Content uploaded by David Marten
Author content
All content in this area was uploaded by David Marten on Apr 17, 2023
Content may be subject to copyright.
FLOATECH
D2.4. Full report on the estimated
reduction of uncertainty in comparison
to the stateoftheart codes OpenFAST
and DeepLines Wind™
31/10/2022
Francesco Papi, Alessandro Bianchini (UNIFI)
Giancarlo Troise, Gerardo Mario Mirra (SEAPOWER)
David Marten, Joseph Saverin, Robert Behrens de Luna (TUB)
MarieLaure Ducasse, Jonathan Honnet (SAIPEM)
Ref. Ares(2022)9015219  30/12/2022
FLOATECH – EUH2020 Grant Agreement N°101007142
FLOATECH D2.4: estimated reduction of uncertainty in comparison to the stateoftheart codes
2/136
Document track details
Project acronym FLOATECH
Project title
Optimization of floating wind turbines using innovative
control techniques and fully coupled opensource
engineering tool
Starting date
01.01.2021
Duration
36 months
Programme
H2020EU.3.3.2.  Lowcost, lowcarbon energy supply
Call identifier
H2020LCSC32020RESRIA
Grant Agreement No
101007142
Deliverable Information
Deliverable number
2.4
Work package number
2
Deliverable title
Full report on the estimated reduction of uncertainty
in comparison to the stateoftheart codes OpenFAST
and DeepLines Wind™
Lead beneficiary
UNIFI  SEAPOWER
Main Author
F. Papi
Due date
31/12/2022
Actual submission date
30/12/2022
Type of deliverable
Report
Dissemination level
Public
FLOATECH – EUH2020 Grant Agreement N°101007142
FLOATECH D2.4: estimated reduction of uncertainty in comparison to the stateoftheart codes
3/136
Version management
Document history and validation
Version
Name
Date
Comment
V 0.1
G. Troise, G. M.
Mirra (SEPOWER)
31/11/2022 First partial draft
V 0.2
F. Papi (UNIFI)
16/12/2022
First complete draft
V 0.3
A. Bianchini (UNIFI),
D. Marten (TUB), M.L
Ducasse, J. Honnet
(SAIPEM)
23/12/2022 Internally Revised draft
V 0.4
S. Auburn (ECN)
28/12/2022
Revised Draft
V 1.0
F. Papi, A. Bianchini
(UNIFI)
29/12/2022 Final document
All information in this document only reflects the author's view. The European Commission is not responsible for any use that
may be made of the information it contains.
FLOATECH – EUH2020 Grant Agreement N°101007142
FLOATECH D2.4: estimated reduction of uncertainty in comparison to the stateoftheart codes
4/136
Background: about the FLOATECH project
The FLOATECH project is a Research and Innovation Action funded by the European Union’s H2020
programme aiming to increase the technical maturity and the cost competitiveness of floating offshore
wind (FOW) energy. This is particularly important because, due to the limitations of available installation
sites onshore, offshore wind is becoming crucial to ensure the further growth of the wind energy sector.
The project is implemented by a European consortium of 5 public research institutions with relevant skills
in the field of offshore floating wind energy and 3 industrial partners, two of which have been involved
in the most recent developments of floating wind systems.
The approach of FLOATECH can be broken down into three actions:
The development, implementation and validation of a userfriendly and efficient design engineering
tool (named QBladeOcean) performing simulations of floating offshore wind turbines with an
unprecedented combination of aerodynamic and hydrodynamic fidelity. The advanced modelling
theories will lead to a reduction of the uncertainties in the design process and an increase of turbine
efficiency.
The development of two innovative control techniques (i.e., Active Wavebased feedforward Control
and the Active Wake Mixing) for Floating Wind Turbines and floaters, combining wave prediction
and anticipation of induced platform motions. This is expected to improve the performance of each
machine and to minimize wake effects in floating wind farms, leading to a net increase in the annual
energy production of the farm.
The economic analysis of these concepts to demonstrate qualitatively and quantitatively the impact
of the developed technologies on the Levelized Cost of Energy (LCOE) of FOW technology.
In addition to the technological and economic impacts, the project is expected to have several impacts
at societal, environmental and political levels, such as: public acceptance, due to no noise and visibility
issues of FOWT; very low impact on biodiversity and wildlife habitat because no piles are needed be to
installed into the seabed; the use of less material and space thanks to an environmentally friendly design;
the promotion of the installation of FOW in transitional water depths (3050 m), as the costs for FOW at
those locations will become more competitive compared to the fixed bottom foundations.
FLOATECH – EUH2020 Grant Agreement N°101007142
FLOATECH D2.4: estimated reduction of uncertainty in comparison to the stateoftheart codes
5/136
Table of contents
1 EXECUTIVE SUMMARY ____________________________________________________________________________________________ 8
2 INTRODUCTION ___________________________________________________________________________________________________ 9
3 CHARACTERISTICS AND ASSUMPTIONS OF THE COMPARED SOFTWARE CODES _______________________________ 10
3.1 QBLADEOCEAN ___________________________________________________________________ 11
3.2 OPENFAST ________________________________________________________________________ 11
3.3 DEEPLINES _______________________________________________________________________ 12
4 MODEL DEFINITIONS _____________________________________________________________________________________________ 12
4.1 SOFTWIND _______________________________________________________________________ 13
4.1.1 Numerical model setup ___________________________________________________________ 14
4.2 NREL 5MW OC4 ___________________________________________________________________ 16
4.2.1 Numerical model setup ___________________________________________________________ 16
4.3 DTU 10MW HEXAFLOAT _____________________________________________________________ 17
5 CHARACTERISTICS OF THE SIMULATION DATASET ______________________________________________________________ 18
5.1 SIMULATED DESIGN LOAD CASES (DLC) _________________________________________________ 19
5.1.1 Turbulent wind __________________________________________________________________ 19
5.1.2 Wind shear _____________________________________________________________________ 21
5.2 ENVIRONMENTAL CONDITIONS _______________________________________________________ 22
5.2.1 Environmental condition database __________________________________________________ 22
5.3 CALCULATION OF ENVIRONMENTAL INPUTS _____________________________________________ 22
5.4 REFERENCE SYSTEMS AND OUTPUTS ___________________________________________________ 23
5.5 DATA AVAILABILITY ________________________________________________________________ 24
6 COMPUTATIONAL TIME __________________________________________________________________________________________ 25
7 STATISTICAL ANALYSIS OF NORMAL OPERATING CONDITIONS _________________________________________________ 26
7.1 SOFTWIND MODEL STATISTIC COMPARISON ____________________________________________ 29
7.2 OC4 MODEL STATISTIC COMPARISON __________________________________________________ 34
7.3 HEXAFLOAT MODEL STATISTIC COMPARISON ____________________________________________ 40
8 EXTREME LOADS _________________________________________________________________________________________________ 45
8.1 METHODS AND TOOLS ______________________________________________________________ 45
8.2 SOFTWIND MODEL EXTREME VALUE ANALYSIS __________________________________________ 46
8.3 OC4 MODEL EXTREME VALUE ANALYSIS ________________________________________________ 59
8.4 HEXAFLOAT MODEL EXTREME VALUE ANALYSIS __________________________________________ 70
9 FATIGUE ANALYSIS ______________________________________________________________________________________________ 81
9.1 METHODS AND TOOLS ______________________________________________________________ 81
FLOATECH – EUH2020 Grant Agreement N°101007142
FLOATECH D2.4: estimated reduction of uncertainty in comparison to the stateoftheart codes
6/136
9.2 COMPARISON OF DAMAGE EQUIVALENT LOADS FOR THE SOFTWIND MODEL __________________ 83
9.2.1 Lifetime DELs comparison (SOFTWIND) _______________________________________________ 83
9.2.2 DELs grouped by wind speed (SOFTWIND) ____________________________________________ 87
9.3 COMPARISON OF DAMAGE EQUIVALENT LOADS FOR THE OC4 MODEL ________________________ 92
9.3.1 Lifetime DELs comparison (OC4) ____________________________________________________ 92
9.3.2 DELs grouped by wind speed (OC4) __________________________________________________ 95
9.4 COMPARISON OF DAMAGE EQUIVALENT LOADS FOR THE HEXAFLOAT MODEL ________________ 102
9.4.1 Lifetime DELs comparison (HEXAFLOAT) _____________________________________________ 102
9.4.2 DELs grouped by wind speed (HEXAFLOAT) ___________________________________________ 105
10 OBSERVATIONS ________________________________________________________________________________________________ 111
10.1 SOFTWIND MODEL ________________________________________________________________ 111
10.1.1 Statistic data analysis ___________________________________________________________ 111
10.1.2 Maximum values _______________________________________________________________ 114
10.1.3 Damage Equivalent Loads ________________________________________________________ 115
10.2 OC4 MODEL _____________________________________________________________________ 116
10.2.1 Statistic data analysis ___________________________________________________________ 117
10.2.2 Maximum values _______________________________________________________________ 118
10.2.3 1 Hz Damage Equivalent Loads ____________________________________________________ 119
10.3 HEXAFLOAT _____________________________________________________________________ 119
10.3.1 Statistic data analysis ___________________________________________________________ 119
10.3.2 Maximum values _______________________________________________________________ 120
10.3.3 Damage Equivalent Loads ________________________________________________________ 121
11 CONCLUSIONS __________________________________________________________________________________________________ 121
12 REFERENCES ____________________________________________________________________________________________________ 125
13 APPENDIX A – ADDITIONAL 1HZ DELS __________________________________________________________________________ 129
13.1 SOFTWIND ______________________________________________________________________ 129
13.2 OC4 ____________________________________________________________________________ 133
13.3 HEXAFLOAT _____________________________________________________________________ 135
FLOATECH – EUH2020 Grant Agreement N°101007142
FLOATECH D2.4: estimated reduction of uncertainty in comparison to the stateoftheart codes
7/136
List of acronyms and abbreviations
Acronym / Abbreviation Meaning / Full text
FOW Floating Offshore Wind
FOWT Floating Offshore Wind Turbine
LCOE Levelized Cost of Energy
OF OpenFAST
DL DeepLines
QB QBladeOcean
DOF Degree of Freedom
TT Fx Tower Top foreaft force
TT Fy Tower Top sideside force
TT Mx Tower Top sideside moment
TT My Tower Top foreaft moment
TB Fx Tower Base foreaft force
TB Fy Tower Base sideside force
TB Mx Tower Base sideside moment
TB My Tower Base foreaft moment
BR Mxb Blade root edgewise bending moment
BR Myb Blade root flapwise bending moment
BR Mxc Blade root inplane bending moment
BR Myc Blade root outofplane bending moment
FEA Finite Element Analysis
DEL Damage Equivalent Load
DLC Design Load Case
FLOATECH – EUH2020 Grant Agreement N°101007142
FLOATECH D2.4: estimated reduction of uncertainty in comparison to the stateoftheart codes
8/136
1 EXECUTIVE SUMMARY
This document is a deliverable of the FLOATECH project, funded under the European Union’s
Horizon 2020 research and innovation programme under grant agreement No 101007142.
In work package 2 (WP2) a detailed validation and verification of the capabilities of QBlade
Ocean is carried out. A detailed description of the models used in the validation is provided in
Deliverable 2.1 [1] and results of the validation that was carried out are presented in Deliverable
2.2 [2]. Some modifications to the models used in Task 2.1 have been adopted in Task 2.2, as
described in deliverable D2.3, in order to have more realistic results when full scale FOWT
models are considered. This document is based on the work described in the previous
deliverables and aims to report a comprehensive codetocode comparison, thus supporting a
quantification study of the uncertainty in the predicted FOWT behaviour. Starting from the data
gathered in the database collected through deliverable D2.3, containing Floating Offshore
Wind Turbine (FOWT) calculations in various design load situations, computed with three
different codes, a comprehensive comparison of simulation results is then reported. Three
FOWT concepts are simulated in a variety of design load cases (DLCs), as explained in the D2.3
deliverable document.
The current document reports the results of the analysis procedure applied to the gathered
simulation data, according to the following scheme:
Presentation of significant aggregate data obtained through postprocessing of the
dataset described in [3]. In particular:
o Global statistics (mean value, maximum and minimum value) of relevant motion
and load related data;
o Extreme loads calculated according to standardized procedure (IEC extreme loads
as described in [4]);
o Fatigue loads, represented by Damage Equivalent Loads (DELs) for acting on the
main FOWT components;
Comparison of the previously reported results between the three codes for each of the
considered FOWT models;
Discussion of the outcomes of the comparison, attempting to find, wherever possible,
reasonable explanations for the differences between codes and quantification the
uncertainty related to the use of stateoftheart numerical codes for the prediction of
FOWTs behaviour.
FLOATECH – EUH2020 Grant Agreement N°101007142
FLOATECH D2.4: estimated reduction of uncertainty in comparison to the stateoftheart codes
9/136
2 INTRODUCTION
QBladeocean is a FOWT simulation tool that was partly developed withing the FLOATECH
H2020 project (www.floatechproject.org). The advanced wind turbine simulation tool Qblade
was expanded in work package 1 to be able to model FOWTs [5,6] and validated in work
package 2 [1,2]. The tool includes advanced physical models, such as a LiftingLine Free Vortex
Wake (LLFVW) module for the solution of unsteady aerodynamic loads and a nonlinear
multibody FEA module for the solution of structural dynamics. Hydrodynamic modelling is also
state of the art: potentialflow or striptheory based models can be modelled and advanced
features such as wave stretching models and explicit buoyancy computations have been
implemented. The scope of this work is twofold: firstly to show the level of uncertainty in load
prediction that can be expected in a complex computation such as the simulation of a FOWT in
complex environmental conditions. This is achieved by discussing the differences between the
numerical models. Aligning the outputs of the three numerical codes involved in this comparison
required significant amount of work and discrepancies are found to be dependent not only on
model theory but also on model setup and on output conventions and export. The second
objective is to investigate the differences that can be expected in the prediction of FOWT
dynamics and loads when using a higherfidelity tool such as QbladeOcean with respect to
other numerical models, and how they may be tied to the underlying physics.
In the next sections we will report the results of the postprocessing analysis of a simulation
dataset generated in this work package for three Floating Offshore Wind Turbine (FOWT)
models using three different codes with different modelling assumptions.
A comparative survey of the models used in this study is presented in sec. 3, together with a
brief description of the characteristics of the code. The FOWT models, already introduced in
previous work package deliverables [1–3], are briefly described in section 4.
Section 5 describes the simulation settings, as defined in terms of the assumed meteorological
and operating conditions, indicating the considered Design Load Cases (DLC). The selected
simulation set has been defined according to a subset of the IEC standard design requirements,
considering the environmental conditions of a European offshore site (west of Barra Isle).
In section 7, 8 and 9, a graphic representation of the comparison is presented, considering thee
different aspects: a comparison of global statistic data for the DLC 1.2 simulation subsets, a
comparison of the maximum values of the some interesting response parameters, a comparison
of the 1 Hz Damage Equivalent loads (DELs) representing the fatigue actions on main turbine
components. The main discussion points are summarized in section 10 and 11.
FLOATECH – EUH2020 Grant Agreement N°101007142
FLOATECH D2.4: estimated reduction of uncertainty in comparison to the stateoftheart codes
10/136
3 CHARACTERISTICS AND ASSUMPTIONS OF THE COMPARED SOFTWARE CODES
A short overview of the codes used in the generation of the reference dataset, gathered in
deliverable D2.3 [2], will be presented in the next subsections. Table 1 provides a summary of
the modelling capabilities of the software packages.
Table 1: Overview of software tools (adapted from [2])
QBladeOcean OpenFAST Deeplines WindTM
Open Source? Yes Yes No
Graphical user interface? Yes No Yes
Distribution Free online Free online Commercial
Structural model: Blades
(MB Multibody)
MB, Corotational
formulation, EulerBernoulli
beams.
MB, Modal reduction, Geometrically
exact beam theory
MB, Corotational formulation,
EulerBernoulli beams.
Structural model: Tower Corotational formulation,
EulerBernoulli beams. Modal reduction. Corotational formulation, Euler
Bernoulli beams.
Structural model:
Substructure
Corotational formulation,
EulerBernoulli beams.
(floating, fixed bottom)
CraigBampton method (fixed bottom).
Corotational formulation, Euler
Bernoulli beams. (floating, fixed
bottom)
Aerodynamics: Onblade
Blade elements. Multi
polar. Dynamic stall: Øye,
BL, Gormont, ATE Flap.
Tower model.
Blade elements. Dynamic stall: Øye, B
L, BoeingVertol. Tower model.
Blade elements. Dynamic stall:
Øye, Risø, BoeingVertol. Tower
model.
Aerodynamics: Rotor/wake Unsteady BEM, Freevortex
wake, Vortex particle
BEM, Dynamic inflow. Freevortex
wake (OLAF) Steady BEM, Dynamic Inflow.
Hydrodynamics
Full Morison approach, Lin.
Potential flow. 2nd Order
QTF
Full Morison approach, Lin. Potential
flow. 2nd Order QTF
Full Morison approach, Lin.
Potential flow. 2nd Order QTF
Hydrostatics Linear buoyancy, explicit
buoyancy Linear buoyancy, explicit buoyancy Explicit buoyancy
Environmental: Wind field Turbsim, Hubheight files. Turbsim, Hubheight files. Turbsim, Hubheight files.
Environmental: Wave field
Regular, Irregular
(numerous spectra),
Multidirectional, Imported
Regular, Irregular (numerous spectra),
Multidirectional, Import as time series.
Regular, Irregular (numerous
spectra), Multidirectional,
Import as time series.
Controller formats Bladed, TUB, DTU Bladed Bladed
FLOATECH – EUH2020 Grant Agreement N°101007142
FLOATECH D2.4: estimated reduction of uncertainty in comparison to the stateoftheart codes
11/136
3.1 QBLADEOCEAN
QBladeOcean (hereafter QB) is a multiphysics code, covering the complete range of aspects
required for the aeroservohydroelastic simulation of horizontal and vertical axis wind turbines.
QB uses a Lifting Line Free Vortex Wake method (LLFVW) for aerodynamic calculations. In this
method, the rotor wake is explicitly modelled through Lagrangian vortex elements. This results
in a more accurate and detailed spatial and temporal representation of the rotor induction,
when compared to Blade Element Momentum (BEM) approaches, and fully resolves the velocity
distribution behind the rotor.
QB models the structural dynamics using a multibody formulation. The components of the
multibody model can be either point masses or rigid and/or flexible EulerBernoulli beam
elements in a corotational formulation. The structural model uses a corotational beam element
formulation implemented within the multiphysics solver Project CHRONO [7]. This allows for
the treatment of highly nonlinear deflections of beam elements.
A full report of the additional development of QB done within the FLOATECH project can be
found in [5]. The QB user training manual can be found in [6].
3.2 OPENFAST
OpenFAST (hereafter OF) is a stateoftheart multiphysics code, developed by the National
Renewable Energy Laboratory (NREL) to model horizontal axis wind turbines. Like QB, the code
is opensource and is able to model onshore and offshore wind turbines, both bottomfixed and
floating. All the calculations presented in this report are performed with OpenFAST v3.0.0.
The wind turbine’s wake induction is modelled using bladeelementmomentum theory (BEM),
with corrections for high thrust, tip and root losses and misaligned inflow. Corrections for
dynamic stall are used, in the form of the BeddoesLeishman dynamic stall model. Dynamic wake
effects are modelled using the approach proposed by Øye [8]. This model is able to account for
the fact that changes in blade induction do not apply instantly to the entire wake by applying a
filtering procedure to the axial and tangential induced velocities at each blade section.
The floating support substructures are modelled as 6DOF rigid bodies. Their interaction with
the sea is computed using HydroDyn [9], OF’s hydrodynamics module. Similarly to QB,
hydrodynamics can be modelled using potential flow theory, a Morison equation approach or a
mix of the two. For test cases presented in this report, potential flow theory is used, with the
addition of the Morison equation to model quadratic drag.
FLOATECH – EUH2020 Grant Agreement N°101007142
FLOATECH D2.4: estimated reduction of uncertainty in comparison to the stateoftheart codes
12/136
The structural dynamics are solved using the modalbased structural module ElastoDyn [10]. The
structural deformations are calculated as linear superposition of the predetermined mode
shapes, allowing for a very fast solution with limited degrees of freedom to solve for [11].
Moorings are modelled using MoorDyn [12], a dynamic lumpedmass mooring line model. The
model accounts for internal axial stiffness and damping forces, weight and buoyancy forces,
hydrodynamic forces from Morison's equation, and vertical springdamper forces from contact
with the seabed.
3.3 DEEPLINES
DeepLinesTM (hereafter DL) is a commercial integrated software solution to perform inplace and
installation analyses of various offshore structures, based on the finite element method. It allows
for design of flexible risers, power cables and umbilical, pipelines, mooring systems, towed
systems and simulations of marine operations. DL is part of the marine software solutions suite
developed by Principia and IFP Energies Nouvelles. This numerical tool is based on the program
Flexan, initially developed for flexible risers and used since 1980.
To address the needs for an overall design tool raised by the development of offshore wind
turbines platforms, a new module DeepLines WIND (hereafter DL) was created in 2011. DL is
designed to perform fullycoupled dynamic finite element analysis. It accounts for combined
effects of aerodynamic loads on the blades, active pitch control, hydrodynamic loads on the
floating platform and dynamic mooring loads. Several BEM models are implemented in an
external .dll library to calculate the aerodynamic loads. DL is able to simulate both horizontal
and vertical axis wind turbines. The aerodynamic model in DL uses a BEM approach coupled
with a dynamic inflow model. In this case no unsteady blade aerodynamics model has been
applied.
4 MODEL DEFINITIONS
A brief summary of the model configuration data, already presented in a previous deliverable
[1–3], will be reported here for the sake of completeness. The main characteristics of each of
the three tested models will be presented in the next subsections. Not all three codes, with the
exception of QB, have been tested with all three test cases. A summary of the codes and their
application for the different test cases is shown in Table 2.
FLOATECH – EUH2020 Grant Agreement N°101007142
FLOATECH D2.4: estimated reduction of uncertainty in comparison to the stateoftheart codes
13/136
Table 2: Summary of compared codes for each test case
Code
DeepLines Wind (DL) OpenFAST (OF) QBladeOcean (QB)
Test case
NREL 5MW OC4
SOFTWIND 10MW
HEXAFLOAT
10MW
4.1 SOFTWIND
This section describes the DTU 10MW RWT mounted on the SOFTWIND spar platform,
henceforth called the 10MW SOFTWIND test case. The SOFTWIND platform is a sparbuoy
platform originally developed and tested by Ecòle Centrale Nantes (ECN) [13] and was used in
WP2.1 of the current project [1,2]. The wind turbine rotor is described in [14], while the semi
submersible and mooring line layout is described in [1,13]. The model used in the current study
is derived from the definition that was used in WP2.1 [1,2], however some adaptations have
been done so that it can be used in a full set of design load calculations. These modifications
are described in [6] but will be briefly summarized here. The model that was used in WP2.1
featured platform mass and inertia distribution as close as possible to the experimental article
described in [13]. In the experimental model, batteries and control electronics are housed inside
the spar, altering the mass and inertia distribution. The model used in WP2.2 features a lower
center of gravity to increase stability in severe seas, making the test case more representative
of an actual floating platform. The tower properties have also been changed. In fact, stiffening
of the tower was considered necessary to avoid a 3P excitation of the structure, as the natural
frequencies of the tower used in WP2.1 are located in the 3P range for some wind speeds.
Therefore, the tower originally developed by OlavOlsen1 for the OOStar platform in the
LifeS50+ Horizon 2020 project [15] is used in the current model. A schematic representation of
the turbine mounted on the spar floater is shown in Figure 1.
1 The OOStar Wind Floater has been developed by Dr.Techn. Olav Olsen (OO) since 2010 and is the property of
Floating Wind Solutions AS. OO has approved that the public model from Lifes50+ can be used for the research
activities within FLOATECH. The model shall not be used for other purposes unless it is explicitly approved by OO.
FLOATECH – EUH2020 Grant Agreement N°101007142
FLOATECH D2.4: estimated reduction of uncertainty in comparison to the stateoftheart codes
14/136
Figure 1: SOFTWIND turbine model in QB.
4.1.1 Numerical model setup
The numerical setup choices for the SOFTWIND test case in the three numerical codes (DL, OF,
QB) used in this work, mostly carry over from those presented in Deliverable 2.2 of the
FLOATECH project [2]. The increase in tower mass and change in platform characteristics that
was described in section 4.1 however required a retuning of the models.
Aerodynamics the setup described in [2] is left unchanged for QB and DL. For OF however, the
aerodynamic definition was updated and AeroDyn v15 [16] was used in place of the legacy
AeroDyn v14. This allowed for improved accuracy in modelling the prebend of the blades.
Moreover, a dynamic induction model can now be used in OF, allowing for this numerical model
to empirically capture dynamic induction effects. The model developed by Øye is used [8], as
described in [17]. The model introduces a timelag in the calculated induction factors, empirically
accounting for the “wake memory” effect. It depends on two time constants and . In the
OF model these constants are left equal to their default values, computed based on rotor
averaged axial induction [17].
Turbine definition tower structural properties and floater mass and inertia properties were
changed in accordance with section 4.1. For the DL model however, this resulted in large
discrepancies in the response of the structure from a dynamic standpoint, with large differences
in freedecay tests being noticed. To better align the DL model to OF and QB, the following
was done:
FLOATECH – EUH2020 Grant Agreement N°101007142
FLOATECH D2.4: estimated reduction of uncertainty in comparison to the stateoftheart codes
15/136
1. A transverse drag coefficient (Ca) of 0.5 is applied to the spar structure through strip
theory. As described in [2], a hybrid potential flow – strip theory approach was used in all
three codes. In particular, Morrison’s equation is used to account for quadratic drag in
the models, while radiation damping, added mass and linear excitation force are included
through an externally computed potential flow solution. In the DL model however, to
align the surge and sway freedecay natural frequencies, an added mass coefficient of
0.5 was added in the striptheory solution, effectively counting added mass twice respect
to OF and QB;
2. Roll and Pitch floater inertias are decreased approximately 40% to align the natural
frequencies in roll and pitch;
3. As discussed in [2], mooring line lengths were tuned in each numerical models as a
compromise between natural frequency in surge and mean mooring line tension. It is
worth pointing out here that the compromise found for DL led to slightly longer lines and
thus lower tensions than the tunings that were done for OF and QB.
Table 3: Masses, inertias and striptheory transverse and axial coefficients for SOFTWIND model in the three numerical
models
variable OF QB DL units
RNA mass 676804 676950 676755 kg
CoG RNA 116.67  118.42 m
Tower mass 1.2598E+6 1.2568 E+6 1.1969 E+6 kg
CoG Tower 41.03 38.31 49.814 m
Platform mass 1.9919E+7 1.9919E+7 1.9919E+7 kg
CoG Ptfm 74.92 74.92 74.89 m
Ixx Ptfm 9.75E+9 9.75E+9 5.87E+9 kg*m2
Iyy Ptfm 9.75E+9 9.75E+9 5.87E+9 kg*m2
Izz Ptfm 9.3E+8 9.3E+8 9.31E+8 kg*m2
FOWT mass 2.1855E+7 2.2792 E+7 2.1793 E+7 kg
CoG FOWT 62.30 61.91 67.51 m
Ca axial 0 0 0 
Cd axial 5 5 8 
Ca trasv 0 0 0.5 
Cd trasv 0.3 0.3 0.3 
A summary of the masses, inertias and striptheory coefficients that were applied to the
SOFTWIND floater are summarized in Table 3.
FLOATECH – EUH2020 Grant Agreement N°101007142
FLOATECH D2.4: estimated reduction of uncertainty in comparison to the stateoftheart codes
16/136
4.2 NREL 5MW OC4
This section describes the NREL 5MW RWT mounted on the DeepCWind semisubmersible
platform, henceforth called the 5MW OC4 test case. This wind turbine model was extensively
used in the OC4 codetocode comparison [18] and hence the model definition is publicly
available [19]. The wind turbine rotor and tower are described in [20], while the semi
submersible and mooring line layout is described in [19].
Figure 2: Visualization of the 5MW OC5 FOWT model in QB in regular waves.
With respect to the OC5 model that was used in WP 2.1 [2] the floater geometry is unchanged,
while the tower, nacelle and blades are those of the NREL 5MW RWT [20], as opposed to the
model scale tower and blades used in OC5 [21]. The model defined within this study follows the
definition in [19].
4.2.1 Numerical model setup
The OC4 model is setup with the same approach that was used for OC5: aerodynamics are
modelled with BEM in OF and LLFVW in QB, structural dynamics are accounted for in OF with
a modal approach and in QB with a multibody FEM model, Linear Potential Morrison Drag
(LPDM) approach is used for hydrodynamics in both codes, as well as a dynamic mooring cable
model.
In OF, the public OC4 definition that is found on GitHub (https://github.com/OpenFAST/r
test/tree/main/gluecodes/openfast/5MW_OC4Semi_WSt_WavesWN) was used as model
definition. Blade Element Momentum theory (BEM) augmented with a dynamic induction model
proposed by Øye is used [8]. Bladelevel unsteadiness is modelled using a BeddoesLeishman
type dynamic stall model. This aerodynamic model will be referred to as DBEM (DynamicBEM).
In agreement with standard industrial practices, corrections for root and tip losses, nonuniform
FLOATECH – EUH2020 Grant Agreement N°101007142
FLOATECH D2.4: estimated reduction of uncertainty in comparison to the stateoftheart codes
17/136
nonaligned inflow and high induction are also included in the aerodynamic model. Details
regarding the specific BEM implementation can be found in [22]. Flexible tower and blades are
modelled using ElastoDyn using a modal formulation, where the blades and tower deformations
are computed by superimposing precomputed modal deformations.
Hydrodynamics are modelled using a combination of potential flow and strip theory that is used
to compute quadratic drag. The drag coefficients computed with the approach described in [19]
are used in both OF and QB. After comparing freedecay tests, lower linear damping was noted
for OF respect to QB. However, given the excellent agreement that was noted in unsteady wave
tests, it was preferred to not tune the models to improve linear damping agreement and instead
maintain the definitions identical in the interest of the codetocode comparison. For the
potential flow solution, WAMITcomputed hydrodynamic coefficients provided during the OC4
projects have been used for fistorder loads. Secondorder potentialflow load quadratic
transfer functions (QTFs) are computed using NEMOH, which was recently extended to be able
to compute QTFs within work package 1 of the FLOATECH project
(https://gitlab.com/lheea/Nemoh). This has allowed us to model second order loads for the
range of wave headings (150 to 150°) that are used in this project (section 5). With respect to
hydrodynamics, QB features two main improvements over the current OF version; firstly, the
hydrostatic buoyancy force is discretely calculated based on the displaced volume by the
floating structure. We refer to this approach as explicit buoyancy. To avoid a doubleaccounting
of the buoyancy forces through the potential flow excitation force and this explicit buoyancy
approach, instead of evaluating buoyancy for the instantaneous sea level we calculate the
buoyancy with respect to the mean sea level (MSL). In contrast buoyancy is modelled in OF as
a constant force and a linear stiffness to account for deviations from the equilibrium. Secondly,
Wheeler kinematic stretching method is applied in QB to approximate the water kinematics
above MSL. Further details can be found in [2].
4.3 DTU 10MW HEXAFLOAT
The DTU 10MW Hexafloat model description can be found in Deliverables 2.1 and 2.2 of the
FLOATECH project [1,2]. The substructure of the Hexafloat 10 MW FOWT is developed by
Saipem® and is a twopart floater depicted in Figure 3. It consists of a floating hexagonal
structure and a counterweight linked together with six tendons. No changes have been made
to the Hexafloat model with respect to the definition shown in D2.1 [1] and D2.2 [2].
FLOATECH – EUH2020 Grant Agreement N°101007142
FLOATECH D2.4: estimated reduction of uncertainty in comparison to the stateoftheart codes
18/136
Figure 3: The Hexafloat 10 MW FOWT aerohydroelastic model in QB.
5 CHARACTERISTICS OF THE SIMULATION DATASET
The objective of the current exercise is to perform a codetocode comparison in realistic
environmental conditions. To achieve this objective, three preliminary actions are necessary,
and are illustrated in this section:
i. Definition of a series of combination of operating conditions and environmental
conditions (Design Load Cases – DLCs) that are relevant for extreme and fatigue loading
on FOWTs
ii. Definition of longterm environmental conditions to use in the DLCs defined in i)
iii. Definition of a procedure to match wind and wave inputs amongst the compared codes
to ensure that Time series can be compared, in addition to more commonly presented
statistical comparisons
All three actions are illustrated in this section. In particular, the first action is explained in section
5.1, while section 5.2 explains the procedure followed to compute a longterm environmental
representation of an offshore site of choice. Finally, the procedure to ensure environmental
inputs match in QB, OF and DL is explained in section 5.3. The reference systems that are used
FLOATECH – EUH2020 Grant Agreement N°101007142
FLOATECH D2.4: estimated reduction of uncertainty in comparison to the stateoftheart codes
19/136
in this analysis are indicated in section 5.4. The analysed datasets are public, and their public
location is specified in section 5.5
5.1 SIMULATED DESIGN LOAD CASES (DLC)
The DLCs that make up the current database have been selected in order to provide a good
estimation of fatigue and extreme loads whilst limiting as much as possible the total number of
simulations. In fact, considering the full design space is not needed since full turbine certification
is out of the scope of this project, while it was important to include those functioning conditions
that could make differences between the modeling approaches more evident. The list of
selected DLCs does not include cases where fault events are simulated. This simplifies future
comparisons using this dataset.
The final list contains the same DLCs that were simulated in the design of the IEA 15MW RWT
[23,24], with the addition of DLC 1.2 for fatigue loads. Except for fault cases, which are not
included in the current list, they match the load cases simulated by Jonkman in [25]. In the latter
reference, the authors state that these DLCs are selected to cover essential designdriving
situations, which is the same objective of the current document. An overview of the DLCs that
were simulated is shown in Table 4.
5.1.1 Turbulent wind
The number of used turbulent windwave seeds was partially limited. In fact, when conducting
an extensive comparison such as the one in this study memory requirements to store the full
field wind fields that are needed for the coupled simulations can become relevant. Therefore,
the following approach is adopted:
In DLC 1.2, simulations with 0° and +10° yaw misalignment share the same wind field.
This DLC is used for fatigue load calculations and therefore is less sensitive to seed
dependent transient events.
In DLC 6.1 and 6.3, two simulations per wind/wave misalignment and yaw error will be
performed. Therefore, a total of two wind fields will be generated. Simulations are
differentiated by yaw angle.
The total number of wind fields required for the full DLC calculation is shown in table 4. As
explained in section 5.3, significant effort was put in to ensure that the environmental inputs
match as much as possible in the compared codes, thus lowering the number of required seeds
to have a good statistical comparison.
FLOATECH – EUH2020 Grant Agreement N°101007142
FLOATECH D2.4: estimated reduction of uncertainty in comparison to the stateoftheart codes
20/136
Table 4: DLCs in codetocode comparison
DLC
Wind
Speed
[m/s]
Sea
Condition
Significant
Wave
Height [m]
Peak
Period [s] Wave Heading [°]
Sea
Currents
Condition
Total N°
sims
1.2 525 NSS 19 814 150°150° None 504
1.3
5
NSS
1.589253 9.788171 0
None
9
7 1.79685 9.992163 0 9
9 2.066758 10.24619 0 9
11 2.396227 10.54261 0 9
13 2.78555 10.87777 0 9
15 3.237992 11.25141 0 9
17 3.757419 11.66434 0 9
19 4.346084 12.1166 0 9
21 5.004369 12.60752 0 9
23 5.731637 13.1363 0 9
25 6.52699 13.70255 0 9
1.4
9.5
NSS
2.066758 10.24619 0
None
2
11.5 2.396227 10.54261 0 2
13.5 2.78555 10.87777 0 2
1.6
5
SSS
8.472305 15.05372 0
None
9
7 9.182249 15.53933 0 9
9 9.698228 15.89071 0 9
11 9.970977 16.07605 0 9
13 10.05924 16.13597 0 9
15 10.12899 16.1833 0 9
17 10.35074 16.33371 0 9
19 10.80451 16.64112 0 9
21 11.48977 17.10469 0 9
23 12.35813 17.69156 0 9
25 13.34401 18.35793 0 9
6.1 36.92 ESS 16.42 18.68 30°/0°/30° None 18
6.3 31.9 ESS 11.93 15.95 30°/0°/30° None 18
6.2 36.92 ESS 16.42 18.68 45°/90°/135°/180° None 12
A schematic representation of the wind field box that is used in the simulations is shown in
Figure 4. The same wind fields are shared by the three test cases, whether they use the NREL
5MW or the DTU 10MW rotor. Mean wind speed is imposed at 100 m above mean sea water
level (MSWL), in consistency with the ERA5 hindcast database. Because of wind shear, wind
speed will be slightly lower than the imposed value at 100m for the OC4 test case and slightly
higher for SOFTWIND and HEXAFLOAT test cases, as if these test machines were installed at
FLOATECH – EUH2020 Grant Agreement N°101007142
FLOATECH D2.4: estimated reduction of uncertainty in comparison to the stateoftheart codes
21/136
the selected location (section 5.2). Turbulent wind fields are generated using TurbSim v2.0.0
[26,27], with IECKaimal spectral model and spectral coherence model.
Table 5: number of wind seeds required for DLC calculation.
DLC seeds per
ws n° ws tot sims
1.3
9
11
99
1.2
252
1.6
9
11
99
6.1
2
1
2
6.2
2
1
2
6.3
2
1
2
TOTAL 454
Figure 4: schematic view of wind field box (black), DTU 10MW rotor (red) and NREL 5MW rotor (black).
5.1.2 Wind shear
Wind shear is modelled using a power law. In normal conditions, a shear exponent of 0.14 is
used, as specified in IEC 614001 [4]. This exponent applies to all simulations with Normal
Turbulence Model (NTM) and Extreme Turbulence Model (ETM). In extreme conditions, a shear
exponent of 0.12 is used, as specified in the LifeS50+ design basis [26]. The latter exponent
applies to simulations with the Extreme Wind Model (EWM).
FLOATECH – EUH2020 Grant Agreement N°101007142
FLOATECH D2.4: estimated reduction of uncertainty in comparison to the stateoftheart codes
22/136
5.2 ENVIRONMENTAL CONDITIONS
Calculations in the current dataset are performed for an offshore installation site west of the
island of Barra (Scotland).
The West of Barra site is located on the European continental shelf and therefore, water depths
are limited to about 120130 m. According to [27], average depth at the site is 95 m and within
the identified installation area depth varies between 56 and 118 m. Water depth in the point
where metocean conditions were sampled for the FLOATECH project is 123 m according to
[28]. Importantly for WP2, this depth allows for the sampling of middepth wave characteristics,
representative of sites where FOWT wind parks are planned to be installed. As a matter of fact,
if we look at the only two operating commercial FOWT wind farms in Europe at the current
date, the Hywind Scotland [29] and WindFloat ATLANTIC [30] wind farms, reported water depth
is 90 and 100 m [29,30] respectively. For the generation of the current dataset, water depth is
ignored, and the West of Barra site is used only to extract wave and wind characteristics. The
considered water depth is defined based on the nominal installation depth of the three test
cases that were considered: 200 m for the NREL 5MW OC4 and DTU 10MW SOFTWIND models
and 250 m for the DTU 10MW HEXAFLOAT model. Moreover, no currents are considered in
the calculations. A water density of 1025 kg/m3 is assumed.
5.2.1 Environmental condition database
The combination of wind speed, significant wave height and windwave misalignment (this latter
point being a distinctive feature of the analysis, since it is not always accounted for) are defined
on a Design Load Case basis with the procedure described in [31]. Starting from hindcast wind
and wave data for the West of Barra site, a joint probabilistic model of the four environmental
variables (wind speed, significant wave height, peak spectral period, windwave misalignment)
is derived. The model is then used to compute environmental contours that define the extreme
metocean conditions (DLC 1.6, 6.1, 6.2, 6.3). The expected or normal metocean conditions are
instead used in DLCs 1.2 and 1.3.
The complete dataset is openly available at the following link:
https://doi.org/10.5281/zenodo.6972014
5.3 CALCULATION OF ENVIRONMENTAL INPUTS
Once the DLCs are defined (section 5.1) and the longterm representation of the installation site
has been computed (section 5.2), turbulent wind fields and irregular wave fields for each
simulation must be defined. To ensure that Time series can be compared, providing insight on
FLOATECH – EUH2020 Grant Agreement N°101007142
FLOATECH D2.4: estimated reduction of uncertainty in comparison to the stateoftheart codes
23/136
how the different models in the compared codes are able to reproduce physics that lead to, for
instance, extreme component loads, care has been put into ensuring that QB, OF and DL share
the same inputs.
Therefore, identical turbulent wind boxes are generated using the same version of TurbSim and
imported in the three codes. This resulted in identical wind fields for QB and OF. For DL a time
shift is noted. This is partially due to the fact that DL exports wind speed at the RNA, while OF
and QB export wind speed at the (0,0,HubHeight) position. Therefore, when the FOWT moves,
a shift in the output will be present. However, this does not completely explain the timeshift.
The observed time shift also depends on how the turbulent wind boxes are imported into the
three codes and DL seems to differ from OF and QB in this regard. This implies that the time
series in DL are not always punctually comparable to those coming from OF and QB.
Wave Time series were generated in DL and then imported in QB and OF, where they are
decomposed into their frequency components using FFT. Thus, wave fields are identical in the
three codes.
5.4 REFERENCE SYSTEMS AND OUTPUTS
A complete list of the available outputs can be found in Deliverable D2.3, amongst which this
work builds upon. The main sensors that are compared herein include blade root loads in blade
reference system, shaft loads, yawbearing loads, tower base loads and mooring line tensions.
A brief description of the reference systems in which the outputs are expressed is provided
here:
Tower bottom reference system (t): ref. system moves with tower base. If no platform
displacement, xaxis directed downwind and zaxis directed upward. Yaxis normal to x
and z axis.
Towertop reference system (p): same as tower base reference system but deforms with
tower top
Coned reference system (c): zaxis directed along blade pitch axis. Yaxis parallel to blade
chord is blade pitch and twist are zero. Xaxis normal to z and y axes.
Blade reference system (b): same as coned reference system but pitches with blade. In
DL loads in the “b” reference system do not pitch with the blades.
Shaft reference system (s): xaxis directed along rotor shaft. Zaxis directed upwards
(normal to xaxis). Yaxis normal to x and zaxis.
FLOATECH – EUH2020 Grant Agreement N°101007142
FLOATECH D2.4: estimated reduction of uncertainty in comparison to the stateoftheart codes
24/136
5.5 DATA AVAILABILITY
The complete dataset is available at https://doi.org/10.5281/zenodo.7254241
Some simulations in OF and DL experienced crashes and could not be completed within the
timeframe of the work package. The crashes are noted only for a limited number of cases. The
fact that the dataset is incomplete for OF and DL in some cases, however, may skew extreme
and fatigue loads in comparison with the other codes. An overview of the status of the
complete/incomplete simulations is shown in Table 6.
Table 6. Status of datasets for each code
code
DLC
status notes
OC4
OF
1.2
504/504
1.3
99/99
1.4
06/06
1.6
99/99
6.1
18/18
6.2
16/16
6.3
18/18
QB
1.2
504/504
1.3
99/99
1.4
06/06
1.6
99/99
6.1
18/18
6.2
16/16
6.3
18/18
SOFTWIND
OF
1.2
504/504
1.3
99/99
1.4
06/06
1.6
99/99
6.1
18/18
6.2
8/16 Missing Seeds: 10003 (mis 30 yaw 45), 10003 (mis 30 yaw 90),
10003 (mis 30 yaw 45), 10003 (mis 30 yaw 90).
6.3
18/18
QB
1.2
504/504
1.3
99/99
1.4
06/06
1.6
99/99
6.1
18/18
6.2
16/16
6.3
18/18
DL 1.2
498/504
Missing Seeds: 8 (y10), 11 (y0 & y10), 14 (y10), 15 (y10), 14 (y0)
FLOATECH – EUH2020 Grant Agreement N°101007142
FLOATECH D2.4: estimated reduction of uncertainty in comparison to the stateoftheart codes
25/136
1.3
86/99 Missing Seeds: 663, 664, 665, 672, 673, 674, 681, 682, 683, 690, 691, 692, 698.
1.4
06/06
1.6
99/99
6.1
12/18 Missing Seeds: all (yaw 10)
6.2
12/16 Missing Seeds: all (yaw 45)
6.3
12/18 Missing Seeds: all (yaw 20)
HEXAFLOAT
QB
1.2
504/504
1.3
99/99
1.4
06/06
1.6
99/99
6.1
18/18
6.2
16/16
6.3
18/18
DL
1.2
487/504
Missing Seeds: 218 (y0), 221 (y0), 222 (y0 & y10), 227 (y10), 230 (y0 & y10), 232
(y0 & y10), 234 (y10), 237 (y0 & y10), 238 (y0 & y10), 239 (y0), 241 (y0 & y10)
1.3
99/99
1.4
06/06
1.6
99/99
6.1
18/18 Missing Seeds: all (yaw 10).
6.2
16/16 Missing Seeds: all (yaw 45).
6.3
18/18
Missing Seeds: all (yaw 20).
6 COMPUTATIONAL TIME
The main scope of this document is the comparison of the performance of FOWT simulation
codes in terms of the physics that they are able to solve. However, for a more complete picture
of the capabilities of the compared softwares, some information regarding computational
requirements will be given herein. It must be noted that this information is valid strictly for the
test cases and setups that they are relative to and generally valid.
For a DLC 1.2 simulation of the SOFTWIND testcase, that is simulated with all three codes (4000
s of simulated time):
QbladeOcean is run on a desktop workstation equipped with a 12th Gen Intel® Core™
i712700KF processor and an NVIDIA GeForce RTX 3070 Ti graphics card. 20 simulations
are run in parallel and average wall clock time per simulation is approximately 10400 s.
OpenFAST is run on a desktop workstation equipped with a i9 10th Gen Core™
processor. 12 simulations are run in parallel and average wall clock time per simulation is
approximately 1140 s.
FLOATECH – EUH2020 Grant Agreement N°101007142
FLOATECH D2.4: estimated reduction of uncertainty in comparison to the stateoftheart codes
26/136
DeepLines Wind™ is run on a cloud computing service with Intel Xeon Processors
(Skylake), @ 2.4 Ghz. 34 simulations are run on a 36core server, or 16 simulations are run
on a 14core server. Average wall clock time is 24900 s per simulation. Standard hard
drives are used for storage. This is relevant as I/O time is often significant, and
calculations will run faster if SSDs are used.
The wake is solved with a mediumfidelity LLFVW model in QB as opposed to a lowfidelity
DBEM model that is used in OF and DL. Such a wake model requires GPU acceleration to be
run efficiently, hence the GPU in the QB workstation is mentioned. SOFTWIND simulations did
not include second order wave loads. The use of Quadratic Transfer Functions (QTFs) can add
significant computational time to QB simulations. QB runs were performed using QBlade
Enterprise Edition (EE) that allows for multiple simulations to be run in parallel, a feature that is
not available in the opensource Community Edition (CE). In comparison to the other codes, OF
requires much less computational time to run. It must be noted however that this code was run
with lower fidelity modal based structural dynamics model and lower fidelity dynamicBEM
aerodynamics.
7 STATISTICAL ANALYSIS OF NORMAL OPERATING CONDITIONS
The simulation datasets, previously generated, undergo the following procedure to determine
a set of significant statistic parameters:
A subset of the database comprising only the results of DLC 1.2 (related to normal
operation in normal turbulent wind and normal sea state conditions) is considered. The
choice of this subset is motivated by the fact that this subset is the largest among the
considered DLC, thus providing a statistically significant set of results with homogeneous
environmental conditions;
The simulations in the data set are processed using a widespread tool developed by
NREL for DLC postprocessing named MLife [32], mainly intended to fatigue analyses, but
also able to generate statistics. The output of this processing phase is represented by a
set of fundamental statistical data (mean, standard deviation, maximum, minimum) for
each simulated condition defined by average wind speed and sea state (corresponding
to single simulation file);
A postprocessing script is used to report the calculated statistics grouping the data
based on wind speed. The grouped data are reported using a boxplot style in order to
give a more concise representation of the large dataset. Simulation results are binned
according to wind speed with a speed interval of 2 m/s, between 5 m/s and 25 m/s, thus
FLOATECH – EUH2020 Grant Agreement N°101007142
FLOATECH D2.4: estimated reduction of uncertainty in comparison to the stateoftheart codes
27/136
considering the following bins: (4, 6] < (6, 8] < (8, 10] < (10, 12] ... (18, 20] < (20, 22] <
(22, 24] < (24, 26]. For each bin, the box centre represents the mean value of the binned
data, the boxplot height represents twice the mean standard deviation, while the
whiskers represent the maximum and minimum value in the bin.
The data considered for the statistical comparison are collected in several groups, based on
their meaning. Table 7 summarizes the analysed quantities.
FLOATECH – EUH2020 Grant Agreement N°101007142
FLOATECH D2.4: estimated reduction of uncertainty in comparison to the stateoftheart codes
28/136
Table 7. Variable considered for statistical analysis.
Data group Variable considered for statistical analysis
Tower base loads Tower base longitudinal (foreaft) force,
Tower base lateral (sideside) force,
Tower base longitudinal (foreaft) moment,
Tower base lateral (sideside) moment,
Tower top loads Tower top longitudinal (foreaft) force,
Tower top lateral (sideside) force,
Tower top longitudinal (foreaft) moment,
Tower top lateral (sideside) moment,
Control related data Generator Power
Generator Torque
Rotor speed
Blade pitch angle
Platform motion data Platform surge displacement
Platform sway displacement
Platform heave displacement
Platform roll displacement
Platform pitch displacement
Platform yaw displacement
Rotor aerodynamics data Aerodynamic thrust
Aerodynamic torque
Blade data Blade tip outofplane (OoP) deflection
Blade tip inplane (IP) deflection
Blade root edgewise moment in blade pitching ref. system
Blade root flapwise moment in blade pitching ref. system
Mooring data Fairlead tension on the mooring lines
Fairlead tensions for OC4 and SOFTWIND (3 mooring lines)
Tendon tensions for HEXAFLOAT (6 lines)
Hydrodynamic data
Hydrostatic force
Horizontal 1st order hydrodynamic force
Horizontal 2nd order hydrodynamic force
(reported for OC4)
FLOATECH – EUH2020 Grant Agreement N°101007142
FLOATECH D2.4: estimated reduction of uncertainty in comparison to the stateoftheart codes
29/136
7.1 SOFTWIND MODEL STATISTIC COMPARISON
The following figures report a comparison of the main aggregate statistical parameters, as
previously defined, for the SOFTWIND model.
Platform motion statistics are shown in Figure 5. Focusing on foreaft sensors such as platform
surge and pitch, which are the ones most influenced by aerodynamic loads in addition to wave
loads, similar overall trends can be noted. However, both OF and DL show larger means,
standard deviations and maximum values than QB. Similar considerations are valid for the Yaw
Degree of Freedom (DOF). Regarding sideside motion on the other hand, all three codes are
very similar. Platform heave shows the most differences, mostly in mean value. These differences
are attributed to different model tuning and carry over from the previous work conducted within
this work package, and discussed in Deliverable D2.2 [2].
Figure 5: SOFTWIND aggregate statistics in DLC 1.2 grouped by wind speed. Platform motion data. Mean (black
horizontal dash), standard deviation (box edges), minmax range (whiskers).
FLOATECH – EUH2020 Grant Agreement N°101007142
FLOATECH D2.4: estimated reduction of uncertainty in comparison to the stateoftheart codes
30/136
Global performance and control sensors are compared in Figure 6. Overall, the three codes
behave well, but as was the case for foreaft platform motion, larger standard deviation can be
seen for OF and DL than for QB in generator power, generator torque and rotor speed.
Differences in mean blade pitch can be seen in Figure 6. These are likely due to small differences
in steadystate aerodynamic forces for the models, as discussed in D2.2 [2].
Figure 6: SOFTWIND aggregate statistics in DLC 1.2 grouped by wind speed. Control related data. Mean (black
horizontal dash), standard deviation (box edges), minmax range (whiskers).
FLOATECH – EUH2020 Grant Agreement N°101007142
FLOATECH D2.4: estimated reduction of uncertainty in comparison to the stateoftheart codes
31/136
Figure 7: SOFTWIND aggregate statistics in DLC 1.2 grouped by wind speed. Tower base loads. Mean (black
horizontal dash), standard deviation (box edges), minmax range (whiskers).
Tower top and tower base load statistics are shown in Figure 7 and Figure 8, respectively. The
signature of rotor torque can clearly be seen in the tower top sideside bending moment (Mx)
but it is negligible in the tower base sideside moment.
FLOATECH – EUH2020 Grant Agreement N°101007142
FLOATECH D2.4: estimated reduction of uncertainty in comparison to the stateoftheart codes
32/136
Figure 8: SOFTWIND aggregate statistics in DLC 1.2 grouped by wind speed. Tower top loads. Mean (black
horizontal dash), standard deviation (box edges), minmax range (whiskers).
Figure 9: SOFTWIND aggregate statistics in DLC 1.2 grouped by wind speed. Rotor aerodynamics data. Mean
(black horizontal dash), standard deviation (box edges), minmax range (whiskers).
FLOATECH – EUH2020 Grant Agreement N°101007142
FLOATECH D2.4: estimated reduction of uncertainty in comparison to the stateoftheart codes
33/136
Figure 10: SOFTWIND aggregate statistics in DLC 1.2 grouped by wind speed. Blade data.
Mean (black horizontal dash), standard deviation (box edges), minmax range (whiskers).
Statistics of the fairlead tensions, measured at the deltaconnection, is shown in Figure 11.
Fairleads 2 and 3 are directed upwind, 120° apart, while fairlead 1 is directed downwind, parallel
to wind heading. As shown from the whiskers in Figure 11, the mooring lines never go slack
during normal operation.
Figure 11: SOFTWIND aggregate statistics in DLC 1.2 grouped by wind speed. Mooring line tensions. Mean
(black horizontal dash), standard deviation (box edges), minmax range (whiskers).
Moreover, tensions are consistent for OF and QB, but an offset of approximately 10% in mean
value is noted for DL. This difference is attributed to the different tuning of this model that was
done during setup.
FLOATECH – EUH2020 Grant Agreement N°101007142
FLOATECH D2.4: estimated reduction of uncertainty in comparison to the stateoftheart codes
34/136
7.2 OC4 MODEL STATISTIC COMPARISON
A comparison of the chosen set of aggregate statistics is reported for the OC4 model. The
following figures report a comparison of the main statistical parameters, as previously defined,
for the SOFTWIND model.
Mean platform motion statistics are shown in Figure 12. Good agreement between QB and OF
in mean values can be noted. However, slightly greater standard deviations and peak loads can
be noted for OF.
Figure 12: OC4 aggregate statistics in DLC 1.2 grouped by wind speed. Platform motion data.
Mean (black horizontal dash), standard deviation (box edges), minmax range (whiskers).
Control sensors are compared in Figure 13. Greater mean generator power, torque and rotor
speed can be noted below rated wind speed for QB. On the other hand, mean blade pitch is
higher aboverated wind speed for OF. Both factors indicate slight differences in mean
aerodynamic loads between the two codes. In particular, the LLFVW module inside QB appears
to predict greater torque than the DBEM algorithm in OF, an observation that was also noted
in [33].
FLOATECH – EUH2020 Grant Agreement N°101007142
FLOATECH D2.4: estimated reduction of uncertainty in comparison to the stateoftheart codes
35/136
Figure 13: OC4 aggregate statistics in DLC 1.2 grouped by wind speed. Control related data.
Mean (black horizontal dash), standard deviation (box edges), minmax range (whiskers).
Tower top loads can be seen in Figure 15. As for the SOFTWIND test case, the influence of
rotor torque on TT Mx can be clearly seen in Figure 15. Significantly larger maximum values of
this metric can be seen for OF. The reason that leads to this difference is worth explaining, as it
depends on complex interactions between aerodynamics, structural dynamics, control and
hydrodynamics. In Figure 14 time series in proximity of the minimum value of TT Mx recorded
in OF in DLC 1.2 at 17 m/s mean wind speed (Figure 14 (af)) and the time series in in proximity
of the maximum value of TT Mx recorded in OF in DLC 1.2 at 19 m/s mean wind speed (Figure
14 (af)) are shown. In both cases, the link between generator torque and TT Mx is apparent, as
the instantaneous values of these two sensors are quite similar. When both the minimum and
maximum values for OF are recorded, an abrupt change in generator torque can be noted. In
fact, as rotor speed decreases, the generator abruptly transitions from aboverated to below
rate torque. For the OC4 testcase, an aggressive minimum pitch saturation schedule is imposed
in the ROSCO controller. The same settings as in the public example that can be found in the
ROSCO repository (https://github.com/NREL/ROSCO/blob/main/Test_Cases/NREL
5MW/DISCON.IN) are used. Minimum pitch saturation consists in imposing a minimum blade
FLOATECH – EUH2020 Grant Agreement N°101007142
FLOATECH D2.4: estimated reduction of uncertainty in comparison to the stateoftheart codes
36/136
pitch angle as a function of the measured wind speed. It is used to perform peak shaving [34]
and/or to prevent excessive reductions in blade pitch in response to fluctuations in wind speed
due to turbulence, to lower aerodynamic peak loads. Once the minimum pitch angle as a
function of the 1second lowpass filtered velocity is reached, the generator transitions to
belowrated operation to maintain rotor speed at it’s rated value. The cause of rotor speed
decrease can be traced back to decreasing aerodynamic torque, as wind speed decreases
because of turbulence and platform pitch increases due to the instantaneous wave field,
lowering relative velocity on the rotor even further. Because of the difference in blade pitch
between QB and OF (as shown in Figure 13), pitch saturation is reached earlier by QB, that
transitions to belowrated torque earlier than OF.
Figure 14: Time series near minimum TT My (left, DLC 1.2 seed 193, mean wind speed 17 m/s) and maximum TT
Mx (right, DLC 1.2 seed 223, mean wind speed 19 m/s) during normal operation (DLC 1.2). (top to bottom)
platform pitch, nacelle foreaft velocity, blade pitch, generator torque, wind X velocity.
FLOATECH – EUH2020 Grant Agreement N°101007142
FLOATECH D2.4: estimated reduction of uncertainty in comparison to the stateoftheart codes
37/136
Moreover, once the transition occurs for OF, a relatively large oscillation at the first sideside
tower natural frequency can be noted. On the other hand, the step change in generator torque
does not initiate such an oscillation in QB, which in turn predicts lower peak TT Mx. Although
damping ratios were not compared when the numerical models were setup, 1% of critical
damping ratio is specified in OF, while a Rayleigh damping coefficient of 0.00127 is specified in
QB, leading to similar damping ratios at the tower natural frequency. In the other testcases
where a less aggressive minimum pitch saturation schedule is imposed, such a phenomenon is
not noted.
Overall good statistical agreement is noted for tower base loads in Figure 16, indicating that
both codes capture similar physics, however slightly greater peak loads for OF can be noted
once again. Rotorlevel aerodynamic loads are shown in Figure 17. While rotor torque is
comparable, aerodynamic thrust is slightly higher for QB.
Figure 15: OC4 aggregate statistics in DLC 1.2 grouped by wind speed. Tower top loads.
Mean (black horizontal dash), standard deviation (box edges), minmax range (whiskers).
FLOATECH – EUH2020 Grant Agreement N°101007142
FLOATECH D2.4: estimated reduction of uncertainty in comparison to the stateoftheart codes
38/136
Finally, mooring line loads are shown in Figure 19 and hydrodynamic forces in the surge DOF
are shown in Figure 20. The mooring line configuration for this test case feature three lines 120°
apart, with line two being directed directly upwind and the other two downwind. Although the
two codes, once again, agree well, higher mean tensions in the downwind lines can be seen for
OF. This is inline with the lower mean surge that was noted in Figure 12.
As for hydrodynamic loads in Figure 20, sound agreement can be noted in 1st order surge force
and hydrostatic force in heave. Very good agreement in secondorder forces in surge can also
be seen.
Figure 16: OC4 aggregate statistics in DLC 1.2 grouped by wind speed. Tower base loads.
Mean (black horizontal dash), standard deviation (box edges), minmax range (whiskers).
FLOATECH – EUH2020 Grant Agreement N°101007142
FLOATECH D2.4: estimated reduction of uncertainty in comparison to the stateoftheart codes
39/136
Figure 17: OC4 aggregate statistics in DLC 1.2 grouped by wind speed. Rotor aerodynamics data. Mean (black
horizontal dash), standard deviation (box edges), minmax range (whiskers).
Figure 18: OC4 aggregate statistics in DLC 1.2 grouped by wind speed. Blade data.
Mean (black horizontal dash), standard deviation (box edges), minmax range (whiskers).
FLOATECH – EUH2020 Grant Agreement N°101007142
FLOATECH D2.4: estimated reduction of uncertainty in comparison to the stateoftheart codes
40/136
Figure 19: OC4 aggregate statistics in DLC 1.2 grouped by wind speed. Mooring line tensions.
Mean (black horizontal dash), standard deviation (box edges), minmax range (whiskers).
Figure 20: OC4 aggregate statistics in DLC 1.2 grouped by wind speed. Hydrodynamic loads. Mean (black
horizontal dash), standard deviation (box edges), minmax range (whiskers).
7.3 HEXAFLOAT MODEL STATISTIC COMPARISON
A comparison of the chosen set of aggregate statistics is reported for the 10MW Hexafloat
model. The following figures show a statistical comparison of the main parameters, as previously
defined, for the SOFTWIND and OC4 models. As discussed in section 4, for this case only results
from the codes QB and DL are available.
Statistics of platform displacements are shown in Figure 21. The differences with respect to the
previous 10MW SOFTWIND and 5MW OC4 models (sections 7.1 and 7.2) is apparent in the
platform surge DOF, that shows peaks around rated wind speed in excess of 100m, much
greater than the approximately 25m for the SOFTWIND platform and 12m for the OC4 platform.
The platform offset is mainly driven by the design of the mooring system which is a “light”
catenary design for the in 10MW Hexafloat model and is not attributable to the floater itself.
FLOATECH – EUH2020 Grant Agreement N°101007142
FLOATECH D2.4: estimated reduction of uncertainty in comparison to the stateoftheart codes
41/136
The variation in mean heave as a function of wind speed is also something that the other two
test cases did not exhibit. This is linked to the platform surge offset: higher surge displacement
leads to more tension in the mooring lines and a lower mean heave position. Although slightly
higher mean surge can be noted for DL in aboverated wind speed, overall the two codes
appear to be in very good agreement.
Figure 21: HEXAFLOAT aggregate statistics in DLC 1.2 grouped by wind speed. Platform motion data. Mean
(black horizontal dash), standard deviation (box edges), minmax range (whiskers).
Figure 22: HEXAFLOAT aggregate statistics in DLC 1.2 grouped by wind speed. Aerodynamic thrust along low
speed shaft for QB and projected in wind heading direction for DL. Mean (black horizontal dash), standard
deviation (box edges), minmax range (whiskers).
FLOATECH – EUH2020 Grant Agreement N°101007142
FLOATECH D2.4: estimated reduction of uncertainty in comparison to the stateoftheart codes
42/136
Performance and control sensor statistics in DLC 1.2 are shown in Figure 23. Mean power
appears to be slightly higher for QB, as can be seen in the generator power statistics for below
rated operation and in the blade pitch statistics for aboverated operation. Moreover, higher
variations in blade pitch and in rotor speed at high wind speeds can be seen for DL.
Figure 23: HEXAFLOAT aggregate statistics in DLC 1.2 grouped by wind speed. Control related data. Mean
(black horizontal dash), standard deviation (box edges), minmax range (whiskers).
Tower top and tower base loads are shown in Figure 25 and Figure 24 respectively. Regarding
towertop loads, large differences can be seen in bending moments. In particular, sideside
bending moment (Mx) is in good agreement between the two codes for mean values (driven by
rotor torque, as discussed previously), but large differences in maximum/minimum values can
be seen. For foreaft tower top bending moment (Mx) no agreement in standard deviations or
extremes can be seen. The reasons for this discrepancy are currently unknown and are being
investigated.
FLOATECH – EUH2020 Grant Agreement N°101007142
FLOATECH D2.4: estimated reduction of uncertainty in comparison to the stateoftheart codes
43/136
Figure 24: HEXAFLOAT aggregate statistics in DLC 1.2 grouped by wind speed. Tower base loads. Mean (black
horizontal dash), standard deviation (box edges), minmax range (whiskers).
Figure 25: HEXAFLOAT aggregate statistics in DLC 1.2 grouped by wind speed. Tower top loads. Mean (black
horizontal dash), standard deviation (box edges), minmax range (whiskers).
FLOATECH – EUH2020 Grant Agreement N°101007142
FLOATECH D2.4: estimated reduction of uncertainty in comparison to the stateoftheart codes
44/136
Figure 26: HEXAFLOAT aggregate statistics in DLC 1.2 grouped by wind speed. Blade data. Mean (black
horizontal dash), standard deviation (box edges), minmax range (whiskers).
Figure 27: HEXAFLOAT aggregate statistics in DLC 1.2 grouped by wind speed. Tendon tensions. Mean (black
horizontal dash), standard deviation (box edges), minmax range (whiskers).
FLOATECH – EUH2020 Grant Agreement N°101007142
FLOATECH D2.4: estimated reduction of uncertainty in comparison to the stateoftheart codes
45/136
8 EXTREME LOADS
8.1 METHODS AND TOOLS
Wind turbine components need to be designed to withstand extreme loads predicted in
numerical tools such as the ones compared in this study. Therefore, accurate estimation of
extreme loads is essential for a costeffective and safe wind turbine design. For this reason, the
design codes benchmarked in this study are compared also on an extreme load basis.
Table 8. Variable considered for extreme load analysis.
Group Variable
Platform motion data Platform surge displacement
Platform sway displacement
Platform heave displacement
Platform roll displacement
Platform pitch displacement
Platform yaw displacement
Blade 1 Tip deflections Blade tip outofplane deflection (B1 Tip DX)
Blade tip inplane deflection (B1 Tip DY)
Shaft actions Rotor thrust
Shaft transvers force
Generator data Generator Power
Generator Torque
Aerodynamic data Aero Thrust
Aero Torque
Tower top loads Tower top foreaft force
Tower top sideside force
Tower base loads Tower base foreaft force
Tower base sideside force
Mooring lines Mooring line tension (SOFTWIND & OC4)
Tendon Tension (HEXAFLOAT)
Extreme loads were calculated using MExtremes [35], an NREL tool for the extraction of peak
loads according to IEC 614001 indications. Maximum values are estimated for all the
considered variables, for each wind speed bin from 5 to 25 m/s, according to the following
procedure:
The peak value from each file, corresponding to a given set of operating and
environmental conditions is determined;
For each DLC simulations are binned based on their mean wind speed;
FLOATECH – EUH2020 Grant Agreement N°101007142
FLOATECH D2.4: estimated reduction of uncertainty in comparison to the stateoftheart codes
46/136
Two set of extreme values are then determined:
o the absolute maximum among all the values in all of the bins (the diamonds in the
following figures);
o the value closest to but greater than the mean of the maximums in each bin,
calculated in accordance to the definition in the MExtremes manual [35] (the bars
in the following figures).
The latter method is similar to the indications in IEC 614001, Annex I [4], and is intended to
remove any outliers in the extreme values. Data is collected in groups based on the location of
the considered loadsensors, as described in Table 8.
For each test case, blade root edgewise and flapwise bending moments and tower base fore
aft and sideside bending moments are analyzed in detail, as they provide a good summary of
the loading on the rotor and on the entire structure. Other key metrics are presented briefly.
Further comments and observations for all three test cases are reported in section 10.
8.2 SOFTWIND MODEL EXTREME VALUE ANALYSIS
A comparison between the data described in the previous section is reported for the
SOFTWIND model. As explained in detail in section 4 (Table 2), for this test case a threeway
comparison between OF, DL and QB is performed. For each examined sensor the extreme
values recorded for each of the numerical models is reported. The value closest to the mean of
maximums, calculated in accordance to the definition in the MLife theory guide [32] is reported
in the bars, while the absolute extreme without averaging is shown in the diamonds. In depth
analysis of the timeseries for some key load sensors is provided to give more insight into the
physics leading to the extreme events.
Figure 28: SOFTWIND spar floater aggregate extreme value analysis. Platform motion data, maxima.
FLOATECH – EUH2020 Grant Agreement N°101007142
FLOATECH D2.4: estimated reduction of uncertainty in comparison to the stateoftheart codes
47/136
In Figure 28, the maximum platform displacements are shown. For these metrics, similar trends
are noted for the compared numerical codes, although better agreement between OF and QB
is shown. This could in part be due to the different modelling choices for the DL model, as
discussed in section 4.1.1. Moreover, maximum displacements are located in the same DLC for
OF and QB. Flapwise (My) and edgewise (Mx) blade root bending moment aggregated
maximum values are shown in Figure 29. Flapwise maximum root bending moment is located
in DLC 1.6 for all three codes. This metric is slightly higher for OF than it is for QB, while DL is
approximately 10% lower than QB. For all three codes maximum values are similar for all three
blades, suggesting that good statistical convergence of the maximum loads was reached. As
for edgewise root bending moment (BR Mx), DL and QB predict the maximum to be located in
a mix of DLC 1.3, 1.4 and 1.6, indicating that all three DLCs are similar in this load sensor. On
the other hand, maximum edgewise root bending moment is in DLC 6.3 for all three blades for
OF. A large difference between the mean of max and the nonaveraged extreme load can be
seen. As shown in the following, this is caused by a resonance.
Figure 29: SOFTWIND spar floater aggregate extreme value analysis. (top) Blade root moment data in blade
ref. system, maxima. (bottom) Blade root moment data in blade nonpitching ref. system, maxima.
FLOATECH – EUH2020 Grant Agreement N°101007142
FLOATECH D2.4: estimated reduction of uncertainty in comparison to the stateoftheart codes
48/136
Timeseries in correspondence of peak blade root flapwise loads are shown for OF and QB in
Figure 30. In Figures (af) the recorded maximum QB load is recorded, in figures (gl) maximum
OF load is recorded. For both time instants, in addition to the Time series of TB My, Time series
of other sensors are shown; platform pitch to give an indication of gravitational loading on the
structure, RNA foreaft acceleration for an indication of inertial loads on the structure, blade
pitch and rotor speed to gauge the operating conditions of the machine, aerodynamic thrust
for an indication of aerodynamic loading on the rotor. Finally, wave height and wind speed at
hub height are also reported. For this test case, the peaks are located in DLC 1.6 for all three
codes. In both the timeinstants analysed in Figure 30, peak loading occurs in correspondence
of a severe wavetrain (Figure 30 (c, i)). By comparing Figure 30 (ag) and (ek), good correlation
between rotor thrust and flapwise bending moment can be noted. This is not surprising as
flapwise loads are typically driven by aerodynamic loading. Gravitational and inertial loads
appear to be less influential for blade loads, confirming the trends that were noted in [36]. In
fact, modern multiMW wind turbine blades are relatively lightweight with respect to the other
structural components such as the nacelle and tower. Therefore, although the gravitational and
inertial loads driven by structural motion play a role, they are not as significant as those acting
on the tower. Nonetheless, the platform motion contributes to the unsteady loading on the
rotor, as foreaft platform motion introduces variations in the apparent wind speed. As shown
in Figure 30 (b) and (d) rotor speed correlates well with the pitch motion, which in turn drives
blade pitch variations. Comparing QB and OF in Figure 30 (d), both rotor speed and blade pitch
are very similar. Rotor thrust and torque in Figure 30 (d) show how, once again, QB and OF are
very similar.
In both seeds shown in Figure 30, rotor thrust goes from a positive peak to a negative one
between 3000 and 3100s, as a consequence of platform motion and blade pitch that rapidly
rises from feather to more than 20°. Even in such dynamic inflow conditions, the DBEM wake
model in OF is very close to the LLFVW model in QB in the prediction of global rotor
aerodynamic loads (thrust and torque) as well as blade root bending moments. As for DL,
platform dynamics (platform pitch and nacelle foreaft acceleration) appear to be well predicted
around the peak wave event that occurs around 3000 s, however differences with respect to QB
and OF can be seen in the control signals and in the blade root flapwise bending moment. Such
differences may be driven by the difference in instantaneous wind speed shown in Figure 30 (f)
FLOATECH – EUH2020 Grant Agreement N°101007142
FLOATECH D2.4: estimated reduction of uncertainty in comparison to the stateoftheart codes
49/136
Figure 30: Peak flapwise root bending moment (My) for QB (left, DLC1.6 seed 1039) and OF (right, DLC1.6 seed
1042). (b) Platform Pitch, (c) nacelle foreaft acceleration and wave elevation (right axis, dashed line), (d) blade
pitch and rotor speed (right axis, dashed lines), (e) rotor aero thrust and torque (right axis, dashed line), (f)
wind speed. Max(My) is also in DLC1.6, seed 1039 for DL, but at different timestep.
Blade root edgewise extreme load Time series for QB and OF are shown in Figure 31. For QB
and DL (not shown for brevity), maximum edgewise load is located in DLC1.4, and, as expected,
is located in correspondence of the Extreme Operating Gust with Direction Change (ECD)
event. In correspondence with the transient event, due to the combination of high wind speed
and yaw angle, the turbine shuts down. The shutdown procedure is simulated by imposing a
pitchtofeather maneuverer starting at 506 s with a 10 °/s pitch rate. In DL, although the
pitching manoeuvre starts at 506 s, the blade reaches 90° pitch angle at 523 s, later than QB
FLOATECH – EUH2020 Grant Agreement N°101007142
FLOATECH D2.4: estimated reduction of uncertainty in comparison to the stateoftheart codes
50/136
Figure 31: Peak edgewise root bending moment (Mx) for QB (left, DLC1.4 seed 637 ECD) and OF (right, DLC6.3
seed 10000 wave misalignment 0°, yaw 20°). (b) Platform Pitch, (c) nacelle foreaft acceleration and wave
elevation (right axis, dashed line), (d) blade pitch and rotor speed (right axis, dashed lines), (e) rotor aero thrust
and torque (right axis, dashed line), (f) wind speed. Max(Mx) in DLC1.4, seed 638 ECD+ for DL.
and OF. This delay is caused by differences in the way an override pitch procedure can be
imposed in DL. Regardless of the differences, the shutdown procedure coupled with the
transient event triggers similar behaviour in the three analysed codes. Some response at the
blade and tower natural frequencies can be seen in the edgewise root bending moment and
nacelle foreaft acceleration in OF. Traces of this can also be seen in DL and QB but to a much
lesser extent. Highamplitude oscillations at the edgewise natural frequency can be seen in
Figure 31, in DLC 6.3 where the peak loads of OF are recorded. Such instabilities were noted
mostly in OF computation, sometimes even leading to crashes and incomplete simulations. The
FLOATECH – EUH2020 Grant Agreement N°101007142
FLOATECH D2.4: estimated reduction of uncertainty in comparison to the stateoftheart codes
51/136
cause of such instabilities can likely be attributed to the simpler modalbased structural model
in OF.
Figure 32: SOFTWIND spar floater aggregate extreme value analysis. (left) Tower base force data, maxima.
(right) Tower base moment data, maxima.
Tower base foreaft and sideside forces and bending moments are shown in Figure 32. Fore
aft force (TB Fx) and bending moment (TB My) are located in DLC 1.6 with the exception of TB
My for DL which is located in DLC 6.1. With this exception, foraft loads are in good agreement
between the codes, with QB often falling between OF and DL. Sideside tower base forces are
located in DLC 6.2 for all three codes, indicating that parked conditions in severe yaw
misalignment are the most severe for this loadsensor and test case combination. The dynamics
that lead to foreaft peak bending moments can be analysed in more detail in Figure 33. Peak
load is shown on the left in DLC 1.6 for OF and QB. In this load case, although the shift in mean
windspeed is present for DL, all three codes agree quite well. Peak tower load occurs when
platform pitch is at its peak and nacelle foreaft acceleration is at its minimum, indicating a
strong contribution to this load of inertial and gravitational loads. It must be noted that such
loads are particularly significant for this test case as the tower developed for the OOStar
platform in the Lifes50+ H2020 project [15] was used, which is particularly heavy. Once again,
despite the severe oscillations in aerodynamic thrust and torque, good agreement is seen
between OF and QB, even when thrust is negative. As expected, rotor thrust and torque present
a timelag with respect to pitch motion and nacelle acceleration, as these metrics depend on
relative wind velocity. Therefore, peak tower base bending moment occurs when thrust force is
FLOATECH – EUH2020 Grant Agreement N°101007142
FLOATECH D2.4: estimated reduction of uncertainty in comparison to the stateoftheart codes
52/136
not at its peak, but rather, differently from an onshore wind turbine, its value is approximately
1e6 N (about 50% of peak).
Figure 33: Peak foreaft tower base bending moment (My) for QB and OF (left, DLC1.6 seed 1092) and DL (right,
DLC6.1 seed 10001 misalignment 30° yaw 0°). (b) Platform Pitch, (c) nacelle foreaft acceleration and wave
elevation (right axis, dashed line), (d) blade pitch and rotor speed (right axis, dashed lines), (e) rotor aero thrust
and torque (right axis, dashed line), (f) wind speed.
The DLC 6.1 Time series where peakloads occur for DL are shown in Figure 33 (right). During
this peakload event, although tower base, inertial and gravitational loads remain closely tied
together, all three codes behave very differently.
FLOATECH – EUH2020 Grant Agreement N°101007142
FLOATECH D2.4: estimated reduction of uncertainty in comparison to the stateoftheart codes
53/136
Figure 34: Peak edgewise sideside tower base bending moment (Mx) for QB (left, DLC6.2 seed 10002
misalignment 30° yaw 135°) and OF & DL (right, DLC6.2 seed 10002 misalignment 30° yaw 180°). (b) Platform
Pitch, (c) nacelle foreaft acceleration and wave elevation (right axis, dashed line), (d) blade pitch and rotor
speed (right axis, dashed lines), (e) rotor aero thrust and torque (right axis, dashed line), (f) wind speed.
Time series where sideside peak bending moments are shown in Figure 34 for OF and QB (left)
and for DL (right). Once again platform roll – since this is a sideside load the sideside oscillation
of the platform is plotted – and tower base load correlates nicely, as does nacelle sideside
acceleration for QB and DL. For OF, nacelle acceleration is exported in a reference system that
yaws with the nacelle, and therefore this output does not correspond to the one shown for QB
and OF for this test case. Comparing the cases where the peak for OF and DL (right) and QB
(left) are recorded, it can be noted that the dynamics of the system are very similar. The two
cases differ only for yaw angle setting, and wave and wind Time series are the same. This seems
FLOATECH – EUH2020 Grant Agreement N°101007142
FLOATECH D2.4: estimated reduction of uncertainty in comparison to the stateoftheart codes
54/136
to hint to the fact that in parked conditions rotor aerodynamic loads have little influence on
system dynamics. Significant differences in rotor speed between OF and QB can be seen.
Unfortunately, we do not have value for DL to compare to as if control is deactivated, as is the
case when the turbine is parked, control sensors cannot be output and are instead replaced by
zeros. Finally, comparing Figure 34 (a) and (g) peak in TB Mx for OF appears to be in Figure 34
(a), not in Figure 34 (g). Indeed, the peak is higher in Figure 34 (a), however, due to the maximum
averaging process described in section 8.1, this value is considered an outlier for OF.
Figure 35: SOFTWIND spar floater aggregate extreme value analysis. Blade tip deflections in blade coned (non
pitching) ref. system, maxima.
Figure 36: SOFTWIND spar floater aggregate extreme value analysis. Shaft loading data, maxima.
FLOATECH – EUH2020 Grant Agreement N°101007142
FLOATECH D2.4: estimated reduction of uncertainty in comparison to the stateoftheart codes
55/136
Figure 37: SOFTWIND spar floater aggregate extreme value analysis. (right) Generator data, maxima. (left)
Aerodynamic data, maxima. Aero Thrust is defined as total aerodynamic force along rotor shaft for OF and QB,
and total aerodynamic force projected along wind heading direction for DL. Large outlier in DL due to numerical
instability in DLC 1.6 seed 1082.
Figure 38: SOFTWIND spar floater aggregate extreme value analysis. (left) Tower top forces, maxima. (right)
Tower top moments, maxima.
A comparison of the extreme values, as estimated by the considered codes, for the SOFTWIND
model is reported in Table 9.
FLOATECH – EUH2020 Grant Agreement N°101007142
FLOATECH D2.4: estimated reduction of uncertainty in comparison to the stateoftheart codes
56/136
Table 9. SOFTWIND model extreme value comparison.
Type
Label
Units
Name QB
QB value
Name OF
OF value
Name DL
DL value
OFQB diff. (%)
DLQB diff. (%)
PtfmSurge Min Surge m 6.1
1.06E+1
1.4
1.20E+1
1.4
1.46E+1
13.6
38.2
Max Surge m 1.6
1.92E+1
1.6
2.12E+1
6.1
2.70E+1
10.3
40.5
PtfmSway Min Sway m 6.2
1.27E+1
6.2
1.25E+1
1.4
9.01E+0
1.7
29.3
Max Sway m 6.2
1.16E+1
6.2
1.40E+1
1.4
9.36E+0
21.1
19.2
PtfmHeave Min Heave m 6.2
7.54E+0
6.1
5.78E+0
6.2
1.00E+1
23.3
33.3
Max Heave m 6.1
1.49E+0
6.1
3.12E+0
6.2
4.89E+0
109.4
227.5
PtfmRoll Min Roll ° 6.2
7.53E+0
6.2
8.44E+0
6.2
5.53E+0
12.1
26.6
Max Roll ° 6.2
7.67E+0
6.2
9.37E+0
6.2
5.46E+0
22.1
28.8
PtfmPitch Min Pitch ° 6.1
7.81E+0
1.4
7.18E+0
6.2
1.23E+1
8.1
57.7
Max Pitch ° 1.6
7.71E+0
1.6
7.84E+0
6.1
1.36E+1
1.6
77.0
PtfmYaw Min Yaw ° 1.3
4.60E+0
6.2
8.75E+0
1.6
6.56E+0
90.2
42.7
Max Yaw ° 6.2
4.66E+0 6.2
1.17E+1
1.3
5.72E+0
151.1
22.7
TipDxc1 Min B1 Tip DX m 1.4
8.86E+0
1.4
8.93E+0
6.2
3.06E01
0.9
96.5
Max B1 Tip DX m 1.6
1.39E+1
1.6
1.48E+1
6.1
2.15E01
6.7
98.4
TipDyc1 Min B1 Tip DY m 6.1
5.04E+0
6.1
5.10E+0
6.2
1.31E+0
1.3
73.9
Max B1 Tip DY m 6.1
4.21E+0
1.4
4.08E+0
6.2
1.31E+0
3.0
68.8
RootMxb1 Min B1 Mx kNm
1.6
2.04E+4
6.3
2.68E+4
6.2
2.10E+4
31.5
3.1
Max B1 Mx kNm
1.6
1.90E+4
6.3
3.19E+4
1.3
2.27E+4
67.8
19.5
RootMyb1 Min B1 My kNm
1.4
3.48E+4
1.4
2.95E+4
1.4
3.31E+4
15.4
5.1
Max B1 My kNm
1.6
5.21E+4
1.6
5.39E+4
1.6
4.68E+4
3.6
10.1
RootMxb2 Min B2 Mx kNm
1.6
1.92E+4
6.3
2.77E+4
6.1
2.04E+4
44.1
6.1
Max B2 Mx kNm
1.6
1.92E+4
6.3
3.03E+4
1.3
2.20E+4
57.5
14.6
RootMyb2 Min B2 My kNm
1.4
3.09E+4
1.4
4.43E+4
1.4
3.36E+4
43.4
8.8
Max B2 My kNm
1.6
5.23E+4
1.6
5.33E+4
1.6
4.68E+4
2.0
10.4
RootMxb3 Min B3 Mx kNm
1.6
1.99E+4
6.3
3.01E+4
6.2
2.05E+4
51.4
3.2
Max B3 Mx kNm
1.4
2.49E+4
6.3
2.70E+4
1.4
2.95E+4
8.3
18.4
RootMyb3 Min B3 My kNm
1.4
3.34E+4
1.4
4.27E+4
1.4
3.43E+4
27.7
2.5
Max B3 My kNm
1.6
5.35E+4
1.6
5.55E+4
1.6
4.62E+4
3.8
13.6
RootMxc1 Min B1 Mx kNm
6.1
2.13E+4
6.2
2.51E+4
6.1
2.00E+4
18.0
6.1
Max B1 Mx kNm
6.2
2.66E+4
1.3
2.50E+4
1.6
2.43E+4
6.1
8.8
RootMyc1 Min B1 My kNm
1.4
4.03E+4
1.4
3.50E+4
1.6
3.76E+4
13.2
6.8
Max B1 My kNm
1.6
5.41E+4
1.6
5.29E+4
1.6
4.47E+4
2.1
17.3
RootMxc2 Min B2 Mx kNm
6.2
2.60E+4
6.1
2.09E+4
1.4
2.05E+4
19.8
21.2
Max B2 Mx kNm
6.1
2.25E+4
1.3
2.42E+4
1.6
2.45E+4
7.9
9.0
RootMyc2 Min B2 My kNm
1.4
3.18E+4
1.4
4.42E+4
1.6
3.88E+4
38.8
22.0
Max B2 My kNm
1.6
5.24E+4
1.6
5.33E+4
1.6
4.36E+4
1.7
16.7
FLOATECH – EUH2020 Grant Agreement N°101007142
FLOATECH D2.4: estimated reduction of uncertainty in comparison to the stateoftheart codes
57/136
RootMxc3 Min B3 Mx kNm
6.2
2.44E+4
6.1
2.07E+4
6.2
1.96E+4
15.2
19.6
Max B3 Mx kNm
6.1
2.23E+4
1.3
2.51E+4
1.4
2.93E+4
12.5
31.6
RootMyc3 Min B3 My kNm
1.4
4.02E+4
1.4
4.75E+4
1.6
3.81E+4
18.1
5.3
Max B3 My kNm
1.6
5.29E+4
1.6
5.42E+4
1.6
4.27E+4
2.3
19.4
LSShftFxa Min Rot Th kN 1.6
1.27E+3
1.6
1.45E+3
1.4
1.12E+3
13.9
12.1
Max Rot Th kN 1.6
3.35E+3
1.6
3.37E+3
1.6
2.88E+3
0.8
13.8
GenTq Min Gen Tq kNm
1.6
0.00E+0
1.3
0.00E+0
6.1
0.00E+0
N/A N/A
Max Gen Tq kNm
1.6
2.21E+2
1.6
2.21E+2
1.3
2.00E+2
0.1
9.6
GenPwr Min Gen Pwr kW 1.6
0.00E+0
1.3
0.00E+0
1.4
0.00E+0
N/A N/A
Max Gen Pwr kW 1.6
1.52E+4
1.6
1.59E+4
1.6
1.51E+4
4.2
1.2
RtAeroFxh Min Aero Th N 6.1
1.31E+5
1.4
1.37E+6
1.4
9.92E+5
940.6
655.5
Max Aero Th N 1.6
2.72E+6
1.6
2.78E+6
1.6
3.57E+6
2.2
31.6
RtAeroMxh
Min Aero Tq Nm 1.4
2.38E+7
1.4
2.54E+7
1.3
0.00E+0
6.7
100.0
Max Aero Tq Nm 1.6
4.29E+7
1.6
4.41E+7
1.3
0.00E+0
2.6
100.0
YawBrFxp Min TT Fx kN 1.6
3.22E+3
1.6
3.51E+3
6.1
4.13E+3
9.1
28.5
Max TT Fx kN 1.6
4.64E+3
1.6
4.80E+3
6.1
5.33E+3
3.4
14.7
YawBrFyp Min TT Fy kN 6.2
3.38E+3
6.3
3.42E+3
6.2
2.53E+3
1.1
25.4
Max TT Fy kN 6.2
3.41E+3
6.2
3.70E+3
6.2
2.44E+3
8.5
28.3
YawBrMxp Min TT Mx kNm
6.2
9.63E+3
6.3
2.90E+4
6.2
9.26E+3
200.7
3.8
Max TT Mx kNm
6.2
1.81E+4
6.3
1.66E+4
1.6
1.42E+4
8.3
21.2
YawBrMyp Min TT My kNm
1.6
3.06E+4
1.4
4.74E+4
1.4
3.89E+4
54.7
27.0
Max TT My kNm
1.6
3.38E+4
1.6
4.03E+4
1.6
3.53E+4
19.1
4.3
TwrBsFxt Min TB Fx kN 6.1
8.26E+3
1.6
6.22E+3
1.6
6.46E+3
24.8
21.8
Max TB Fx kN 1.6
9.07E+3
1.6
8.99E+3
1.6
9.23E+3
0.9
1.8
TwrBsFyt Min TB Fy kN 6.2
7.87E+3
6.2
8.09E+3
6.2
5.52E+3
2.8
29.9
Max TB Fy kN 6.2
7.71E+3
6.2
8.57E+3
6.2
5.47E+3
11.2
29.1
TwrBsMxt Min TB Mx kNm
6.2
5.33E+5
6.2
5.67E+5
6.2
3.84E+5
6.4
27.9
Max TB Mx kNm
6.2
5.41E+5
6.2
5.33E+5
6.2
3.90E+5
1.5
27.9
TwrBsMyt Min TB My kNm
6.1
5.94E+5
1.6
4.66E+5
6.1
6.56E+5
21.4
10.4
Max TB My kNm
1.6
6.73E+5
1.6
6.41E+5
6.1
8.40E+5
4.7
24.7
FAIRTEN1 Min FAIRTEN1
kN 1.6
6.62E+2
1.6
6.20E+2
6.1
5.52E+0
6.4
100.8
Max FAIRTEN1
kN 6.1
5.18E+3
1.6
3.87E+3
6.1
4.64E+3
25.2
10.4
FAIRTEN2 Min FAIRTEN2
kN 6.1
1.35E+3
6.2
5.69E+2
6.2
4.54E+0
57.9
99.7
Max FAIRTEN2
kN 6.2
6.05E+3
6.2
5.49E+3
6.2
4.94E+3
9.2
18.4
FAIRTEN3 Min FAIRTEN3
kN 6.1
1.30E+3
6.2
5.85E+2
6.2
3.07E+1
55.2
97.6
Max FAIRTEN3
kN 6.2
6.04E+3
6.2
6.85E+3
6.1
5.84E+3
13.4
3.3
An indication of the average differences between the codes is shown in Table 10, separately for
overpredictions and underpredictions.
FLOATECH – EUH2020 Grant Agreement N°101007142
FLOATECH D2.4: estimated reduction of uncertainty in comparison to the stateoftheart codes
58/136
Table 10. Average codetocode differences, SOFTWIND model
OFQB DLQB
average
+
diff 24.06% 51.76%
average

diff 14.21% 29.22%
FLOATECH – EUH2020 Grant Agreement N°101007142
FLOATECH D2.4: estimated reduction of uncertainty in comparison to the stateoftheart codes
59/136
8.3 OC4 MODEL EXTREME VALUE ANALYSIS
An analysis of extreme loads recorded for the OC4 model is shown in this section. Meanofmax
values (IEC 614001 appendix I) are shown in the bars, while absolute maximum values with no
averaging are shown in the corresponding diamonds. Indetail Time series analysis are provided
for key loadsensor for an indepth analysis of the physics leading to the peak loads on the
components.
Figure 39: OC4 spar floater aggregate extreme value analysis. Platform motion data, maxima.
Peaks in platform motions appear to agree well, with the exception of surge, which however
remains in the same DLC for both OF and QB (Figure 39).
Blade root flapwise, edgewise, inplane and outofplane root loads are shown in Figure 40.
Flapwise loads are in very good agreement. OF underestimated QB approximately 24%
depending on which blade is considered. DLC 1.3 (normal operation in extreme turbulence)
generates the highest flapwise loads for both codes on all three blades. Edgewise loads on the
other hand are not in good agreement. As will be shown in detail in the following this is due to
an edgewise blade resonance in parked conditions for OF. This can be noted in Figure 40
looking at the outliers for OF, that are very high. Because the outliers are referred to DLC 6.2,
where the turbine is parked and the blades are pitched to 90°, they are also apparent in the
outofplane peak loads. Time series relative to the recorded peaks in blade root flapwise
bending moment (BR Myb) in QB and OF are shown in Figure 41 (af) and (gl), respectively. As
mentioned previously, for this test case peak BR Myb is recorded in DLC 1.3, in simulations near
rated, with 11 m/s mean wind speed.
FLOATECH – EUH2020 Grant Agreement N°101007142
FLOATECH D2.4: estimated reduction of uncertainty in comparison to the stateoftheart codes
60/136
Figure 40: OC4 spar floater aggregate extreme value analysis. (top) Blade root moment data in blade ref.
system, maxima. (bottom) Blade root moment data in blade nonpitching ref. system, maxima.
BR Myb is mainly influenced by aerodynamic loads and thus depends on the relative inflow
velocity. Because of wind shear, the local azimuth of each blade will influence loads significantly
and since QB and OF predict different rotor speeds, blade azimuth will be different in the two
codes. Figure 41 (d,j) shows how rotor speed is higher for QB when wind speed dips below
rated. This characteristic is noted in all three models (Figure 6, Figure 13, Figure 23) and also in
previous codetocode comparisons in onshore conditions [33]. In practice, this implies that the
Tip Speed Ratio (TSR) is higher for QB, leading to higher aerodynamic thrust (Figure 41 (e, k)).
The higher thrust seems to be influencing platform pitch, that is on average higher for QB thrust
(Figure 41 (b, h)). Despite this, overall, blade root bending moments are very similar in
magnitude, as if we consider the maximum value of BR My on all three blades OF is only 1%
higher than QB (Table 12).
FLOATECH – EUH2020 Grant Agreement N°101007142
FLOATECH D2.4: estimated reduction of uncertainty in comparison to the stateoftheart codes
61/136
Figure 41: Peak flapwise blade root bending moment (My) for QB (left, DLC1.3 seed 638) and OF (right, DLC1.3
seed 634). (b) Platform Pitch, (c) nacelle foreaft acceleration and wave elevation (right axis, dashed line), (d)
blade pitch and rotor speed (right axis, dashed lines), (e) rotor aero thrust (f) wind speed.
Time histories in correspondence with blade root edgewise peak loads are shown in Figure 42.
For this sensor, the DLC where peak loading occurs is not the same: DLC 1.4 for QB and DLC
6.2 for OF. These correspond respectively to operation with extreme direction change and
parked in Extreme Sea State (ESS) with grid loss. Focusing on DLC 1.4 first, QB and OF behave
very similarly in terms of rotor speed, blade pitch and aerodynamic thrust. The strong variation
in aerodynamic thrust in correspondence with the transient wind gust event and subsequent
shutdown is predicted quite well by the DBEM routine in OF if compared to the higher fidelity
LLFVW model in QB. Once the rotor reaches a full stop, high frequency edgewise oscillations in
blade root bending moment can be seen for OF in Figure 42 (a). In QB these oscillations are
FLOATECH – EUH2020 Grant Agreement N°101007142
FLOATECH D2.4: estimated reduction of uncertainty in comparison to the stateoftheart codes
62/136
much better damped. The same can be said for the tower. In fact, looking at tower foreaft
acceleration, it is apparent how OF exhibits large and less damped oscillations at the tower’s
natural frequency (Figure 42 (c)). These oscillations only become relevant once the rotor is
parked and aerodynamic foreaft damping is missing. Now analysing DLC 6.2 (Figure 42 (gl)),
a clear edgewise blade resonance in OF can be seen. This leads to peak loads that are nearly
three times those recorded in QB. Instability at the system’s natural frequencies was noted
multiple times in OF (Figure 31) and also influences fatigue loads (Figure 80) and is most likely
linked to the fact that OF uses a lowerfidelity modal based structural model.
Figure 42: Peak edgewise blade root bending moment (Mx) for QB (left, DLC1.4 seed 638) and OF (right, DLC6.2
seed 10002 mis 30° yaw 45°). (b) Platform pitch, (c) nacelle foreaft acceleration and wave elevation (right axis,
dashed line), (d) blade pitch and rotor speed (right axis, dashed lines), (e) rotor aero thrust (f) wind speed.
FLOATECH – EUH2020 Grant Agreement N°101007142
FLOATECH D2.4: estimated reduction of uncertainty in comparison to the stateoftheart codes
63/136
Figure 43: OC4 spar floater aggregate extreme value analysis. (left) Tower base force data, maxima. (right)
Tower base moment data, maxima.
Predicted tower base peak loads are very similar in the two codes (Figure 43). As shown in Table
11, the foreaft force (TB Fx) is 7% lower for OF, the tower base sideside bending moment (TB
Mx) is 18% lower, and the tower base sideside force (TB Fy) and foreaft bending moments (TB
My) are 0.4% an 1.2% lower, respectively.
Time series in proximity of the extreme recorded TB My are shown in Figure 44. In both codes
the extreme value is recorded in Severe Sea State (SSS) conditions (DLC 1.6) at an average wind
speed of 23 m/s. In this condition, the higher rotor speed for QB that was visible in Figure 41
(d, j) is not present, and aerodynamic thrust is quite similar in the two codes. Similar to the
Softwind test case, the TB My signal correlates quite well with the platform pitch Time series;
this is an indication that hydrodynamic loading is the main driver of tower base bending moment
on this test case too. In both Figure 41 (a) and (g) oscillations around the tower’s natural
frequency of approximately 0.4 Hz can also be noted, more marked for OF than QB.
FLOATECH – EUH2020 Grant Agreement N°101007142
FLOATECH D2.4: estimated reduction of uncertainty in comparison to the stateoftheart codes
64/136
Figure 44: Peak foreaft tower base bending moment (My) for QB (left, DLC1.6 seed 1098) and OF (right, DLC1.6
seed 1095) (b) Platform pitch, (c) nacelle foreaft acceleration and wave elevation (right axis, dashed line), (d)
blade pitch and rotor speed (right axis, dashed lines), (e) rotor aero thrust (f) wind speed.
Finally, Time series in proximity of peak TB Mx are shown in Figure 45. For both codes peak
loads are recorded in parked conditions, in presence of grid loss (DLC 6.2), and large platform
motions appear to be the main contributor to the peak loading. Very good agreement for this
test case in parked conditions is noted between QB and OF in both Figure 44 and Figure 45, a
slight exception being the increased variability in aerodynamic thrust that is noted for QB in
Figure 45. It must be noted however that with respect to the operating conditions (Figure 41),
the absolute values of rotor thrust are quite small.
FLOATECH – EUH2020 Grant Agreement N°101007142
FLOATECH D2.4: estimated reduction of uncertainty in comparison to the stateoftheart codes
65/136
Figure 45: Peak sideside tower base bending moment (Mx) for QB (left, DLC6.2 seed 10003 mis 30° yaw 90°)
and OF (right, DLC6.2 seed 10002 mis 30° yaw 90°) (b) Platform pitch, (c) nacelle foreaft acceleration and
wave elevation (right axis, dashed line), (d) blade pitch and rotor speed (right axis, dashed lines), (e) rotor aero
thrust (f) wind speed.
FLOATECH – EUH2020 Grant Agreement N°101007142
FLOATECH D2.4: estimated reduction of uncertainty in comparison to the stateoftheart codes
66/136
Figure 46: OC4 spar floater aggregate extreme value analysis. Blade tip deflections in blade nonpitching ref.
system, maxima.
Figure 47: OC4 spar floater aggregate extreme value analysis. (left) Generator data, maxima. (middle)
Aerodynamic data, maxima. (right) Total thrust force along rotor shaft.
FLOATECH – EUH2020 Grant Agreement N°101007142
FLOATECH D2.4: estimated reduction of uncertainty in comparison to the stateoftheart codes
67/136
Figure 48: OC4 spar floater aggregate extreme value analysis. Tower top loading data, maxima.
Figure 49: OC4 spar floater aggregate extreme value analysis. Mooring data, maxima.
A comparison of the extreme values, as estimated by the considered codes, for the OC4 model
is reported in Table 11.
FLOATECH – EUH2020 Grant Agreement N°101007142
FLOATECH D2.4: estimated reduction of uncertainty in comparison to the stateoftheart codes
68/136
Table 11. OC4 model extreme value comparison.
Type
Label
Units
DLC QB
QB value
DLC OF
OF value
OFQB diff. (%)
PtfmSurge Min Surge m 6.1
6.95E+00 6.1
8.96E+00 28.87
Max
Surge m 6.1
1.71E+01 6.1
1.56E+01 9.22
PtfmSway Min Sway m 6.2
7.24E+00 6.2
7.17E+00 0.85
Max
Sway m 1.4
6.52E+00 1.4
6.70E+00 2.77
PtfmHeave Min Heave m 6.2
1.11E+01 6.2
1.02E+01 8.22
Max
Heave m 6.2
1.12E+01 6.2
1.14E+01 2.05
PtfmRoll Min Roll ° 6.2
5.03E+00 6.2
4.51E+00 10.30
Max
Roll ° 6.2
5.06E+00 6.2
4.51E+00 10.97
PtfmPitch Min Pitch ° 6.1
9.60E+00 6.1
8.58E+00 10.69
Max
Pitch ° 6.2
9.45E+00 6.2
9.24E+00 2.16
PtfmYaw Min Yaw ° 6.1
8.34E+00 6.1
8.44E+00 1.13
Max
Yaw ° 6.2
4.97E+00 6.2
4.66E+00 6.21
TipDxc1 Min B1 Tip DX m 1.4
9.06E+00 1.4
8.92E+00 1.52
Max
B1 Tip DX m 1.6
7.35E+00 1.3
7.07E+00 3.75
TipDyc1 Min B1 Tip DY m 6.1
3.95E+00 6.2
4.02E+00 1.75
Max
B1 Tip DY m 1.4
3.56E+00 1.4
4.81E+00 35.19
RootMxb1 Min B1 Mx kNm
1.6
5.71E+03 6.2
1.94E+04 239.80
Max
B1 Mx kNm
1.4
9.14E+03 6.2
1.82E+04 99.50
RootMyb1 Min B1 My kNm
1.4
1.72E+04 1.4
1.63E+04 5.21
Max
B1 My kNm
1.3
1.37E+04 1.3
1.43E+04 4.28
RootMxb2 Min B2 Mx kNm
1.3
5.53E+03 6.2
1.80E+04 226.25
Max
B2 Mx kNm
1.4
6.60E+03 1.4
8.79E+03 33.20
RootMyb2 Min B2 My kNm
1.4
1.50E+04 1.4
1.30E+04 13.13
Max
B2 My kNm
1.3
1.41E+04 1.3
1.36E+04 4.05
RootMxb3 Min B3 Mx kNm
1.3
5.67E+03 6.2
1.33E+04 134.21
Max
B3 Mx kNm
1.3
6.06E+03 6.2
1.49E+04 146.46
RootMyb3 Min B3 My kNm
1.4
1.50E+04 1.4
1.65E+04 9.78
Max
B3 My kNm
1.3
1.39E+04 1.3
1.41E+04 1.67
RootMxc1 Min B1 Mx kNm
6.2
8.35E+03 1.4
6.95E+03 16.86
Max
B1 Mx kNm
1.3
8.39E+03 6.2
9.78E+03 16.51
RootMyc1 Min B1 My kNm
1.4
1.88E+04 1.4
1.74E+04 7.52
Max
B1 My kNm
1.6
1.36E+04 1.3
1.42E+04 4.75
RootMxc2 Min B2 Mx kNm
6.1
6.84E+03 6.2
8.75E+03 27.95
Max
B2 Mx kNm
1.3
8.52E+03 1.4
1.08E+04 26.71
RootMyc2 Min B2 My kNm
1.4
1.59E+04 1.4
1.55E+04 2.60
Max
B2 My kNm
1.3
1.41E+04 1.3
1.36E+04 3.74
RootMxc3 Min B3 Mx kNm
6.2
7.01E+03 6.2
8.90E+03 26.94
Max
B3 Mx kNm
1.3
8.41E+03 1.3
8.67E+03 3.10
FLOATECH – EUH2020 Grant Agreement N°101007142
FLOATECH D2.4: estimated reduction of uncertainty in comparison to the stateoftheart codes
69/136
RootMyc3 Min B3 My kNm
1.4
1.46E+04 1.4
1.70E+04 16.64
Max
B3 My kNm
1.3
1.39E+04 1.3
1.35E+04 2.89
LSShftFxa Min Rot Th kN 1.4
5.48E+02 1.4
5.42E+02 1.18
Max
Rot Th kN 1.4
1.08E+03 1.4
1.05E+03 2.58
GenTq Min Gen Tq kNm
1.4
0.00E+00 1.3
0.00E+00 N/A
Max
Gen Tq kNm
1.6
4.31E+01 1.3
4.31E+01 0.00
GenPwr Min Gen Pwr kW 1.4
0.00E+00 1.4
9.67E+02 N/A
Max
Gen Pwr kW 1.6
6.71E+03 1.6
6.96E+03 3.69
RtAeroFxh Min Aero Th N 1.4
6.18E+05 1.4
6.03E+05 2.44
Max
Aero Th N 1.4
8.88E+05 1.4
8.53E+05 3.92
RtAeroMxh Min Aero Tq Nm 1.4
9.86E+06 1.4
1.02E+07 3.62
Max
Aero Tq Nm 1.6
1.20E+07 1.6
1.13E+07 5.56
YawBrFxp Min TT Fx kN 6.1
1.31E+03 6.1
9.93E+02 24.26
Max
TT Fx kN 1.6
1.42E+03 1.6
1.44E+03 1.47
YawBrFyp Min TT Fy kN 6.2
8.04E+02 6.2
7.27E+02 9.57
Max
TT Fy kN 6.2
6.92E+02 6.2
6.88E+02 0.66
TwrBsFxt Min TB Fx kN 6.1
1.69E+03 6.1
1.26E+03 25.15
Max
TB Fx kN 6.2
2.03E+03 6.2
1.89E+03 6.94
TwrBsFyt Min TB Fy kN 6.2
1.16E+03 6.2
1.01E+03 13.16
Max
TB Fy kN 6.2
9.84E+02 6.2
9.80E+02 0.38
TwrBsMxt Min TB Mx kNm
6.2
6.54E+04 6.2
6.26E+04 4.19
Max
TB Mx kNm
6.2
7.94E+04 6.2
6.46E+04 18.54
TwrBsMyt Min TB My kNm
6.1
1.20E+05 6.1
8.96E+04 25.41
Max
TB My kNm
1.6
1.32E+05 1.6
1.31E+05 1.20
FAIRTEN1 Min FAIRTEN1
kN 6.2
4.68E+01 6.2
1.44E+02 208.32
Max
FAIRTEN1
kN 6.2
2.12E+03 6.2
2.36E+03 11.14
FAIRTEN2 Min FAIRTEN2
kN 1.6
0.00E+00 6.1
5.98E01 N/A
Max
FAIRTEN2
kN 6.1
6.00E+03 6.1
5.70E+03 4.96
FAIRTEN3 Min FAIRTEN3
kN 6.2
1.11E+02 6.2
1.41E+02 28.04
Max
FAIRTEN3
kN 6.1
2.22E+03 6.1
2.35E+03 5.68
An indication of the average differences between the codes for OC4 model can be represented
in Table 12, separately for overpredictions and underpredictions.
Table 12. Average codetocode differences, OC4 model
OFQB
average + diff.
44.06%
average  diff. 8.21%
FLOATECH – EUH2020 Grant Agreement N°101007142
FLOATECH D2.4: estimated reduction of uncertainty in comparison to the stateoftheart codes
70/136
8.4 HEXAFLOAT MODEL EXTREME VALUE ANALYSIS
A comparison between the data described in the previous section is reported for the Hexafloat
model. As shown in Figure 50, extreme values of platform motion are in good agreement, in
both magnitude and DLC in which they are recorded.
Figure 50: HEXAFLOAT spar floater aggregate extreme value analysis. Platform motion data, maxima.
Blade root bending moments are also generally in good agreement. With respect to QB
however, DL shows an underprediction of approximately 8% in flapwise loads and an
overprediction of approximately 4.5% in edgewise loads (Table 12). Flapwise loads, which are
the most influenced by aerodynamics, peak in DLC 1.6 for QB and in a mix of DLC 1.3 and 1.6
for DL, indicating that the two DLCs are very close in blade extreme loads for this test case.
Time series where peak loads are recorded are shown in Figure 52 (left) for DL and (right) for
QB. In both cases the platform pitch signals show the system is oscillating around its natural
frequency with a period of approximately 55 s. A timeshift in the oscillation can be seen, with
QB lagging DL. This is likely due to the approximately 12s shift in windspeed that can be seen
in Figure 52 (f). These slow naturalfrequency oscillations cause variations in relative windspeed
which cause the slowvarying oscillation in flapwise moments observable both in DLC 1.3 and in
DLC 1.6. Regarding Figure 52 (left), apart from the mentioned timeshift the two codes behave
similarly, and a similar peak value is recorded in both cases.
FLOATECH – EUH2020 Grant Agreement N°101007142
FLOATECH D2.4: estimated reduction of uncertainty in comparison to the stateoftheart codes
71/136
Figure 51: HEXAFLOAT spar floater aggregate extreme value analysis. (top) Blade root moment data in blade
ref. system, maxima. (bottom) Blade root moment data in blade nonpitching ref. system, maxima.
This changes in DLC 1.6 Figure 52 (right), where maximum for QB is recorded. In this case the
platform pitch combines negatively with the wave elevation for QB. In fact, a wave train of
relatively high intensity hits the turbine around 1010 s – 1040 s, when the platform is near its
peak value of pitch for QB. On the other hand, in DL platform pitch is already decreasing when
the wave train hits, therefore relative windspeed is high and, to counteract this, the blades are
pitched out slightly, and therefore the wave train causes lower flapwise loads.
FLOATECH – EUH2020 Grant Agreement N°101007142
FLOATECH D2.4: estimated reduction of uncertainty in comparison to the stateoftheart codes
72/136
Figure 52: Peak flapwise root bending moment (My) for DL (left, DLC1.3 seed 632) and QB (right, DLC1.6 seed
1045). (b) Platform Pitch, (c) nacelle foreaft acceleration and wave elevation (right axis, dashed line), (d) blade
pitch and rotor speed (right axis, dashed lines), (e) rotor aero thrust, (f) wind speed.
Blade root edgewise peak loads are located in DLC 6.1 for DL (Figure 53, left) and in DLC 1.4
for QB (Figure 53, right). In both cases the dynamics are well reproduced by the codes, platform
pitch is similar and nacelle acceleration appears to be roughly in phase and at a similar frequency
to the incoming waves. The phase shift in platform pitch that can be seen in DLC 1.4 (Figure 53,
right) is due to the difference in the shutdown procedure in the two codes (Figure 53 (e)), which
also leads to different maximum edgewise loads being predicted.
FLOATECH – EUH2020 Grant Agreement N°101007142
FLOATECH D2.4: estimated reduction of uncertainty in comparison to the stateoftheart codes
73/136
Figure 53: Peak edgewise root bending moment (My) for DL (left, DLC6.1 seed 10001 mis30 y10) and QB (right,
DLC1.4 seed 619). (b) Platform Pitch, (c) nacelle foreaft acceleration and wave elevation (right axis, dashed
line), (d) blade pitch and rotor speed (right axis, dashed lines), (e) rotor aero thrust, (f) wind speed.
Tower base extreme foreaft bending moments are shown in Figure 54. Good agreement in
shear forces both in magnitude and DLC can be noted. The same can be said for bending
moments although the DLC where the peak sideside bending moment is not the same. Time
series for the foreaft tower base loads are shown in Figure 55. In both cases peak loading occurs
in DLC 1.6. Once again lowfrequency oscillations in pitch can be seen. These oscillations
however do not appear to influence the nacelle foreaft acceleration signals, since the timelag
between QB and DL that is observable in the former is not observable in the latter. Nacelle
acceleration appears to be driven by oscillations at the wave frequency. As it was the case for
the other test cases (SOFTWIND, OC4), the peak loads occur when the foreaft acceleration and
FLOATECH – EUH2020 Grant Agreement N°101007142
FLOATECH D2.4: estimated reduction of uncertainty in comparison to the stateoftheart codes
74/136
the platform pitch are in an unfavourable combination, maximising inertial and gravitational
loads. However, mean pitch appears to influence the foreaft bending moment less than in the
previously discussed test cases, as can be noted by the weaker correlation between foreaft
bending moment and platform pitch. This is possibly due to the fact that in this test case the
standard onshore tower for the DTU 10MW RWT is used which is much lighter than the OO
Star platform tower used in the SOFTWIND test case. Finally, although the control signals are
quite different  because of the different platform motion and consequently different relative
inflow velocity – rotor thrust is quite similar.
Figure 54: HEXAFLOAT spar floater aggregate extreme value analysis. (left) Tower base force data, maxima.
(right) Tower base moment data, maxima.
FLOATECH – EUH2020 Grant Agreement N°101007142
FLOATECH D2.4: estimated reduction of uncertainty in comparison to the stateoftheart codes
75/136
Figure 55: Peak foreaft tower base bending moment (My) for DL (left, DLC1.6 seed 1038) and QB (right, DLC1.6
seed 1035). (b) Platform Pitch, (c) nacelle foreaft acceleration and wave elevation (right axis, dashed line), (d)
blade pitch and rotor speed (right axis, dashed lines), (e) rotor aero thrust, (f) wind speed.
Tower base sideside extreme loads are shown in Figure 56 for DL (left) and QB (right). In both
cases they are recorded in DLC 6.2. In absence of significant rotor aerodynamic loads, platform
roll and sideside bending moment signals show much better alignment. In fact, the rotor is
yawed 90° out of the wind with feathered blades and thus contributes only marginally to side
side loads. For QB (Figure 56, right) peak loading is caused by a lowfrequency oscillation,
around the roll natural frequency. On the other hand, for DL peak loads are generated by a
much higher frequency oscillation.
FLOATECH – EUH2020 Grant Agreement N°101007142
FLOATECH D2.4: estimated reduction of uncertainty in comparison to the stateoftheart codes
76/136
Figure 56: Peak sideside tower base bending moment (Mx) for DL (left, DLC6.2 seed 10000 mis30 y90) and QB
(right, DLC6.2 seed 10003 mis30 y90). (b) Platform Roll, (c) nacelle sideside acceleration and wave elevation
(right axis, dashed line), (d) blade pitch and rotor speed (right axis, dashed lines), (e) rotor aero thrust, (f) wind
speed.