POLITECNICO DI TORINO

FACOLTA’ DI INGEGNERIA

Corso di Laurea Magistrale in Ingegneria Energetica e Nucleare

MASTER THESIS

Implementation and validation of a hydrodynamic analysis

model in the BEMUse library for floating platforms

SUPERVISORS CANDIDATE
Ing. Giovanni Bracco Vittorio Martini
MSc. Joseph Saverin

ADVISOR
Prof. Giuliana Mattiazzo

October 2019
 Abstract

ABSTRACT

The ambitious goal set at the Climate Conference in Paris in 2015 (COP21) to try and contain
the world’s average temperature increment below 1.5 °C with respect to the pre-industrial era
requires the introduction of more efficient and productive technologies for the exploitation of
renewable energy sources in all their forms.
Among these technologies, wind power is at the moment one of the most reliable and advanced,
as it’s been effectively operating for decades now. Wind turbines are currently scattered all
around the world, both on land and at sea. Finding a suitable installation site on land is becoming
increasingly difficult in a lot of countries, though, a factor that has increased the interest in
offshore solutions in recent years.
Unfortunately, current technology has quite strict requirements when it comes to the choice of
a suitable site, specifically on the water depth. To increase the exploitable surface for wind
energy production, the concept of floating offshore wind turbines (FOWTs) has recently gained
increasing interest in the sector. As the name suggests, wind turbines are in this case mounted
on floating platforms instead of being anchored to the seafloor.
This solution increases enormously the sea surface suitable for wind farms, but at the same time
requires accurate simulations of the behavior and performance of the turbines, when subjected
to forces and motions caused by both waves and wind.
This analysis requires precise and reliable hydrodynamic analysis softwares like WAMIT and
Ansys AQWA. At the Technische Universität Berlin, a new such model called BEMUse has
been developed and needs to be validated.
In this thesis, the results given by BEMUse are compared with the ones obtained with the two
softwares mentioned before for a set of geometries ranging from a simple hemisphere to
prototypes of platforms of great interest for commercial installations. The comparison will
examine four crucial hydrodynamic parameters: added mass, radiation damping, exciting
forces, and Response Amplitude Operators (RAOs).
If the results agree, BEMUse can be considered validated and can be used as an alternative to
expensive softwares like WAMIT

I
II
 Summary

SUMMARY
ABSTRACT ................................................................................................................................ I
LIST OF FIGURES ................................................................................................................... V
LIST OF TABLES ..................................................................................................................VII
1. INTRODUCTION ............................................................................................................ 1
1.1 Wind energy today ....................................................................................................................... 1
1.2 Floating offshore wind turbines ................................................................................................... 5
1.2.1 Advantages and disadvantages ........................................................................................................... 6
1.2.2 Floater types........................................................................................................................................ 8
1.2.3 Mooring systems ................................................................................................................................. 8
1.2.4 Anchoring systems ............................................................................................................................ 10
1.2.5 Economic considerations .................................................................................................................. 10
1.3 The need for hydrodynamic models ........................................................................................... 13
2. THEORY ........................................................................................................................ 15
2.1 Description of the problem ......................................................................................................... 15
2.2 First order problem ..................................................................................................................... 16
2.3 Boundary value problem ............................................................................................................ 17
2.4 Integral equations ....................................................................................................................... 18
2.5 Green’s theorem and distribution of singularities ...................................................................... 19
2.5.1 The Green function ........................................................................................................................... 22
2.6 First order force .......................................................................................................................... 23
2.7 Hydro-static and -dynamic force and moment ........................................................................... 23
2.7.1 Coordinate system............................................................................................................................. 23
2.7.2 Coordinate transform ........................................................................................................................ 23
2.7.3 Pressure integration........................................................................................................................... 24
2.7.4 Hydrostatic force and moment of O (1) ............................................................................................ 25
2.8 Linear force and moment ........................................................................................................... 26
2.9 Irregular frequencies .................................................................................................................. 27
2.9.1 Removal of irregular frequencies ...................................................................................................... 28

3. DESCRIPTION OF BEMUse ........................................................................................ 31

3.1 Method of solution ..................................................................................................................... 31
3.2 Geometry .................................................................................................................................... 32
3.3 Solid body transformation .......................................................................................................... 32
3.4 Dipole potential and velocity ..................................................................................................... 33
3.5 Influence coefficients ................................................................................................................. 34
3.6 Boundary conditions .................................................................................................................. 36
3.7 Linear system ............................................................................................................................. 38
3.8 Discretization ............................................................................................................................. 39
3.9 From potential to forces ............................................................................................................. 41
3.9.1 Added mass and damping ................................................................................................................. 41
3.9.2 Exciting forces .................................................................................................................................. 42
3.9.3 Response Amplitude Operators ........................................................................................................ 42

III
 Summary

4. WAMIT COMPARISON ............................................................................................... 45

4.1 Cylinder ...................................................................................................................................... 48
4.1.1 Added mass ....................................................................................................................................... 49
4.1.2 Damping ........................................................................................................................................... 50
4.1.3 Exciting forces and moments ............................................................................................................ 51
4.1.4 Response Amplitude Operators ........................................................................................................ 52
4.2 Box ............................................................................................................................................. 54
4.2.1 Added mass ....................................................................................................................................... 55
4.2.2 Damping ........................................................................................................................................... 56
4.2.3 Exciting forces and moments ............................................................................................................ 57
4.2.4 Response Amplitude Operators ........................................................................................................ 58
4.3 Tension leg platform .................................................................................................................. 60
4.3.1 Added mass ....................................................................................................................................... 62
4.3.2 Damping ........................................................................................................................................... 63
4.3.3 Exciting forces and moments ............................................................................................................ 64
4.3.4 Response Amplitude Operators ........................................................................................................ 65

5. AQWA COMPARISON ................................................................................................. 67

5.1 Hemisphere ................................................................................................................................ 67
5.1.1 Added mass ....................................................................................................................................... 68
5.1.2 Damping ........................................................................................................................................... 69
5.1.3 Exciting forces and moments ............................................................................................................ 70
5.1.4 Response Amplitude Operators ........................................................................................................ 71
5.3 Cylinder ...................................................................................................................................... 72
5.2.1 Added mass ....................................................................................................................................... 73
5.2.2 Damping ........................................................................................................................................... 74
5.2.3 Exciting forces and moments ............................................................................................................ 75
5.2.4 Response Amplitude Operators ........................................................................................................ 76
5.3 Box ............................................................................................................................................. 77
5.3.1 Added mass ....................................................................................................................................... 78
5.3.2 Damping ........................................................................................................................................... 79
5.3.3 Exciting forces and moments ............................................................................................................ 80
5.3.4 Response Amplitude Operators ........................................................................................................ 81
5.4 Oc3 ............................................................................................................................................. 82
5.4.1 Added mass ....................................................................................................................................... 84
5.4.2 Damping ........................................................................................................................................... 85

6. CONCLUSION ............................................................................................................... 87
6.1 WAMIT comparison .................................................................................................................. 87
6.2 AQWA comparison .................................................................................................................... 87
6.3 Future work ................................................................................................................................ 88
7. BIBLIOGRAPHY ........................................................................................................... 91
ACKNOWLEDGMENTS ........................................................................................................ 93

IV
 List of figures

LIST OF FIGURES

1.1 Global cumulative wind capacity rise in the period 2001-2017. ...................................... 1
1.2 Global cumulative and annual wind power installation forecast. .................................... 2
1.3 Planetary boundary layer’s graphical representation. ..................................................... 3
1.4 Global cumulative and annual offshore wind capacity. .................................................... 4
1.5 Graphical representation of a floating wind turbine. ....................................................... 5
1.6 Sea depth around the coasts of Europe. ............................................................................ 6
1.7 Average wind speed 50 m above ground level. ................................................................. 7
1.8 Graphical representation of catenary and taut-leg moorings. ......................................... 9
1.9 CAPEX breakdown for FOWTs and comparison with fixed offshore turbines. .............. 11
1.10 LCOE comparison between FOWTs and fixed offshore turbines. .................................. 12
1.11 LCOE comparison between various energy sources....................................................... 12
2.1 Surfaces of integration for Green’s theorem................................................................... 20
3.1 Paneling of a hemisphere generated with BEMUse........................................................ 31
3.2 Generation of the vector normal to the panel’s surface in BEMUse. ............................. 40
4.1 Cylindrical geometry generated with BEMUse. ............................................................. 46
4.2 Parallelepipedal geometry generated with BEMUse. ..................................................... 46
4.3 TLP geometry generated with BEMUse. ......................................................................... 47
4.4 Cylinder with colorized panels generated with BEMUse. .............................................. 48
4.5 Added mass coefficient comparison, cylinder. ................................................................ 49
4.6 Damping coefficient comparison, cylinder. .................................................................... 50
4.7 Exciting forces and moments comparison, cylinder........................................................ 51
4.8 RAOs comparison, cylinder............................................................................................. 52
4.9 Box with colorized panels generated with BEMUse. ...................................................... 54
4.10 Added mass coefficient comparison, box. ....................................................................... 55
4.11 Damping coefficient comparison, box. ............................................................................ 56
4.12 Exciting forces and moments comparison, box. .............................................................. 57
4.13 RAOs comparison, box. ................................................................................................... 58
4.14 Geometry of the TLP. ...................................................................................................... 60
4.15 TLP with colorized panels generated with BEMUse. ..................................................... 61
4.16 Added mass coefficient comparison, TLP. ...................................................................... 62
4.17 Damping coefficient comparison, TLP............................................................................ 63
4.18 Exciting forces and moments comparison, TLP. ............................................................. 64
4.19 RAOs comparison, TLP. .................................................................................................. 65
5.1 Mesh grid generated by AQWA. ...................................................................................... 67
5.2 Added mass comparison, hemisphere. ............................................................................ 68

V
 List of figures

5.3 Damping comparison, hemisphere. ................................................................................. 69

5.4 Exciting forces and moments comparison, hemisphere. ................................................. 70
5.5 RAOs comparison, hemisphere. ...................................................................................... 71
5.6 Added mass comparison, cylinder. .................................................................................. 73
5.7 Damping comparison, cylinder. ...................................................................................... 74
5.8 Exciting forces and moments comparison, cylinder........................................................ 75
5.9 RAOs comparison, cylinder............................................................................................. 76
5.10 Added mass comparison, box. ......................................................................................... 78
5.11 Damping comparison, box. ............................................................................................. 79
5.12 Exciting forces and moments comparison, box. .............................................................. 80
5.13 RAOs comparison, box. ................................................................................................... 81
5.14 OC3 floating platform generated by BEMUse ................................................................ 83
5.15 Added mass comparison, OC3. ....................................................................................... 84
5.16 Damping comparison, OC3............................................................................................. 85

VI
 List of tables

LIST OF TABLES

Table 1.1. Exploitable potential wind resource with FOWTs. ................................................... 6

Table 4.1. Geometrical and physical properties of the three geometries. ............................... 45
Table 4.2. Normalization factors for the four parameters evaluated. ...................................... 47
Table 4.3. Geometrical and physical properties of the TLP. ................................................... 60
Table 5.1. Physical and geometrical parameters of the OC3 platform. .................................. 82

VII
VIII
 1 - Introduction

1. INTRODUCTION

1.1 WIND ENERGY TODAY

Ever since the initial stages of the now impelling transition to green energy, wind power has
always represented a vital source of clean and renewable energy, one of the main actors in the
much-needed reduction of fossil fuels’ predominance in the world energy mix.
Starting from the late 1980s, when the first wind farm started generating electric power in New
Hampshire, USA, the industry has been on a continuous search for new markets and installation
sites around the world [1].
In the last 20 years, the sector has experienced continuous and remarkable growth, with an
overall installed capacity, in 2017, of 539,123 MW globally. The graph below gives an excellent
overlook of this evolution during that time.

Figure 1.1 The global installed wind capacity [MW] has continuously risen over the last 17 years.

Despite the fact that in the last couple of years the annual installed capacity (new turbines built)
has experienced a slight reduction, as reported in the Global Wind Report of 2017 ‘wind power
is in a rapid transition to becoming a fully commercialized, unsubsidized technology,
successfully competing in the marketplace against massively subsidized fossil and nuclear
incumbents.’[1]. This kind of transition involves going through a period of adjustments, but
already today wind is the most competitively priced green technology in most markets, and the
rise of more advanced grid management and affordable storage devices is shaping what will be
the first commercial fossil-free power sector.

1
 1 - Introduction

The market forecasts carried out for the near future depict a promising scenario, as shown in
the following figure.

Figure 1.2. The installed capacity should grow in the near future, despite a reduction in new installations.

The main issue for wind power, like for many other renewable energy sources, lies on the
discontinuity of the source, as there is no place on earth where the wind blows consistently all
year long. Additionally, currently operating turbines can only produce power when the wind
speed value lies between certain limits, typically 4 to 25 m/s. This is to prevent potential
damages or excessive strains on the delicate structure. For the same reason, turbulent winds are
also to be avoided.
The dominant trend has pushed towards the construction of increasingly taller towers, larger
rotors and consequently more powerful turbines, which now reach stunning dimensions. To
give an example, the newest prototype presented by General Electric for offshore applications
is the Haliade-X, that will be 260 meters high, with 107-meter long blades and 220-meter rotor
diameter, with a rated output of 12 MW [2].
This tendency has both economic and technical reasons. Larger structures allow to reduce costs,
a fundamental concept of economy of scale, therefore allowing the produced electricity to be
cheaper and hence more competitive on the energy market. From a technical perspective, taller
turbines with larger rotors can exploit the strong, constant and steady high-altitude winds, which
are not affected by the planetary boundary layer.

2
 1 - Introduction

This concept, of capital importance in the wind power industry, is visually explained in the
following figure.

Figure 1.3. Graphical representation of the planetary boundary layer concept.

Morphological roughness, irregularity of the terrain, the presence of tall trees, buildings and
mountains are all factors that tend to decrease the wind speed near the earth surface, increasing
at the same time the turbulence of the flow with worsening effects both on the performance and
the durability of the turbine. These issues make it increasingly challenging to find suitable
installation sites for wind farms, which need large surfaces of low and flat terrain which, if not
in hardly accessible areas, are often already used for farming or crop production.
All these reasons together lead to the result that despite its enormous potential, wind energy can
currently rely on an average capacity factor of around 25 to 40% of the rated power [3]. To
make things worse, during the years governments and wind power companies have faced
vehement protests by the people living near the planned installation sites, who lamented noise
and aesthetic pollution caused by the turbines.
Given all the issues illustrated so far, it is easy to understand why, after just a few years after
the construction of the first wind farm, engineers and companies have focused on how to install
turbines on the sea instead of on land. The reasons for that are straightforward, as the ocean
offers broad, flat surfaces with no obstacles, thus reducing the detrimental effect of the planetary
boundary layer. Moreover, winds blowing over the sea and oceans are usually stronger and
steadier, allowing turbines to reach higher capacity factor values. Finally, it is estimated that
75% of world energy demands are concentrated in coastal areas [4].

3
 1 - Introduction

The first operational offshore wind farm started its production in 1991 off the coast of Denmark
and, given its good productivity, numerous new projects followed and in the last twenty years
the installed European offshore capacity has increased, often helped by government incentives
and high industrial investments, as shown in the next graph.

Figure 1.4. Both the cumulative and annual offshore capacity [MW] have grown enormously in the last 17 years.

Nowadays, numerous wind farms are scattered off the coast of Germany and Denmark in the
North Sea.
As the technology advanced and the related costs decreased, larger wind farms have been
installed at increasing distances from the coast to exploit the highly constant winds. At the same
time, turbines got larger and more efficient, with some models specifically engineered for
offshore applications. Given the proven potential of offshore wind power, it can seem surprising
that only very few areas around the world have been considered suitable for the installation of
wind farms, with a strong predominance of northern European countries. Moreover, turbines
are installed near the coast, where the winds are notoriously weaker and less constant than in
high-sea locations.
The main reason for that lies in the necessity to anchor the base of the structure to the seafloor,
thus limiting the available sites to shallow waters, close to the shore. The installation near the
coast allows, on the other hand, to contain the costs and the losses for the transmission of electric
energy to the users. Another drawback is that being near the coast, the turbines are still visible
by the people living there, attracting the already mentioned protests about the visual impact of
such installations. Nowadays, only depths up to ca. 50 m are suitable with modern techniques,
which are all very expensive and require specifically engineered vessels and machinery.

4
 1 - Introduction

1.2 FLOATING OFFSHORE WIND TURBINES

Being able to exploit the enormous reserves of energy present in deeper waters would probably
represent a decisive step towards the transition to a fossil-free energy sector. This urgency has
led to the development, in the last few years, of a new concept: the Floating Offshore Wind
Turbines (FOWTs). As the name suggests, the turbine in this case is not directly anchored to
the seafloor, but instead is installed on a platform that makes it float and at the same time tries
to minimize the motion caused by the action of wind and waves on the structure. The whole
installation is kept in position by a mooring system anchored to the seafloor. In this way,
locations with waters up to several hundred meters deep become suitable for the installation of
wind farms.
In the figure, the basic concept of floating wind turbines is clearly illustrated.

Figure 1.5. FOWTs are just anchored on the seafloor, making them easier to install.

The vision of large-scale FOWTs was first introduced by Professor William E. Heronemus back
in 1972, but it was only after the commercial wind industry was well established that the topic
was taken up again [5].

5
 1 - Introduction

1.2.1 Advantages and disadvantages

In 2015 Carbon Trust, an independent company that carries out carbon reduction evaluations
for organizations and governments, estimated that the energy produced by wind turbines in
deep waters in the North Sea alone could meet the EU's electricity consumption four times over
[6]. The table below reports some other staggering conclusions present in the same study.

Table 1.1. The potential wind resource exploitable with FOWTs is enormous.

These numbers give an idea of the immense potential of the FOWT concept.
Moreover, the exploitable sea surface wouldn’t be difficult to reach: the first of the following
two figures shows the sea depth around the coasts of Europe, while the second one the average
wind speed 50 meters over the sea surface in the same area. The superposition of regions in the
two pictures with shallow waters and strong winds is undoubtedly of great interest and confirms
the vast potential for FOWTs.

Figure 1.6. Most of Europe, particularly in the north, is surrounded by shallow waters.

6
 1 - Introduction

Figure 1.7. The farther away from the coast, the stronger the winds at a 50 m altitude.

It’s also worth noting how the increased exploitable sea surface is not FOWTs’ only advantage.
As previously stated, working far from the coastline and its morphological irregularities reduces
the wind turbulence on the turbine, increasing the expected lifespan of the farms up to 30 years
[4], with respect to the ca. 20 expected for the currently operating technology.
Additionally, floating platforms allow for simplified and more flexible deployment, as more
operations required for their construction and installation can be carried out on land, and then
the assembled structure can be easily towed in position without requiring the specialized vessels
needed today. Their greater mobility also allows them to be easily swapped or moved to port in
case some repairs are necessary. Furthermore, an increased installation depth usually means
also an increased distance from the coast. Wind farms will be installed outside the coastal Zone
of Visual Influence, being invisible from shore and thus eliminating any protest. The size of
this area, also known as ZVI, can be easily calculated through a simple equation, reported by
Sclavounos [4]

𝐿 = √2𝐻𝑅 (1.1)
Where L is the distance from shore for a turbine to be invisible, H is the maximum height of
the turbine’s blade tip, and R is Earth’s radius. To give a quick example, L = 45 km in case of
a 155-meter-high blade tip.

7
 1 - Introduction

1.2.2 Floater types

The extensive research that is being carried out on FOWTs still hasn't figured out which is the
best design for the floating platform to put beneath the turbine. Multiple designs have been
proposed by various companies, each having their pros and cons. Without analyzing every
model present on the market today, a sufficiently accurate overview can be performed
describing the three main types of platforms that are more promising to this day.
- Spar-buoy types: it’s the simplest design, characterized by a very long and thin cylinder-
like platform, often more than 100 m long. Its center of mass is moved far below the
center of buoyancy, which creates a strong restoring moment when the structure is out
of its equilibrium point. Due to its simple design, it’s easy to build and provides
excellent stability, even though the notable draft causes difficulties during the assembly,
transportation, and installation phases and requires depths of more than 100 meters. One
of the most promising models of this type is the OC3 platform by Hywind.
- Semi-submersible Platform: this kind of platform uses the hydrostatic stiffness of
substructures piercing the waterplane area in positions far from the center axis of the
tower. In this way, the restoring stiffness needed for stability is provided. This design
often requires a large and heavy structure to maintain proper balance, but the reduced
draft allows for more flexible applications and an easier installation.
- Tension-Leg Platform: in this case, the mooring is actively involved in the stabilization
of the structure. This design is inherently unstable because the center of gravity is
situated above the center of buoyancy. Therefore, the mooring provides the righting
moment by being connected to off-axis fairleads, and from there transfer the wind
turbine loads into the anchors below. The shallow draft and tension stability allow for a
smaller and lighter structure, while the stresses on the tendons and anchors system are
higher than in the previous designs.

1.2.3 Mooring systems

As for the platform geometry and characteristics, extensive research is currently underway to
find the best mooring and anchoring systems and materials for commercial applications. Extra-
heavy-duty cables are required, as they need to fix the platforms in place while enduring
mechanical loads from both waves and wind, and chemical corrosion from seawater [7].
The choice of the mooring method has such an impact on the overall architecture of the platform
that FOWTs can be classified according to the system they use [5]. The most commonly used
are the catenary moorings, the taut-leg moorings and vertical tension legs, which are considered
a subset of the taut-leg moorings.

8
 1 - Introduction

The first two systems are shown in the following figure.

Figure 1.8. A graphical comparison of catenary moorings (left) and taut-leg moorings (right).

The pros and cons of each system are briefly reported. Some of the information is taken from
[5].
- Catenary moorings: the most significant advantages of this kind of system are the
relatively low cost of the anchors and the possibility to be deployed in shallow waters.
On the other hand, the vertical tension of the anchor line is often insufficient to prevent
overturning, because of the weight and the strong horizontal forces acting so far above
the center of buoyancy. This distribution of weights means that additional ballast must
be added below it, or buoyancy must be widely distributed. Catenary moored platforms
present a significant portion of the structure above the waterline, which means a higher
wave loading acting on it. Therefore, in general, this kind of platform subjects the
turbine to a broader base motion in all directions, increasing the complexity of the
system integration.
- Taut-leg moorings: the more the water depth increases, the more these systems become
advantageous over the previous type. This is because their footprint (the surface of
seafloor necessary for anchoring) is smaller, needing shorter mooring lines. If the taut
legs are installed in a vertical orientation, the footprint becomes even smaller, but the
high vertical forces consequently developed require more complex anchoring systems.
Vertical moorings allow a larger portion of the structure to be submerged, minimizing
wave action while maintaining the platform very stable.
In deciding which system to use, it’s often a trade-off between the added complexity introduced
by platform dynamics and the associated turbine cost, and the added complexity and expenses
for the anchoring system.

9
 1 - Introduction

1.2.4 Anchoring systems

The main factor influencing the choice of a certain kind of anchoring system is the bottom soil,
because its shear strength is the primary mechanism for resisting the forces applied. Also the
soil weight is a major factor to consider. As expected, the deeper the anchor can be embedded,
the higher the quantity of affected soil and, consequently, the greater the holding capacity.
The direction of the applied force also influences the holding capacity of an anchor. If the force
is applied parallel to the bottom, as in the case of catenary moorings, there is no need for a deep
embedment, and less precision is required in their placement. Thus, installation is less
expensive, although platforms using catenary mooring experience more significant motions in
every direction compared to other designs.
On the other hand, vertical load anchors depend on deep embedment to affect a large soil
surface, increasing the cost of installation. They also need a more accurate placement.
Describing in detail all the types of mooring systems used goes beyond the scope of this thesis.
It’s sufficient to say that the factors determining the choice of the anchor vary so widely that,
in most cases, the technical solution is specifically designed for the bottom conditions of the
site.
Beside the kind of installation required, other factors influencing the final cost of the mooring
system are the material cost of the anchor and mooring lines. The challenge regarding the
economic aspect of anchors is to find a relatively inexpensive system with a high vertical load
capacity which, at the same time, is also easy to install.

1.2.5 Economic considerations

Before digging deep in an economic comparison between fixed-bottom and floating-platform
wind turbines, it's essential to keep in mind that while cost data for the former technology are
relatively robust and related to a fully commercialized sector, the available data for the latter is
much more uncertain and referring prevalently to prototypes and pre-commercial installations.
Like for every new technology, in the initial stages of conception and development, costs tend
to quickly increase, with a successive decrease in the phases of optimization and
industrialization.
This premise is essential, as the cost of early prototypes does not reflect the actual costs that
can be expected once designs are optimized, new technologies are introduced, and serial
fabrication methods are implemented; in other words, once all the benefits of scale effects and
mass production kick in.
One of the most comprehensive economic evaluations of FOWT was carried out in 2014 by
four researchers at the Norwegian University of Life Sciences [8] and was used to gather the
data reported in this chapter. In their work, several promising prototypes of FOWTs are
compared with the two most diffused solutions adopted for bottom-fixed turbines.
First, it’s interesting to compare the Capital Expenditure (CAPEX) for a typical fixed-bottom
and floating offshore wind projects.

10
 1 - Introduction

The figure below not only reports absolute monetary values, expressed in $/MW, but it also
breaks down how these costs are divided among the various aspects of the installation. The data
referring to bottom-fixed designs are the last two bars to the right.

Figure 1.9. FOWTs CAPEX is almost competitive with that of fixed turbines.

As expected, the foundation and installation costs are lower for the FOWT, because the
structure can be assembled onshore and then towed to the site even in rough sea conditions,
which usually halt the works for traditional offshore projects. Obviously, mooring costs are
absent in the case of bottom-fixed turbines. In both cases, most of the CAPEX flows into the
turbine itself. The incidence is higher for floating projects because they require a more detailed
match-making process to couple a turbine and its base as effectively as possible, while bottom-
fixed platforms are said to be ‘turbine agnostic’ [6].
Summarizing, a higher CAPEX is expected for floating solutions, but what could be decisive
for the success of FOWTs are the cost savings that could be achieved during the operation of
the turbine, named Operational Expenditure (OPEX). There are important reductions expected
in this sector, especially for major repairs like the gearbox substitution. This is because carrying
out these tasks on fixed platforms requires jack-up or dynamic positioning vessels, which are
very expensive. On the other hand, most FOWTs are designed so that they can be disconnected
from the moorings and towed to the nearest port, where easier maintenance can be performed.
Combining the CAPEX and OPEX, together with the expected energy output of the project,
one can calculate the cost competitiveness of FOTWs expressed by the Levelized Cost of
Energy (LCOE, in $/MWh), which at the end of the day is the parameter determining the success
or failure of a specific technology. That’s because it shows if a technology can produce energy
at a competitive price for the market.
Early studies have shown that floating wind turbines will be able to achieve results comparable
to those of bottom-fixed technology real soon, mostly because of higher yields (the ratio
between the effective output and the nominal one) thanks to faster, more constant winds on
high-sea.

11
 1 - Introduction

The figure below seems to confirm this prediction; again, bottom-fixed designs are on the far
right.

Figure 1.10. LCOE for the reference wind farm for each concept with values for best- and worst-case scenarios.

Taking the average values of this plot and comparing them with the ones available for other
energy sources gives additional reasons to be optimistic about the future of the technology. As
the figure below shows, LCOE for FOWTs is comparable to the ones calculated both for
conventional (like nuclear power) and unconventional (like fuel cells) energy sources, as Lazard
categorizes them [9].

Figure 1.11. Comparing with data of fig. 1.10, soon FOWTs will be competitive on the market.

12
 1 - Introduction

1.3 THE NEED FOR HYDRODYNAMIC MODELS

Despite their high potential for future development, FOWTs force engineers to face a whole
new set of challenges related to the behavior of the floating platform and the overlying turbine,
once they are free to move following the motion of the waves and subject to the force of the
wind. Having a rotor that oscillates has a detrimental effect on the performances of the turbine
since its plane is often not perpendicular to the incoming wind.
Therefore, the hydrodynamic behavior of the platform in the supposed operative conditions
requires to be thoroughly analyzed. Since these evaluations cannot be carried out on real-scale
models due mainly to high costs, there is the need to use hydrodynamic analysis models and
softwares to calculate the forces and moments acting on the platform.
These softwares evaluate how the incoming waves influence the motion of the platform and the
intensity of the forces and moments acting on its surface. Most of these softwares use the
Boundary Element Method (BEM), or Panel Method, which is a numerical technique that uses
systems of partial differential equations formulated into the integral boundary form. BEM codes
employ the Green's functions to transform a flow problem into a problem of source distribution
on the body surface [10]. When used in the hydrodynamic context, BEM solves separately the
scattered and radiated velocity potentials that arise from the interaction of a harmonic linear
wavefield with a body located within that field [11].
BEM is the most common method used in hydrodynamic context because, while limited by the
linear nature of potential flow theory, the speed of the numerical simulation is much higher than
other simulation methods like Computation Fluid Dynamics. This makes BEM a common
choice for early-stage device development [10].
Several commercial softwares have been developed to implement BEM in the field of
hydrodynamic diffraction simulations. Such systems include the well-established commercial
software WAMIT or Ansys AQWA, and some more recently released, open-source solvers like
NEMOH.
A new hydrodynamic analysis model has been developed at the Hermann-Föttinger-Institut at
the Technische Universität Berlin as part of the QBlade software, which is an open-source wind
turbine calculation and modeling software, an easy-to-use and flexible tool that allows the user
to simulate the performances of a customized wind turbine accurately. Adding BEMUse’s
library will allow the precise analysis of a FOWT, as it will let the user evaluate the influence
that the motion caused by the sea waves has on the overall performance of the installation.
However, since BEMUse is a recently developed model, its results must be compared to the
ones of other older and widely used hydrodynamic analysis softwares, to make sure it works
correctly. Only then BEMUse can be included in QBlade. This thesis has the aim to carry out
the comparison between BEMUse and both WAMIT and AQWA softwares to check if the
results for the most important hydrodynamic coefficients coincide. If they do, BEMUse can be
considered validated and, therefore, ready to be implemented in QBlade.
Before proceeding with the comparison, though, an overview of BEMUse’s theoretical
background is provided in the next chapter.

13
14
 2 - Theory

2. THEORY

This chapter will give a brief overview of BEMUse’s theoretical basis, which is, in fact, the
same as WAMIT’s. WAMIT Theory Manual [12] thoroughly illustrates the theory behind the
hydrodynamic model; thus, the majority of what’s written in this chapter was taken from that
source.
BEMUse is designed to solve the boundary value problem for the interaction of water-waves
both in finite- and infinite- water depth, given a prescribed geometry of the body. The model
uses quadrilateral panels to carry out this task.
Viscous effects of the fluid are neglected, and therefore the flow field is considered potential
without circulation.
Another fundamental assumption is that the body stays at its mean position, and in case it’s not
fixed, the oscillatory amplitude of the body motion is of the same order as the wave amplitude.
The boundary value problem is reorganized into integral equations using the wave source
potential as a Green function. This integral equation is then solved via the panel method for the
unknown velocity potential or the source strength on the body surface. The latter is needed to
then calculate the fluid velocity on the body surface.

2.1 DESCRIPTION OF THE PROBLEM

As stated before, the flow is assumed to be potential, free of separation or lifting effects and it’s
governed by the velocity potential ϕ (x, t) which satisfies Laplace’s equation in the fluid domain:
∇2 𝜙 = 0 (2.1)
Where t is the time, and x = (x, y, z) denotes the Cartesian coordinates of a point in space. In all
cases, the undisturbed free surface is the z = 0 plane, and the fluid domain is in z < 0.
The gradient of the velocity potential gives the fluid's velocity
𝜕𝜙 𝜕𝜙 𝜕𝜙
𝑉(𝑥, 𝑡) = ∇𝜙 = 𝑖+ 𝑗+ (2.2)
𝜕𝑥 𝜕𝑦 𝜕𝑧
And the pressure is then calculated with Bernoulli’s equation
𝜕𝜙 1
𝑝(𝑋, 𝑡) = −𝜌 ( + ∇𝜙 ∙ ∇𝜙 + 𝑔𝑧) (2.3)
𝜕𝑡 2
Where ρ is the fluid density, and g is the gravitational acceleration.
The velocity potential satisfies the nonlinear free-surface condition:
𝜕 2𝜙 𝜕𝜙 𝜕𝜙 1
2
+𝑔 + 2∇𝜙 ∙ ∇ + ∇𝜙 ∙ ∇(∇𝜙 ∙ ∇𝜙) = 0 (2.4)
𝜕𝑡 𝜕𝑧 𝜕𝑡 2

15
 2 - Theory

Applied on the exact free surface

1 𝜕𝜙 1
𝜁(𝑥, 𝑦) = − ( + ∇𝜙 ∙ ∇𝜙) (2.5)
𝑔 𝜕𝑡 2 𝑧=𝜁

Assuming a perturbation solution in terms of a small wave slope of the incident waves, the
velocity potential can be expanded as:
𝜙(𝑥, 𝑡) = 𝜙 (1) (𝑥, 𝑡) + 𝜙 (2) (𝑥, 𝑡) + ⋯ (2.6)
And the same happens to the motion amplitude of the body if it is not fixed:
𝜉 = 𝜉 (1) + 𝜉 (2) + ⋯ (2.7)
Given a wave spectrum, it is usually assumed that it is expressed as a linear superposition of
first-order incident waves of different frequencies. Thus, the local first-order potential for the
wave-body interaction can be represented by a sum of components having circular frequency
ωj > 0

𝜙 (1) (𝑥, 𝑡) = 𝑅𝑒 ∑ 𝜙𝑗 (𝑥)𝑒 𝑖𝜔𝑗𝑡 (2.8)

𝑗

Where ϕj (x) is the complex velocity potential, which is independent of time; the real part of the
time-harmonic solution is physically relevant.
In Eq. (2.8) ϕj (x) indicates the first-order solution in the presence of the incident wave with
frequency ωj and wave heading βj, while the directional spreading of the incident waves is not
shown explicitly.

2.2 FIRST ORDER PROBLEM

In this chapter, the first-order boundary value problem and its solution procedure are illustrated.
From now on, the subscript j indicating the frequency component will be omitted, as only one
particular value of ω is considered.
The velocity potential for the first-order incident wave is defined by
𝑖𝑔𝐴
𝜙𝐼 = 𝑍(𝜅𝑧)𝑒 −𝑖𝜅(𝑥𝑐𝑜𝑠𝛽+𝑦𝑠𝑖𝑛𝛽) (2.9)
𝜔
It represents a plane-progressive wave of circular frequency ω and wave heading angle β
(calculated starting from the positive x-axis). A is the complex wave amplitude.
The function Z identifies the depth-dependence of the flow and has different forms in the case
of infinite or finite water depth. In the former case, it's given by
𝑍(𝜅𝑧) = 𝑒 𝜅𝑧 (2.10)
𝜔2
And 𝜅 = . In the latter, instead
𝑔

cosh(𝜅(𝑧 + ℎ))
𝑍(𝜅𝑧) = (2.11)
cosh(𝜅𝑧)

16
 2 - Theory

And this time κ is the real root of the dispersion relation

𝜔2
𝜅 tanh 𝜅ℎ = (2.12)
𝑔
The scattering velocity potential ϕS denotes the disturbance to the incident wave due to the
presence of the body in its fixed position. The linearity of the problem allows it to be
distinguished from the interference due to the motion of the body, which is represented by the
radiation potential ϕR. The total velocity potential is given by
𝜙 = 𝜙𝐼 + 𝜙𝑆 + 𝜙𝑅 = 𝜙𝐷 + 𝜙𝑅 (2.13)
Where the diffraction potential ϕD is defined as the sum of ϕI and ϕS.
The radiation potential is a linear combination of the components corresponding to the modes
of motion such that
6

𝜙𝑅 = 𝑖𝜔 ∑ 𝜉𝑘 𝜙𝑘 (2.14)
𝑘=1

Where ξk is the complex amplitude of the oscillatory motion in mode k of the six degrees of
freedom, and ϕk the corresponding unit-amplitude radiation potential.
These modes are referred to as surge, sway, heave, roll, pitch and yaw in the increasing order
of j.

2.3 BOUNDARY VALUE PROBLEM

The first-order velocity potential ϕ, together with each of its components, satisfies the Laplace
equation in the fluid domain
∇2 𝜙 = 0 (2.15)
The linear free-surface condition
𝜙𝑧 − 𝜈𝜙 = 0 (2.16)
On z = 0, and a condition on the sea bottom
∇ϕ → 0 (2.17𝑎)
As 𝑧 → −∞, or
𝜙𝑧 = 0 (2.17𝑏)
On 𝑧 → −ℎ, for infinite- and finite-water depth, respectively.
Moreover, the scattering and the radiation potentials are subject to a radiation condition stating
that the wave energy associated with the disturbance due to the body is carried away from the
body in all directions in the far field.
Finally, the conditions on the body surface complete the description of the boundary value
problem.

17
 2 - Theory

They take the form

𝜕𝜙𝑘
= 𝑛𝑘 (2.18)
𝜕𝑛
And
𝜕𝜙𝑆 𝜕𝜙𝐼
=− (2.19)
𝜕𝑛 𝜕𝑛
Where (n1, n2, n3) = n and (n4, n5, n6) = x x n. The vector of unitary value n is normal to the
body boundary, and it’s assumed it points out of the fluid domain. X is the position of a point
on the body boundary. From (2.13), it follows that
𝜕𝜙𝐷
=0 (2.20)
𝜕𝑛

2.4 INTEGRAL EQUATIONS

The boundary value problem is then solved by the integral equation method. Therefore, the
velocity potential ϕk on the body boundary is obtained from the equation
𝜕𝐺(𝜉; 𝑥)
2𝜋𝜙𝑘 (𝑥) + ∬ 𝑑𝜉𝜙𝑘 (𝜉) = ∬ 𝑑𝜉𝑛𝑘 𝐺(𝜉; 𝑥) (2.21)
𝑆𝐵 𝜕𝑛𝜉 𝑆𝐵

Where SB denotes the body boundary.

The corresponding equation for the diffraction velocity potential ϕD is
𝜕𝐺(𝜉; 𝑥)
2𝜋𝜙𝐷 (𝑥) + ∬ 𝑑𝜉𝜙𝐷 (𝜉) = 4𝜋𝜙𝐼 (𝑥) (2.22)
𝑆𝐵 𝜕𝑛𝜉
Instead, the scattering potential can be obtained from
𝜕𝐺(𝜉; 𝑥) 𝜕𝜙𝐼
2𝜋𝜙𝑆 (𝑥) + ∬ 𝑑𝜉𝜙𝑆 (𝜉) = ∬ 𝑑𝜉 (− ) 𝐺(𝜉; 𝑥) (2.23)
𝑆𝐵 𝜕𝑛𝜉 𝑆𝐵 𝜕𝑛
And the diffraction potential follows from (2.13). From a computational point of view, (2.22)
is more efficient than (2.23) due to the simpler form of the right-hand side.
The velocity may be computed from the particular derivatives of Green’s integral equations,
(2.21) to (2.23). However, when these equations are solved using a low-order panel method like
in BEMUse, they cannot predict the velocity accurately on or close to the body boundary.
Therefore, the fluid velocity is computed based on source formulation. The integral equation
for the source strength σk corresponding to the radial potential ϕk takes the form
𝜕𝐺(𝜉; 𝑥)
2𝜋𝜎𝑘 (𝑥) + ∬ 𝑑𝜉𝜎𝑘 (𝜉) = 𝑛𝑘 (2.24)
𝑆𝐵 𝜕𝑛𝑥
And that of σS corresponding to the scattering potential ϕS
𝜕𝐺(𝜉; 𝑥) 𝜕𝜙𝐼
2𝜋𝜎𝑆 (𝑥) + ∬ 𝑑𝜉𝜎𝑘 (𝜉) =− (2.25)
𝑆𝐵 𝜕𝑛𝑥 𝜕𝑛

18
 2 - Theory

The velocity of the fluid on the body boundary or in the fluid domain due to ϕk or ϕS is then
obtained from

∇𝜙(𝑥) = ∇ ∬ 𝑑𝜉𝜎(𝜉)𝐺(𝜉; 𝑥) (2.26)

𝑆𝐵

The fluid velocity due to the incident wave is evaluated directly from (2.9).
Integral equations (2.21) to (2.25) are solved by the panel method, where a high number of
quadrilateral panels represent the wetted body surface. The unknowns are assumed to be
constant on each panel, and the integral equation is enforced at the centroid of each of them.
As an example, the discrete form of the equation (2.23) takes the form
𝑁𝐸𝑄𝑁 𝑁𝐸𝑄𝑁
𝜕𝐺(𝜉; 𝑥𝑘 ) 𝜕𝜙𝐼 (𝑥𝑛 )
2𝜋𝜙𝑆 (𝑥𝑘 ) + ∑ 𝜙𝑆 (𝑥𝑛 ) ∫ 𝑑𝜉 = ∑ − ∫ 𝑑𝜉𝐺(𝜉; 𝑥𝑘 ) (2.27)
𝑆𝑛 𝜕𝑛𝜉 𝜕𝑛 𝑆𝑛
𝑛=1 𝑛=1

Where NEQN is the total number of panels (unknowns), and xk are the coordinates of the
centroid of the k-th panel.

2.5 GREEN’S THEOREM AND DISTRIBUTION OF SINGULARITIES

Given the basilar importance of the Green function in the solution of hydrodynamic diffraction
and radiation problems, a more in-depth overview of the topic is necessary. For this scope,
Newman’s book Marine Hydrodynamics [13] provides sufficient information.
Certain immersed bodies could be represented by a point dipole. Given the effect of these
singularities on the surrounding fluid, it seems plausible that a larger number of them could be
used to represent more complicated body shapes. For the cases analyzed here, a distribution on
the body surface only is usually enough.
Let’s consider two solutions of Laplace’s equation in a volume V of fluid bounded by a closed
surface S. Denoting these two potentials by φ and ϕ and applying the divergence theorem we
get

𝜕𝜑 𝜕𝜙
∬ [𝜙 − 𝜑 ] 𝑑𝑆 = ∭ ∇ ∙ (𝜙∇𝜑 − 𝜑∇𝜙)𝑑𝑉 =
𝜕𝑛 𝜕𝑛
𝑆 𝑉

= ∭(𝜙∇2 𝜑 + ∇𝜙 ∙ ∇𝜑 − 𝜑∇2 𝜙 − ∇𝜑 ∙ ∇𝜙)𝑑𝑉 = 0 (2.28)

𝑉

This significant result is a form of Green’s theorem that will be utilized later. Now let’s consider
the consequence of replacing ϕ by the potential of a source.
For this analysis, the source strength is m = 1. Of more significance is the position of this unit
source, which must be carefully specified. The source point is therefore defined as ξ = (ξ, η, ζ)
as the position of the source in the coordinates x = (x, y, z). For a unit source, the potential at
the field point x is given by
1 1 1
𝜙= = ( ) [(𝑥 − 𝜉)2 + (𝑦 − 𝜂)2 + (𝑧 − 𝜁)2 ]−2 (2.29)
4𝜋𝑟 4𝜋
19
 2 - Theory

As easily predictable, the value of (2.29) is unchanged if the source point and the field point are
interchanged. At the same time, (2.29) is a solution of Laplace’s equation with respect to ξ as
well as x. Thus, in utilizing this source potential in Green’s theorem (2.28), it’s possible to
integrate with respect to either coordinate system.
In the subsequent derivation, the integration of (2.28) is performed for the coordinates of the
source point ξ. This requires the potential ϕ and the normal derivative 𝜕𝜙/𝜕𝑛 to be defined with
respect to (ξ, η, ζ) by a simple change of the dummy variable of integration. Physically, what’s
being performed is the integration over a continuous distribution of sources and normal dipoles
that are located on the surface S, with a fixed value of the field point x.
Substituting (2.29) in (2.28) requires caution, for the source potential does not satisfy the
Laplace equation at the singular point r = 0, and thus (2.28) is not valid when the source point
is situated within V. This difficulty can be avoided by surrounding the source point by a small
sphere of radius r = ε with surface Sε as shown in the figure below.

Figure 2.1. In (a), the field point is interior to S; in (b), it’s on the boundary surface and Sε is a hemisphere.

Then S + Sε is a closed surface surrounding the volume of the fluid interior to S, but exterior to
Sε; within the volume, (2.29) is regular. Thus, (2.28) can be replaced

1 𝜕 1 1 𝜕𝜙
∬ [𝜙 − ] 𝑑𝑆 = 0 (2.30)
4𝜋 𝜕𝑛 𝑟 𝑟 𝜕𝑛
𝑆+𝑆𝜀

Or

1 𝜕 1 1 𝜕𝜙 1 𝜕 1 1 𝜕𝜙
∬ [𝜙 − ] 𝑑𝑆 = − ∬ [𝜙 − ] 𝑑𝑆 (2.31)
4𝜋 𝜕𝑛 𝑟 𝑟 𝜕𝑛 4𝜋 𝜕𝑛 𝑟 𝑟 𝜕𝑛
𝑆 𝑆𝜀

In the limit r → ε, the contribution from the integral over Sε in (2.31) can be evaluated under
the assumption that the velocity potential and its normal derivative on Sε are both regular. The
area of Sε is 4πr2, while the normal derivative of 1/r is −𝜕/𝜕𝑟(1/𝑟) = 1/𝑟 2 . Thus, the first
term in the integrand on the right-hand side of (2.31) is singular in proportion to 1/r2, and when
it is multiplied by the area 4πr2, a finite limit will result as ε → 0, whereas the weaker singularity
of the second term will give no contribution.

20
 2 - Theory

For sufficiently small ε, the potential ϕ may be assumed constant and taken outside of the
integral sign, so that the final limiting value of the right-hand side of (2.31) becomes

1 𝜕 1
− 𝜙(𝑥, 𝑦, 𝑧) ∬ 𝑑𝑆 = −𝜙(𝑥, 𝑦, 𝑧) (2.32)
4𝜋 𝜕𝑛 𝑟
𝑆𝜀

Thus, if (x, y, z) is inside S,

1 𝜕 1 1 𝜕𝜙
𝜙(𝑥, 𝑦, 𝑧) = − ∬ [𝜙 − ] 𝑑𝑆 (2.33)
4𝜋 𝜕𝑛 𝑟 𝑟 𝜕𝑛
𝑆

Equation (2.33) can be regarded as a representation of the velocity potential in terms of a

standard dipole distribution of moment ϕ and a source distribution of strength −𝜕𝜙/𝜕𝑛
distributed over the boundary surface S. Thus, in general, the flow can be represented by a
suitable distribution of dipoles and sources on S.
If the point (x, y, z) is located on the surface S, the surface Sε is chosen to be a small hemisphere
that indents the original surface S inside the source point, as in the figure above. The
contribution form of this hemisphere is just half that given by (2.32), so (2.33) is replaced by

1 𝜕 1 1 𝜕𝜙
𝜙(𝑥, 𝑦, 𝑧) = − ∬ [𝜙 − ] 𝑑𝑆 (2.34)
2𝜋 𝜕𝑛 𝑟 𝑟 𝜕𝑛
𝑆

Here, the surface integral must be defined to exclude the immediate vicinity of the singular
point, i.e. the locally-plane infinitesimal surface bounded by the intersection of S and the
hemisphere Sε. This situation is analogous to a principal value integral, except that in (2.34) the
precise shape of the excluded infinitesimal area is not essential.
Equation (2.34) is frequently used for constructing the velocity potential due to the motion of a
ship hull or other moving bodies. Generally, the normal derivative 𝜕𝜙/𝜕𝑛 is known on the
body, so that (2.34) is an integral equation for the determination of the unknown potential, and
it may be solved by numerical techniques like the one used by BEMUse.
In many situations, however, the body may move in a fluid bounded by other boundaries such
as the free surface, the fluid bottom, or possibly lateral boundaries such as canal walls. In each
of these cases, additional boundary conditions are imposed, and there is often a computational
advantage in solving (2.34) if the source potential is modified to satisfy the same boundary
conditions as ϕ. In this context the Green function
1
𝐺(𝑥, 𝑦 , 𝑧; 𝜉, 𝜂, 𝜁) = + 𝐻(𝑥, 𝑦 , 𝑧; 𝜉, 𝜂, 𝜁) (2.35)
𝑟
can be substituted for the source potential in (2.33-2.34) since (2.28) is valid for the contribution
from the regular and generic function H, which satisfies the Laplace equation. Thus, with the
Green function defined by (2.35), we can state that
0
𝜕𝐺 𝜕𝜙
∬ (𝜙 − 𝐺 ) 𝑑𝑆 = {−2𝜋𝜙(𝑥, 𝑦, 𝑧)} (2.36)
𝜕𝑛 𝜕𝑛
𝑆 −4𝜋𝜙(𝑥, 𝑦, 𝑧)

21
 2 - Theory

For (x, y, z) outside, on, or inside the closed surface S. The regular function H can be chosen to
suit any additional boundary conditions that may be imposed. If H can be found with the
property that
𝜕𝐺
=0 (2.37)
𝜕𝑛
on the boundary surfaces of the fluid, the unknown term in the integrand of (2.36) vanishes.
With this choice of the Green function, (2.36) provides an explicit solution for the potential in
terms of the prescribed normal velocity on the boundaries. In the next chapter, a more detailed
analysis of the Green function used by BEMUse is given.

2.5.1 The Green function

The Green function G (x; ξ), also known as the wave source potential, is the velocity potential
at the point x due to a point source of strength -4π located at the point ξ. It satisfies the free-
surface and radiation conditions, and in infinite water depth it is defined by Wehausen and
Laitone [14]
1 1 2𝜈 ∞ 𝑒 𝑘(𝑧+𝜁)
𝐺(𝑥; 𝜉) = + ′ + ∫ 𝑑𝑘 𝐽 (𝑘𝑅) (2.38)
𝑟 𝑟 𝜋 0 𝑘−𝜈 0
𝑟 2 = (𝑥 − 𝜉)2 + (𝑦 − 𝜂)2 + (𝑧 − 𝜁)2 (2.39)
2
𝑟 ′ = (𝑥 − 𝜉)2 + (𝑦 − 𝜂)2 + (𝑧 + 𝜁)2 (2.40)
Where J0(x) is the Bessel function of zero order, while r and r’ are the geometrical distances
between the source and dipole (the former), and between the reflection of the source above sea
level and the dipole (the latter). In finite depth, it’s defined by
∞ (𝑘 + 𝜈) cosh 𝑘(𝑧 + ℎ) cosh 𝑘(𝜁 + ℎ) −𝑘ℎ
1 1
𝐺(𝑥; 𝜉) = + ′′ + 2 ∫ 𝑑𝑘 𝑒 𝐽0 (𝑘𝑅) (2.41)
𝑟 𝑟 0 𝑘 sinh 𝑘ℎ − 𝜈 cosh 𝑘ℎ
(𝑟 ′′ )2 = (𝑥 − 𝜉)2 + (𝑦 − 𝜂)2 + (𝑧 + 𝜁 + 2ℎ)2 (2.42)
In both expressions (2.37) and (2.40), the Fourier k-integration is indented above the pole on
the real axis to enforce the radiation condition.
In the equation (2.27), the influence due to the continuous distribution of the Rankine part of
the wave source potential on a quadrilateral panel is evaluated with the algorithms described by
Newman (1985)[15]. The remaining wave part of the Green function is assessed based on the
algorithms contained in Newman (1992) [16]. The integration of the last part over a panel is
performed using the single-point Gaussian quadrature. When the field and source points are
close together, and in the vicinity of the free surface, G takes a form
1 1
𝐺(𝑥; 𝜉) = + ′ − 2𝜈𝑒 𝜈(𝑧+𝜁) (log(𝑟 ′ + |𝑧 + 𝜁|) + (𝛾 − 𝑙𝑜𝑔2) + 𝑟 ′ + 𝑂(𝑟 ′ 𝑙𝑜𝑔 𝑟 ′ )) (2.43)
𝑟 𝑟
Where γ is the Euler constant. The logarithmic singularity in (2.43) cannot be evaluated
accurately using Gauss’s quadrature in this situation. The algorithm to assess the influence of
the logarithmic singularity distributed over a panel can be found in Newman and Sclavounos
(1988) [17]. When the panel is on the free surface, special attention is required for the evaluation
of Green function, as is discussed in Newman (1993) [18] and Zhu (1994) [19].

22
 2 - Theory

2.6 FIRST ORDER FORCE

The expression for the first-order force is derived from direct integration of the fluid pressure
over a body boundary. By making use of Green’s theorem, part of the forces can be obtained
without solving the scattering potential.

2.7 HYDRO-STATIC AND -DYNAMIC FORCE AND MOMENT

2.7.1 Coordinate system

Three coordinate systems are considered: X = (X, Y, Z) is a global coordinate system with Z =
0 being the undisturbed free surface, and the positive Z-axis points upward; x = (x, y, z) is a
body-fixed coordinate system, z also is positive upwards when the body is at rest; the third
coordinate system is fixed in space and coincides with “x at rest”. It is denoted by x’ = (x’,y’,z’).
It’s important to note that X and x’ are inertial reference frames, while x is not. The origin of x’
may be displaced from the free-surface and Z0 denotes the Z-coordinate of the origin of x’.

2.7.2 Coordinate transform

The position vectors in the x'- and x- coordinate systems are related to each other by a linear
transformation
𝑥′ = 𝜉 + 𝑇𝑡 (2.44)
And the normal vector by
𝑛′ = 𝑇 𝑡 (2.45)
Tt is the transpose of T = T3T2T1 and T1, T2 and T3 take forms
1 0 0
𝑇1 = (0 cos 𝛼1 sin 𝛼1 )
0 − sin 𝛼1 cos 𝛼1
cos 𝛼2 0 − sin 𝛼2
𝑇2 = ( 0 1 0 ) (2.46)
sin 𝛼2 0 cos 𝛼2
cos 𝛼3 sin 𝛼3 0
𝑇2 = (− sin 𝛼3 cos 𝛼3 0)
0 0 1
ξ = (ξ1, ξ2, ξ3) and α = (α1, α2, α3) are, respectively, the translational and rotational displacements
of the x-coordinate system with respect to x’. They also represent the motion amplitudes of the
body for surge-sway-heave and roll-pitch-yaw.

23
 2 - Theory

2.7.3 Pressure integration

All BEMUse’s derivations (pressure, velocity potential, and body motion amplitudes) are a
function of time, even though the time t does not appear explicitly.
The pressure at x’ is given by Bernoulli’s equation
1
𝑃(𝑥 ′ ) = −𝜌 [𝜙𝑡 (𝑥 ′ ) + ∇𝜙(𝑥 ′ ) ∙ ∇𝜙(𝑥 ′ ) + 𝑔(𝑧 ′ + 𝑍0 )] (2.47)
2
The hydrostatic and hydrodynamic force and moment are obtained from the integration of the
pressure over the instantaneous wetted surface

𝐹 = ∬ 𝑃(𝑥 ′ )𝑛′ 𝑑𝑆 (2.48)

𝑆′𝐵

And

𝑀 = ∬ 𝑃(𝑥 ′ )(𝑥 ′ × 𝑛′ )𝑑𝑆 (2.49)

𝑆′ 𝐵

Where SB’ denotes the instantaneous wetted body boundary.

The pressure on the exact body surface (x’ ∈ SB’) may be approximated by Taylor expansion
with respect to the mean body surface (x ∈ SB’)
𝑃(𝑥 ′ ) = 𝑃(𝑥) + [𝜉 + (𝑇 𝑡 − 𝐼)𝑥] ∙ ∇𝑃(𝑥) + ⋯ (2.50)
Substituting (2.50) into (2.44-2.45) we have
𝑥 ′ = 𝑥 + 𝜉 (1) + 𝛼 (1) × 𝑥 + 𝐻𝑥 + 𝜉 (2) + 𝛼 (2) × 𝑥 + 𝑂(𝐴3 ) (2.51)
𝑛′ = 𝑛 + 𝛼 (1) × 𝑛 + 𝐻𝑛 + 𝛼 (2) × 𝑛 + 𝑂(𝐴3 ) (2.52)
The cross product of (2.51) and (2.52) then takes a form

𝑥 ′ × 𝑛′ = 𝑥 × 𝑛 + 𝜉 (1) × 𝑛 + 𝛼 (1) × (𝑥 × 𝑛) + 𝜉 (1) × (𝛼 (1) × 𝑛) +

+ 𝐻(𝑥 × 𝑛) + 𝜉 (2) × 𝑛 + 𝛼 (2) × (𝑥 × 𝑛) + 𝑂(𝐴3 ) (2.53)
From (2.47), (2.50) and (2.53), the pressure is expressed by the values on the mean body
position
(1) (1) (1) (1)
𝑃(𝑥 ′ ) = −𝜌{𝑔(𝑧 + 𝑍0 ) + [𝜙𝑡 (𝑥) + 𝑔(𝜉3 + 𝛼1 𝑦 − 𝛼2 𝑥)] +
1 (1)
+ [ ∇𝜙 (1) (𝑥) ∙ ∇𝜙 (1) (𝑥) + (𝜉 (1) + 𝛼 (1) × 𝑥) ∙ 𝜙𝑡 (𝑥) + 𝑔𝐻𝑥 ∙ ∇𝑧] + (2.54)
2
(2) (2) (2) (2)
+ [𝜙𝑡 (𝑥) + 𝑔(𝜉3 + 𝛼1 𝑦 − 𝛼2 𝑥)]} + 𝑂(𝐴3 )

24
 2 - Theory

H here is the second-order component of Tt, and it takes a form

1 (1) 2 (1) 2
− ((𝛼2 ) + (𝛼3 ) ) 0 0
2
(1) (1) 1 (1) 2 (1) 2
𝐻= 𝛼1 𝛼2 − ((𝛼1 ) + (𝛼3 ) ) 0 (2.55)
2
(1) (1) (1) (1) 1 (1) 2 (1) 2
( 𝛼1 𝛼3 𝛼2 𝛼3 − ((𝛼1 ) + (𝛼2 ) ))
2
Substituting the expansions (2.50-2.54) into (2.48-2.49), we have the series expansion of the
integrals (2.48-2.49).

2.7.4 Hydrostatic force and moment of O (1)

These are the buoyancy forces and moments when the body is at rest and are expressed by

𝐹 = −𝜌𝑔 ∬ (𝑧 + 𝑍0 )𝑛𝑑𝑆 (2.56𝑎)

𝑆𝐵

𝑀 = −𝜌𝑔 ∬ (𝑧 + 𝑍0 )(𝑥 × 𝑛)𝑑𝑆 = − 𝜌𝑔 ∬ [(𝑧 + 𝑍0 )𝑥] × 𝑛𝑑𝑆 (2.56𝑏)

𝑆𝐵 𝑆𝐵

The following relations are invoked frequently

− ∬ 𝜓𝑛𝑑𝑆 = ∭ ∇𝜓𝑑𝑉
𝑆 𝑉

− ∬ 𝑛 × 𝜓𝑑𝑆 = ∭ ∇ × 𝜓𝑑𝑉 (2.57)

𝑆 𝑉

∇ × (𝑧 + 𝑍0 )𝑥 = 𝑘 × 𝑥
Here S is a closed surface consisting of SB and the waterplane area Awp. V denotes the volume
of the body.
Using the relations (2.57), the force and moment are expressed in familiar forms
𝐹 = 𝜌𝑔𝑉𝑘 (2.58𝑎)
𝑀 = 𝜌𝑔𝑉(𝑦𝑏 𝑖 − 𝑥𝑏 𝑗) (2.58𝑏)
Where xb and yb are the coordinates of the center of buoyancy.
In (2.58a-2.58b), i, j, and k are the unit vectors in x’ coordinate system.

25
 2 - Theory

2.8 LINEAR FORCE AND MOMENT

The linear force and moment are found from

(1)
𝐹 = −𝜌 ∬ 𝑛𝜙𝑡 𝑑𝑆 − 𝜌𝑔 ∬ (𝛼 (1) × 𝑛)(𝑧 + 𝑍0 )𝑑𝑆 +
𝑆𝐵 𝑆𝐵

(1) (1) (1)

−𝜌𝑔 ∬ 𝑛(𝜉3 + 𝛼1 𝑦 − 𝛼2 𝑥)𝑑𝑆 (2.59𝑎)
𝑆𝐵

(1) (1) (1) (1)

𝑀 = − 𝜌 ∬ (𝑥 × 𝑛) 𝜙𝑡 𝑑𝑆 − 𝜌𝑔 ∬ (𝑥 × 𝑛)(𝜉3 + 𝛼1 𝑦 − 𝛼2 𝑥)𝑑𝑆 +
𝑆𝐵 𝑆𝐵

−𝜌𝑔 ∬ (𝜉 (1) × 𝑛)(𝑧 + 𝑍0 )𝑑𝑆 − 𝜌𝑔 ∬ [𝛼 (1) × (𝑥 × 𝑛)](𝑧 + 𝑍0 )𝑑𝑆 (2.59𝑏)

𝑆𝐵 𝑆𝐵

Where the first terms are the hydrodynamic force and moment, and the others are the hydrostatic
ones. Following the decomposition (2.13), we consider component potentials such that
ϕ(1) = ϕI(1) + ϕS(1) + ϕR(1) = ϕD(1) + ϕR(1). The hydrodynamic forces and moments are divided into
two components: the “wave exciting force” due to ϕD(1) and the force due to ϕR(1) expressed as
the added mass and damping coefficients.
The integrals to evaluate the hydrostatic pressure can be simplified by applying (2.57) and their
variations with the results

(1) (1) (1) (1)

𝐹 = −𝜌 ∬ 𝜙𝑡 𝑛𝑑𝑆 − 𝜌𝑔𝐴𝑤𝑝 (𝜉3 + 𝛼1 𝑦𝑓 − 𝛼2 𝑥𝑓 )𝑘 (2.60𝑎)
𝑆𝐵

(1)
𝑀 = −𝜌 ∬ (𝑥 × 𝑛)𝜙𝑡 𝑑𝑆 +
𝑆𝐵
(1) (1) (1) (1) (1)
−𝜌𝑔[−𝑉𝜉2 + 𝐴𝑤𝑝 𝑦𝑓 𝜉3 + (𝑉𝑧𝑏 + 𝐿22 )𝛼1 − 𝐿12 𝛼2 − 𝑉𝑥𝑏 𝛼3 ]𝑖
(1) (1) (1) (1) (1)
−𝜌𝑔[−𝑉𝜉1 + 𝐴𝑤𝑝 𝑥𝑓 𝜉3 − 𝐿12 𝛼1 + (𝑉𝑧𝑏 + 𝐿11 )𝛼2 − 𝑉𝑦𝑏 𝛼3 ]𝑗 (2.60𝑏)

Where Lij is the second moment over the waterplane area. For example, 𝐿12 = ∬𝐴 𝑥𝑦𝑑𝑆.
𝑤𝑝
xf and yf are the coordinates of the center of floatation.
Following (2.8), the force can be represented by a discrete spectrum (the moment takes an
identical form and is omitted here)

𝐹 = 𝑅𝑒 ∑ 𝐹𝑗 𝑒 𝑖𝜔𝑗𝑡 (2.61)
𝑗

26
 2 - Theory

2.9 IRREGULAR FREQUENCIES

The Green function satisfies the free surface condition everywhere on the plane of the free
surface, both outside and inside the body. Consequently, for bodies that intersect that surface,
a discrete spectrum of irregular frequencies exists, where the solutions of the boundary integral
equations either do not exist or are non-unique. The problem of irregular frequencies and its
solution was thoroughly examined in Lee, Newman, and Zhu (1996) [20].
These irregular frequencies coincide with the eigenfrequencies of non-physical wave motions
inside the body, where the same free surface condition is imposed on the interior free surface,
and a homogeneous Dirichlet boundary condition is imposed on the body surface.
The connection between the irregular frequencies and the interior Dirichlet problem follows by
noting that this problem can be solved in general by either the potential or source formulation,
the one BEMUse uses.
However, the corresponding fluid motion inside the body, subject to a Dirichlet condition,
would have a set of resonance frequencies at which non-trivial homogeneous solutions exist.
The potential and source formulations for the exterior boundary value problem break down at
the same frequencies owing to the relation of their kernels, so that of the integral equation for
the interior Dirichlet problem. At the irregular frequencies, the Fredholm determinants of the
two formulations vanish, and thus the solutions must be either non-unique or non-existent. The
same set of irregular frequencies applies to the potential and source formulations and the
second-order solution, since each of the corresponding integral equations has the same kernel
or its transpose.
In the discrete problem (a consequence of the panel method), the condition number of the linear
system increases near the irregular frequencies, indicating that the system is ill-conditioned in
the vicinity of these frequencies, where the numerical solutions are wrong. The width of the
wrong solution in terms of frequency interval may be reduced increasing the number of panels
on the body, although this has a worsening effect on the computational cost of the calculations;
moreover, it's often impossible to predict which frequencies generate these irregularities.
Therefore, it's necessary to develop modified integral equations which avoid the effects of the
irregular frequencies in practical applications.
Since the occurrence of the irregular frequencies is associated with the interior Dirichlet
problem, their effects in waters of infinite or finite depth are similar.
Green’s theorem yields a Fredholm integral equation for the unknown ϕ(x) over the domain Sb
𝜕𝐺(𝑥; 𝜉)
2𝜋𝜙(𝑥) + ∬ 𝜙(𝜉) 𝑑𝜉 = ∬ 𝑞𝑏 (𝜉)𝐺(𝑥; 𝜉)𝑑𝜉 + 𝐹(𝑥) (2.62)
𝑆𝑏 𝜕𝑛𝜉 𝑆𝑏

Where
1
𝐹(𝑥) = ∬ 𝑞 (𝜉)𝐺(𝑥; 𝜉)𝑑𝜉 (2.63)
𝑔 𝑆𝑓 𝑓

The normal vector n points out of the fluid domain, and Sf denotes the free surface exterior to
the body.

27
 2 - Theory

While it’s generally assumed that solutions of the boundary value problem for the potential ϕ
are unique, in the unbounded fluid domain exterior to the body, it can be shown that non-trivial
homogeneous solutions of (2.62) exist at the irregular frequencies. The existence of these
solutions is associated with the non-physical portion of the free surface domain interior to the
body in the definition of the Green function.
There are two methods to solve the problem of irregular frequencies, the Modified Green
Function method, and the Extended Boundary Condition method. Since BEMUse uses the
latter, the next paragraph will solely focus on that one.

2.9.1 Removal of irregular frequencies

The Extended Boundary method is used by WAMIT as well, and the procedure is explained
well in its Theory Manual [12].
The extended boundary integral equations for ϕk are
𝜕𝐺(𝑥; 𝜉) 𝜕𝐺(𝑥; 𝜉)
2𝜋𝜙𝑘 (𝑥) + ∬ 𝜙𝑘 (𝜉) 𝑑𝜉 + ∫ 𝜙 ′ 𝑘 (𝜉) 𝑑𝜉 = ∬ 𝑛𝑘 𝐺(𝑥; 𝜉)𝑑𝜉 (2.64𝑎)
𝑆𝑏 𝜕𝑛𝜉 𝑆𝑓 𝜕𝑛𝜉 𝑆𝑏

𝜕𝐺(𝑥; 𝜉) 𝜕𝐺(𝑥; 𝜉)
−4𝜋𝜙𝑘′ (𝑥) + ∬ 𝜙𝑘 (𝜉) 𝑑𝜉 + ∫ 𝜙 ′ 𝑘 (𝜉) 𝑑𝜉 = ∬ 𝑛𝑘 𝐺(𝑥; 𝜉)𝑑𝜉 (2.64𝑏)
𝑆𝑏 𝜕𝑛𝜉 𝑆𝑓 𝜕𝑛𝜉 𝑆𝑏

Where ϕ’k is an artificial velocity potential defined in the interior domain. Equations (2.64a)
and (2.64b) are for x on Sb and Si, respectively. These equations are solved simultaneously for
ϕ k on Sb and ϕ’k on Si. ϕ’k is discarded after their solution since only ϕk on Sb is physically
relevant.
The equations for the total diffraction potential ϕD are
𝜕𝐺(𝑥; 𝜉) 𝜕𝐺(𝑥; 𝜉)
2𝜋𝜙𝐷 (𝑥) + ∬ 𝜙𝐷 (𝜉) 𝑑𝜉 + ∫ 𝜙 ′ 𝐷 (𝜉) 𝑑𝜉 = 4𝜋𝜙𝐼 (𝑥) (2.65𝑎)
𝑆𝑏 𝜕𝑛𝜉 𝑆𝑓 𝜕𝑛𝜉

𝜕𝐺(𝑥; 𝜉) 𝜕𝐺(𝑥; 𝜉)
−4𝜋𝜙𝐷′ (𝑥) + ∬ 𝜙𝐷 (𝜉) 𝑑𝜉 + ∫ 𝜙 ′ 𝐷 (𝜉) 𝑑𝜉 = 4𝜋𝜙𝐼 (𝑥) (2.65𝑏)
𝑆𝑏 𝜕𝑛𝜉 𝑆𝑓 𝜕𝑛𝜉

The scattering potential has the same form as the radiation potential but −𝜕𝜙𝐼 /𝜕𝑛 replaces nk
on the right-hand side of the equations (2.64a) and (2.64b).
The extended boundary integral equation for the source formulation of the radiation or
scattering problem takes a form
𝜕𝐺(𝑥; 𝜉) 𝜕𝐺(𝑥; 𝜉)
2𝜋𝜎(𝑥) + ∬ 𝜎(𝜉) 𝑑𝜉 + ∫ 𝜎 ′ 𝐷 (𝜉) 𝑑𝜉 = 𝑔(𝑥) (2.66𝑎)
𝑆𝑏 𝜕𝑛𝜉 𝑆𝑓 𝜕𝑛𝜉
𝜕𝐺(𝑥; 𝜉)
𝜕𝐺(𝑥; 𝜉) ′(𝜉)
𝜕𝑛𝜉
−4𝜋𝜎𝐷′ (𝑥) + ∬ 𝜎(𝜉) 𝑑𝜉 + ∫ 𝜎 𝑑𝜉 = −𝑉(𝑥) (2.66𝑏)
𝑆𝑏 𝜕𝑛𝜉 𝑆𝑓

Where 𝑔(𝑥) = 𝑛𝑘 or 𝑔(𝑥) = −𝜕𝜙𝐼 /𝜕𝑛 for the radiation (scattering) problem.

28
 2 - Theory

A detailed analysis of the derivation of these equations is described, among others, by Zhu
(1994) [19]. There, the author explains that because the kernel of the modified source integral
equations is the transpose of those in the modified potential integral equations, the
homogeneous solutions of the source integral equations should be trivial. Therefore, the
extended boundary condition method in source formulation can also remove the irregular
frequencies.
The solution of the integrals implies that the potential and velocity inside the body is zero.
When we apply the extensions to the original boundary integral equations, the inside potential
property is enforced in the new boundary integral equations, thus the potential inside the body
is homogeneous for the integral equation of Extended Boundary Condition method. Therefore,
the left-hand side of the equations (2.66a) and (2.66b) must be set equal to zero.

29
30
 3 – Description of BEMUse

3. DESCRIPTION OF BEMUse

As previously stated, BEMUse is a hydrodynamic analysis model created at the Hermann-

Föttinger Institut at the Technische Universität Berlin and developed inside QBlade software.
While the previous chapter illustrated the theory on which the model lays on, in this chapter an
overview of how BEMUse operates is given.

3.1 METHOD OF SOLUTION

To achieve the complete hydrodynamic analysis of a wind turbine platform, BEMUse uses the
Boundary Element Method (BEM). This, often referred to as a low-order panel method, allows
us to discretize the solving partial differential equations using the definitions for the
perturbation potentials for surface dipoles distributions and source distributions, and
numerically solve for a finite number of degrees of freedom, i.e. panels. From the solution, the
velocity vector field and the pressure distribution can be obtained. From the surface pressure
distribution, the forces and moments acting on the submerged body are then calculated [21].
The critical property of the panel method is that the solution of the flow equations (obviously
in 3D space) is wholly determined by the solution on the surface of the body only. This makes
the geometric modeling of the problem a relatively simple task, as the geometry is described in
terms of a body consisting of structured patches of surface grid, the so-called panels in the
method. The figure below shows an example of paneling for a hemispherical platform, as
generated by BEMUse.

Figure 3.1. A hemisphere discretized with quadrilateral flat panels by BEMUse

31
 3 – Description of BEMUse

In a low order panel method, the error in the solution is of O(h) and thus linearly dependent on
a characteristic panel size h.
Dirichlet boundary conditions are enforced at a finite number of discrete points, so-called
collocation points that are located right below the surface of the body panel centroids. Some
characteristics of the approach followed are:
- Structured grids of quadrilateral panels represent body surface geometries, and the
arbitrary translational and rotational motions of the geometry are specified in a
hierarchical setup, which will be described afterward.
- Each panel on the body surface is assigned a constant strength dipole distribution and a
constant strength source distribution.
- The perturbation velocity vectors at the panel corner points are determined by a 2D
version of the Gradient Theorem.
- The hydrodynamic force and moment on each panel are determined through the
integration of the surface pressure distribution, obtained by a bi-linear interpolation of
all contributions in the Bernoulli equation that define the pressure. A summation then
yields the force and moment acting on the whole body.

3.2 GEOMETRY
The surface of the body is discretized in a structured surface grid. The grid cells on the surface
of the platform, called panels, are in the boundary integral discretization.
The cross-product of the vectors determines the direction of the vector normal to the surface
and based on the centroid of each panel through diagonally opposite panel corner points.
𝑛⃗ = (𝑥𝑖+1,𝑗 − 𝑥𝑖,𝑗+1 ) × (𝑥𝑖+1,𝑗+1 − 𝑥𝑖,𝑗 ) (3.1)
Where i, j are the indices of each grid node, as BEMUse generates them in a structured order.
The value of 𝑛⃗ is then normalized to give the unit vector 𝑛̅𝑚𝑖𝑑 = 𝑛⃗/|𝑛⃗|. The panel area is
⃗⃗⃗ /2.
approximated by a flat, non-curved panel surface representation and equal to |𝑛|

3.3 SOLID BODY TRANSFORMATION

For the arbitrary motion of the geometry, the technique described in van Garrel (2003) [22] is
used. At each time step, the new constellation of the geometry is determined. Solid body
rotations and translations are performed using 4 x 4 transformation matrices expressed in a
homogeneous coordinate system.
These matrices, which operate on homogeneous coordinates, are composed of the three
components of a Cartesian vector supplemented with a 4th component equal to 1. A position
vector (𝑥, 𝑦, 𝑧)𝑇 is thus extended in the homogeneous form to (𝑥, 𝑦, 𝑧, 1)𝑇 .

32
 3 – Description of BEMUse

The rotation matrix R for rotation of a position vector 𝑥 about a general axis through the origin
with direction 𝜐̅ = (𝜐1 , 𝜐2 , 𝜐3 )𝑇 through an angle θ, interpreted in a right-hand rule sense, reads
𝜐1 𝜐1 𝑑 + 𝑐 𝜐1 𝜐2 𝑑 − 𝜐3 𝑠 𝜐1 𝜐3 𝑑 + 𝜐2 𝑠 0
𝜐 𝜐 𝑑 + 𝜐3 𝑠 𝜐2 𝜐2 𝑑 + 𝑐 𝜐2 𝜐3 𝑑 − 𝜐1 𝑠 0
𝑅=( 2 1 ) (3.2)
𝜐3 𝜐1 𝑑 − 𝜐2 𝑠 𝜐3 𝜐2 𝑑 + 𝜐1 𝑠 𝜐3 𝜐3 𝑑 + 𝑐 0
0 0 0 1
Where
𝑐 = 𝑐𝑜𝑠𝜃
𝑠 = 𝑠𝑖𝑛𝜃 (3.3)
d = 1 − 𝑐𝑜𝑠𝜃
And
𝜐̅ ∙ 𝜐̅ = 1 (3.4)
The rotation of a position vector 𝑥 is now accomplished by a matrix multiplication that gives a
new position vector 𝑥𝑛𝑒𝑤 = 𝑅𝑥 .
Similarly, the translation of a position vector 𝑥 can be performed through multiplication with
the translation matrix T such that 𝑥𝑛𝑒𝑤 = 𝑇𝑥. The translation matrix is formed by
1 0 0 𝑡1
0 1 0 𝑡2
𝑇=( ) (3.5)
0 0 1 𝑡3
0 0 0 1
Where (𝑡1 , 𝑡2 , 𝑡3 )𝑇 is the Cartesian translation vector.
The transformation matrix for a translation in the reverse direction −(𝑡1 , 𝑡2 , 𝑡3 )𝑇 is the inverse
T-1 of matrix T. Rotation about a general axis not through the origin can be performed by a
compound transformation matrix A, which is composed of a translation of the axis to the origin
by T, a rotation R, and a translation T-1 back to the original position:
𝑥𝑛𝑒𝑤 = 𝑇 −1 (𝑅(𝑇𝑥 )) = 𝐴𝑥 (3.6)
For a set of position vectors 𝑥, it is more efficient to compute the compound transformation
matrix A first and perform the matrix multiplication with all position vectors next.

3.4 DIPOLE POTENTIAL AND VELOCITY

The velocity potential induced at point 𝑥 by a panel with surface Sj and dipole distribution 𝜇(𝑦)
is given by

1 𝑛̅ ∙ 𝑟
𝜙𝜇 (𝑥) = − ∬ 𝜇 3 𝑑𝑆 (3.7)
4𝜋 𝑟
𝑆𝑗

Where
𝑟 = 𝑥 − 𝑦, 𝑟 = |𝑟|, 𝑎𝑛𝑑 𝑦 ∈ 𝑆𝑗 (3.8)

33
 3 – Description of BEMUse

The usual approach taken in low-order panel methods is to use a flat surface approximation for
the panel geometry, for which analytical results exist for the integral in equation (3.7) [23]. In
general, however, flat panels lead to gaps between the panels in the surface approximation of a
curved surface, that grow larger with increasing surface curvature and twist.
Panel method is very useful in the case of very large domains, like a floating platform, where a
Finite Element approximation would not be practical [24]. This is because only the boundary
of the domain needs to be discretized.
Another advantage is that in some applications like the one of interest for this study, the
physically relevant data are given not by the solution in the interior of the domain, but rather
by the boundary values of the solution or its derivatives. These data can be obtained directly
from the solution of boundary integral equations, whereas boundary values obtained from FEM
solutions are, in general, not very accurate [25].
To conclude this chapter, a brief overview of the steps followed by a Panel Method software in
case of hydrodynamic diffraction problems is reported [26]:
1) Use Green’s theorem to derive integral equations for velocity potentials on the body
boundary
2) Discretize the body surface by a large number N of panels
3) The sources and dipole moments are assumed constant on each panel total of N
unknowns
4) The potential is evaluated at the centroid of each panel and set equal to the normal
incident potential
5) Solve the system of equations
6) Compute required forces and moments

3.5 INFLUENCE COEFFICIENTS

The source potential, or Green function, is the fundamental analysis of wave-induced motions
and forces acting on floating vessels. It satisfies the linearized boundary condition at the free
surface and the radiation condition. BEMUse’s numerical model is based on the distribution of
sources and dipoles, located on the submerged portion of the body surface. This procedure can
be justified by Green’s theorem and requires the solution of an integral equation in the domain
of the body surface for the velocity potential.
The Green function can be expressed as the sum of the free-space singularity (Rankine source)
and a component that accounts for the free surface. The latter part can be further decomposed
into a non-oscillatory local flow component and a wave component, which is dominant in the
far-field. This one, in its basic decomposition, is defined in terms of special real functions:
Bessel functions J0 and J1 of a single variable.
The non-oscillatory local flow components in the expressions for the Green function and its
gradient are real functions of two variables and are defined by single integrals. The two
variables are the field point (identified by the Cartesian coordinate system (x, y, z)) and the
source point marked by (ξ, η, ζ).

34
 3 – Description of BEMUse

Near-field and far-field approximations to these components are used to remove the near-field
singularities and to reduce the unbounded region to a finite one, making it possible to use
polynomial approximations. These are very well-suited for highly efficient computation.
In its conventional form, taken from Wehausen and Laitone (1960) [14], the source potential is
defined by the expression
∞
1
2 ]−2
𝑘 + 𝐾 𝑘(𝑧+𝜁)
𝐺= [𝑅 2 + (𝑧 − 𝜁) +∫ 𝑒 𝐽0 (𝑘𝑅)𝑑𝑘 (3.9)
0 𝑘−𝐾
Where J0 denotes the Bessel function of the first kind, order zero, and the contour of integration
passes above the pole k to satisfy the radiation condition of outgoing waves at infinity. K is
defined by gravity and frequency as 𝐾 = 𝜔2 /𝑔, which corresponds to the wavenumber in the
infinite-depth case. R is instead the radius expressing the horizontal coordinate defined by the
magnitude of the horizontal vectors with components (x – ξ, y – η)
In Liang, Wu, and Noblesse (2018) [27] it is expressed in a more compact way as
1 1
4𝜋𝐺 = − − + 𝐿 + W (3.10)
𝑟 𝑑
Where L represents a non-oscillatory local flow component, and W is a wave component. This
basic decomposition is not unique.
The gradient of the Green function is given by
𝑧−𝜁 𝜈 1
4𝜋𝐺𝑧 = 3
+ 3 + 𝐿𝑧 + 𝑊 𝑤ℎ𝑒𝑟𝑒 𝐿𝑧 = − + 𝐿 (3.11𝑎)
𝑟 𝑑 𝑑
ℎ ℎ
4𝜋𝐺ℎ = 3 + 3 + 𝐿ℎ + 𝑊ℎ (3.11𝑏)
𝑟 𝑑
𝑥−𝜉 𝑥−𝜂
4𝜋𝐺𝑥 = 𝐺ℎ 𝑎𝑛𝑑 4𝜋𝐺𝑦 = 𝐺ℎ (3.11𝑐)
ℎ ℎ
Where the subscript z, h and x identify the direction of derivation. To calculate the linear and
mean drift wave loads on the platform's surface, a constant panel boundary element method
based on combined source and dipole distributions is used. Irregular frequencies are removed
via the extension of the flow region to the waterplane area. Specifically, these two integral
equations are solved:
1 𝜕𝐺(𝑥, 𝜉) 𝜕𝐺(𝑥, 𝜉) 𝜕𝜙(𝜉)
𝜙(𝑥) + ∬ 𝜙(𝜉) 𝑑𝑆 + ∬ 𝜇(𝜉) 𝑑𝑆 = ∬ 𝐺(𝑥, 𝜉) 𝑑𝑆 (3.12)
2 Σ𝐻 𝜕𝑛𝜉 Σ𝐼 𝜕𝑛𝜉 Σ𝐻 𝜕𝑛𝜉

𝜕𝐺(𝑥, 𝜉) 𝜕𝐺(𝑥, 𝜉) 𝜕𝜙(𝜉)

𝜇(𝑥) + ∬ 𝜙(𝜉) 𝑑𝑆 + ∬ 𝜇(𝜉) 𝑑𝑆 = ∬ 𝐺(𝑥, 𝜉) 𝑑𝑆 (3.13)
Σ𝐻 𝜕𝑛𝜉 Σ𝐼 𝜕𝑛𝜉 Σ𝐻 𝜕𝑛𝜉
On ΣH and ΣI, respectively.
They are needed to determine the flow potential ϕ on the hull surface ΣH of the floating body
and the density µ of the dipole distribution over the interior free surface ΣI. The singular
Rankine source component 1/r + 1/d in (3.10) and its gradient in (3.11a-3.11c) are analytically
integrated over a flat panel, and the weakly-singular local flow component L and its gradient in
(3.10) or (3.11a-3.11c) are numerically integrated via a Gaussian quadrature rule with four
Gaussian points in the nearfield and one point in the far-field [27].

35
 3 – Description of BEMUse

These two integral equations are imposed on each panel of the body surface, generating a linear
system of N equations, where N is the number of panels. This linear system can be expressed
in matrix form:
𝜙 = 𝐴−1 𝐵𝜙 ′ (3.14)
Where A is an N by N matrix whose coefficients are the Green’s function derivative as expressed
in the first integral of equation (3.12), ϕ is the vector of unknowns - the values of the velocity
potential at the centroid of each panel – and B is another N by N matrix containing the values
of the Green function at each centroid. Finally, ϕ’ is another vector, containing the Green’s
function derivative for the direction normal to the panel.

3.6 BOUNDARY CONDITIONS

To solve the linear system, another fundamental element is the imposition of appropriate
boundary conditions to the problem. As Newman (1999) [13] states, two different types of
conditions must be discussed: a kinematic one corresponding to a statement regarding the
velocity of the fluid on the boundary, and a dynamic boundary condition corresponding to an
account of the forces on the boundary.
A kinematic condition is appropriate on any boundary surface with a specified geometry and
position. Whether the boundary is a fixed and rigid body (most straightforward case,
corresponding to the cases studied in this thesis), or one moving with prescribed velocity U
through the fluid, the physically relevant condition for the flow at the boundary is that the
normal component 𝑉 ∙ 𝑛 of the velocity must be equal to the normal velocity 𝑈 ∙ 𝑛 of the
boundary surface itself. In other words, no fluid can flow through the boundary surface. This
condition, also known as zero penetration condition, can be expressed in terms of the velocity
potential as
𝜕𝜙
=𝑈∙𝑛 =0 (3.15)
𝜕𝑛
Where 𝜕/𝜕𝑛 denotes the derivative in the direction of the unitary normal n directed out of the
fluid.
In the absence of viscous shear stresses (inviscid flow), only the normal component of the fluid
velocity is prescribed on the boundary.
As the derivative of the velocity potential is the fluid velocity, this condition can also be written
as
𝜕𝜙
𝜙 ∙ 𝑛⃗ = (3.16)
𝜕𝑛
This means that the fluid's velocity component normal to the surface of the body must be equal
to 0 for each panel. This condition assures that the flow does not penetrate the surface of the
body.
The kinematic boundary condition (3.15) contains precisely the right amount of information
regarding the fluid motion. On the other hand, there will be problems where the position and
velocity of the boundary are unknown; in particular, on the free surface where waves occur and
the form of the wave motion and free surface elevation is not known a priori. In this instance,

36
 3 – Description of BEMUse

the kinematic boundary condition (3.15) remains valid, but the velocity U of the boundary is
unknown. Therefore, additional information must be provided, in the form of the free surface
boundary condition, expressed as
𝜕 2𝜙 𝜕𝜙 𝜕𝜙 1
2
+𝑔 + 2∇𝜙 ∙ ∇ + ∇𝜙 ∙ ∇(∇𝜙 ∙ ∇𝜙) = 0 (3.17)
𝜕𝑡 𝜕𝑧 𝜕𝑡 2
And applied on the exact free surface:
1 𝜕𝜙 1
𝜁(𝑥, 𝑦) = − ( + ∇𝜙 ∙ ∇𝜙) (3.18)
𝑔 𝜕𝑡 2 𝑧=𝜁

This formula can be significantly simplified under the assumptions of small wave slope and
harmonic solution as follows:
𝜕𝜙
− 𝜈𝜙 = 0 (3.19)
𝜕𝑧
Here, ν represents the non-dimensional wave number of the problem (𝜈 = 𝜔2 /𝑔 in the infinite-
depth case, where ω is the wave frequency).
The definition of the boundary conditions wouldn’t be complete without a statement of what
happens far from the body, i.e. the asymptotic behavior when r → ∞, with r the distance from
the evaluation point. Thus, the radiation condition is imposed, stating that the disturbance
potential vanishes for r going to infinity.
Apart from these fundamental boundary conditions required by BEMUse to carry out its
calculations, there are other more general conditions which need to be respected when working
with hydrodynamic modeling softwares.
The pressure is constant across the free surface interface: p = patm on z = η.
𝜕𝜙 1 2
𝑝 = −𝜌 { − 𝑉 − 𝑔𝑧} + 𝑐(𝑡) = 𝑝𝑎𝑡𝑚 (3.20)
𝜕𝑡 2
With the choice a suitable integration constant, c(t) = patm, the boundary condition on z = η
becomes
𝜕𝜙 1 2
𝜌{ + 𝑉 + 𝑔𝜂} = 0 (3.21)
𝜕𝑡 2
Once a particle is on the free surface, it always remains there. Similarly, the normal velocity of
a particle on the surface is equal to the normal velocity of the surface itself.
𝑧𝑝 = 𝜂(𝑥𝑝 , 𝑡) (3.22)
𝜕𝜂 𝜕𝜂
𝑧𝑝 + 𝛿𝑧𝑝 = 𝜂(𝑥𝑝 + 𝛿𝑥𝑝 , 𝑡 + 𝛿𝑡) = 𝜂(𝑥𝑝 , 𝑡) + 𝛿𝑥𝑝 + 𝛿𝑡
𝜕𝑥 𝜕𝑡
On the free surface, where zp = η, we can reduce the above equation to
𝜕𝜂 𝜕𝜂
𝛿𝑧𝑝 = 𝑢𝛿𝑡 + 𝛿𝑡 (3.23)
𝜕𝑥 𝜕𝑡

37
 3 – Description of BEMUse

And substitute δzp = wδt and δxp = uδt to demonstrate that the normal velocity follows the
particle:
𝜕𝜂 𝜕𝜂
𝑤=𝑢 + (3.24)
𝜕𝑥 𝜕𝑡
on z = η.
On an impermeable body boundary B (x, y, z, t) = 0, the velocity of the fluid normal to the body
must be the same as the body velocity in that direction:
𝜕𝜙
𝜈 ∙ 𝑛̂ = 𝜙 ∙ 𝑛̂ = ⃗ (𝑥, 𝑡) ∙ 𝑛̂(𝑥, 𝑡) = 𝑈𝑛
=𝑈 (3.25)
𝜕𝑡
On B = 0.
Alternatively, a particle P on B remains always on B; i.e. B is a material surface.
As an example, if P is on B at some time t = t0 such that
𝐵(𝑥, 𝑡0 ) = 0, 𝑡ℎ𝑒𝑛 𝐵(𝑥, 𝑡0 ) = 0 𝑓𝑜𝑟 𝑎𝑙𝑙 𝑡, (3.26)
So that if we were to follow P, then B = 0 always. Therefore:
𝐷𝐵 𝜕𝐵
= + (∇ϕ ∙ ∇)B = 0 𝑜𝑛 B = 0 (3.27)
𝐷𝑡 𝜕𝑡
As an example, a flat bottom at z = -H is considered:
𝜕𝜙
= 0 𝑜𝑛 𝑧 = −𝐻 (3.28)
𝜕𝑧

3.7 LINEAR SYSTEM

For numerical evaluations, the integral equation should be avoided due to its high computational
cost for every panel created on the body surface.
It's important to note that the body surface, to be accurately reproduced with panels, requires a
high number of them. This means that as an average, each surface is divided into approximately
102 to 104 panels, each one of them requiring a solving equation.
The corresponding linear system of equations is characterized by a square matrix of complex
coefficients with the same dimension, which must be solved using linear algebra.
A brief overview of how the matrix and vectors are generated has been previously given. Now,
a more in-depth analysis of the single values composing the matrix follows.
Matrix A in equation (3.14) contains the derivative of the Green’s function to the direction
normal to the panel, each row comprising the value of this derivative for the panel currently
under inspection, while each column refers to a different panel on the surface. This is the basis
of the dipole concept.
𝐺11 ⋯ 𝐺1𝑁
𝐴=[ ⋮ ⋱ ⋮ ] (3.29)
𝐺𝑁1 ⋯ 𝐺𝑁𝑁

38
 3 – Description of BEMUse

The vector ϕ is the unknown, and after the solution of the linear system will contain the value
of the potential on the centroid of each panel.
The matrix B instead contains the value of the Green function evaluated with the same
procedure followed for A; therefore, each row refers to a certain centroid, and each column
measures the effect of the other panels on it. Like A, matrix B has the form:
𝐺11 ⋯ 𝐺1𝑁
𝐵=[ ⋮ ⋱ ⋮ ] (3.30)
𝐺𝑁1 ⋯ 𝐺𝑁𝑁
Finally, the vector ϕ’ contains the values of the normal derivative of the velocity potential, again
calculated at each centroid. This vector is known because the velocity field around the body is
defined. It thus has the form:
𝜕𝜙1
𝜕𝑛
𝜙′ = … (3.31)
𝜕𝜙𝑁
[ 𝜕𝑛 ]
Among these four components, only the vector ϕ is unknown, while the rest is known.
Therefore, the linear system can be solved calculating the inverse of matrix A, using equation
(3.14).
The solution obtained is thus the vector containing the values of the velocity potential on the
centroid of each panel. This procedure, which is the most computationally expensive step in the
solution of the problem, allows BEMUse to obtain the value of the potential, which is then
needed to calculate the forces and the pressure acting on the body surface.
The linear system generated implementing the solution procedure described in Chapter 2 is
solved by an iterative method (or block iterative method, or Gauss elimination method).
In most cases, the iterative method converges in 10-15 iterations and is the most efficient way
to solve the linear system. On the other hand, using this method all or part of the matrix may be
stored on the hard disk and then recovered at each iteration.
There are, though, a few problems for which the iterative method is slowly convergent or non-
convergent due to bad conditioning of the linear system, like in the case of barges with a shallow
draft or multiple bodies separated by small gaps. The iterative method may also be slowly
convergent in the case of the linear system for the extended boundary integral equation. In this
circumstance, other methods are more efficient.

3.8 DISCRETIZATION
Boundary Element Method is often referred to as the Panel Method because it requires the
surface of the body under examination to be discretized in a finite number of diffraction panels
to carry out the analysis. As previously stated, to get accurate simulations hundreds or even
thousands of panels are needed, and they need to be constructed consistently to solve the linear
system subsequently.
BEMUse, like many other hydrodynamic analysis softwares, uses quadrilateral panels.

39
 3 – Description of BEMUse

They are created starting from the discrete grid points whose coordinates are determined by the
parametric equations that define the geometry of the body. The level of refinement of the grid
can be freely set by the user, although it must always be kept in mind that a higher number of
panels increases the precision of the results, but on the other hand quickly increases the
computational cost of the analysis.
Once all the points have been defined, the adjacent ones need to be grouped to form the panels.
BEMUse does so in a consistent way, i.e. following the same counterclockwise order. This is
required because the order in which the points are accounted for determines the direction of the
normal vector coming out from the centroid of the panel. All the vectors have to indeed point
away from the fluid, i.e. inside the body. In the following figure, a clear example of the result
obtained through this procedure can be seen.

Figure 3.2. All the vectors normal to the panel surface point away from the fluid, i.e. inside the body.

Often, to increase the mesh’s refinement near borders or changes in geometry, where the
velocity potential is expected to undergo the most significant variations, cosine spacing is used.
This allows achieving the objective without increasing the total number of grid points and,
consequently, panels.
When generating the grid points, it’s essential to be careful not to create a point twice in the
same spot because that could create some problems in the handling of irregular frequencies.
This risk is particularly present in the case of symmetrical geometry, where the points on the
axis separating the quadrants need to be taken care of with particular attention.
It’s important to note that the panels created by BEMUse are linear panels, which means that
they are flat, with their vertices on the same geometric plane. This is the main reason why such
a high number of panels is needed. Curved surfaces would be poorly approximated with a less
refined discretization.
As already stated, BEMUse evaluates the velocity potential on the centroid of each panel, while
other softwares do that on the grid points of the body.
40
 3 – Description of BEMUse

3.9 FROM POTENTIAL TO FORCES

Calculating the velocity potential is necessary for then being able to calculate the forces and
pressures acting on the body. To get there, BEMUse uses the equations written in Chapter 2. In
particular, the total pressure working on the surface of the body is calculated through the
Bernoulli equation
𝜕𝜙 1
𝑝(𝑥, 𝑡) = −𝜌 ( + ∇𝜙 ∙ ∇𝜙 + 𝑔𝑧) (3.32)
𝜕𝑡 2
There are although numerous other parameters of great interest when dealing with floating
platforms and hydrodynamics in general.

3.9.1 Added mass and damping

First there is added mass, which is defined as the inertia added to a system because a body
moves or deflects some volume of the fluid surrounding it when subjected to forces, thus
accelerating and decelerating. Since it identifies inertia, it's expressed in 𝑘𝑔 in the case of
translational motions and 𝑘𝑔 ∙ 𝑚2 for rotational ones. In alternative added mass can also be
expressed in dimensionless form, becoming the added mass coefficient. To obtain it, it’s
sufficient to divide the added mass times the displaced fluid mass.
Then there is damping, which in physics is generally described as the influence on an oscillatory
system which reduces, restricts or prevents its oscillation. Damping is directly proportional to
the relative velocity between the two bodies (platform and water, in this case). In the case of
fluids, we refer to radiation damping which is mostly due to the viscosity of the water. It is
expressed in 𝑘𝑔/𝑠 for translational motions and in 𝑘𝑔 ∙ 𝑚2 /𝑠 for rotational ones. Also in this
case, a dimensionless damping coefficient can be expressed as the ratio of the calculated
damping and the mass of displaced water.
The added mass and damping are obtained integrating the value of the radiation potential over
the body:
𝑖
𝐴𝑖𝑗 − 𝐵 = ∬ 𝑛𝑖 𝜙𝑟,𝑗 𝑑𝑆 (3.33)
𝜔 𝑖𝑗 𝑆

Where the subscript i indicates the particular direction of the force and j the coordinate in which
the surface oscillation occurs. A11, therefore, designates the x-forces caused by an oscillation in
the x-direction. This kind of motion is commonly referred to as surge motion. In the y-direction
it's called sway, and in z-direction it's heave. For rotational motions, rotation around the x-, y-
and z-axis are called roll, pitch and yaw, respectively.
In WAMIT the non-dimensional added mass and damping coefficients are obtained as follows
𝐴𝑖𝑗 𝐵𝑖𝑗
𝐴̅𝑖𝑗 = 𝐵̅𝑖𝑗 = (3.34 𝑎 𝑎𝑛𝑑 𝑏)
𝜌𝐿𝑘 𝜌𝐿𝑘 𝜔
Where k = 3 for both i, j being equal to 1, 2 or 3; k = 4 for i = 1, 2, 3 and j = 4, 5, 6 or i = 4, 5,
6 and j = 1, 2, 3; k = 5 for both i, j = 4, 5, 6.

41
 3 – Description of BEMUse

3.9.2 Exciting forces

The periodic forces on the body which arise due to the diffraction of the incoming harmonic
wave are called exciting forces. To calculate them, the potential of the incoming wave is needed.
In the infinite-depth case, it’s expressed as
𝑖𝑔𝐴 𝜈𝑧 𝑖𝜈(𝑥 𝑐𝑜𝑠𝛽+𝑥𝑦 𝑠𝑖𝑛𝛽)
𝜙𝐼 = 𝑒 𝑒 (3.35)
𝜔
Where A is the wave’s amplitude, and β is its heading. Making use of Haskind relations for the
diffraction potential, the forces can be expressed as a function of the radiation potential defined
previously for the oscillating body:
𝜕𝜙𝑖
𝑋𝐼 = −𝑖𝜔𝜌 ∬ (𝑛𝑖 𝜙𝑖 − 𝜙 ) 𝑑𝑆 (3.36)
𝑆 𝜕𝑛 𝑗
This allows to avoid the calculation of the diffraction potential on the surface, but simply to
manipulate the results of the radiation potential. The exciting forces differ according to the
direction of the incoming waves. Therefore, the results obtained must always be expressed
indicating the angle which they refer to. This angle is usually measured starting from the
positive x-axis. In this text, an angle of 45° was always used.
Exciting forces are expressed in 𝑁/𝑚 for translational motions and in 𝑁𝑚/𝑚 for rotational
ones, but they are also found in their non-dimensional form
𝑋𝑖
𝑋̅𝑖 = (3.37)
𝜌𝑔𝐴𝐿𝑚
Where m = 2 for i = 1, 2, 3 and m = 3 for i = 4, 5, 6.

3.9.3 Response Amplitude Operators

The Response Amplitude Operator, or RAO, is an indication of the degree of movement
induced in a floating body due to a passing hydrodynamic wave. RAOs can be calculated for
each degree of freedom of the body, both translational and rotational. Like the exciting force,
the direction of the incoming wave must be defined. As before, 45° was always the choice. The
formula used to calculate each RAO is taken from Baghfalaki, Das, and Das (2012) [28]:
𝑋𝑖 (𝜔) 𝐹𝑖 (𝜔)
𝑍𝑖 (𝜔) = = (3.38)
𝐷𝑖 −𝜔 2 (𝐴𝑖𝑖 (𝜔) + 𝑀𝑖𝑖 ) + 𝑖𝜔𝐵𝑖𝑖 (𝜔) + 𝐶𝑖𝑖
Where Xk (ω) is the motion in the frequency domain, and Dk is the corresponding wave
amplitude. Fk is the exciting force, Akk is the value of added mass, Mkk the mass and inertia
matrix, Bkk is the damping coefficient, Ckk is the value of the hydrostatic restoring matrix and ω
is the frequency of the waves.

42
 3 – Description of BEMUse

The mass and inertia matrix contains the value of mass and rotational inertia of the body under
inspection and is defined as
𝑚 0 0 0 𝑚𝑧𝑔 −𝑚𝑦𝑔
0 𝑚 0 −𝑚𝑧𝑔 0 𝑚𝑥𝑔
0 0 𝑚 𝑚𝑦𝑔 −𝑚𝑥𝑔 0
𝑀𝑖𝑖 = (3.39)
0 −𝑚𝑧𝑔 𝑚𝑦𝑔 𝐼𝑥 −𝐼𝑥𝑦 −𝐼𝑥𝑧
−𝑚𝑧𝑔 0 −𝑚𝑥𝑔 −𝐼𝑦𝑥 𝐼𝑦 −𝐼𝑦𝑧
[−𝑚𝑦𝑔 𝑚𝑥𝑔 0 −𝐼𝑧𝑥 −𝐼𝑧𝑦 𝐼𝑧 ]
On the other hand, the hydrostatic restoring matrix is defined as follows in the WAMIT User
Manual [29]
0 0 0 0 0 0
0 0 0 0 0 0
0 0 𝐶3,3 𝐶3,4 𝐶3,5 0
𝐶𝑖𝑖 = (3.40)
0 0 0 𝐶4,4 𝐶4,5 𝐶4,6
0 0 0 0 𝐶5,5 𝐶5,6
[0 0 0 0 0 0 ]
Where

𝐶(3, 3) = 𝜌𝑔 ∬ 𝑛3 𝑑𝑆 (3.41𝑎)
𝑆𝑏

𝐶(3, 4) = 𝜌𝑔 ∬ 𝑦𝑛3 𝑑𝑆 (3.41𝑏)

𝑆𝑏

𝐶(3, 5) = −𝜌𝑔 ∬ 𝑥𝑛3 𝑑𝑆 (3.41𝑐)

𝑆𝑏

𝐶(4, 4) = 𝜌𝑔 ∬ 𝑦 2 𝑛3 𝑑𝑆 + 𝜌𝑔𝑉𝑧𝑏 − 𝑚𝑔𝑧𝑔 (3.41𝑑)

𝑆𝑏

𝐶(4, 5) = −𝜌𝑔 ∬ 𝑥𝑦𝑛3 𝑑𝑆 (3.41𝑒)

𝑆𝑏

𝐶(4, 6) = −𝜌𝑔𝑉𝑥𝑏 + 𝑚𝑔𝑥𝑔 (3.41𝑓)

𝐶(5, 5) = 𝜌𝑔 ∬ 𝑥 2 𝑛3 𝑑𝑆 + 𝜌𝑔𝑉𝑧𝑏 − 𝑚𝑔𝑧𝑔 (3.41𝑔)

𝑆𝑏

𝐶(5, 6) = −𝜌𝑔𝑉𝑦𝑏 + 𝑚𝑔𝑦𝑔 (3.41ℎ)

V is the volume of the body, ρ is the density of the water, and m is the mass of the body. The
coordinates with subscript b refer to the center of buoyancy of the body, the ones with subscript
g to its center of mass.
To improve the computational performance of the software, the surface integrals are calculated
as the discrete sum of the values on each panel. x and y are thus the coordinates of each centroid,
and n is the vector normal to the surface, with its three components.

43
 3 – Description of BEMUse

These values are then normalized as follows:

𝐶̅ (3, 3) = 𝐶(3, 3)⁄𝜌𝑔𝐿2 (3.42𝑎)
𝐶̅ (3, 4) = 𝐶(3, 4)⁄𝜌𝑔𝐿3 (3.42𝑏)
𝐶̅ (3, 5) = 𝐶(3, 5)⁄𝜌𝑔𝐿3 (3.42𝑐)
𝐶̅ (4, 4) = 𝐶(4, 4)⁄𝜌𝑔𝐿4 (3.42𝑑)
𝐶̅ (4, 5) = 𝐶(4, 5)⁄𝜌𝑔𝐿4 (3.42𝑒)
𝐶̅ (4, 6) = 𝐶(4, 6)⁄𝜌𝑔𝐿4 (3.42𝑓)
𝐶̅ (5, 5) = 𝐶(5, 5)⁄𝜌𝑔𝐿4 (3.42𝑔)
𝐶̅ (5, 6) = 𝐶(5, 6)⁄𝜌𝑔𝐿4 (3.42ℎ)
Where L is the characteristic length of the body.
RAOs have the dimension of 𝑚/𝑚 for translational motions because they measure the ratio
between the motion of the body and the amplitude of the wave, while rotational ones are
expressed in °/𝑚, as they evaluate the ratio between the angular motion of the body and, again,
the amplitude of the wave. As for the other parameters calculated, also RAOs have their non-
dimensional form
𝑍𝑖
𝑍̅𝑖 = (3.43)
𝐴
𝐿𝑛
Where n = 0 for i = 1, 2, 3 and n = 1 for i = 4, 5, 6.

44
 4 – WAMIT Comparison

4. WAMIT COMPARISON

In the process of validating BEMUse, the result obtained with it must be very similar or, at best,
equal to the ones given by already well-established hydrodynamic softwares. For this
comparison to be as thorough and accurate as possible, the values of the four hydrodynamic
coefficients given by BEMUse are compared to the ones of two other well-known softwares
like WAMIT and AQWA. The comparison with WAMIT is essential because BEMUse’s
theoretical basis is the same. The comparison with AQWA is also of great interest, as it is the
hydrodynamic analysis tool inside the ANSYS Workbench, one of the most widely used
engineering softwares in the world. Only the close accordance with results between all the
outputs can confirm the accuracy of BEMUse.
The first comparison was carried out against the ‘WAMIT-MOSES Hydrodynamic Analysis
Comparison Study’ (2000) by the Hull Engineering Department [30]. In their study, the
researchers compared the results obtained with the diffraction and radiation simulation software
WAMIT to the ones given by MOSES, a software that was originally developed for installation
simulation, fatigue, and global motion analysis. Since the latter has the same 3-D diffraction
and radiation module as the former, they thought it could be used for the same kind of platform
studies with excellent results.
As the interest in the use of MOSES is modest for the scope of this thesis, only the results given
by WAMIT were considered.
The analysis was performed on three simple geometries: a cylinder, a parallelepipedal box, and
the ISSC Tension Leg Platform.
For all of them, precise indications of dimensions, mass, and number of panels used for the
discretization were given, making it easy to replicate the analysis with BEMUse. All the data
are reported in the table below. The column ‘CoG’ sets the depth at which the Center of Gravity
is set for each geometry. It’s expressed in meters in the z-direction.

Type Length (m) Beam/ Draft Displacement CoG (m) N of panels

Diameter (m) (m) (Metric-Tons)
Cylinder 200.0 40.0 200.0 256,011.0 200.0 1120
Box 200.0 40.0 28.0 229,645.0 28.0 1264
TLP 51.6 51.6 35.0 52,761.0 35.0 512
Table 4.1. Geometrical and physical data of the three geometries.

45
 4 – WAMIT Comparison

The geometry and the mesh used are shown in the following figures, taken from BEMUse’s
GUI.

Figure 4.1. The cylinder generated by BEMUse. The cosine spacing is visible on the lateral surface.

Figure 4.2. The box, a parallelepipedal barge. Again, cosine spacing is evident on all sides.

46
 4 – WAMIT Comparison

Figure 4.3. The Tension Leg Platform. Only the section below the water is generated.

The wave periods used in the numerical calculation were selected from 4 seconds to 42 seconds
for a total of 20 steps separated by an increment of two seconds. This interval covers a wide
range of waves.
The hydrodynamic coefficients (added mass and damping), the wave exciting forces, and the
response amplitude operators were extracted from BEMUse and compared with WAMIT
results. To do so, as these had undergone a normalization with respect to mass and wave
amplitude, so had to do BEMUse’s.
The normalization for each parameter is reported in the following table. A is the amplitude of
the incoming wave, which is set equal to 1 for all the simulations.

Resulting Components Units Normalized variable

Added mass coefficient Mass Aij/Mass
Damping coefficient Mass/Time SQRT(Bij/Mass)
Wave exciting forces Tons Tons/A
Wave exciting moments Tons*Meter Tons*Meter/A
Linear motion RAO Meter Meter/A
Angular motion RAO Degree Degree/A
Table 4.2. Normalization factors for the four parameters evaluated.
For all the values, a wave heading of 45° starting from the x-axis was considered.

47
 4 – WAMIT Comparison

4.1 CYLINDER
The number of panels for the cylinder is the same as for the WAMIT experiment, 1120. This
means that its base was discretized in 32 azimuthal sections and 9 radial ones. The lateral
surface is divided into 32 sections. To achieve a more refined grid near the free surface and on
edges, a cosine spacing scheme was used, like the researchers in Hull seem to have done.
The mass of the cylinder is calculated multiplying its displacement times the density of
seawater, which is equal to ρ = 1025 kg/m3 . The result is m = 2.55958e+08 kg. The center
of mass, which also corresponds to the center of buoyancy, is placed at half of the cylinder’s
draft, on its central axis.
For the definition of the moments of inertia Ixx, Iyy and Izz, as no precise information was given
in the paper, the parallel axis theorem was used to translate the values referring to rotations
around the axis of symmetry to rotations around the free water surface. These are the values
obtained.
1 𝐿 2
𝐼𝑥𝑥 = 𝐼𝑦𝑦 = 𝑚(3𝑅 2 + 𝐿2 ) + 𝑚 ( ) = 3.43837𝑒 + 12 𝑘𝑔 ∙ 𝑚2 (4.1)
12 2
1
𝐼𝑧𝑧 = 𝑚𝑅 2 = 5.11917𝑒 + 10 𝑘𝑔 ∙ 𝑚2 (4.2)
2
Where R and L are the radius and the height of the cylinder, respectively.
In the figure below, the different colors of the panels illustrate the different values of the
velocity potential on the surface of the cylinder for a given wave period. The red color identifies
higher values, blue lower ones. In the next few pages, instead, the results for this geometry are
reported.

Figure 4.4. The colorized panels give an immediate idea of the potential distribution.

48
 4 – WAMIT Comparison

4.1.1 Added mass

Surge, A11 Sway, A22

1.2 1.2

Added mass coefficient [-]

Added mass coefficient [-]

1 1
0.8 0.8
0.6 0.6
0.4 0.4
0.2 0.2
0 0
0 10 20 30 40 0 10 20 30 40
Period [s] Period [s]

WAMIT BEMUse WAMIT BEMUse

Heave, A33 Roll, A44

0.1 120
Added mass coefficient [-]

Added mass coefficient [-]

0.08 100
80
0.06
60
0.04
40
0.02 20
0 0
0 10 20 30 40 0 10 20 30 40
Period [s] Period [s]

WAMIT BEMUse WAMIT BEMUse

Pitch, A55 Yaw, A66

120 1
Added mass coefficient [-]

Added mass coefficient [-]

100 0.8
80
0.6
60
0.4
40
20 0.2

0 0
0 10 20 30 40 0 10 20 30 40
Period [s] Period [s]

WAMIT BEMUse WAMIT BEMUse

Figure 4.5. Added mass coefficient comparison, cylinder.

49
 4 – WAMIT Comparison

4.1.2 Damping

Surge, B11 Sway, B22

0.12 0.12

Damping coefficient [-]

Damping coefficient [-]

0.1 0.1
0.08 0.08
0.06 0.06
0.04 0.04
0.02 0.02
0 0
0 10 20 30 40 0 10 20 30 40
Period [s] Period [s]

WAMIT BEMUse WAMIT BEMUse

Heave, B33 Roll, B44

5 12
Damping coefficient [-] * 10-4

Damping coefficient [-]

4 10
8
3
6
2
4
1 2
0 0
0 10 20 30 40 0 10 20 30 40
Period [s] Period [s]

WAMIT BEMUse WAMIT BEMUse

Pitch, B55 Yaw, B66

12 1
Damping coefficient [-]

Damping coefficient [-]

10 0.8
8
0.6
6
0.4
4
2 0.2

0 0
0 10 20 30 40 0 10 20 30 40
Period [s] Period [s]

WAMIT BEMUse WAMIT BEMUse

Figure 4.6. Damping coefficient comparison, cylinder.

50
 4 – WAMIT Comparison

4.1.3 Exciting forces and moments

Surge, F11 Sway, F22

2000 2000

Exciting force [N/m]

Exciting force [N/m]

1500 1500

1000 1000

500 500

0 0
0 10 20 30 40 0 10 20 30 40
Period [s] Period [s]

WAMIT BEMUse WAMIT BEMUse

Heave, F33 Roll, F44

1000 120
Exciting moment [Nm/m] * 103
Exciting force [N/m]

800 100
80
600
60
400
40
200 20
0 0
0 10 20 30 40 0 10 20 30 40
Period [s] Period [s]

WAMIT BEMUse WAMIT BEMUse

Pitch, F55 Yaw, F66

120 7
Exciting moment [Nm/m] * 103

Exciting moment [Nm/m] * 10-4

100 6
5
80
4
60
3
40
2
20 1
0 0
0 10 20 30 40 0 10 20 30 40
Period [s] Period [s]

WAMIT BEMUse WAMIT BEMUse

Figure 4.7. Exciting forces and moments comparison, cylinder.

51
 4 – WAMIT Comparison

4.1.4 Response Amplitude Operators

Surge, Z11 Sway, Z22

0.6 0.6
0.5 0.5

RAO [m/m]
RAO [m/m]

0.4 0.4
0.3 0.3
0.2 0.2
0.1 0.1
0 0
0 10 20 30 40 0 10 20 30 40
Period [s] Period [s]

WAMIT BEMUse WAMIT BEMUse

Heave, Z33 Roll, Z44

0.5
15 0.4
RAO [m/m]

RAO [°/m]

0.3
10
0.2
5
0.1

0 0
0 10 20 30 40 0 10 20 30 40
Period [s] Period [s]

WAMIT BEMUse WAMIT BEMUse

Pitch, Z55 Yaw, Z66

0.5 0.1

0.4 0.08
RAO [°/m]

RAO [°/m]

0.3 0.06

0.2 0.04

0.1 0.02

0 0
0 10 20 30 40 0 10 20 30 40
Period [s] Period [s]

WAMIT BEMUse WAMIT BEMUse

Figure 4.8. RAOs comparison, cylinder.

52
 4 – WAMIT Comparison

As shown, the values of almost all the hydrodynamic coefficients are in close accordance with
WAMIT results. To get these excellent correspondence, though, the simple normalization
according to table (4.2) was often not enough. Thus, a case-by-case summary of the operations
performed on the coefficients is considered necessary:
- For added mass, the translational motions were already according excellently after the
prescribed normalization, so no further action was needed.
- The rotational motions' added mass coefficient needed to be still divided by a factor 102
to agree with WAMIT. This additional normalization can be explained according to
formula (3.34a), where for roll, pitch and yaw L's exponent is 5 instead of 3. This also
means that for the cylinder L = 10.
- For the damping coefficient, the reference paper prescribed the square root of the
damping divided by the mass. For surge, heave and sway, though, the square root was
not necessary as the results agreed nicely without it.
- The exciting forces in all six motions needed to be divided by an additional factor 10.
- The corrections needed by the RAOs were even more critical, as the translational
motions needed to be multiplied by a factor 103 and the rotational ones by a factor 104.
This is probably due to the fact that the reference paper expresses the mass of the
cylinder in tons and not in kilograms. Anyways, this only explains the factor 103, not
the 104.
Apart from these additional corrections, there still are some criticalities in the comparison. The
exciting force for the heave motion doesn’t agree at all with WAMIT and, consequently, neither
does the heave’s RAO. There are then further problems concerning the RAOs. For roll, pitch,
and yaw, no agreement has been found between BEMUse and WAMIT. Another small issue
concerns the exciting forces in surge and sway, where the value at T = 4 s is negative, which is
physically impossible.
Regardless of these flaws, the results for the cylinder and, as reported in the next chapters, for
the other geometries, were all more than satisfactory, especially considering the early stage of
BEMUse’s development.

53
 4 – WAMIT Comparison

4.2 BOX
The surface of the parallelepipedal box is discretized with 1264 panels, again using a cosine
spacing scheme. To reach that number, the length (in the x-direction) required 32 discrete
points, the beam (y-direction) 12, and the draft 10.
Again, the mass of the barge is equal to the displaced volume of water, m = 2.296e+08 kg. The
center of mass and buoyancy are again overlapping and in the center of the geometry.
The moments of inertia were calculated using the parallel axis theorem once more to move the
axis of rotation from the center of gravity to the center of the surface at the water surface.
Therefore
1 𝐷 2
𝐼𝑥𝑥 = 𝑚(𝐵2 + 𝐷2 ) + 𝑚 ( ) = 7.35332𝑒 + 11 𝑘𝑔 ∙ 𝑚2 (4.3)
12 2
1 𝐷 2
𝐼𝑦𝑦 = 𝑚(𝐿 + 𝐷 + 𝑚 ( ) = 6.12267𝑒 + 08 𝑘𝑔 ∙ 𝑚2
2 2) (4.4)
12 2
1
𝐼𝑧𝑧 = 𝑚(𝐿2 + 𝐵 2 ) = 7.95947𝑒 + 11 𝑘𝑔 ∙ 𝑚2 (4.5)
12
Where L, B, and D are the length, beam and draft of the barge, respectively.
After the colorized figure below, the four coefficients calculated for the six motions are
reported.

Figure 4.9. Box with colorized panels, generated by BEMUse.

54
 4 – WAMIT Comparison

4.2.1 Added mass

Surge, A11 Sway, A22

0.25 2.5

Added mass coefficient [-]

Added mass coefficient [-]

0.2 2

0.15 1.5

0.1 1

0.05 0.5

0 0
0 10 20 30 40 0 10 20 30 40
Period [s] Period [s]

WAMIT BEMUse WAMIT BEMUse

Heave, A33 Roll, A44

1.2 20
Added mass coefficient [-]

Added mass coefficient [-]

1
15
0.8
0.6 10
0.4
5
0.2
0 0
0 10 20 30 40 0 10 20 30 40
Period [s] Period [s]

WAMIT BEMUse WAMIT BEMUse

Pitch, A55 Yaw, A66

50 80
Added mass coefficient [-]
Added mass coefficient [-]

70
40
60
30 50
40
20 30
20
10
10
0 0
0 10 20 30 40 0 10 20 30 40
Period [s] Period [s]

WAMIT BEMUse WAMIT BEMUse

Figure 4.10. Added mass coefficient comparison, box.

55
 4 – WAMIT Comparison

4.2.2 Damping

Surge, B11 Sway, B22

0.1 1
Damping coefficient [-]

Damping coefficient [-]

0.08 0.8

0.06 0.6

0.04 0.4

0.02 0.2

0 0
0 10 20 30 40 0 10 20 30 40
Period [s] Period [s]

WAMIT BEMUse WAMIT BEMUse

Heave, B33 Roll, B44

0.14 10
Damping coefficient [-]
Damping coefficient [-]

0.12
8
0.1
0.08 6
0.06 4
0.04
2
0.02
0 0
0 10 20 30 40 0 10 20 30 40
Period [s] Period [s]

WAMIT BEMUse WAMIT BEMUse

Pitch, B55 Yaw, B66

14 50
Damping coefficient [-]

Damping coefficient [-]

12
40
10
8 30
6 20
4
10
2
0 0
0 10 20 30 40 0 10 20 30 40
Period [s] Period [s]

WAMIT BEMUse WAMIT BEMUse

Figure 4.11. Damping coefficient comparison, box.

56
 4 – WAMIT Comparison

4.2.3 Exciting forces and moments

Surge, F11 Sway, F22

2 6

Exciting force [N/m] * 103

Exciting force [N/m] * 103

5
1.5
4
1 3
2
0.5
1
0 0
0 10 20 30 40 0 10 20 30 40
Period [s] Period [s]

WAMIT BEMUse WAMIT BEMUse

Heave, F33 Roll, F44

8 6
Exciting force [N/m] * 103

Exciting moment [Nm/m] * 104

7 5
6
5 4
4 3
3 2
2
1 1
0 0
0 10 20 30 40 0 10 20 30 40
Period [s] Period [s]

WAMIT BEMUse WAMIT BEMUse

Pitch, F55 Yaw, F66

12 4
Exciting moment [Nm/m] * 104

Exciting moment [Nm/m] * 105

10 3.5
3
8 2.5
6 2
4 1.5
1
2 0.5
0 0
0 10 20 30 40 0 10 20 30 40
Period [s] Period [s]

WAMIT BEMUse WAMIT BEMUse

Figure 4.12. Exciting forces and moments comparison, box.

57
 4 – WAMIT Comparison

4.2.4 Response Amplitude Operators

Surge, Z11 Sway, Z22

0.8 0.8
0.7 0.7
0.6 0.6

RAO [m/m]
RAO [m/m]

0.5 0.5
0.4 0.4
0.3 0.3
0.2 0.2
0.1 0.1
0 0
0 10 20 30 40 0 10 20 30 40
Period [s] Period [s]

WAMIT BEMUse WAMIT BEMUse

Heave, Z33 Roll, Z44

2 0.35
0.3
1.5 0.25
RAO [m/m]

RAO [°/m]

0.2
1
0.15
0.5 0.1
0.05
0 0
0 10 20 30 40 0 10 20 30 40
Period [s] Period [s]

WAMIT BEMUse WAMIT BEMUse

Pitch, Z55 Yaw, Z66

2.5 0.4
0.35
2
0.3
RAO [°/m]
RAO [°/m]

1.5 0.25
0.2
1 0.15
0.1
0.5
0.05
0 0
0 10 20 30 40 0 10 20 30 40
Period [s] Period [s]

WAMIT BEMUse WAMIT BEMUse

Figure 4.13. RAOs comparison, box.

58
 4 – WAMIT Comparison

As for the cylinder, some additional corrections had to be done on the normalization prescribed
by the paper of reference. To make things worse, these other operations were, in some cases,
different from the ones done in the cylinder test case. Here’s what has been done:
- Added mass coefficients agree for surge, sway and heave. Meanwhile, to make roll,
pitch and yaw agree, the square root of the expected normalization needs to be used.
- On the contrary, for damping the rotational motions were agreeing as expected, while
the translational modes didn’t need the square root, which was therefore removed.
- To make the exciting forces agree, not only had the result to be yet again divided by a
factor 10, but due to some problems with the sign, the absolute value was needed. These
corrections were common to all the degrees of motion.
- For the RAOs, like in the case of the cylinder surge, sway and heave needed to be
multiplied by a factor 103, while roll, pitch and yaw by 104. The explanation for that is
the same as for the cylinder.
Again, not all the coefficients agree with WAMIT. Like in the previous test case, the most
significant issues concern exciting forces and RAOs, but it’s worth noting how the RAO for
heave agrees much better for the box than for the cylinder. This is a direct consequence of a
closer agreement between the exciting forces. The biggest issues remain the RAOs for the
rotational motions, which don’t agree at all.

59
 4 – WAMIT Comparison

4.3 TENSION LEG PLATFORM

Of particular interest for the industry of FOWTs is the third test case, the ISSC tension leg
platform, as it is one of the most promising designs for floating platforms in wind power
applications. The complete geometry of the ISSC is shown in the figure below, taken from
Eatock Taylor and Jefferys (1986) [31].

Figure 4.14. The geometry of the Tension Leg Platform.

The graphical representation on BEMUse, shown in figure (4.3), is different because only the
section below the water surface is considered. In the same publication, there is also a table
reporting all the physical and geometrical parameters of the TLP.

Parameter Value
Spacing between column centers 86.25 m
Column radius 8.44 m
Pontoon width 7.50 m
Pontoon height 10.50 m
Draft 35.00 m
Displacement 54.5 x 106 kg
Weight 40.5 x 106 kg
Roll moment of inertia 82.37 x 109 𝑘𝑔 ∙ 𝑚2
Pitch moment of inertia 82.37 x 109 𝑘𝑔 ∙ 𝑚2
Yaw moment of inertia 98.07 x 109 𝑘𝑔 ∙ 𝑚2
Vertical position of CoG above keel 38.00 m
Table 4.3. Geometrical and physical properties of the TLP.

60
 4 – WAMIT Comparison

Interestingly, for these simulations the center of mass, which coincides again with the center of
buoyancy, is located in the center of the geometry 38 m above the bottom of the keel, which
means above the water level. The platform weights 40.5𝑒 + 6 𝑘𝑔, with a displacement of
54.5𝑒 + 6 𝑘𝑔. The moments of inertia are, respectively, 𝐼𝑥𝑥 = 𝐼𝑦𝑦 = 82.37𝑒 + 9 𝑘𝑔 ∙ 𝑚2
and 𝐼𝑧𝑧 = 98.07𝑒 + 9 𝑘𝑔 ∙ 𝑚2 .
Because of slight differences in the method for creating the panels, their exact number couldn’t
be replied. However, 520 instead of 512 were used, which assures a close similarity between
the simulations. In the figure below, the colorized geometry as generated by BEMUse is
reported, followed once again by the complete comparison of coefficients.

Figure 4.15 TLP with colorized panels, generated by BEMUse.

61
 4 – WAMIT Comparison

4.3.1 Added mass

Surge, A11 Sway, A22

1.2 1.2
Added mass coefficient [-]

Added mass coefficient [-]

1 1
0.8 0.8
0.6 0.6
0.4 0.4
0.2 0.2
0 0
0 10 20 30 40 0 10 20 30 40
Period [s] Period [s]

WAMIT BEMUse WAMIT BEMUse

Heave, A33 Roll, A44

0.7 40
Added mass coefficient [-]
Added mass coefficient [-]

0.6 35
0.5 30
25
0.4
20
0.3
15
0.2 10
0.1 5
0 0
0 10 20 30 40 0 10 20 30 40
Period [s] Period [s]

WAMIT BEMUse WAMIT BEMUse

Pitch, A55 Yaw, A66

40 60
Added mass coefficient [-]
Added mass coefficient [-]

35 50
30
40
25
20 30
15 20
10
10
5
0 0
0 10 20 30 40 0 10 20 30 40
Period [s] Period [s]

WAMIT BEMUse WAMIT BEMUse

Figure 4.16. Added mass coefficient comparison, TLP.

62
 4 – WAMIT Comparison

4.3.2 Damping

Surge, B11 Sway, B22

0.35 0.35
Damping coefficient [-]

Damping coefficient [-]

0.3 0.3
0.25 0.25
0.2 0.2
0.15 0.15
0.1 0.1
0.05 0.05
0 0
0 10 20 30 40 0 10 20 30 40
Period [s] Period [s]

WAMIT BEMUse WAMIT BEMUse

Heave, B33 Roll, B44

0.02 7
Damping coefficient [-]

Damping coefficient [-]

6
0.015 5
4
0.01
3
0.005 2
1
0 0
0 10 20 30 40 0 10 20 30 40
Period [s] Period [s]

WAMIT BEMUse WAMIT BEMUse

Pitch, B55 Yaw, B66

7 30
Damping coefficient [-]

Damping coefficient [-]

6 25
5
20
4
15
3
10
2
1 5
0 0
0 10 20 30 40 0 10 20 30 40
Period [s] Period [s]

WAMIT BEMUse WAMIT BEMUse

Figure 4.17. Damping coefficient comparison, TLP.

63
 4 – WAMIT Comparison

4.3.3 Exciting forces and moments

Surge, F11 Sway, F22

1000 1000

Exciting force [N/m]

Exciting force [N/m]

800 800

600 600

400 400

200 200

0 0
0 10 20 30 40 0 10 20 30 40
Period [s] Period [s]

WAMIT BEMUse WAMIT BEMUse

Heave, F33 Roll, F44

700 2
Exciting moment [Nm/m] * 104
Exciting force [N/m]

600
500 1.5
400
1
300
200 0.5
100
0 0
0 10 20 30 40 0 10 20 30 40
Period [s] Period [s]

WAMIT BEMUse WAMIT BEMUse

Pitch, F55 Yaw, F66

2 0.025
Exciting moment [Nm/m] * 104

Exciting moment [Nm/m]

0.02
1.5
0.015
1
0.01
0.5
0.005

0 0
0 10 20 30 40 0 10 20 30 40
Period [s] Period [s]

WAMIT BEMUse WAMIT BEMUse

Figure 4.18. Exciting forces and moments comparison, TLP.

64
 4 – WAMIT Comparison

4.3.4 Response Amplitude Operators

Surge, Z11 Sway, Z22

1.4 1.4
1.2 1.2
1 1

RAO [m/m]
RAO [m/m]

0.8 0.8
0.6 0.6
0.4 0.4
0.2 0.2
0 0
0 10 20 30 40 0 10 20 30 40
Period [s] Period [s]

WAMIT BEMUse WAMIT BEMUse

Heave, Z33 Roll, Z44

8 3
7 2.5
6
RAO [m/m]

2
RAO [°/m]

5
4 1.5
3 1
2
1 0.5
0 0
0 10 20 30 40 0 10 20 30 40
Period [s] Period [s]

WAMIT BEMUse WAMIT BEMUse

Pitch, Z55 Yaw, Z66

3 0.01
2.5
0.008
2
RAO [°/m]

RAO [°/m]

1.5 0.006

1 0.004
0.5
0.002
0
0 10 20 30 40
0
Period [s] Period [s]

WAMIT BEMUse WAMIT BEMUse

Figure 4.19. RAOs comparison, TLP.

65
 4 – WAMIT Comparison

Again, some modifications with respect to the original normalization were needed:
- As for the box, the added mass coefficient for the rotational motions agrees only if the
square root is calculated on the ratio of added mass and the mass of the body
- Again, for the damping coefficient the translational motions do not need the square root,
while roll, pitch and yaw do.
- The values of the exciting force need to be divided by ten for all 6 degrees of motion.
- As for the other two test cases, the translational RAOs need to be further multiplied by
103, 104 for the rotational ones.
Similarly to the other two geometries, the added mass and damping coefficients agree almost
perfectly between BEMUse and WAMIT, and so do the exciting forces except for the heave
motion. Some issues arise with the RAOs, as the peaks seen for surge and sway at period
𝑇 = 36 𝑠 in the WAMIT simulation cannot be found in BEMUse. Like for the box, there is
good accordance for the heave, while no similarity can be seen for roll, pitch and yaw.
Summarizing, for all three geometries the added mass and damping coefficients calculated by
BEMUse coincide almost perfectly with the results given by WAMIT. There were indeed some
additional operations to do on the values to make them agree, but since the authors of the paper
used as reference do not clearly explain all the steps followed for their study, it is difficult to
know why these operations are needed. The agreement between the two softwares is excellent
also for the exciting forces, with the notable exception of the heave motion. The RAO for surge
and sway also agree beautifully, while the other motions are flawed with errors that was not
possible identify.
Overall, also given that BEMUse is a relatively new software and this validation is the first test
of such kind carried on the model, the results obtained are considered more than satisfying.
After the comparison with WAMIT, another range of case-studies was considered useful to
increase BEMUse's trustworthiness further. Therefore, using another well-affirmed
hydrodynamic analysis software like Ansys AQWA, a new set of experiments was carried out.
The results are contained in the next chapter.

66
 5 – AQWA Comparison

5. AQWA COMPARISON

Considering how extensively Ansys is used in engineering, having BEMUse’s result to agree
with AQWA’s would be of great importance. Therefore, various test cases were studied.
This software has its 3D-CAD with an excellent GUI, with which it's relatively easy to generate
the desired geometry. After creating the body surface, physical parameters need to be added.
After that, the program automatically generates the quadrilateral panels; the user cannot control
this process, and therefore the number of panels on the body will be different from the one in
BEMUse. Besides, the meshing is generated with a different process, as shown in the figure
below. Unlike what happened with WAMIT, the results calculated by AQWA are not
normalized.

Figure 5.1. The mesh generated by AQWA is completely different with respect to BEMUse’s.

These differences cannot be avoided and could be the cause of slight differences in the results.

5.1 HEMISPHERE
The first case-study analyzed is a simple hemisphere with a 1 m radius. This geometry is not of
practical interest, but it has the great advantage of also having an analytical solution that can be
used to check the accuracy of the analysis. For the hemisphere, a new set of angular frequencies
ω has been chosen, going from 0 to 10 ∙ 𝑔 𝑟𝑎𝑑/𝑠, where g is gravity. This is because, for such
a small geometry, these values of ω are the most meaningful to study.
The results of the two softwares for the hemispherical body shown in figure (3.1) agree nicely
once again.

67
 5 – AQWA Comparison

5.1.1 Added mass

Figure 5.2. Added mass comparison, hemisphere.

68
 5 – AQWA Comparison

5.1.2 Damping

Figure 5.3. Damping comparison, hemisphere.

69
 5 – AQWA Comparison

5.1.3 Exciting forces and moments

Figure 5.4. Exciting forces and moments comparison, hemisphere.

70
 5 – AQWA Comparison

5.1.4 Response Amplitude Operators

Figure 5.5. RAOs comparison, hemisphere.

71
 5 – AQWA Comparison

For added mass, as hoped, the curves both agree with the analytical solution quite accurately,
especially for surge, sway and heave. The rotational motions show some similarities in the
shape of the curves, although the peaks and lows occur at different frequencies and the
numerical values have varying distances between the two softwares.
Also for damping the results are fitting, as the translational motions agree almost perfectly,
except for some unexpected irregularities in both curves, occurring at different frequencies.
These are probably due to an imperfect irregular frequency removal procedure. The situation is
different for the rotational motions, as the curves have a similar shape but do not coincide. This
is probably a dimensional problem, as the values in AQWA were probably multiplied by a
factor which was not possible to identify in this comparison.
For the exciting forces, the situation is quite different, as the curves show some similarities, but
the comparison can only be deemed unsatisfactory for all directions of motion. The main issue
here is that, apart from the evident differences in shape, the curves generated by BEMUse all
reach negative values at some point in the plot, which is physically impossible.
The RAOs are again quite similar between the two softwares, but only the translational motions
have been compared. The peak is visible in the plots at 𝜔 = 0 𝑟𝑎𝑑/𝑠 for AQWA’s curve is
absent in BEMUse’s because the analysis of the value 0 does not give a result there, so the first
point evaluated was at 𝜔 = 1 𝑟𝑎𝑑/𝑠. This is because the comparison with WAMIT has already
shown there is something wrong in BEMUse with roll, pitch and yaw. Moreover, as the RAOs
are calculated starting from the other three parameters, it has been considered meaningless to
carry out this last comparison given the issues encountered previously, especially with the
exciting moments.
Having proven that the two softwares give similar results, more complex geometries can be
confidently analyzed.

5.3 CYLINDER
The next case-study is the same cylinder analyzed in the comparison with WAMIT. This choice
was made first of all because comparing the three softwares on the same body is of high interest,
and then because a cylinder that big is an excellent intermediate step between a pure geometrical
test case and a commercial scale FOWT’s platform, which will be examined in the next
simulation.
For this simulation and the next the incoming waves had 20 different angular frequencies, going
from T = 4 s to T = 42 s, exactly like for the WAMIT comparison.
The same physical parameters as before (mass, moments of inertia, coordinated of the CoG)
were set as input, and the results are shown in the next plots.

72
 5 – AQWA Comparison

5.2.1 Added mass

Figure 5.6. Added mass comparison, cylinder.

73
 5 – AQWA Comparison

5.2.2 Damping

Figure 5.7. Damping comparison, cylinder.

74
 5 – AQWA Comparison

5.2.3 Exciting forces and moments

Figure 5.8. Exciting forces and moments comparison, cylinder.

75
 5 – AQWA Comparison

5.2.4 Response Amplitude Operators

Figure 5.9. RAOs comparison, cylinder.

76
 5 – AQWA Comparison

The results are quite surprising. For surge, sway and heave, the added mass plots show the same
evolution for the two curves, but the numerical values obtained by BEMUse are approximately
10% lower than those of AQWA. This is probably due to different handling of the top of the
geometry, which is not considered by AQWA but discretized in BEMUse. For roll, pitch and
yaw, the differences in the two curves are more marked, with significant differences also in the
shape.
The translational motions for damping agree almost perfectly between the two, while the
rotational ones differ markedly both in the shape and values of the curves. The difference in
outline is quite interesting though, as the two curves show similar behavior, but they peak at
different frequencies.
The exciting forces that agree are only those for surge and sway, as heave has the unknown
issues already encountered in WAMIT comparison, while roll, pitch and yaw do not coincide
at all.
RAOs do instead agree almost perfectly for translational motions, while again, rotational ones
haven’t been compared.

5.3 BOX
The box already studied in the WAMIT comparison is studied with AQWA as well. All the
geometrical and physical parameters are the same as before.
If the previous comparison had been more successful and the accordance more precise, the
parallelepipedal geometry would not have been analyzed, considering the work redundant.
Instead, as the issues encountered were more critical than expected, it has been thought better
to compare the box as well, instead of turning the attention to other platforms of more practical
interest. This choice, unfortunately, excluded from this thesis the analysis of the ISWEC
platform, which is not a floating platform designed for wind turbines, but an innovative concept
of a wave energy converter developed by a private company with a close collaboration with the
Politecnico di Torino, which directly produces electricity from the motion of the waves.
This decision was made not only because of the reason just explained, but also because the data
available for an analysis were not precise enough.
Anyway, the parallelepipedal box allows a further comparison between the three softwares,
which is something probably more useful for the validation of a new hydrodynamic model.
The results for the four parameters considered are shown in the next plots.

77
 5 – AQWA Comparison

5.3.1 Added mass

Figure 5.10. Added mass comparison, box.

78
 5 – AQWA Comparison

5.3.2 Damping

Figure 5.11. Damping comparison, box.

79
 5 – AQWA Comparison

5.3.3 Exciting forces and moments

Figure 5.12. Exciting forces and moments comparison, box.

80
 5 – AQWA Comparison

5.3.4 Response Amplitude Operators

Figure 5.13. RAOs comparison, box.

81
 5 – AQWA Comparison

For added mass, the results for surge, sway and heave agree closely between the two softwares.
Unfortunately, there are no explanations for the behavior of the curve in the roll graph, while
pitch and yaw show again a close agreement in form and values.
The same can be said for damping as well, but the differences between BEMUse and AQWA
are more important than for added mass. In this case, the peak reached by the former in each
plot is much lower than the one of the latter. Once again, the roll is the motion which gives
more troubles.
Unfortunately, for the exciting forces none of the six motions agree, and the differences are
marked. None of the curves agree with each other, with unexpected peaks and lows on all of
them. Quite interestingly, though, this time it’s the AQWA curves that show more irregularities.
Anyways, further analysis should be carried out to try to understand what’s wrong.
On the other hand, the RAOs for translational motions agree almost entirely, which is surprising
as they derive directly from the other parameters. Again, only surge, sway and heave were
studied.

5.4 OC3
Finally turning the attention on a test case of more practical importance, the choice fell on the
OC3 platform for FOWT applications, as it is both easy to model - the shape is similar to a
cylinder - and a promising design for commercial floating turbines.
Indeed, the OC3 platform is a spar-buoy type of platform 120 m high with a larger diameter of
4.7 m and the smaller of 3.7 m. The geometrical parameters of the platform are reported in the
next table.

Physical parameter Value

Depth to platform base below seawater level 120 m
Depth to top of taper below seawater level 4m
Depth to bottom of taper below seawater level 12 m
Platform diameter above taper 6.5 m
Platform diameter below taper 9.4 m
Platform mass, including ballast 7 466 330 kg
Center of Mass location below seawater level 89.9155 m
Platform roll inertia about CoM 4 229 230 000 𝑘𝑔 ∙ 𝑚2
Platform pitch inertia about CoM 4 229 230 000 𝑘𝑔 ∙ 𝑚2
Platform yaw inertia about CoM 164 230 000 𝑘𝑔 ∙ 𝑚2
Table 5.1. Physical and geometrical parameters of the OC3 platform.

For this geometry, instead of the usual comparison with AQWA plots obtained directly on my
calculator, the results obtained by BEMUse were compared against the paper published by
Jonkman (2007) [32], which used AQWA himself in his study. This choice was made because
getting the same results as the ones published by one of the leading experts on offshore
platforms would mean a much more reliable validation than that obtained so far.

82
 5 – AQWA Comparison

In the figure below, as done for all the other geometries, the geometry generated by BEMUse
is reported. Except for the tapered part on top, the platform is basically a cylinder.

Figure 5.14. OC3 floating platform generated by BEMUse.

In this paper, only the results for added mass and damping were published, so the comparison
could only be made for those parameters. The frequencies analyzed are different than the
previous cases, as the values of the normalized frequency 𝜈 = 𝜔2 /𝑔 go from 0 to 5. Also in
this analysis, the values of both added mass and damping are not normalized.
The results of the comparison are shown in the next couple of pages.

83
 5 – AQWA Comparison

5.4.1 Added mass

Figure 5.15. Added mass comparison, OC3.

84
 5 – AQWA Comparison

5.4.2 Damping

Figure 5.16. Damping comparison, OC3.

85
 5 – AQWA Comparison

The first thing to make clear is that the reason why some of Jonkman’s curves have sharp angles
instead of smooth bends on the graphs is the way they were extracted from the plots published
on the paper. The small figures, where all three translational motions on one and the rotational
on the other were plotted together, didn’t allow a precise acquisition of data. This is particularly
evident in the heave plot for both added mass and damping.
The results of the comparison are satisfying. For added mass, the two curves show close
similarity in shape, even though the numerical values obtained with BEMUse are slightly
smaller than the reference ones for all the degrees of motion, but mainly for the translational
ones. Various tries to solve this discrepancy couldn’t find the reason why this difference is so
marked.
In the damping plots, the two curves are much closer to each other on all the graphs. BEMUse’s
curves always have lower peaks, but the difference is less evident. For both surge and sway, at
frequencies close to 3.5 and 4.5 rad/s, there are some unexpected irregularities on the BEMUse
curves. These, due to some issues on the irregular frequency removal algorithm, are absent in
Jonkman’s plot.
Apart from these small issues, the comparison with the cited paper can be considered successful,
even though only added mass and damping were considered.

86
 6 - Conclusion

6. CONCLUSION

As reported and explained in the previous chapters, this thesis had the objective to prove the
validity and accuracy of BEMUse.
At the end of the tests, that target is successfully achieved, even though not completely.

6.1 WAMIT COMPARISON

Regarding the agreement of the BEMUse’s results with WAMIT’s, the curves are, for the most
part, remarkably similar to each other, and they are even closer than initially expected. Added
mass, damping, exciting forces except for heave motion, and the translational motions of RAOs
agree almost perfectly for the three geometries under investigation. Some concerns remain on
the four remaining parameters, which do not agree at all.
The reason for the errors in the rotational motions of RAOs is unfortunately unclear. RAOs are
calculated directly from the other three parameters studied, as equation (3.38) states. Since all
three of them are correct, it’s uncertain why the BEMUse’s results are so different.
It has necessarily got something to do with the definition of both the inertia matrix M and the
hydrostatic restoring matrix C. Unfortunately, in the paper from which the test cases were taken,
no explanation at all is given on how these two matrices were set. On BEMUse, on the other
hand, they are configured precisely how the WAMIT User Manual [29] prescribes.
The definition of the two matrices is undoubtedly part of the problem, but it does not explain
the sudden peaks that the researchers in Hull [30] found in their simulations, for all three
geometries. This is because neither M nor C depend on the wave frequency ω, a factor that is
necessary to explain these peaks.
The issues concerning the exciting force for the heave motion are unclear as well. For that,
though, no explanatory hypothesis could be formulated.
Concerns also derive from the extensive unforeseen corrections needed to make the results of
the two softwares agree. They were much broader than prescribed on the paper and, worst of
all, were not consistent among the three geometries under investigation. This leads to the
suspect that the results in the article were misreported, or at least the graphs have, in some cases,
issues with the order of magnitude of the results. Another possibility is that there was some
additional normalization made without a statement.

6.2 AQWA COMPARISON

The comparison with AQWA was less successful than the first one. The results never coincide
as closely as they do with WAMIT, even though there are good similarities in the shape of the
curves, which agree almost entirely in their evolution in both period and frequency in most
cases.
Moreover, the numerical difference between the two is consistent, with BEMUse’s results being
often approximately up to 10% lower than AQWA’s. One possible explanation of this could be

87
 6 - Conclusion

the different way the two softwares treat the upper surface of the geometries studied, i.e. the
surface at z = 0, the free water surface.
AQWA does not consider it since the geometries it uses are thin hollow surfaces obtained
cutting solid bodies.
BEMUse, on the other hand, discretizes and creates panels on the top surface of each geometry,
which then obviously are considered in the calculations. Subsequently, since the most
significant contribution to the hydrodynamic parameters under investigation comes from near
the free water surface, the difference between the two results surely depends on this.
For the same reason, as the mesh in AQWA is automatically generated and cannot be controlled
by the user, the greater contribution near the body’s edges and the water surface cannot be
adequately taken into account as it is impossible to increase the mesh refinement near these
particularly important areas.
Finally, it can be noted how the plots generated by AQWA show in some cases the effects of a
non-optimal removal of irregular frequencies. These sudden peaks are not to be seen in
BEMUse, as its removal procedure is, like the one in WAMIT, much more efficient.
The comparison with Jonkman (2007) [32] is considered successful and of particular
importance, as it validates BEMUse on a geometry with practical applications in the field of
FOWTs, against a paper published by one of the leading experts at NREL, the National
Renewable Energy Laboratory in the USA.
Summarizing, the validation of BEMUse as a reliable hydrodynamic analysis software can be
considered successful, as the results compare quite nicely with both WAMIT and AQWA. This
is just the first step though, as there is plenty of work to be done before the library can be
successfully and confidently incorporated inside QBlade. In the next chapter, the future
objectives to be achieved in the development of BEMUse are briefly illustrated.

6.3 FUTURE WORK

The work of validation carried out in this thesis is just the first step in the development of a
well-established and acknowledged hydrodynamic analysis model. However, the encouraging
results of this study demonstrate the potential of BEMUse as an open-source, easy-to-use
alternative to expensive softwares like AQWA and WAMIT.
Before being able to include it in QBlade, giving the users the possibility to simulate the
behavior and the performance of FOWTs like they nowadays do with fixed-bottom turbines,
BEMUse needs to undergo significant additional development. In this chapter, the next steps in
its development are briefly assessed.
First, the issues faced during this work of validation need to be corrected. It’s fundamental to
understand why the exciting force for the heave motion is always wrong, and especially where
is the issue with the RAOs for the rotational motions. Until then, this whole work of validation
cannot be considered completely successful.
For the moment, all the simulations are carried out in the infinite-depth configuration, which
does not consider the effect of the seafloor on the velocity potential. This is because the finite-

88
 6 - Conclusion

depth case requires another form of the Green function, which is much more challenging to
implement on a software.
Numerous researchers in recent years have faced this issue, and various solutions have been
proposed. Liu, Iwashita, and Hu (2015) [33] suggests using different equations depending on
the ratio between the characteristic length of the body to be studied and the water depth h. The
four equations become progressively challenging to implement numerically as h increases, and
the ratio consequently decreases.
Newman (1985) [15], on the other hand, proposes one single equation to be used, no matter
how deep the water is. The problem is that for reasons of computational efficiency, he
transforms the integral of the Green function in a sum, whose rate of convergence depends
again on the ratio L/h. This series is practically useless for small values of L/h since each
summand contains a logarithmic singularity when L/h tends to 0. Moreover, it’s been proven
that the number of terms required in the sum for a given accuracy is proportional to h/L, which
means that the computational cost of the calculation increases rapidly.
Chen (2018) [34] recently proposed a new and improved formulation of the sum computed by
Newman, but the issues are not solved, and therefore, the computational cost is still very high.
Additionally, Chen proposed a sum formulation also for the derivative of the Green’s functions
both in the horizontal direction and in the vertical one. When trying to implement Chen’s sums
in BEMUse, the results were quite correct even though they didn’t converge to 0 when the wave
angular frequency ω was 0.
For the moment, BEMUse only calculates first-order forces and moments. Once the previous
two improvements will be completed, the second-order problem also needs to be implemented
in the model. Second-order waves are generated from the quadratic interaction of two linear
wave components in the discrete spectrum. The second-order problem is much more
challenging to study and implement because of the increased complexity of the solving
equations.
There are a lot of other features that need to be added in BEMUse before it could confidently
be used in place of other hydrodynamic modeling softwares. This chapter only gave an idea of
where the work at Hermann-Föttinger-Institut will focus in the near future to soon provide a
reliable, efficient and free-to-use hydrodynamic analysis model.
Humanity cannot afford to wait anymore to change the energy sources on which to rely on, and
my hope is that BEMUse will soon help with the research on FOWTs, which have the potential
to become one of the leading technologies in the urgent transition to a fossil-free world.

89
90
 Bibliography

7. BIBLIOGRAPHY

[1] Global Wind Energy Council. 2017. “Global Wind Energy Report: Annual Market Update
2017.”
[2] https://www.ge.com/renewableenergy (Website). 2017. “Haliade-X Offshore wind turbine
platform.” https://www.ge.com/renewableenergy/wind-energy/offshore-wind/haliade-x-
offshore-turbine
[3] Boccard, N. 2009. “Capacity Factor of Wind Power Realized Values vs. Estimates.” Energy
Policy 37(7):2679–88.
[4] Sclavounos P. D. 2007. “Floating Wind Turbines.” Lecture presentation. Massachusetts
Institute of Technology, USA.
[5] Musial, W., Butterfield, S., and Boone, A. 2004. “Feasibility of Floating Platform Systems
for Wind Turbines.” Collection of ASME Wind Energy Symposium Technical Papers AIAA
Aerospace Sciences Meeting and Exhibit (November 2003):476–86.
[6] Golbert, D. R., Blanco, P. J., Clausse, A., and Feijóo, R. A. 2016. “Corrigendum to ‘On the
Search of More Stable Second-Order Lattice-Boltzmann Schemes in Confined Flows’ [J.
Comp. Phys. 294 (2015) 605-618] Doi: 10.1016/j.Jcp.2015.03.065.” Journal of
Computational Physics 311(June):374.
[7] https://pmiind.com (Website). Burger T. 2016. “Pros and Cons of Floating Platforms in
Marine Renewable Energy.” https://pmiind.com/pros-cons-floating-platforms-marine-
renewable-energy/
[8] Myhr, A., Bjerkseter, C., Ågotnes, A., and Nygaard, T. A. 2014. “Levelised Cost of Energy
for Offshore Floating Wind Turbines in a Lifecycle Perspective.” Renewable Energy
66:714–28.
[9] Lazard. 2018. “Lazard’s Levelized Cost of Storage Analysis — Version 12.0.” Lazard’s
Levelized Cost of Storage - Version 12.0 (November):0–19.
[10] Babarit, A., Delhommeau, G. 2015. “Theoretical and Numerical Aspects of the Open
Source BEM Solver NEMOH.” Proceedings of the 11th European Wave and Tidal Energy
Conference. (September 2015):1–12.
[11] Penalba, M., Kelly, T., and Ringwood, J. V. 2017. “Using NEMOH for Modelling Wave
Energy Converters : A Comparative Study with WAMIT.” 12th European Wave and Tidal
Energy Conference 10.
[12] Lee, C. H. 1995. “WAMIT Theory Manual.” WAMIT Theory Manual 1–19.
[13] Newman, J. N. 1999. “Marine Hydrodynamics.” Marine Hydrodynamics 3–5. Maple-Vail,
Inc. Massachussets Institute of Technology. USA.
[14] Wehausen, J. V. and Laitone, E. V. 1960. “Surface waves.” Handbuch der Physik: 9, 446-
778. Springer, Berlin.
[15] Newman, J. N. 1985. “Algorithms for the Free-Surface Green Function.” Journal of
Engineering Mathematics 19(1):57–67.
[16] Newman, J. N. 1992. “The approximation of free-surface Green functions”. Wave
Asymptotics. P. A. Martin & G. R. Wickham, editors, Cambridge University Press.

91
 Bibliography

[17] Newman, J. N., Sclavounos, P. D. 1988. “The computation of wave loads on large offshore
structures.” 5th International Conference on the Behaviour of Offshore Structures.
Trondheim.
[18] Newman, J. N. 1993. “Note on bodies and panels in the free surface.” Unpublished report.
[19] Zhu, X. 1994. “Irregular Frequency Removal from the Boundary Integral Equation for the
Wave-Body Problem.” MSc Thesis. Dept. of Ocean Eng., Massachussetts Institute of
Technology. USA.
[20] Lee, C. H., Newman, J. N., and Zhu, X. 1996. “An Extended Boundary Integral Equation
Method for the Removal of Irregular Frequency Effects.” International Journal for
Numerical Methods in Fluids 23(7):637–60.
[21] van Garrel, A. 2016. “Multilevel panel method for wind turbine rotor flow simulations.”
PhD Thesis. University of Twente.
[22] van Garrel, A. 2003“Development of a Wind Turbine Aerodynamics Simulation Module.”
ECN-C--03-079. Energy research Centre of the Netherlands, 2003.
[23] Newman, J. N. 1986. “Distributions of sources and normal dipoles over a quadrilateral
panel.” Journal of Engineering Mathematics, 20(2):113–126.
[24] Laforce, T. and June, C. A. 2006. “PE281 Boundary Element Method Course Notes.” Tara
(June):1–12.
[25] Costabel, M. 1987. “Principles of Boundary Element Methods.” Computer Physics Reports
6(1–6):243–74.
[26] Edwards, E. N.d. “Introduction to the Linear Panel Method.” Lecture notes.
[27] Liang, H., Wu, H., and Noblesse, F. 2018. “Validation of a Global Approximation for
Wave Diffraction-Radiation in Deep Water.” Applied Ocean Research 74(May):80–86.
[28] Baghfalaki, M., Das, S. K., and Das, S. N. 2012. “Analytical Model to Determine Response
Amplitude Operator of a Floating Body for Coupled Roll and Yaw Motions and
Frequency-Based Analysis.” International Journal of Applied Mechanics 4(4).
[29] WAMIT, Inc. n.d. “WAMIT 6.2 User Manual.” WAMIT 6.2 User Manual 4.1–4.8.
[30] McDermott, R.J. 2000. “WAMIT-MOSES Hydrodynamic Analysis Comparison Study.”
(July):1–30.
[31] Eatock Taylor, R., and Jefferys, E. R. 1986. “Variability of Hydrodynamic Load
Predictions for a Tension Leg Platform.” Ocean Engineering 13(5):449–90.
[32] Jonkman, J. M. 2007. “Dynamics Modeling and Loads Analysis of an Offshore.” NREL
(November).
[33] Liu, Y., Iwashita, H., and Hu, C. 2015. “A Calculation Method for Finite Depth Free-
Surface Green Function.” International Journal of Naval Architecture and Ocean
Engineering 7(2):375–89.
[34] Chen, Z. 2018. “New Formulation of the Finite Depth Free Surface Green Function.”

92
 Acknowledgements

ACKNOWLEDGMENTS

First, I want to thank my supervisor at TU Berlin, Joseph Saverin, for the chance he gave me.
Working on my thesis in such a prestigious university was the perfect ending to my academic
career. Your help and support were fundamental during the nine months we worked together,
and I genuinely appreciate all the time you spent tutoring me.
Special thanks are also due to Ing. Giovanni Bracco, who was supportive and present from
Torino every time I needed him. Not only did you allow me to live this abroad experience, but
as a final gift you helped me find the perfect job for the beginning of my working career, and I
will be forever thankful for that.
After six hard years of studying, there are plenty of people that I want to thank for sharing this
important period of my life with me.
And how can I not start with my parents? Dear Papi and Mami, your unconditional support has
been invaluable. You’ve always let me organize and live my life as I wanted to, allowing me to
make my mistakes and learn from them; but in the time of need, I always knew I could count
on your help and support. You have two very different ways to show it, but I know you are
there for me no matter what. You even backed me up when a crazy idea that involved guns and
lethal animals recently brought me to South Africa. Even in my wildest dreams, I could not
have hoped for better parents.
Dear Gogo, you worked hard and took bold choices to get where you are now. I’m incredibly
happy and proud of what you are achieving. You taught me a lot about what it takes to reach
your dreams, and I’m thankful for all the help and the advice you manage to give me even living
far away. I’m sure that you will get wherever you want to go with your life, and I will always
be by your side.
Dear Anna, all the hard work and the efforts you put on a daily basis are truly inspirational.
You manage to succeed in a tough field of study while spending so much time and energy on
your biggest passion. Maybe I don’t say it too often, but I’m so proud of the young woman
you’ve become, and you can always count on my help when you need it, never forget that.
Thank you for teaching me that for our biggest passions, all the sacrifices are more than worth
doing; and thanks for all the big and little things you do for me, I notice them all, even if
sometimes I fail to show it.
There are then so many good friends in my life, incredible people that make my life incredible.
Thank you, Alice, for all the late-night talks in my car, discussing life, travels, and many other
crazy ideas. Thank you for always being present in my life, no matter how far from each other
we were. Our friendship is truly one that will survive distances and time unscathed.
Thanks to Simone and Giorgio, my best friends. I don’t think there is much to add; I can’t wait
to build new unforgettable memories with you.
Thanks to Stefano, Emanuele, Marco, and all the other members of ‘La Plaza’. You tried so
hard to make me fail my exams, and I’m happy you didn’t succeed in that. But you are the best
group for a fun night or a nightmare car trip. together, we rarely get where we thought we
would, but we always end up with a cool story to tell afterward.

93
 Acknowledgements

Thanks to the members of the legendary ‘GruppOtto’, you all made my last year at Politecnico
by far the best one. A very special mention is due to Marta. Even though we will never end up
being Instagram superstars partying in Porto Cervo, I’m so happy you decided to leave the
beautiful Rome to come to the much nicer Torino. But I’m even happier you chose to pursue a
career in engineering, instead of one in plant caregiving: that would have been a complete
disaster.
Thank you, Stefano, because as you never fail to remind me, I would have never graduated
without all your help, your wisdom, and, most importantly, your notes. You just started the
most incredible and rewarding period of your life: Margherita is so lucky to have you as a father.
No matter how many obstacles life throws at the three of you, you will go through everything.
Finally, I want to thank all the kids I’ve trained during these six years. No matter how bad my
day had been, or how tired I was, you never failed to remind me that life can be much simpler,
if you look at it with the smile and the lightheartedness of a child. I’m firmly convinced you
taught me way more than what I taught you.
There are so many other extraordinary people in my life, great friends with whom I enjoy
sharing a beer, an ice cream, or just some time together. My life is much richer with you in it.
Thank you for sharing this journey with me, and thank you for being here with me today. I can’t
wait to see what life has in store for us next.

Vittorio

94