. (PhD
Thesis, CeSOS)
IMT- Rustad, Anne Marthine Modelling and Control of Top Tensioned Risers.
2007-23 (PhD Thesis, CeSOS)
IMT- Johansen, Vegar Modelling flexible slender system for real-time

2007-24 simulations and control applications
IMT- Wroldsen, Anders Sunde Modelling and control of tensegrity structures. (PhD
2007-25 Thesis, CeSOS)
IMT- Aronsen, Kristoffer Høye An experimental investigation of in-line and

2007-26 combined inline and cross flow vortex induced
vibrations. (Dr. avhandling, IMT)
IMT- Gao, Zhen Stochastic Response Analysis of Mooring Systems

2007-27 with Emphasis on Frequency-domain Analysis of
Fatigue due to Wide-band Response Processes (PhD
Thesis, CeSOS)
IMT- Thorstensen, Tom Anders Lifetime Profit Modelling of Ageing Systems

2007-28 Utilizing Information about Technical Condition.
(Dr.ing. thesis, IMT)
IMT- Berntsen, Per Ivar B. Structural Reliability Based Position Mooring.

2008-29 (PhD-Thesis, IMT)
IMT- Ye, Naiquan Fatigue Assessment of Aluminium Welded Box-

2008-30 stiffener Joints in Ships (Dr.ing. thesis, IMT)
IMT- Radan, Damir Integrated Control of Marine Electrical Power

2008-31 Systems. (PhD-Thesis, IMT)
IMT- Thomassen, Paul Methods for Dynamic Response Analysis and

2008-32 Fatigue Life Estimation of Floating Fish Cages.
(Dr.ing. thesis, IMT)
IMT- Pákozdi, Csaba A Smoothed Particle Hydrodynamics Study of Two-

2008-33 dimensional Nonlinear Sloshing in Rectangular
Tanks. (Dr.ing.thesis, IMT/ CeSOS)
IMT- Grytøyr, Guttorm A Higher-Order Boundary Element Method and

2007-34 Applications to Marine Hydrodynamics.
(Dr.ing.thesis, IMT)
IMT- Drummen, Ingo Experimental and Numerical Investigation of

2008-35 Nonlinear Wave-Induced Load Effects in
Containerships considering Hydroelasticity. (PhD
thesis, CeSOS)
IMT- Skejic, Renato Maneuvering and Seakeeping of a Singel Ship and

2008-36 of Two Ships in Interaction. (PhD-Thesis, CeSOS)
IMT- Harlem, Alf An Age-Based Replacement Model for Repairable

2008-37 Systems with Attention to High-Speed Marine
Diesel Engines. (PhD-Thesis, IMT)
IMT- Alsos, Hagbart S. Ship Grounding. Analysis of Ductile Fracture,

2008-38 Bottom Damage and Hull Girder Response. (PhD-
thesis, IMT)
IMT- Graczyk, Mateusz Experimental Investigation of Sloshing Loading and

2008-39 Load Effects in Membrane LNG Tanks Subjected to
Random Excitation. (PhD-thesis, CeSOS)
IMT- Taghipour, Reza Efficient Prediction of Dynamic Response for

2008-40 Flexible amd Multi-body Marine Structures. (PhD-
thesis, CeSOS)
IMT- Ruth, Eivind Propulsion control and thrust allocation on marine

2008-41 vessels. (PhD thesis, CeSOS)
IMT- Nystad, Bent Helge Technical Condition Indexes and Remaining Useful
2008-42 Life of Aggregated Systems. PhD thesis, IMT
IMT- Soni, Prashant Kumar Hydrodynamic Coefficients for Vortex Induced
2008-43 Vibrations of Flexible Beams, PhD
thesis, CeSOS
IMT- Amlashi, Hadi K.K. Ultimate Strength and Reliability-based Design of

2009-43 Ship Hulls with Emphasis on Combined Global and
Local Loads. PhD Thesis, IMT
IMT- Pedersen, Tom Arne Bond Graph Modelling of Marine Power Systems.
2009-44 PhD Thesis, IMT
IMT- Kristiansen, Trygve Two-Dimensional Numerical and Experimental

2009-45 Studies of Piston-Mode Resonance. PhD-Thesis,
CeSOS
IMT- Ong, Muk Chen Applications of a Standard High Reynolds Number

2009-46 Model and a Stochastic Scour Prediction Model for
Marine Structures. PhD-thesis, IMT
IMT- Hong, Lin Simplified Analysis and Design of Ships subjected

2009-47 to Collision and Grounding. PhD-thesis, IMT
IMT- Koushan, Kamran Vortex Induced Vibrations of Free Span Pipelines,

2009-48 PhD thesis, IMT
IMT- Korsvik, Jarl Eirik Heuristic Methods for Ship Routing and Scheduling.
2009-49 PhD-thesis, IMT
IMT- Lee, Jihoon Experimental Investigation and Numerical in

2009-50 Analyzing the Ocean Current Displacement of
Longlines. Ph.d.-Thesis, IMT.
IMT- Vestbøstad, Tone Gran A Numerical Study of Wave-in-Deck Impact usin a

2009-51 Two-Dimensional Constrained Interpolation Profile
Method, Ph.d.thesis, CeSOS.
IMT- Bruun, Kristine Bond Graph Modelling of Fuel Cells for Marine
2009-52 Power Plants. Ph.d.-thesis, IMT
IMT Holstad, Anders Numerical Investigation of Turbulence in a Sekwed

2009-53 Three-Dimensional Channel Flow, Ph.d.-thesis,
IMT.
IMT Ayala-Uraga, Efren Reliability-Based Assessment of Deteriorating Ship-

2009-54 shaped Offshore Structures, Ph.d.-thesis, IMT
IMT Kong, Xiangjun A Numerical Study of a Damaged Ship in Beam Sea

2009-55 Waves. Ph.d.-thesis, IMT/CeSOS.
IMT Kristiansen, David Wave Induced Effects on Floaters of Aquaculture

2010-56 Plants, Ph.d.-thesis, CeSOS.
IMT Ludvigsen, Martin An ROV-Toolbox for Optical and Acoustic

2010-57 Scientific Seabed Investigation. Ph.d.-thesis IMT.
IMT Hals, Jørgen Modelling and Phase Control of Wave-Energy

2010-58 Converters. Ph.d.thesis, CeSOS.
IMT Shu, Zhi Uncertainty Assessment of Wave Loads and

2010- 59 Ultimate Strength of Tankers and Bulk Carriers in a
Reliability Framework. Ph.d. Thesis, IMT/ CeSOS
IMT Shao, Yanlin Numerical Potential-Flow Studies on Weakly-
2010-60 Nonlinear Wave-Body Interactions with/without
Small Forward Speed, Ph.d.thesis,CeSOS.
IMT Califano, Andrea Dynamic Loads on Marine Propellers due to
2010-61 Intermittent Ventilation. Ph.d.thesis, IMT.
IMT El Khoury, George Numerical Simulations of Massively Separated
2010-62 Turbulent Flows, Ph.d.-thesis, IMT
IMT Seim, Knut Sponheim Mixing Process in Dense Overflows with Emphasis
2010-63 on the Faroe Bank Channel Overflow. Ph.d.thesis,
IMT
IMT Jia, Huirong Structural Analysis of Intect and Damaged Ships in

2010-64 a Collission Risk Analysis Perspective. Ph.d.thesis
CeSoS.
IMT Jiao, Linlin Wave-Induced Effects on a Pontoon-type Very

2010-65 Large Floating Structures (VLFS). Ph.D.-thesis,
CeSOS.
IMT Abrahamsen, Bjørn Christian Sloshing Induced Tank Roof with Entrapped Air
2010-66 Pocket. Ph.d.thesis, CeSOS.
IMT Karimirad, Madjid Stochastic Dynamic Response Analysis of Spar-

2011-67 Type Wind Turbines with Catenary or Taut Mooring
Systems. Ph.d.-thesis, CeSOS.
IMT - Erlend Meland Condition Monitoring of Safety Critical Valves.

2011-68 Ph.d.-thesis, IMT.
IMT – Yang, Limin Stochastic Dynamic System Analysis of Wave

2011-69 Energy Converter with Hydraulic Power Take-Off,
with Particular Reference to Wear Damage
Analysis, Ph.d. Thesis, CeSOS.
IMT – Visscher, Jan Application of Particla Image Velocimetry on

2011-70 Turbulent Marine Flows, Ph.d.Thesis, IMT.
IMT – Su, Biao Numerical Predictions of Global and Local Ice

2011-71 Loads on Ships. Ph.d.Thesis, CeSOS.
IMT – Liu, Zhenhui Analytical and Numerical Analysis of Iceberg

2011-72 Collision with Ship Structures. Ph.d.Thesis, IMT.
IMT – Aarsæther, Karl Gunnar Modeling and Analysis of Ship Traffic by

2011-73 Observation and Numerical Simulation. Ph.d.Thesis,
IMT.
Imt – Wu, Jie Hydrodynamic Force Identification from Stochastic

2011-74 Vortex Induced Vibration Experiments with Slender
Beams. Ph.d.Thesis, IMT.
Imt – Amini, Hamid Azimuth Propulsors in Off-design Conditions.

2011-75 Ph.d.Thesis, IMT.
IMT – Nguyen, Tan-Hoi Toward a System of Real-Time Prediction and

2011-76 Monitoring of Bottom Damage Conditions During
Ship Grounding. Ph.d.thesis, IMT.
IMT- Tavakoli, Mohammad T. Assessment of Oil Spill in Ship Collision and

2011-77 Grounding, Ph.d.thesis, IMT.
IMT- Guo, Bingjie Numerical and Experimental Investigation of Added

2011-78 Resistance in Waves. Ph.d.Thesis, IMT.
IMT- Chen, Qiaofeng Ultimate Strength of Aluminium Panels, considering

2011-79 HAZ Effects, IMT
IMT- Kota, Ravikiran S. Wave Loads on Decks of Offshore Structures in

2012-80 Random Seas, CeSOS.
IMT- Sten, Ronny Dynamic Simulation of Deep Water Drilling Risers

2012-81 with Heave Compensating System, IMT.
IMT- Berle, Øyvind Risk and resilience in global maritime supply

2012-82 chains, IMT.
IMT- Fang, Shaoji Fault Tolerant Position Mooring Control Based on

2012-83 Structural Reliability, CeSOS.
IMT- You, Jikun Numerical studies on wave forces and moored ship
2012-84 motions in intermediate and shallow water, CeSOS.
IMT- Xiang ,Xu Maneuvering of two interacting ships in waves,

2012-85 CeSOS
IMT- Dong, Wenbin Time-domain fatigue response and reliability

2012-86 analysis of offshore wind turbines with emphasis on
welded tubular joints and gear components, CeSOS
IMT- Zhu, Suji Investigation of Wave-Induced Nonlinear Load

2012-87 Effects in Open Ships considering Hull Girder
Vibrations in Bending and Torsion, CeSOS
IMT- Zhou, Li Numerical and Experimental Investigation of

2012-88 Station-keeping in Level Ice, CeSOS
IMT- Ushakov, Sergey Particulate matter emission characteristics from

2012-90 diesel enignes operating on conventional and
alternative marine fuels, IMT
IMT- Yin, Decao Experimental and Numerical Analysis of Combined

2013-1 In-line and Cross-flow Vortex Induced Vibrations,
CeSOS
IMT- Kurniawan, Adi Modelling and geometry optimisation of wave

2013-2 energy converters, CeSOS
IMT- Al Ryati, Nabil Technical condition indexes doe auxiliary marine

2013-3 diesel engines, IMT
IMT- Firoozkoohi, Reza Experimental, numerical and analytical investigation

2013-4 of the effect of screens on sloshing, CeSOS
IMT- Ommani, Babak Potential-Flow Predictions of a Semi-Displacement

2013-5 Vessel Including Applications to Calm Water
Broaching, CeSOS
IMT- Xing, Yihan Modelling and analysis of the gearbox in a floating
2013-6 spar-type wind turbine, CeSOS
IMT-7- Balland, Océane Optimization models for reducing air emissions

2013 from ships, IMT
IMT-8- Yang, Dan Transitional wake flow behind an inclined flat plate-
2013 ----Computation and analysis, IMT
IMT-9- Abdillah, Suyuthi Prediction of Extreme Loads and Fatigue Damage

2013 for a Ship Hull due to Ice Action, IMT
IMT-10- Ramìrez, Pedro Agustìn Pèrez Ageing management and life extension of technical
2013 systems-
Concepts and methods applied to oil and gas
facilities, IMT
IMT-11- Chuang, Zhenju Experimental and Numerical Investigation of Speed

2013 Loss due to Seakeeping and Maneuvering. IMT
IMT-12- Etemaddar, Mahmoud Load and Response Analysis of Wind Turbines

2013 under Atmospheric Icing and Controller System
Faults with Emphasis on Spar Type Floating Wind
Turbines, IMT
IMT-13- Lindstad, Haakon Strategies and measures for reducing maritime CO2
2013 emissons, IMT
IMT-14- Haris, Sabril Damage interaction analysis of ship collisions, IMT

2013
IMT-15- Shainee, Mohamed Conceptual Design, Numerical and Experimental
2013 Investigation of a SPM Cage Concept for Offshore
Mariculture, IMT
IMT-16- Gansel, Lars Flow past porous cylinders and effects of biofouling
2013 and fish behavior on the flow in and around Atlantic
salmon net cages, IMT
IMT-17- Gaspar, Henrique Handling Aspects of Complexity in Conceptual Ship

2013 Design, IMT
IMT-18- Thys, Maxime Theoretical and Experimental Investigation of a

2013 Free Running Fishing Vessel at Small Frequency of
Encounter, CeSOS
IMT-19- Aglen, Ida VIV in Free Spanning Pipelines, CeSOS

2013
IMT-1- Song, An Theoretical and experimental studies of wave
2014 diffraction and radiation loads on a horizontally
submerged perforated plate, CeSOS
IMT-2- Rogne, Øyvind Ygre Numerical and Experimental Investigation of a

2014 Hinged 5-body Wave Energy Converter, CeSOS
IMT-3- Dai, Lijuan Safe and efficient operation and maintenance of

2014 offshore wind farms ,IMT
IMT-4- Bachynski, Erin Elizabeth Design and Dynamic Analysis of Tension Leg
2014 Platform Wind Turbines, CeSOS
IMT-5- Wang, Jingbo Water Entry of Freefall Wedged – Wedge motions

2014
and Cavity Dynamics, CeSOS
IMT-6- Kim, Ekaterina Experimental and numerical studies related to the

2014 coupled behavior of ice mass and steel structures
during accidental collisions, IMT
IMT-7- Tan, Xiang Numerical investigation of ship’s continuous- mode

2014 icebreaking in leverl ice, CeSOS
IMT-8- Muliawan, Made Jaya Design and Analysis of Combined Floating Wave
2014 and Wind Power Facilities, with Emphasis on
Extreme Load Effects of the Mooring System,
CeSOS
IMT-9- Jiang, Zhiyu Long-term response analysis of wind turbines with

2014 an emphasis on fault and shutdown conditions, IMT
IMT-10- Dukan, Fredrik ROV Motion Control Systems, IMT

2014
IMT-11- Grimsmo, Nils I. Dynamic simulations of hydraulic cylinder for
2014 heave compensation of deep water drilling risers,
IMT
IMT-12- Kvittem, Marit I. Modelling and response analysis for fatigue design
2014 of a semisubmersible wind turbine, CeSOS

Marit Irene Kvittem PHD

Uploaded by

Copyright:

Available Formats

Marit Irene Kvittem PHD

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Marit Irene Kvittem PHD

Uploaded by

Copyright:

Available Formats

Doctoral theses at NTNU, 2014:337

Marit Irene Kvittem

ISBN 978-82-326-0586-6 (printed version)

Doctoral theses at NTNU, 2014:337

Department of Marine Technology

Modelling and response analysis for

Thesis for the degree of Philosophiae Doctor

Trondheim, November 2014

Norwegian University of Science and Technology

Thesis for the degree of Philosophiae Doctor

Faculty of Engineering Science and Technology

© Marit Irene Kvittem

ISBN 978-82-326-0586-6 (printed version)

Doctoral theses at NTNU, 2014:337

Printed by Skipnes Kommunikasjon as

3 Modelling of floating wind turbines 23

5 Conclusions and recommendations for further work 57

B Abstracts of additional papers 153

C Specifications for a semi-submersible wind turbine with active ballast system157

D Analytical Morison force transfer functions 165

E Standards and guidelines on FLS design 171

Abbreviations RAO response amplitude operator

FE finite element ULS ultimate limit state

FLS fatigue limit state UNI uni-directional

FWT Floating wind turbine Greek characters

LWT land based wind turbine κ(t) retardation function

MBS multi body simulation λ wave length

Mor. Morison ω circular frequency

MSL mean sea level ρa air density

MUL multi-directional ρw seawater density

MWL mean water level σ axial stress

NSS normal sea state τ time lag or shear stress

NTM normal turbulence model Ω rotor speed

The thesis consists of a collection of 4 papers, which are appended in App. A.

1.2 Research context

1.5 Papers and declaration of authorship

P2 Marit I. Kvittem and Torgeir Moan: Time domain analysis procedures

P3 Erin E. Bachynski, Marit I. Kvittem, Chenyu Luan and Torgeir Moan:

1.5.1 Additional papers

P6 Marit I. Kvittem and Torgeir Moan: Effect of mooring line modelling

P7 Erin E. Bachynski, Mahmoud Etemaddar, Marit I. Kvittem, Chenyu

P8 Torgeir Moan, Zhen Gao, Madjid Karimirad, Erin E. Bachynski, Mah-

C1 A numerical tool for coupled analysis of floating wind turbines. Ex-

C2 Development of a simplified method to assess fatigue damage for the

Simulation Reduce number

Figure 1.1: Relations between research questions, papers and contributions.

2.1 Floating wind turbine concepts

(c) Hywind (d) Pelastar

(e) HiPRWind (f) WindFloat

2.1.1 Semi-submersible wind turbines

Figure 2.2: Illustration of ballast system.

2.2 Design and analysis of floating wind turbines

2.2.1 Design load cases

Name DLC Wind Wave Wind/wave Current Water

Power 1.2 NTM NSS COD, No NWLR

Power 2.4 NTM NSS COD, No NWLR

Start up 3.1 NWP NSS(or NWH) COD, No NWLR

Normal 4.1 NWP NSS(or NWH) COD, No NWLR

Parked 6.4 NTM NSS COD, No NWLR