1.

Introduction to the offshore wind energy

The constant growth of the world's population and the evolution of mankind, accompanied by
the quest to improve the quality of life, have led to a steady increase in global energy demand
in recent decades. The high dependence on fossil fuels has led to several problems, such as
climate change, air pollution and the depletion of the planet's natural resources. It is therefore
necessary to search for technologies to exploit renewable energies that allow mankind to meet
its energy needs in a sustainable manner. Among them, wind energy is one of the most
attractive, as it is available everywhere on the planet with a potential far greater than human
energy demand.

1.1. Wind energy overlook

Wind energy is the most mature and developed of the renewable energies. Since ancient times,
it has been used to propel sailing ships or to move the machinery of windmills through its blades,
taking advantage of the kinetic energy of air currents. Since the beginning of the twentieth
century and based on the same principle, it has been used to generate electricity through wind
turbines. In these devices, the rotor blades, hit by the wind, cause a central axis to rotate. This
motion is transferred through a gearbox to an electrical generator which is connected to the
grid. Different configurations of wind turbine exist, with vertical or horizontal axis and different
number of blades, however the chosen configuration in most of the cases is the horizontal, three
bladed wind turbine. Based on its location, wind turbines can be either onshore or offshore [1].

Figure 1. New worldwide wind capacity predictions (2020-2030) [2]

The strong growth of the wind industry in recent years is not surprising. 2021 has been its
second-best year, with growth only 1.8% lower than 2020, despite the situation of the COVID-

1
19 pandemic. A total of 93.6 GW in new installations adds up to a total capacity of 837 GW, with
an annual growth of 12%. Unfortunately, this rate of growth is not enough to achieve a safe and
resilient energy transition: it is expected that by 2030 the installed capacity will not reach two
thirds of what is required for a 1.5ºC and net zero pathway (global carbon neutrality) set by
2050.

The onshore market added 72.5 GW worldwide, which is 18% less than the previous year, due
to the slowdown in the two largest markets, China and the United States. Conversely, Europe,
Latin America and Africa and the Middle East experimented record growth of 19%, 27% and
120%, respectively.

As for the offshore market, it experienced its best year in 2021, with 21.1 GW of installed power
worldwide, approximately three times more than the previous year. New offshore installations
account for 22.5% of new global wind installations. This brings the total offshore capacity to 57
GW, which represents 7% of the global wind capacity [2].

As far as Europe is concerned the total wind power capacity is 236 GW, 12% of which is offshore.
Germany continues to have the largest installed capacity in Europe, followed by Spain, the UK,
France, and Sweden. Figure 2 shows the growth in capacity over the period 2012-2021 in Europe.
It is expected that 116 GW of new wind farms will be installed over the period from 2022-2026.

Figure 2. The growth of total wind energy capacity in Europe, 2012-21 [3]

On 2021, 17 GW were installed, which is not even half of what the EU should be building to be
on track to deliver its 2030 Climate and Energy goals. 81% of the new installations were onshore,
with the UK accounting for most of the new offshore wind installations (figure 3) [3].

2
 Figure 3. New wind installations in Europe in 2021 per country [3]

1.2. Offshore wind energy

Wind turbines can be onshore and offshore, each of them having its own advantages. In the first
case, the advantages are mainly due to economic and logistical reasons: as they are cheaper and
more accessible, they are easier to install, maintain and repair. On the other hand, offshore wind
turbines have two major advantages: firstly, they allow higher capacity factors (the ratio
between the energy generated over a period of time and the energy generated if the turbine
had been operating at full load during the same period). The reason for this is that the offshore
wind resource is of higher quality than onshore wind. As it is not disturbed by the orography and
the irregularities of the terrain, the wind reaches higher speeds, which allow a greater extraction
of energy by making bigger and taller wind turbines. Secondly, the installation of offshore wind
farms significantly reduces the environmental impact, compared to a wind farm on land. The
further from the coast, the lesser the impact.

The first offshore wind turbine was installed in Denmark in 1991. Since then, impressive
advances have been made, both in terms of size and power. Development of new technologies
and manufacturing systems have allowed the tip heigh to grow form 100 m in 2003 (3 MW
turbine) to more than 200 m in 2016 (8 MW turbine), and the swept area to increase by 230%.
Development is now focused on 15-20 MW turbines by 2030 (figure 4). The larger the size of the
blades the larger the area swept by the turbine, allowing therefore to capture more wind and
extract more energy. As will be seen in the next chapter, the International Energy Agency (IEA)
15-MW offshore reference wind turbine will be studied in this work, with a tip height of 270 m.

3
 Figure 4. Evolution of the largest commercially available wind turbines [4]

Growth in the size of offshore wind turbines has meant a challenge for construction and
foundations, in addition to higher capital costs. However, the operation and maintenance costs
are reduced, which in the long run causes the levelized cost of energy (LCOE) to also decrease,
making offshore energy more competitive.

Another important consequence in relation to larger turbines is the increase of the capacity
factor. From 2010, when the average turbine size used in offshore wind farms was 3 MW, to
2018, when the average size was 5.5 MW, annual capacity factors increased from 38% to 43%.
Current offshore wind turbines achieve capacity factors well over 50%. Given the same site
conditions, a large turbine may have an increase in the capacity factor between 2% and 7%,
compared to smaller turbines. However, the capacity factor depends on the quality of the wind
and therefore the location, so not all new wind farm projects have to see higher capacity factors,
although the trend is for it to continue growing as suggested by the data.

In fact, the average capacity factor is a great tool to reflect the quality of the wind resource for
energy production, as it translates the wind speeds in a given area into the average performance
over the course of a year. As can be seen from figure 5, wind is strongly affected by the
geographical position, specifically the attitude. Wind resources are of higher quality near the
poles, as pointed out by the IEA Offshore Wind Outlook 2019: “In Europe, the North Sea, Baltic
Sea, Bay of Biscay, Irish Sea and Norwegian Sea, offshore wind has average annual capacity
factors of around 45-65%, which is higher than the comparable figures for the United States (40-
55%), China (35-45%), and Japan (35-45%). The capacity factor is also high in regions off the
coast of South America and New Zealand (50-65%). Moderate wind speeds resources in India
translate to a 30-40% average capacity factor. The average capacity factor in general is relatively
low in regions nearer to the equator for example in Southeast Asia and parts of western Africa”
[4].

4
 Figure 5. Average simulated capacity factors for offshore wind worldwide [4]

1.3. Floating offshore wind turbines

The downside of the offshore wind turbines is generally the cost and difficulty of installation and
maintenance, as they are less accessible. As the distance from the coast increases, and with it,
the depth of the seabed, bottom-fixed structures become more expensive.

It is estimated that 80% of the world´s offshore wind resource potential lies in waters deeper
than 60 m. However, at depths of 50 m or more, the cost of the fixed structure becomes
prohibitively expensive: in particular, for many countries, bottom-fixed is not an option as their
potential for fixed offshore wind is limited. Floating offshore wind turbines (FOWTs) are the
solution if these countries want to exploit their wind resource.

The first FOWT was installed in 2009 in Norway. After its success, the offshore industry
connected the world’s first commercial floating offshore wind project, Equinor/Masdar’s
Hywind Scotland wind farm, which used 5 SGRE 6MW turbines, in the UK in 2017. The current
largest floating offshore wind site is the 50 MW Kincardine project in Scotland which uses the
Principal Power Windfloat platform and five Vestas V164-9.5 MW turbines [5].

As illustrated below, there are four dominant types of floating wind foundations in use or
development. As will be seen in the next chapter, the study case of this work is the semi-
submersible VolturnUS-S Reference Platform, developed for the IEA Wind 15- MW Offshore
Reference Wind Turbine.

5
Figure 6. Floating offshore designs [6]

6
 2. Fatigue theory
Initially, wind turbines design was conducted using static and quasi-static analyses. This type of
analyses, at best, led to over-designed turbines, and at worst, led to premature failures. Soon it
was discovered that wind turbines are fatigue critical machines, i.e., its design is dictated by
fatigue behaviour. In the specific case of the blades, they are one of the most unique elements
of wind turbines: they must satisfy criteria of minimum weight and cost, while ensuring that last
a very high number of load cycles during their lifetime, which ranges from approximately 20 to
30 years. Turbine blades must withstand load cycles several orders of magnitude higher than
those of an aircraft, as exemplified in figure 7. Composite materials have become the material
of choice for blades after years of its study and design.

Figure 7. Schematic S-N diagram for different fatigue critical structures [7]

Fatigue can be defined as the structural response to cyclic loading. According to the ASTM E206
standard, “Fatigue is a permanent, progressive and localized phenomenon related to a structural
change in a material undergoing stresses and strains changing in time, which can lead to crack
nucleation and / or fracture after an adequate number of cycles”. [8]

Thus, the repeated application of a load can lead to failure even if the maximum load is lower
than the static resistance of the material. Other factors can affect fatigue life: high temperatures
can lower material properties; while oxidation, wear, corrosion and friction can help the
nucleation of the cracks. The fatigue failure is characterised by several stages:

• Crack nucleation due to applied stresses

• Crack propagation along a plane different from the crack nucleation and perpendicular
to the stress direction
• Failure of the material due to the cross-section reduction, no more enough to bear the
maximum load

The crack nucleation happens in elastic stress regime, around a surface defect which acts as a
stress intensification zone. Every time the stress cycles the crack propagates a defined quantity,
which leads to a decrease in the resisting cross section and consequently to an increase in the
maximum stress.

Every time the crack opens and closes it pushes the surface asperities, giving the crack surface a
smooth look. This propagation takes place until the cross section is so small that the stress

7
reaches the elastic limit and the residual resisting cross section fails due to a static failure
mechanism.

Fatigue fracture surfaces have a typical appearance: firstly, there is no plastic strain present
around the failure zone. This is because the failure mechanism is related to microstructural
changes and not to plastic strain. On the other hand, the failure surface is recognizable for having
different smooth and rough areas. The smooth appearance areas are caused by the gradual
fatigue failure and the slow crack propagation resulting in the so-called “sand lines”. The rough
appearance areas are caused by the sudden static failure, that is, the sudden crack propagation.

Figure 8. Fatigue failure [9], [10]

Generally, a distinction is made between high and low cycle fatigue. On high cycle fatigue the
loads are lower than the elastic limit and no plastic strains appear. Normally, it is referred to
when 100.000 load cycles are exceeded. On the other hand, the low cycle fatigue is when the
loads applied are larger than the elastic limit and plastic deformation occurs, and it is referred
to when less than 100.000 load cycles are reached. All fatigue considerations in this work refer
to high cycle fatigue, as wind turbine components are subjected to a very high number of load
cycles: the rotation of the rotor, aerodynamic turbulence, and other factors cause up to 109 load
alternations in the typical 20-year lifetime of a turbine. [11]

2.1. Fatigue parameters

To analyse the fatigue parameters, it is useful to consider a cyclic load applied in a fatigue test.
The stress in whatever point of the structure has a time dependency which can be expressed as
a function:

σ(t) = σ𝑚 + σ𝑎 sin⁡(𝑤𝑡)
Where:

• σ𝑚 is the mean stress

• σ𝑎 is the alternate stress
• 𝑤 is the pulsation

Generally, σ𝑚 and σ𝑎 are not constant. In simple cases they are constant values, as represented
in figure 9:

8
 Figure 9. Characterization of a cyclic load [10]

Where σ𝑚𝑎𝑥 and σ𝑚𝑖𝑛 are the maximum and minimum stress in the cycle, respectively. The
relation between the alternate, mean, minimum and maximum stress is given by:
σ𝑚𝑎𝑥 + σ𝑚𝑖𝑛
σ𝑚 =
2
σ𝑚𝑎𝑥 − σ𝑚𝑖𝑛
σ𝑎 =
2
Other parameters used to describe the fatigue loading are:

• The stress range (Δσ), given by:

Δσ = 2σ𝑎
• The stress ratio (R), which is, by definition, the ratio between the minimum and
maximum stress:
σ𝑚𝑖𝑛
R=
σ𝑚𝑎𝑥
The stress ratio is a useful parameter to characterize the loads. A typical value for the fatigue
test and curves is R = −1, which means that the load is symmetric and alternate with no mean
stress (σ𝑚 = 0). Different cases of stress ratio are illustrated in figure 10.

Figure 10. Different stress ratios [10]

9
 2.2. S-N curve
Standard tests are defined to describe fatigue behaviour and find the fatigue limit. Different
material specimens are tested under cyclic sinusoidal loading with a constant mean value. The
number of cycles to failure (N) and the alternate stress (σ𝑎 ) are recorded in a diagram known as
the Whoeler diagram (also S-N curve, in which a stress S level is required to fail the sample). A
threshold number of cycles is set to interrupt the test if specimens do not break, assigning
therefore a fatigue limit to the material. Data in this diagrams are normally quite scattered. An
example of S-N curve is shown in figure 11.

Figure 11. S-N curve [12]

The S-N behaviour of composite materials at constant R value is typically fit using one of two
equations [7]. The first is a power law of the form:

σ = 𝐶𝑁 −1/𝑚
or alternately,
1
log(σ) = log(C) − log⁡(N)
𝑚
where:

• N is the number of cycles to failure at a stress level σ

• C is a constant
• m is the fatigue exponent (slope of the S-N curve)

In nondimensional form, the equation takes the form:

σ
= C´𝑁 −1/𝑚
σ0
or
σ 1
log ( ) = log(C´) − log⁡(N)
σ0 𝑚
where:

• σ0 is the static strength of the composite

• C’ has a value of 1 when the curve fit to the S-N data set passes through the static
strength at 100 cycles (at static failure in the first fatigue cycle)

10
The second form is given by a logarithmic-linear function of the form:
1
σ=C− log(N) = C − b⁡log(N)
𝑚
or alternately,

10σ = C𝑁 −1/𝑚
Where the inverse of m is typically denoted by b. In nondimensional form, the equation takes
the following form:
σ 1
= C´ − log(N) = C´ − b log(N)
σ0 𝑚
Where C’ also has a value of 1 when the curve fit to the S-N data set passes through the static
strength at 100 cycles. Figure 12 shows the S-N curve for a typical fiberglass composite (material
wind turbine blades are based on).

Figure 12. Normalized S-N for a typical fiberglass composite, R=-1 [7]

2.3. Constant life diagrams

Most of the S-N curves are done based on tension compression tests with no mean stress (R =
−1). To take account the effects of the mean stress it is common to use a constant-life Goodman
Diagram, which represents the locus of all stress states that produce a given fatigue failure.
These curves allow to determine the effect on lifetime changes in the stress on the component
studied. A typical Goodman diagram is illustrated in figure 13.

11
 Figure 13. Typical symmetric Goodman Diagram [7]

In the vertical axis the alternate stress is represented (in this case, normalized by the ultimate
tensile strength). The horizontal axis measures the mean stress (again, normalized). It can be
seen that a constant R ratio plots as a straight line in the diagram. On the other side, the constant
life curves, bounded by the ultimate tensile strength (and ultimate compressive strength as this
case is symmetric), are plotted based on a family of S-N curves. Different constant-life curve can
be plotted: straight lines (simpler, but less accurate), straight-lines segmented, or curves. Figure
14 shows different constant-life curves:

Figure 14. Different Goodman curves [10]

12
 3. System description
The case study of this work is focused on the FOWT given by the IEA 15-MW Offshore Reference
Wind Turbine supported by the VolturnUS-S Semisubmersible Reference Platform.

On July 2020, the IEA published the technical report “Definition of the IEA Wind 15-Megawatt
Offshore Reference Wind Turbine” [13], as an addendum to IEA Wind Task 37 on Wind Energy
Systems Engineering. The report documents the design and performance of the IEA Wind 15-
MW reference wind turbine, jointly developed by the National Renewable Energy Laboratory
(NREL), Technical University of Denmark (DTU), and the University of Maine (UMaine).

After it, another following technical report was published, regarding the semisubmersible
floating platform: “Definition of the UMaine VolturnUS-S Reference Platform Developed for the
IEA Wind 15- Megawatt Offshore Reference Wind Turbine” [14].

The reference floating offshore wind turbine, which is the case study of this work, comprises a
floating semisubmersible platform, a chain catenary mooring system, a floating-specific tower,
and a rotor-nacelle assembly.

Both the IEA 15-MW offshore reference wind turbine and the VolturnUS-S reference platform
are presented below, with special regard to the turbine blades, which is where the fatigue
analysis will be conducted.

Figure 15. IEA-15-240 RWT and VolturnUS-S reference platform [13]

13
 3.1. IEA 15-MW Offshore Reference Wind Turbine
The 15 MW offshore reference turbine has a conventional three-bladed upwind design with a
rotor diameter of 240 m; a 150-m hub height; a variable-speed, collective pitch controller; and
a low-speed, direct-drive generator. The overall parameters of the turbine are shown in the
table 1.

Table 1. Key Parameters for the IEA Wind 15-MW Turbine [13]

Parameter Units Value

Power rating MW 15
Turbine class - IEC Class 1B
Specific rating W/m2 332
Rotor orientation - Upwind
Number of blades - 3
Variable speed
Control -
Collective pitch
Cut-in wind speed m/s 3
Rated wind speed m/s 10.59
Cut-out wind speed m/s 25
Design tip-speed ratio - 9.0
Minimum rotor speed rpm 5.0
Maximum rotor speed rpm 7.56
Maximum tip speed m/s 95

Rotor diameter m 240

Airfoil series - FFA-W3
Hub height m 150
Hub diameter m 7.94
Hub overhang m 11.35
Rotor precone angle deg -4.0
Blade prebend m 4
Blade mass t 65

Drivetrain - Direct drive

Shaft tilt angle deg 6
Rotor nacelle assembly mass t 1,017
deg - degrees rpm - Revolutions per minute
m - meters t – tons
m/s – meters per second W/m2 – watts per square meter

3.2. Blades
The blade length is 117 m, with a total weight of 65 t. The root diameter is 5.2 m and its
maximum chord (5.77 m) is located at approximately 20% span. The most important blade
properties are shown in table 2.

14
Table 2. Blade Properties [13]

Description Value Units

Blade length 117 m
Root diameter 5.20 m
Root cylinder length 2.34 m
Max chord 5.77 m
Max chord spanwise position 27.2 m
Tip prebend 4.00 m
Precone 4.00 deg
Blade mass 65,250 kg
Blade center of mass 26.8 m
Design tip-speed ratio 9.00 -
First flapwise natural
0.555 Hz
frequency
First edgewise natural
0.642 Hz
frequency
Design Cp 0.489 -
Design CT 0.799 -
deg - degrees kg - kilograms
m – meters Hz – Hertz

3.2.1. Blade aerodynamic properties

The blade is designed based on the DTU FFA-W3 series of airfoils, which are shown in figure 16.

Figure 16. DTU FFA-W3 airfoil family used in the IEA Wind 15-MW blade design [13]

The shape of the blade changes from the cylinder cross section at the blade root to the airfoil
shape between the 2%–15% of span, and the maximum chord of 5.77 m is given at 27.2 m of
span (23.3%). It is worth mentioning that due to the huge blade radius, the blade is designed
with a significant prebend away from the tower to provide additional tip clearance, with 4 m
separating the tip chord line from the root.

15
 3.2.2. Blade structural properties
The structural layout of the blade is composed of (figure 17):

Figure 17. Blade internal structure scheme [15]

• Spar caps: There are two of them, located in the pressure and suction side of the airfoil.
As they carry the main loads of the blade, they are made of carbon fiber, to provide as
much stiffness with as little weight as possible.
• Trailing and leading edges reinforcement made of uniaxial glass fiber to provide
additional edgewise stiffness.
• Shear webs: Two of them connect the pressure side and suction side, attached to the
main spars, extending from a 10% to 95% span.
• Foam filler panels: located between the leading-edge and trailing-edge reinforcement
and the spar caps, on both the pressure side and suction side.

The figure 18 shows the internal structure of the blade for different cross sections along the
span. Change on the shape from the cylindrical cross section to the different airfoils can be
noticed.

16
 Figure 18. Blade cross sections at different spans [13]

3.3. VolturnUS-S Reference Platform

The semisubmersible platform is designed to support the IEA-15-MW wind turbine and
comprises a four-column (three-radial and one central) steel attached to the seabed by means
of a three-line chain catenary mooring system with a vertical pretension of 6,065 kN.

The tower is specifically designed for the flotation application: floating offshore wind turbine
towers have higher stiffness requirements than fixed-bottom configurations because of the
increased inertial and gravity loads resulting from platform motion. The design in this case is an
isotropic steel tube weighting 1,263 tons.

17
The platform has a draft of 20 m with a 15-m freeboard to the upper deck of the columns and
displaces 20,206 cubic meters of seawater. Figure 19 shows a representation of the whole
assembly.

Figure 19. Platform view (left) and top-side view (right) [14]

3.4. Controller
The control system of the wind turbine is managed by the ROSCO controller [27] (Reference
Open-Source COntroller for fixed and floating offshore wind turbines), developed by the Delft
University of Technology. While common controllers are developed for specific turbines and are
difficult to modify for use on other turbines, this controller has been developed to provide a
reference from which also non-control engineers can deal with.

The objective of the controller is to maximize power modifying three control parameters: the
orientation of the rotor (yaw system), the orientation of the blades (pith system) and the torque
generated. This is achieved by two methods of actuation: to control the generator power a
variable-speed generator torque controller is used; to regulate rotor speed a collective blade
pitch controller is used.

The behaviour of the control system can be divided in four regions, as shown in figure 20:

• Region 1: corresponds to below cut-in wind speed conditions. In this region, wind speed
is too low to produce power and the generator torque is set to zero to allow the wind
to accelerate the rotor for start-up.
• Region 2: corresponds to below rated conditions. In this region, the objective is the
maximum extraction of the wind energy.
• Region 3: corresponds to above rated conditions. In this region, the power is saturated
to not damage the components.

18
 • Region 4: corresponds to above cut-out wind speed conditions. In this region, turbine
must be turned off because wind speed is too strong. Blades are pitched to reduce thrust
force to zero (feathering).

Figure 20. Controller zones [28]

Both generator torque and blade pitch controllers are PI controllers. Both modules vary
according to the conditions in which the system is. This two modules plus some additional
control modules are briefly discussed below.

3.4.1. Generator Torque Controller and blade pitch controller

For the generator torque controller, four different generator torque controllers are available:
two methods for below wind speed operations and two methods for above wind speed
conditions.

Regarding below rated operations, one of the options to maximize extracted power at each wind
condition is to follow a quadratic law of generator torque with respect to rotor angular speed.
Alternatively, a tip speed ratio tracking to maintain tip speed ratio (TSR) at its optimal value can
be adopted, for which the wind speed can be measured or estimated accurately.

Regarding the two existing methods for above rated conditions, first of them considers a
constant generator torque, while the second strategy considers a constant extracted power
equal to its rated value.

As for the blade pitch controller, its main goal is to keep the rotor angular speed at its rated
value.

3.4.2. Additional Control Modules

In addition to the main modules, ROSCO incorporates some additional modules to improve its
performance:

• Wind speed estimator: This module estimates wind sped used for TSR tracking in the
generator torque controller
• Set Point Smoothing: Near rated operations the generator torque and blade pitch
controllers normally conflict due to incompatible reference rotor speed. This module

19
 shifts the speed reference signal of the inactive controller while the active one works to
avoid this problem.
• Minimum pitch Saturation: Near rated condition thrust force reaches its highest values,
since below rated wind speed is lower and above rated condition blade pitching reduces
that force. In order to limit the loads, this module defines a minimum value of blade
pitch angle which will be used as a saturation limit during control operations.
• Floating offshore wind turbine feedback: This module introduces a new term in the PI
blade pitch controller, with the aim of finding a gains’ value that reduces rotor angular
acceleration to increase the average extracted power and stabilize the structure.

20
 4. Methodology
The objective of this work is to perform a fatigue analysis on the blades of the IEA 15 MW
offshore reference wind turbine, supported by the semisubmersible structure platform, under
realistic load cases. A brief explanation of the methodology that will be followed in the next
chapters is given here:

Firstly, aeroelastic simulations are carried out under specific cases of external conditions (wind
speeds and sea states) to obtain the internal forces of the blade. Different cross sections of the
blade are selected to retrieve the internal cross-sectional forces and moments.

Then, using a postprocessor, each section is divided in multiple elements, in which the resulting
stresses due to the state of loading are calculated. For each of these elements the fatigue
analysis is carried out, where the equivalent damage accumulated after the load cycles is
calculated. The accumulated damage is used to estimate the life of the material, which allows
to understand the behaviour of the blade under the effects of fatigue and even to vary the design
of the blade to obtain a better response.

These steps will be further explained in the next chapters. Figure 21 represents a schematic
description of the workflow followed for structural blade analysis.

Figure 21. Schematic description of the workflow for structural analysis of the wind turbine
blades [16]

Starting from the geometrical model of the blade, a pre-processing phase is necessary to
generate the beam finite elements representing the blade in the aeroelastic analysis. To do so,
a series of cross sections along the blade are analysed and its geometry and properties are
retrieved (figure 22). The resulting is a beam finite elements model which is used to represent
the blades in the wind turbine assembly inside the aeroelastic analysis tool. Finally, based on the
cross-section forces and moments resulting from the aeroelastic simulations it is possible to
recover the detailed three-dimensional stress components, which are used to perform the
fatigue analysis.

21
Figure 22. Methodology to account for 3D structural designs in 1D beam finite element models
[17]

This work covers the aeroelastic analysis and the post-processing of the results do the fatigue
analysis, as the first stages (design of the blade and pre-processing) are already defined. The
geometry, materials and internal structure of the blade are already known, and the work of the
pre-processor is mainly to compute, based on these quantities, the cross-section stiffness and
mass properties, which are then used in the aeroelastic analysis. The developers used SONATA
(Structural Optimization and Aeroelastic Analysis) [17], as the pre-processor.

For the aeroelastic analysis, Beamdyn [18] is used to get the internal forces and moments of the
cross sections. It works as a module inside OpenFAST [19].

For the Post-processing, BECAS (BEam Cross section Analysis Software) [20] was used. BECAS, a
finite element based cross section analysis tool, is a group of Matlab functions used for the
analysis of beam cross sections. It can be used either as a pre-processor, that is, for the
generation of beam finite element models in cross sections of arbitrary geometry; or as a post-
processor, to perform a stress and strain recovery of the cross-section elements, as it is used in
this work.

Wind turbine aeroelastic codes are commonly based on beam theory, as is the case for both
BeamDyn and BECAS, which allow for an accurate yet computationally inexpensive
representation of a general class of three-dimensional beam-like structures. The development
of beam models which accurately describe the behaviour of the blades is challenging, as modern
wind turbine blades feature complex geometry and a mix of different anisotropic materials.
Although 3D finite element models (shell or solid models) provide accurate results, they are
computationally expensive. On the other hand, beam models achieve a great level of accuracy
while being more computationally efficient (figure 23).

22
Figure 23. Beam model (up) and shell model (down) of a wind turbine blade [16]

23
 5. Aeroelastic simulations

5.1. Understanding BeamDyn

The objective of the simulations is to retrieve the internal forces and moments of the blade
under a realistic set of load cases. Also, the deflection of the blade will be analysed to better
understand its behaviour. To do so, BeamDyn is used coupled into OpenFAST as the software
for the aeroelastic simulations. These tools allow to simulate the wind turbine in a dynamic time-
domain scenario. OpenFAST (previously called FAST v8) is a computer-aided engineering tool for
simulating the coupled dynamic response of wind turbines. It joins:

• Aerodynamics models, using wind-inflow data to compute the blade aerodynamic loads
(AeroDyn module)
• Hydrodynamics models, which simulate the incident waves (HydroDyn module)
• Control and electrical system dynamics models, to simulate the controller logic
• Structural dynamics models, which simulate the elasticity of the blades and support
structure applying the aerodynamic, hydrodynamic and gravitational loads (BeamDyn
and ElastoDyn modules)

There are modules to account also for the mooring system (MoorDyn), the tower and structure
motion (ElastoDyn) and wind distributions generation (Turbsim). The coupling between all
models is achieved through a glue code.

BeamDyn is based in the geometrically exact beam theory (GEBT), which supports full geometric
nonlinearity and large deflection (with bending, torsion, shear, and extensional degree-of-
freedom), anisotropic composite material couplings and a reference axis that permits blades
that are not straight, as it is the case of the 15 MW wind turbine blade. Figure 24 represents the
workflow of BeamDyn when coupled with OpenFAST (BeamDyn can also work in a stand-alone
version which will not be dealt with in this work).

Figure 24. Coupled interaction between BeamDyn and FAST [21]

24
Loads are transferred between the different modules. From other modules, BeamDyn receives:

• The blade root node displacements (three translations and three rotations)
• The blade root node velocities (three translational and three rotational)
• The blade root node accelerations (three translational and three rotational)

With this variables, BeamDyn computes:

• The six reaction forces at the root (three forces and three moments)
• The blade displacements, velocities, and accelerations along the beam length

These displacements are then used in AeroDyn to compute the aerodynamic loads distributed
along the length of the blade, which are used in turn as input to BeamDyn to calculate the
internal reaction loads of the blade.

The output values can be requested by the user at a maximum of nine nodes among the
quadrature points (QPs), which are the points that BeamDyn uses internally to perform the
calculations and derive from the finite element nodes (FE nodes). These points range from the
blade root to the blade tip.

5.2. Input files

Two input files are used as input to BeamDyn to perform the simulations, which are explained
here.

5.2.1. Primary input file

Named “IEA-15-240-RWT_BeamDyn”, in this file the blade geometry is defined through a set of
key-point (kp) coordinates and initial twist angles. Each key point is defined by three physical
coordinates (“kp_xr”, “kp_yr” and “kp_zr”) in the IEC standard blade coordinate system,
where Zr is along blade axis from root to tip, Xr directs normally toward the suction side,
and Yr directs normally toward the trailing edge. The blade reference coordinate system is
represented in figure 25.

Figure 25. BeamDyn Blade Geometry - Top: Side View; Middle: Front View (Looking Downwind);
Bottom: Cross Section View (Looking Toward the Tip, from the Root) [21]

For each key point, a structural twist angle (“initial_twist” variable) is given. The structural twist
angle is also following the IEC standard which is defined as the twist about the negative Zl axis.

25
A total of 50 key points are used to define to define the geometry of the blade of the 15 MW
wind turbine. Table 3 shows the key points and their respective initial twists.

Table 3. Key points defining the blade geometry in the Reference Blade Coordinate system

kp_xr (m) kp_yr (m) kp_zr (m) Initial twist (deg)

0.00E+00 0.00E+00 0.00E+00 1.56E+01
1.84E-02 0.00E+00 2.39E+00 1.56E+01
4.43E-02 0.00E+00 4.78E+00 1.53E+01
7.04E-02 0.00E+00 7.16E+00 1.49E+01
1.03E-01 0.00E+00 9.55E+00 1.42E+01
1.35E-01 0.00E+00 1.19E+01 1.34E+01
1.65E-01 0.00E+00 1.43E+01 1.23E+01
1.95E-01 0.00E+00 1.67E+01 1.13E+01
2.19E-01 0.00E+00 1.91E+01 1.03E+01
2.39E-01 0.00E+00 2.15E+01 9.37E+00
2.47E-01 0.00E+00 2.39E+01 8.57E+00
2.50E-01 0.00E+00 2.63E+01 7.85E+00
2.49E-01 0.00E+00 2.87E+01 7.19E+00
2.47E-01 0.00E+00 3.10E+01 6.56E+00
2.45E-01 0.00E+00 3.34E+01 5.93E+00
2.41E-01 0.00E+00 3.58E+01 5.36E+00
2.37E-01 0.00E+00 3.82E+01 4.80E+00
2.32E-01 0.00E+00 4.06E+01 4.31E+00
2.25E-01 0.00E+00 4.30E+01 3.84E+00
2.09E-01 0.00E+00 4.54E+01 3.45E+00
1.86E-01 0.00E+00 4.78E+01 3.08E+00
1.45E-01 0.00E+00 5.01E+01 2.74E+00
9.73E-02 0.00E+00 5.25E+01 2.42E+00
4.10E-02 0.00E+00 5.49E+01 2.11E+00
-2.32E-02 0.00E+00 5.73E+01 1.83E+00
-8.97E-02 0.00E+00 5.97E+01 1.56E+00
-1.60E-01 0.00E+00 6.21E+01 1.30E+00
-2.44E-01 0.00E+00 6.45E+01 1.07E+00
-3.35E-01 0.00E+00 6.69E+01 8.42E-01
-4.49E-01 0.00E+00 6.92E+01 6.38E-01
-5.69E-01 0.00E+00 7.16E+01 4.37E-01
-6.98E-01 0.00E+00 7.40E+01 2.40E-01
-8.33E-01 0.00E+00 7.64E+01 3.61E-02
-9.70E-01 0.00E+00 7.88E+01 -1.73E-01
-1.11E+00 0.00E+00 8.12E+01 -4.16E-01
-1.26E+00 0.00E+00 8.36E+01 -6.89E-01
-1.41E+00 0.00E+00 8.60E+01 -1.00E+00
-1.58E+00 0.00E+00 8.83E+01 -1.32E+00
-1.74E+00 0.00E+00 9.07E+01 -1.62E+00
-1.92E+00 0.00E+00 9.31E+01 -1.88E+00
-2.10E+00 0.00E+00 9.55E+01 -2.06E+00
-2.29E+00 0.00E+00 9.79E+01 -2.16E+00
-2.48E+00 0.00E+00 1.00E+02 -2.18E+00
-2.67E+00 0.00E+00 1.03E+02 -2.15E+00
-2.88E+00 0.00E+00 1.05E+02 -2.10E+00
-3.09E+00 0.00E+00 1.07E+02 -2.02E+00
-3.31E+00 0.00E+00 1.10E+02 -1.89E+00
-3.54E+00 0.00E+00 1.12E+02 -1.71E+00
-3.77E+00 0.00E+00 1.15E+02 -1.48E+00
-4.00E+00 0.00E+00 1.17E+02 -1.24E+00

26
Notice that the key points are entered sequentially (from the root to tip) in accordance with the
Z coordinate, which ranges from 0 m (blade root) to 117 m (blade tip).

In addition to the blade geometry, the following parameters are specified in this file, which are
necessary for the simulations:

• Quadrature: this value specifies the spatial numerical integration scheme, between
Gauss quadrature and trapezoidal quadrature. Trapezoidal quadrature is selected.
• Refine factor: this value is used in trapezoidal quadrature and splits the space between
two input stations into “Refine factor” of segments. An increment in the refinement
factor multiplies the number of quadrature points for a more precise calculation,
however, simulation times are considerably longer. For this reason, a refine factor value
of two is selected.
• Number of members: the blade beam model is composed of several “members” in
contiguous series. As specified in the BeamDyn user guide: “with the LSFE discretization,
a single member and a sufficiently high element order, may well be sufficient. […] With
the LSFE discretization, an increase in accuracy will, in general, be better achieved by
increasing the element order (i.e., p-refinement) rather than increasing the number of
members (i.e., h-refinement). For Gauss quadrature, the element order should be
greater than one” [21]. The element order refers to the order of shape functions for
each finite element, which in this case is set to five. For this reason, the number of
members is set to one.

5.2.2. Blade input file

Named “IEA-15-240-RWT_BeamDyn_blade”, this file defines the cross-sectional properties
(inertial and structural) at various stations along the blade. The number of stations is changed
through the variable “Station_Total” which is set to twenty-one in this case. This results in a total
of forty-one QPs.

Each station is defined trough a non-dimensional parameter which specifies its location along
the local blade reference axis ranging from [0.0, 1.0], referring the first to the root and the
second to the tip, respectively. In each station the stiffness matrix and the mass matrix of the
cross section are given, defined in the local coordinate system along the blade axis.

The stiffness matrix has the following form:

K ShrFlp 0 0 0 0 0
0 K ShrEdg 0 0 0 0
0 0 EA 0 0 0
0 0 0 EIEdg 0 0
0 0 0 0 EIFlp 0
[ 0 0 0 0 0 GJ]
Where:

• KShrEdg is the edge shear stiffnesses

• KShrFlp is the flap shear stiffnesses
• EA is the extension stiffness
• EIEdg is the edge stiffnesses
• EIFlp is the flap stiffnesses

27
 • GJ is the torsional stiffness

The mass matrix is given by:

𝑚 0 0 0 0 −mYcm
0 𝑚 0 0 0 mXcm
0 0 𝑚 mYcm −mXcm 0
0 0 mYcm iEdg −icp 0
0 0 −mXcm −i𝑐𝑝 iFlp 0
[−mY cm mYcm 0 0 0 iplr ]

Where:

• m is the mass density per unit span

• Xcm and Ycm are the local coordinates of the sectional center of mass
• iEdg is the edge mass moments of inertia per unit span
• iFlp is the flap mass moments of inertia per unit span
• iplr is the polar moment of inertia per unit span
• icp is the sectional cross-product of inertia per unit span

The cross-section matrices of the twenty-one sections are already known data.

Table 4 summarizes the variables used in the primary input file and the blade input file for the
simulations:

Table 4. Input files variables

Variable
Value Meaning
name
quadrature 2 Trapezoidal integration method
refine 2 Splits the space between two input stations into two segments
member_total 1 Blade beam model composed of one member
order_elem 5 Order of interpolation (basis) function
kp_total 50 Number of key points used to define the blade geometry
Number of sections at which the cross-sectional properties are
Station_Total 21
defined

5.2.3. Other input parameters

In addition to the BeamDyn input files, some more files are needed by the other modules of
OpenFAST to set the parameters of the simulation:

• Glue code: In the glue code file, named “IEA-15-240-RWT-UMaineSemi.fst”, the

integration time step (DT) can be changed. A smaller time step allows for a more precise
calculation but increases the computational time. It is necessary to find the balance
between the two, for which several test simulations have been carried out, which are
explained below. Final time step selected for the simulations is 0,0005 s.
• AeroDyn: The file for this module is called “IEA-15-240-RWT_AeroDyn15” and allows to
set the model of aerodynamics which will be used to compute the loads on the blade.
• Inflow file: In this module the wind speed can be changed, as well as the type of wind
condition (steady or turbulent). All the simulations in this work use turbulent wind
conditions in order to get more realistic result. The wind speeds are specified below.

28
 • Turbsim: With this module it is possible to generate turbulent wind distributions with
de desired mean wind speed, plus modifying more characteristics on the incident wind.
• HydroDyn: In this module it is possible to set the model of the waves (still or non-still
water) as well as the wave significant height (m) and peak period of the waves.
• ElastoDyn: In this module, the degrees of freedom (DOF) of the platform can be
controlled (fixed or not) as well as the initial blade pitch angle. The precone (-4°) and tilt
(-6°) angles of the wind turbine are fixed here also.

The simulations are run with the help of a matlab script in which the simulation time can be set.

5.3. Setting the simulation parameters

Given that the simulations take a long time and to better understand how some input
parameters affect the simulation results, some test simulations are first run.

The simulation time is set to 120 seconds (which approximately takes half an hour of running
time). This is a small amount of time and is not representative of real situations, but sufficient
for testing purposes. The integration time step is set to 0.0005 seconds in the glue code. Latter
verifications will make sure that the precision with this integration time is enough. Also, the
effect of the waters is neglected and the still water model is used.

To get representative results, the following three wind speeds are chosen:

• 10 m/s
• 14 m/s
• 18 m/s

All of them have been generated using ‘Turbsim’, which creates turbulent wind distributions
with the mean values above mentioned, referred to the hub height. The wind speed
distributions are shown in figure 26.

Figure 26. Wind speed distributions for 10, 14 and 18 m/s, 120s

The table 5 summarizes the parameters adopted for the testing simulations:

29
Table 5. Variables for the test simulations

Variable Value(s) Units

Simulation Time (T) 120 s
Integration time step (dT) 0.0005 s

10
Wind speed (turbulent) 14 m/s
18

Wave conditions Still water -

Among the forty-one quadrature points, which are the points of the blade in which BeamDyn
performs the calculations, BeamDyn can output the data at a maximum of 9 nodes. The position
vector of the quadrature points is specified in the BeamDyn summary file after running the
simulation and is exposed below (table 6). Note that they cover the whole blade span, ranging
from the blade root (Z = 0 m) to the blade tip (Z = 117 m).

Table 6. Quadrature points position vector in the reference blade coordinate system

QP # X (m) Y (m) Z (m)

1 0.00E+00 0.00E+00 0.00E+00
2 2.75E-02 0.00E+00 2.93E+00
3 5.80E-02 0.00E+00 5.85E+00
4 9.01E-02 0.00E+00 8.78E+00
5 1.23E-01 0.00E+00 1.17E+01
6 1.55E-01 0.00E+00 1.46E+01
7 1.85E-01 0.00E+00 1.76E+01
8 2.12E-01 0.00E+00 2.05E+01
9 2.36E-01 0.00E+00 2.34E+01
10 2.54E-01 0.00E+00 2.63E+01
11 2.67E-01 0.00E+00 2.93E+01
12 2.74E-01 0.00E+00 3.22E+01
13 2.73E-01 0.00E+00 3.51E+01
14 2.65E-01 0.00E+00 3.80E+01
15 2.48E-01 0.00E+00 4.10E+01
16 2.21E-01 0.00E+00 4.39E+01
17 1.85E-01 0.00E+00 4.68E+01
18 1.38E-01 0.00E+00 4.97E+01
19 8.08E-02 0.00E+00 5.27E+01
20 1.24E-02 0.00E+00 5.56E+01
21 -6.75E-02 0.00E+00 5.85E+01
22 -1.59E-01 0.00E+00 6.14E+01
23 -2.62E-01 0.00E+00 6.44E+01
24 -3.78E-01 0.00E+00 6.73E+01
25 -5.05E-01 0.00E+00 7.02E+01
26 -6.45E-01 0.00E+00 7.31E+01
27 -7.96E-01 0.00E+00 7.61E+01
28 -9.60E-01 0.00E+00 7.90E+01
29 -1.13E+00 0.00E+00 8.19E+01
30 -1.32E+00 0.00E+00 8.48E+01
31 -1.52E+00 0.00E+00 8.78E+01
32 -1.73E+00 0.00E+00 9.07E+01
33 -1.95E+00 0.00E+00 9.36E+01

30
 34 -2.18E+00 0.00E+00 9.65E+01
35 -2.41E+00 0.00E+00 9.95E+01
36 -2.66E+00 0.00E+00 1.02E+02
37 -2.92E+00 0.00E+00 1.05E+02
38 -3.18E+00 0.00E+00 1.08E+02
39 -3.45E+00 0.00E+00 1.11E+02
40 -3.72E+00 0.00E+00 1.14E+02
41 -4.00E+00 0.00E+00 1.17E+02

The output nodes (the nodes where the results are gotten) are specified in the BeamDyn input
file. They are selected from the root to the tip as evenly distributed as possible (note that the
quadrature points are not evenly distributed along the span). The correspondence between the
output nodes (N) and the quadrature points (QP) and its coordinates are specified in the table
7.

Table 7. Output nodes position

Node # QP # Non-dimensional span Span (m)

1 1 0 0.0000
2 6 0.125 14.6250
3 11 0.25 29.2500
4 16 0.375 43.8750
5 21 0.5 58.8000
6 26 0.625 73.1250
7 31 0.75 87.7500
8 36 0.875 102.3750
9 41 1 117.0000

To better understand the results, the translational DOF of the tip (motions on X, Y and Z axes)
are presented in figures 27, 28 and 29 respectively:

Figure 27. Tip deflection on X axis for 10, 14, 18 m/s, 120s

31
 Figure 28. Tip deflection on Y axis for 10, 14 and 18 m/s, 120s

Figure 29. Tip deflection on Z axis for 10, 14 and 18 m/s

The effects of the deflection are much higher on the X axis (0, 13 m) than in the Y (-2.5, 1.5 m)
and Z (-1.2, 0.2 m) axes. The first instants are dominated by oscillations as the initial blade pitch
is set to 0° and the controller adjust it as can be seen on figure 30. Counter-intuitively, (ignoring
therefore the transient time at the beginning of the simulation) the deflection on the X axis is
slightly bigger for the 10 m/s case and decreases with the windspeed. This is due to the action
of the controller, which increases the blade pitch for higher wind speeds. If the controller was
switched off, the expected results would be the opposite of those obtained.

32
 Figure 30. Pitch angle for 10, 14 and 18 m/s

In the Y and Z directions the deflection is one and two orders of magnitude smaller than in the
X direction, respectively. It is observed that in these axes the deflection increases slightly with
wind speed.

The deflection on X axis (as it is the one that best represents the blade behaviour) is now
presented for the nine output nodes (figure 31):

33
 Figure 31. Deflection in nodes 1-9 for 10, 14 and 18 m/s, 120s

Where the first and ninth nodes correspond to the root and the tip of the blade, respectively.
As expected, the deflection is maximum at the tip and decreases as the span decreases, until it
reaches zero displacement at the root. The shape of the deflection of the nodes as the z-
coordinate increases is similar.

Having these data, it is possible retrieve the maximum deflection of the blade. This happens for
the 10 m/s windspeed case, where the deflection (X axis) reaches 13,24 m at its maximum point,
which happens at t = 25.9 s. This point is represented in figure 32.

34
 Figure 32. Maximum deflection, 10m/s and 120s

With the deflection of the nodes, a matlab script is created to plot the three-dimensional
deflexion of the blade, for every time step. With it, is possible to see how the blade changes
shape over time. The 3D deformed and undeformed shape of the blade at the instant of
maximum deflexion is represented in figure 33, whereas the deflexion in the X and Y axes at
the same moment are plotted in figure 34.

Figure 33. Three-dimensional deflexion, 10m/s and 120s

35
 Figure 34. Maximum deflection on X (left) and Y (left) axes, 10m/s and 120s

To verify that the chosen time step is sufficiently accurate, another 30-second simulation is
performed with a smaller time step of 0.0002s. The results are compared with the previous
results and the tip deflection is shown in figure 35.

Figure 35. Time step comparison of the tip deflection, between 0.0005s and 0.0002s

The deflection for the smaller time step (0.002s, in orange), is the same as the bigger one
(0.0005s in blue) as both graphs overlap, so it is possible to conclude that a time step of 0.0005s
is small enough to get accurate results. It is noted that the simulations with the smaller time
step have taken much longer time to complete (about 5 hours for just 30 seconds of simulation
time), while the one with higher time step achieved the same accuracy while running much
quicker.

36
Bigger time steps have also been tested: convergence is not reached for time steps bigger than
0.0006s. In any case, for values smaller the results are consistent. For this reason, the final time
step chosen to perform the simulations is 0.0005s.

5.4. Design load case

5.4.1. Site environmental conditions

In order to carry out a realistic and representative fatigue analysis and to take account of the
external forces on the blades, a location has to be chosen and the environmental conditions
from this location have to be used in the simulations. The variables involved are:

• Wind speed (V), in meters per second

• Significant height of the waves (Hs), in meters
• Peak period of the waves (Ts), in seconds

In this case the same data as in [22] is used (Sirigu, et al, 2022), as the results are to be compared.
As explained by Sirigu, et al, 2022, the real scatter of the wind, height and period of the waves
is obtained from the shore of Pantelleria, Italy in ten years of data acquisition. The occurrence
of the wind speed in Pantelleria based on these data is shown in figure 36.

Figure 36. Occurrence of the wind speed in Pantelleria, Italy in ten years of data collection [22]

The standard IEC 61400 [23] describes the design load cases for offshore wind turbines. The life
of a wind turbine can be represented by a set of design scenarios that cover the most significant
conditions it may experience throughout its lifetime. These situations include normal design
conditions, but also transport, installation and maintenance conditions. For all of them, the
standard describes a set of design load conditions (DLC) which are divided into:

• Power generation (DLC 1)

• Power generation plus fault occurrence (DLC 2)
• Start-up (DLC 3)
• Normal shut-down (DLC 4)

37
 • Emergency stop (DLC 5)
• Parking (stationary or idling) (DLC 6)
• Parking and fault conditions (DLC 7)
• Transport, assembly, maintenance and repair (DLC 8)

The load case DLC 1.2 (power production in normal turbulence) is used for the fatigue analysis.
This load case is used for fatigue assessment and consists of the simulation of power production
without faults performed for wind speeds in the entire operational range with normal
turbulence (Vin < Vhub < Vout) and a normal sea state. The wind and wave states are evaluated
according to a joint probability distribution and each wind speed (Vhub) is assigned a wave state
(Hs, Tp). The wind speed is discretized in intervals of 0,25 m/s, the wave height in intervals of
0,25 m and the peak period in intervals of 0,25 s.

Table 8 summarises the environmental conditions for setting up the simulation:

Table 8. Wind and sea states used in the simulations

V (m/s) Hs (m) Ts (s)

5 1.2 6.24
7 1.37 6.25
9 1.59 6.20
11 1.89 6.24
13 2.22 6.38
15 2.64 6.62
17 3.19 7.06
19 3.72 7.38
21 4.31 7.85
23 4.71 7.95

Resulting in ten different simulations to be performed. The simulation time is set to 600s (10
min) as recommended by [24].

5.4.2. Applied forces on the blade

Each of the conditions described above produce different forces applied on the blades. With
MOST [25], these forces can be obtained through a lookup table in Matlab. For example, for a
wind speed of 15 m/s, the following normal and tangential pressures (pn and pt, respectively)
act on the blade (figures 37 and 38):

38
 Figure 37. Normal pressure (pn) along blade length

Figure 38. Tangential pressure (pt) along blade length

Where both pn and tn are zero at the root and blade tip and are calculated according to:
1
𝑝𝑛 = 𝜌𝑣 2 𝐶𝑛 𝑙
2
1
𝑝𝑡 = 𝜌𝑣 2 𝐶𝑡 𝑙
2
In every section along the span, where:

• 𝜌 is the air density (1.225 kg/m3)

• 𝑣 is the wind speed
• 𝐶𝑛 is the normal drag coefficient
• 𝐶𝑡 is the tangential drag coefficient
• 𝑙 is the chord of the section

Integrating the pressures along the length of the blade gives the external forces, Fx and Fy
(figures 39 and 40) and moments, Mx and My (figures 41 and 42):

39
 Figure 39. Fx along the blade length

Figure 40. Fy along the blade length

Where it can be seen that both Fx and Fy decrease monotonically from the root of the blade to
the tip. Note that Fx is an order of magnitude larger than Fy. Fx and Fy are calculated according
to:

𝐹𝑥 = ∫ 𝑝𝑛 (𝑧)𝑑𝑧

𝐹𝑦 = ∫ 𝑝𝑡 (𝑧)𝑑𝑧

Where z is the blade span.

40
 Figure 41. Mx along the blade length

Figure 42. My along the blade length

The external moments Mx and My are calculated according to:

𝑀𝑥 = ∫ 𝑝𝑛 (𝑧)⁡𝑧𝑑𝑧

𝑀𝑦 = ∫ 𝑝𝑡 (𝑧)⁡𝑧𝑑𝑧

In this case, My is predominant over Mx. Again, both decrease monotonically from the root to
the tip.

41
 5.5. Simulation results

In this section the results obtained in the simulation are presented. Table 9 summarizes the main
parameters used in the simulation.

Table 9. Simulation inputs

Variable Value(s) Units V (m/s) Hs (m) Ts (s)

Simulation Time (T) 600 s 5 1.2 6.24
Integration time step (dT) 0.0005 s 7 1.37 6.25
Output nodes 9 - 9 1.59 6.20
11 1.89 6.24
13 2.22 6.38
15 2.64 6.62
17 3.19 7.06
19 3.72 7.38
21 4.31 7.85
Total number of simulations: 10 23 4.71 7.95

Each simulation takes between two to three hours to complete. For each wind speed and sea
state, a structure in Matlab is created to store the results. Each structure contains the following
arrays of values in time:

• Time step of the simulation

• Windspeed
• Power
• Rotor speed
• DOF of the platform (surge, sway, heave, roll, pitch and yaw)
• Blade pitch
• DOF of the blade tip
• DOF of the blade nodes
• Internal forces at the blade root (forces and moments)
• Internal forces at the blade nodes (forces and moments)

Each of them with a length of 1.200.000 time-steps (corresponding to 600s/0,0005s). The results
of the simulation are shown below, disregarding the first ten seconds of simulation to overcome
the transient effects. For clarity purposes, only the results of three different wind speeds, 7, 15,
and 21m/s are showed, given that showing all the results at once does not allow to appreciate
them properly. The generated wind speeds are shown in figure 43.

42
 Figure 43. Wind speed

5.5.1. Power
The power developed by the wind turbine is showed in figure 44, during the 600s of simulation.
For 15 and 21m/s wind speeds, the power in centred around the rated power which is 15 MW.
On the other hand, for the case of 7 m/s the extracted power is clearly lower since the wind
speed is less than the nominal wind speed (10.59 m/s).

Figure 44. Power

5.5.2. Rotor speed

The rotor speed is showed in figure 45, during the 600s of simulation. For 15 and 21m/s wind
speeds, the rotor speed in centred around its maximum one which is 7,56 rpm. Regarding the
7m/s case, the rotor speed is centred around its minimum value which is 5 rpm.

43
 Figure 45. Rotor speed

5.5.3. DOF of the platform

Figure 46 shows the DOF of the platform: three translational (surge, sway and heave) and
three rotational (roll, pith and yaw).

44
 Figure 46. DOF of the platform

As can be seen between the translational DOF, the amplitude of the oscillations of the surge is
higher than the ones on the sway and the heave, reaching almost 25 meters. This is because the
surge axis is aligned with the direction of the waves and therefore the oscillations on this axis
are bigger.

The effect of the waves can be clearly seen in the case of the highest wind speed, 21m/s, where
the vertical oscillations (heave) exceed half a metre. In contrast, for lower waves (and therefore
lower wind speeds), the oscillations are less than half this magnitude.

Concerning the rotational DOF, the pitch presents higher oscillations as it axes is perpendicular
to the direction of the waves, reaching almost 4°. The yaw presents oscillations of the same
order of magnitude while the roll oscillations reach a maximum if 1,5°.

5.5.4. Blade pitch angle

Figure 47 shows the pitch angle of the blade during the 600s of simulation.

Figure 47. Blade pitch angle

As explained above in the test simulations, the higher the wind speed, the higher the pitch angle
set by the controller, which is consistent with the results. For the 7m/s case the pitch is

45
practically zero, while for the 15 and 21 m/s cases the pitch angle ranges between 10 and 17
degrees, respectively.

5.5.5. DOF of the blade tip

Figure 48 presents the translational deflections of the blade tip along X axis. Figure 49 presents
the translational deflections of the blade tip along Y and Z axes. Y and Z motions have been
adjusted on time axis for better visualization.

Figure 48. Tip deflection along X

As seen before, the average deflection is higher for medium wind speeds. This is because at low
speeds, the wind thrust force is small and causes little deflection, while at high speeds the
controller increases the pitch angle and the deflection decreases. The intermediate case (15 m/s,
orange) is the one with the highest deflection, as the wind speed is considerable but the pitch
angle is not so large. On the other hand, as the wind speed increases, the oscillations increase:
the 7 m/s case (blue) presents oscillations between [5, 10] m; while the 21 m/s case presents
oscillations between [-8, +10] m.

Figure 49. Tip deflection along Y (left) and Z (right)

As for the other translational deflections, their effects are, in general, much smaller. On the Y
axis the mean deflections and the oscillations are similar for the different wind speeds and range

46
from [-2.5, 2] m approximately. On Z axis the mean deflection is small for all the cases, however
for greater wind speeds it presents negative oscillations up to -1 m.

The maximum deflection between all the case studied is found to happen at a wind speed of 13
m/s, where the maximum deflection reached along X axis is 14,785 m relative to the undeflected
position of the blade. Figure 50 shows a 3D representation of the deflected and undeflected
blade at this exact moment. Figure 51 shows the same instant but referred to X and Y axes.

Figure 50. Maximum deflection. Wind speed is 13 m/s

Figure 51. Maximum deflection along X and Y. Wind speed is 13 m/s

5.5.6. DOF of the blade nodes

Figure 52 presents the translational deflections along X axis of the nine nodes of the blade for
the 15 m/s case. Each coloured line represents a node of the blade: the node one coincides with
the blade root and the node nine coincides with the tip. Figure 53 presents the translational
deflections along Y and Z axis.

47
 Figure 52. Nodes deflection along X. Wind speed is 15 m/s

It can be clearly seen that the shape of the deflection of all the nodes is the same and increases
as the blade span increases. For node 1, at the root, the deflection is zero or practically zero as
the blade is cantilevered to hub. The deflection increases at the different nodes until it reaches
the ninth node, at the tip which has no external motion restrictions.

Figure 53. Nodes deflection along Y (left) and Z (right). Wind speed is 15 m/s

As for the deflection in the Y and Z axes, the trend is the same. The deflection is zero at the root
and increases along the blade until it reaches the tip, where it becomes maximum. The shape of
all the nodes is the same.

5.5.7. Internal forces at the blade root

Figure 54 presents the internal forces and moments in the root section of the blade. The forces,
expressed in Newtons (N), are the edgewise force (or the internal force along X axis), flapwise
force (or internal force along Y axis) and vertical force (or internal force along Z axis, that is the
direction of the blade span). The moments, expressed in Newton-meters (Nm), are the edgewise
moment (or internal moment along X axis), flapwise moment (or internal moment along Y axis)
and the torque moment (or internal moment along Z axis).

48
 Figure 54. Blade root reaction forces

The edgewise, flapwise and vertical forces are of the same order of magnitude. The edgewise
force exhibits larger oscillations and higher mean values as the wind speed increases. It generally
oscillates between the values [0, 10×105] N for the higher wind speed values. As for the flapwise
force, the oscillations are similar for all cases, generally in the range [-6×105, 6×105] N, and
centred at the origin. Finally, the vertical force is where the highest values are reached, close to
20×105 N. In this case, the higher the wind speed, the higher the average vertical force (see how
in the figure, the vertical force in the case of 7m/s is lower than that of 15 m/s, which in turn is
lower than that of 21 m/s). As for the oscillations, they are all of similar intervals ([-2, 12] N for
the 7 m/s cade, and [5, 20] N for the 15 and 21 m/s cases, approximately).

The last three graphs of the figure show the bending moments (flapwise and edgewise) and the
torque moment. The edgewise moment is the one that gives torque to the rotor, while the

49
flapwise moment is the one generated by the wind force. Both are orders of magnitude larger
(107 Nm) than the torque moment (105 Nm). The edgewise moment has similar oscillations for
different wind speeds, centred at the origin and in the range of approximately [-2×107, 2×107]
Nm. In the flapwise moment, the amplitude of the oscillations increases with wind speed,
reaching peak values of 5×107 Nm. For lower speeds, the curve is more constant and changes
less abruptly. Finally, the torque moment generally presents oscillations of more moderate
amplitude, in the range of [-2×105, 2×105] Nm and centred at the origin.

5.5.8. Internal forces at the blade nodes

Figure 55 shows the internal sectional forces (Fx, Fy and Fz) in Newtons, and moments (Mx, My
and Mz) in Newton-meters of the nine different nodes along the blade length. Each coloured
line represents a node of the blade: the node one coincides with the blade root and the node
nine coincides with the tip. The wind speed for the case shown is 15 m/s.

50
 Figure 55. Internal forces at the blade nodes. Wind speed is 15 m/s

The shape of the internal forces for all the nodes is the same and decreases with the blade span.
For node 1 at the blade root, the forces and moments reach the maximum values as is where
the blade is cantilevered. For node 9, all the forces cancel out, as it is the end of the blade.

51
 6. Fatigue analysis
In this section the fatigue analysis methodology is explained and applied. Firstly, on the root of
the blade, and finally on several sections distributed along the blade.

6.1. Fatigue analysis on the blade root

The fatigue analysis on the blade root can be done following a simplified approach due to its
simple geometry and composition: a circular cross-section with a thickness of 100 mm, made
with a traditional triaxial fiberglass composite. The methodology followed in this analysis is the
same approach used by Sirigu, et al, 2022 [22], as results are to be compared.

The root section is shown in figure 56 and its dimensions and material properties are listed on
table 10.

Figure 56. Blade root cross-section

Table 10. Geometry of the blade root cross-section and properties of the composite material [22]

Composite name Vectorply E-LT-5500 (triaxial fiberglass)

Density 1960 kg/m3
Ultimate tensile strength Sut 396 MPa
Fatigue parameter m
39.6 MPa
(Slope of S-N curve)
External diameter 5.2 m
Thickness 100 mm
Area section 1.6022 m2
Inertia section 5.2112 m4

52
 6.1.1. Stress recovery
The first step to perform the fatigue analysis on the blade root is to retrieve the stress on
different points along the section, on the basis of the internal forces and moments that have
been obtained from the simulations. The fatigue analysis only considers the uniaxial load on the
blade, caused by the bending moments (Mx and My) due to the aerodynamic and the
gravitational forces, and the vertical force (Fz) caused by the gravity and the centrifugal forces.
Thus, the internal shear forces (Fx and Fy) are neglected, as well as the torsional moment (Mz).

The resulting axial stress or mechanical stress derived from these forces is computed as:
𝑀𝑥 ∙ 𝑅 ∙ cos⁡(𝛼 + 𝛽) 𝑀𝑦 ∙ 𝑅 ∙ sin⁡(𝛼 + 𝛽) 𝑁
σ= + +
𝐽 𝐽 𝐴
Where:

• Mx is the edgewise moment, i.e., the moment that gives torque to the rotor
• My is the flapwise moment, i.e., the moment caused by the wind thrust force
• N is the axial force (or vertical force, Fz)
• A is the area of the cross section (see table 10)
• J is the inertia of the cross section (see table 10)
• R is the external radius of the blade root circumference
• α is the blade pitch angle
• β is the angle along the blade root circumference

The stress is evaluated on the external circumference of the blade root (as it is where the stress
due to the bending moments is grater), at several points with a discretization of 10 degrees (β
angle). The resulting mechanical stress is shown on figure 57 for the whole simulation time
(except the first ten seconds, as they have been disregarded to ignore the transient) and a closer
look is shown on figure 58. The stress is shown in the point β = 90°, which, as will be seen later,
is the most critical one in the cross section).

Figure 57. Mechanical stress at the blade root. β = 90° and wind speed is 15 m/s

53
 Figure 58. Closer look to the mechanical stress at the blade root, β = 90° and wind speed is 15
m/s

6.1.2. Damage and life calculation

The next step on the fatigue analysis is to compute the damage equivalent load and the expected
fatigue life on the blade section. To do so, the theory explained in the second chapter will be
used.

Starting from the mechanical stress, a Rainflow algorithm is used, according with the ASTM E
1049, to calculate the number of cycles and the alternate and mean stress (σ𝑎 and σ𝑚 ,
respectively) of each of them. The S-N curve equation is shown below:

S𝑒 = S𝑢𝑡 − 𝑚 ∙ log10 (𝑁)

To account for the effect of the mean stress, a constant life Goodman diagram has to be used.
A variety of different curves exist, but for the sake of simplicity, the modified Goodman line will
be used. This curve is linear (as opposed to others, which may be quadratic or with other shapes,
see fatigue theory in chapter two), and is one of the simplest analyses, but a widely used
resource as it generally fits great with the actual fatigue behaviour. The modified Goodman line
is characterized by the following equation:
σ𝑎 σ𝑚
+ =1
S𝑒 S𝑢𝑡
This rule states that the fatigue life at alternating stress σ𝑎 and mean stress σ𝑚 is equal to the
fatigue life at an equivalent zero-mean-stress alternating stress state of S𝑒 . By putting the two
equations together and substituting Se, one can obtain explicitly the N value of cycles that the
material can withstand at each point. The algebraic steps are shown below:
σ𝑎 σ𝑚
+ =1
S𝑢𝑡 − 𝑚 log10(𝑁) S𝑢𝑡
2
S𝑢𝑡 − σ𝑎 ∙ S𝑢𝑡 − σ𝑚 ∙ S𝑢𝑡
log10(𝑁) =
𝑚(S𝑢𝑡 − σ𝑚 )

54
 S2𝑢𝑡 −σ𝑎 ∙S𝑢𝑡 −σ𝑚 ∙S𝑢𝑡
𝑁= 10 𝑚(S𝑢𝑡−σ𝑚)
Where σ𝑎 and σ𝑚 are obtained from the Rainflow algorithm. Once obtained the N value for each
cycle of the load history, the damage equivalent load (DEL) is computed by applying the
Palmgren-Miner method:
𝑛
1
DEL = ∑
𝑁𝑖
𝑖

Where:

• N is the total number of cycles counted by the Rainflow algorithm

• 𝑁𝑖 is the number of cycles the point can withstand for a given cycle characterized by σ𝑚
and σ𝑎

The DEL varies along the different points in the circumference as well as with the wind speed
because it depends on the pitch angle and the ratio between the aerodynamic and gravitational
forces [22]. Figure 59 shows the DEL along the different points on the circumference, for
different wind speeds:

55
 Figure 59. DEL along blade root circumference for 7, 11, 15, 19, 23 m/s wind speeds

For most of the wind speed spectrum (approximately 9 to 21 m/s), the critical angle is β = 90° as
it is the point with the highest DEL. This means that the flapwise moment (My) due to the thrust
of the wind, is predominant over the edgewise moment (Mx) or torque of the generator. For
cases where the wind speed is lower, the reverse is true, and the angles around β = 90° or β =
100° suffer the least DEL. However, the effect of the DEL in these cases is not as noticeable
precisely because the wind speed is low. Finally, when the wind speed is higher, the points that
suffer most from the effect of the loads are those located around β = [40°, 90°]. With this
information, it seems likely that the point with the highest DEL is β = 90°.

The effect of the wind speed on de DEL can be seen on figure 60, which shows the angle β = 90°.

Figure 60. DEL for different wind speeds β = 90°

As expected, lower wind speeds produce less damage on the blade than higher ones. As seen
on the last chapter, for higher wind speeds, in particular, higher than the rated one of the wind
turbine, the mean components of the thrust force and bending moment decrease, but the
amplitude of the oscillations becomes grater. This explains the increase of the DEL for the bigger
wind speeds. This figure is important, as it allows to relate the DEL of each point (in this case,
the critical angle) of the section with the corresponding wind speed.

56
To obtain the service life of the blade root, the actual wind speed has to be taken into account.
For this purpose, the DEL curve in figure 60 is interpolated with the occurrence of the wind speed
(figure 36), which represents the real environmental conditions in Pantelleria.

To calculate the mean DEL over a year with these conditions the next expression is used, which
takes into account that only 590 seconds of conditions where simulated (600 s of which the first
ten have been disregarded):

3600
𝑀𝑒𝑎𝑛⁡𝐷𝐸𝐿 = ∑ ( ∙ 𝐷𝐸𝐿𝑣 ) ∙ (𝑜𝑐𝑐𝑣 ∙ 24 ∙ 365)
𝑣 590

Where:

• 𝐷𝐸𝐿𝑣 is the DEL associated with the wind speed v for the 590 seconds of simulation
• 𝑜𝑐𝑐𝑣 is the occurrence of the wind speed v

The resulting mean DEL over a year is represented on figure 61. As discussed above, the critical
angle of the blade root section is β = 90° which has an average annual damage of 6.0795e-04.

Figure 61. Mean DEL on blade root

The service life, in years can be computed as one over the mean DEL:
1
𝐿=
3600
∑𝑣( ∙ 𝐷𝐸𝐿𝑣 ) ∙ (𝑜𝑐𝑐𝑣 ∙ 24 ∙ 365)
590
The service life of the root section is shown in figure 62. On table 11, the different points of the
section are compared.

57
 Figure 62. Service life on blade root

Table 11. Service life at the blade root

Angle β (degrees) Service life (years)

0° 1837
10° 1799
20° 1796
30° 1807
40° 1827
50° 1853
60° 1841
70° 1761
80° 1687
90° 1644
100° 1749
110° 1898
120° 2053
130° 2099
140° 2112
150° 2059
160° 1986
170° 1929
180° 1852

6.1.3. Results analysis

The service life of the section is the one of the most critical angle, which in this case is 1644
years. This result is very high compared to the average lifetime of a wind turbine, which ranges
from 20 to 30 years. This is due to several reasons. Firstly, the fatigue analysis carried out in this
work is very simple, as it only considers axial stresses, and disregards shear stresses. The shear
stress can also be very high and lead to other types of failure, such as inter-fibre failure. In the
present analysis, which is unidirectional, only tensile/compression failure has been considered.

58
In order to take into account shear and torsional stresses and thus other types of failure, a multi-
axial fatigue analysis should be carried out.

Secondly, the constant life diagram used is linear. This is only an approximation and may
underestimate the effect of the mean stress, so as a next step, a curve that better fits the fatigue
properties of the material could be used.

In the paper “Development of a simplified blade root fatigue analysis for floating offshore wind
turbines” (Sirigu, et al, 2022), different simplified models of the wind turbine are compared
under the same fatigue approach used here, to check the validity of MOST (numerical model for
floating offshore wind developed by the Politecnico di Torino). The paper compares the service
life under three different cases:

• MOST: with three DOF on the platform, a rigid body blade model and simplified
aerodynamics
• FAST (simplified): with three DOF on the platform and a rigid boy blade model
• FAST (complete): with six DOF on the platform, more complex aerodynamics and a
flexible blade model

The only different between the FAST complete model and the one used in this work is the use
of BeamDyn into FAST, to obtain the internal forces and moments on the blades.

Table 12 shows a comparison on the results obtained between the three models previously
described and the one used in this work. As can be seen, as the complexity of the model
increases the service life becomes higher. As stated by Sirigu, et al, 2022, “The complete version
of FAST shows smaller peaks and less oscillations of bending moment (My) due to the flexibility
of the blades” which results in less amplitude in the stress cycle. This explains why the complete
models show a higher fatigue life.

Table 12. Comparison between service life using different models

Model Service life (years)

MOST [22] 591
FAST (simplified) [22] 581
FAST (without BeamDyn) [22] 1215
FAST (with BeamDyn) 1644

As discussed in the paper “Comparison of blade optimisation strategies for the IEA 15MW
reference turbine” (Scott & Graves, 2022) [26], in which a structural analysis of the 15 MW wind
turbine´s blades using the finite element analyses, most critical failure mode is the inter-fibre
failure at 15-20% of the blade length, while the blade root shows lower damage, because the
blade root’s thickness is mainly driven by the high stiffness required for this section. It is
important to emphasise that these results correspond to static material failure analyses and not
to dynamic cases, such as fatigue. However, they can be used as a guide for the study of fatigue
in the rest of the blade.

In the following, some sections between the 10-30% of the blade length are studied with the
same approach to see if results, as expected, will be more realistic.

59
 6.2. Fatigue analysis on blade sections
The change of blade shape from root to airfoil occurs between 2% and 15% of the blade length.
In this transition zone, the loads are still quite high as it is close to the cantilever, and it is likely
that the most critical section for fatigue failure is located here. Taking this into account and the
above mentioned in the previous section, three airfoils are studied in this section, located at
12%, 20% and 28% of the blade length. The software BECAS is used as a postprocessor to get
the stress based on the forces.

6.2.1. Postprocessing of the forces

The data coming from the simulations is stored in different matlab structures for each wind
speed and sea state. It is remembered that the internal forces of the blade available as data,
correspond to nine different sections almost evenly distributed along the blade span. For each
of the sections, the forces are stored in a matrix of dimensions 1.200.000x6, which correspond
to the simulations made with BeamDyn (10 minutes of simulation time with a time step of
0,0005s).

These data are post processed with the help of a matlab script: the first ten seconds of
simulations are disregarded, the dimensions are reduced, and the forces are interpolated to the
desired sections of the blade. The procedure is explained below:

To reduce the computational time for the calculations, it is helpful to reduce the dimensions of
the vectors of the internal forces. This can be done as long as the accuracy or quality of the
results is not reduced. In this case, the downsizing of the force vector does not affect the
precision as the time step used in the simulations is sufficiently small, as shown in figure 63:
figure 63 left shows the original size vector whereas figure 63 right shows the reduced-
dimension vector.

Figure 63. Downscaling of the force vector for the 12% blade length section. Full (left) vs
reduced (right)

The forces are reduced 10 times its dimension in time. While the original vectors have time
dimensions of 1.200.000-time length, the reduced ones are 180.000 long.

The last step is to interpolate the internal forces to the three sections to be studied. From the
available nodes, each at a different blade length, the forces are obtained in the sections of 12%,
20% and 28% of the blade length. Linear interpolation has been used for the sake of simplicity.
Table 13 shows the span of the sections in which the data are available (the nine nodes resulting

60