loaded from http://asmedigitalcollection.asme.org/fluidsengineering/article-pdf/135/12/121103/6189147/fe_135_12_121103.pdf?

casa_token=-W2teFbAAG0AAAAA:uUHxG5Ln8NMvKu-N2lCb3NyN5AwfOYEhG4rLiTHdnFoh6EoNrqd9JcKRP3V64CQou0AkbB_aRQ by University Of Tasmania user on 02 Apri

Ali Nematbakhsh
Graduate Research Assistant
A Nonlinear Computational
Student Mem. ASME
e-mail: [email protected]
Model of Floating Wind Turbines
David J. Olinger The dynamic motion of floating wind turbines is studied using numerical simulations. The
Associate Professor full three-dimensional Navier–Stokes equations are solved on a regular structured grid
Mem. ASME using a level set method for the free surface and an immersed boundary method for the
e-mail: [email protected] turbine platform. The tethers, the tower, the nacelle, and the rotor weight are included
using reduced-order dynamic models, resulting in an efficient numerical approach that
Department of Mechanical Engineering, can handle nearly all the nonlinear hydrodynamic forces on the platform, while imposing
Worcester Polytechnic Institute, no limitation on the platform motion. Wind speed is assumed constant, and rotor gyro-
Worcester, MA 01609 scopic effects are accounted for. Other aerodynamic loadings and aeroelastic effects are
not considered. Several tests, including comparison with other numerical, experimental,
and grid study tests, have been done to validate and verify the numerical approach. The
Gretar Tryggvason response of a tension leg platform (TLP) to different amplitude waves is examined, and
Professor for large waves, a nonlinear trend is seen. The nonlinearity limits the motion and shows
Fellow ASME that the linear assumption will lead to overprediction of the TLP response. Studying the
Department of Mechanical Engineering, flow field behind the TLP for moderate amplitude waves shows vortices during the tran-
University of Notre Dame, sient response of the platform but not at the steady state, probably due to the small Keule-
Notre Dame, IN 46556-5684 gan–Carpenter number. The effects of changing the platform shape are considered, and
e-mail: [email protected] finally, the nonlinear response of the platform to a large amplitude wave leading to slack-
ing of the tethers is simulated. [DOI: 10.1115/1.4025074]

1 Introduction Floating wind turbines were first envisioned by Heronemus [1],

but it was not until the mid-1990s that they were examined further
The use of wind turbines on land to produce electricity is now
in Refs. [2–7]. These studies considered various aspects of the fea-
fairly well established, and wind power is rapidly becoming eco-
sibility, design, and economics of the concept. The overall find-
nomically competitive with conventional carbon-based electricity
ings were that floating turbines are technically feasible with
production. However, places best suitable for land-based wind tur-
existing or near-term technology, that some design challenges still
bines, such as the plains of the midwestern United States, are of-
exist, and that reducing platform and mooring system costs is crit-
ten located far from densely populated areas along the coasts, and
ical for economic viability. The cost does, in turn, depend on the
concerns about the negative visual impact of wind turbines on
wind and wave loading on the floating wind turbine and the
attractive natural landscapes have been raised. Many offshore,
detailed motion of the coupled platform, turbine, and mooring sys-
bottom-mounted wind turbines have been constructed in shallow
tem. Studies focused on modeling the response of different
waters (less than 30 m depth) in Europe. The wind is usually
designs to operational and extreme loads are relatively recent and
stronger and steadier for these installations, but objections about
have used either linear frequency-domain (LFD) analysis or time-
the visual impact of shallow water wind farms on coastal land-
domain dynamics (TDD) models or a combination of both. LFD
scapes and properties have delayed development in the US.
approaches are mainly based on linear or, at most, weakly nonlin-
Placing floating wind turbines far offshore, in deep water,
ear descriptions of the waves, assuming potential flow. Viscous
avoids many of the concerns of land-based and shallow water
effects are neglected, and it is assumed that platform motions are
wind turbines. Interest in floating turbines is therefore increasing,
small compared to platform size, so that radiation, diffraction, and
and a few large-scale prototypes for research purposes have been
hydrostatic effects can be superimposed. Also, wave impact
built. Those include two turbines deployed off the coast of Nor-
(slamming) effects are neglected. In TDD modeling, the forces on
way: one full-scale model by Statoil called Hywind and another
the platform are usually computed using the Morison formula,
one called SWAY, which is 1:6 scale-model prototype. In addi-
where drag and inertial effects are considered as two separate
tion, the Blue H prototype has been tested off the coast of Italy
terms. Inertial effects are included as added mass, and drag forces
and Energias de Portugal and Principle Power Inc. have recently
are included using empirically determined drag laws. The stand-
deployed a full-scale 2-MW WindFloat platform off the coast of
ard Morison formula does not include radiational damping and
Portugal. These installations use turbine platforms that are taken
scattering effects, nor does it account for forces in the heave
largely from the oil and gas industry. The Hywind and SWAY
direction.
systems use a spar buoy, consisting of a large-aspect-ratio (draft/
A brief summary of investigations using the above-mentioned
diameter), buoyant steel tank, stabilized by ballast at the bottom.
models is given below. Utsunomiya et al. [8] studied the optimum
The mooring system for spar buoys usually consists of slack or
shape for spar-type floating wind turbine and Phuc and Ishahara
ballasted catenary cables. Another design, and the one studied in
[9] conducted a feasibility study of new platform designs. Both
this paper, is the tension leg platform (TLP) (with a smaller tank
simulations were conducted by TDD using the Morison formula.
aspect ratio), where vertical, tensioned cables provide the restor-
Since this formula was originally developed for long cylinders,
ing moments for stability. The spar buoy and the tension leg plat-
both papers modified it by adding mass and damping terms in the
forms support a tower, the turbine rotor, and the nacelle. Most
heave direction. Karimirad et al. [10] did a code-to-code compari-
design studies, simulations, and experiments of floating turbines
son between HAWC2, developed by the Riso National laboratory
have focused on a 5-MW turbine.
in Denmark [11], and USFOS/vpOne from USFOS Ltd. in Nor-
way [12]. In both codes, the Morison formula is used to find the
Contributed by the Fluids Engineering Division of ASME for publication in the
JOURNAL OF FLUIDS ENGINEERING. Manuscript received December 4, 2012; final
forces in the direction of wave propagation on the platform and
manuscript received July 24, 2013; published online September 12, 2013. Assoc. heave forces are found by integrating the pressure. Subsequently,
Editor: Zvi Rusak. Karmirad and Moan [13] used the HWC2 code to model the

Journal of Fluids Engineering Copyright V

C 2013 by ASME DECEMBER 2013, Vol. 135 / 121103-1
 loaded from http://asmedigitalcollection.asme.org/fluidsengineering/article-pdf/135/12/121103/6189147/fe_135_12_121103.pdf?casa_token=-W2teFbAAG0AAAAA:uUHxG5Ln8NMvKu-N2lCb3NyN5AwfOYEhG4rLiTHdnFoh6EoNrqd9JcKRP3V64CQou0AkbB_aRQ by University Of Tasmania user on 02 Apri
NREL 5-MW spar buoy floating wind turbine. Two new designs 2 Computational Model and Problem Setup
for floating wind turbines were proposed and analyzed by
Sclavounos et al. [14] in order to reduce the installation cost and In the present work, the focus is on the hydrodynamic forces on
weight of floating wind turbine, and Shim and Kim [15] per- the platform and its response. Those cannot, however, be computed
formed a numerical simulation of mini-TLP platforms by consid- accurately without accounting for static and dynamic effects of the
ering the coupled dynamics of blade-rotor, mooring system, and tower, nacelle and rotor, wind forces, and the forces from the teth-
hydrodynamics effects. An extensive summary of the applicability ers. A fully resolved computational model of the full platform may
and limitation of the LFD and TDD approaches can be found in be possible but would certainly be very demanding on computa-
Jonkman [16], who compared predictions by both models. In tional resources. It is likely, however, that the elasticity of the
Roddier et al. [17], the recently conceived WindFloat platform tower, for example, plays only a small role in determining the
was modeled numerically, and Nielsen et al. [18] performed an response of the platform to wave loading. Similarly, using simpli-
integrated dynamic analysis of spar buoy floating wind turbine fied models for other aspects, such as the wind load and the tethers,
(HYWIND). In all these studies [14–18], the WAMIT code [19], is unlikely to change the results we are interested in to any signifi-
which is based on using a potential function in the frequency do- cant degree. Thus, a hybrid computational model has been devel-
main, is used. However, in Refs. [16–18], the final solution is in oped where the hydrodynamics and the platform motion are fully
the time domain so that other effects, such as viscous dissipation resolved, while the wind loading; the tower, nacelle, and rotor mass
and previously radiated waves (memory effects), can be added by and inertia; rotor gyroscopic effects; and the mooring system are
taking advantage of TDD techniques. The empirical nature of the included using reduced models that are coupled to the hydrody-
Morison formula may lead to limitations for new design shapes of namic simulation. Although the aerodynamic and dynamic models
floating platforms, and these can be seen in Roddier et al. [17], are relatively simple at this point, the results show the potential of inte-
who had to conduct experiments to find the viscous drag forces, grating such models with the hydrodynamic model developed here.
due to the special shape of WindFloat columns. Other examples The computational model consists of the flow solver; tracking
of the modeling of hydrodynamic forces on a floating wind turbine methods for the free surface and solid; inclusion of reduced order
platform, using either the time or the frequency domain, can be models for tower, nacelle, and rotor weight and inertia, consider-
found in Refs. [20–23]. ing rotor gyroscopic effects; and modeling the tethers.
To help overcome the limitations of LFD and currently used Our approach is applicable to most platform shapes, but in the
TDD-type models, here, a time domain model is presented, using results section, it will be used to examine the full-scale TLP
the full Navier–Stokes equations to describe the interactions of shown in Fig. 1. It consists of a cylindrical buoyant tank with con-
large-amplitude waves on a floating platform. Fully nonlinear crete at the bottom to provide the ballast for the platform and the
waves are likely to result in maximum structural loading, which in standard NREL 5-MW wind turbine [30]. The ballast will reduce
turn is likely to drive design decisions for these systems. In this the response of the platform to the incoming ocean winds and
approach, there is no limitation on the platform motion, the free waves and also lower the wind turbine center of gravity. The plat-
surface motion is arbitrary, and there is no dependency on experi- form is tethered to the ocean floor by four prestressed tethers that
mental “corrections.” The floating platform is included using an are attached to spokes sticking out horizontally from the bottom
immersed boundary (IB) method [24,25]. Tether forces, tower and of the platform. The platform and tether orientations with respect
rotor weights, rotor gyroscopic effects, and wind loading on the to the incident waves are shown in Fig. 1. The tethers provide the
rotor are modeled as auxiliary forces in the Navier–Stokes equa- main mechanism for keeping the TLP wind turbine stable, and the
tions. This allows us to limit the fluid solver to a region near the excess buoyancy of the tank keeps the tethers in tension. The
water-air interface, thus significantly reducing the computational spokes are mainly used to decrease the pitch, roll, and yaw
cost. The present numerical formulation focuses on the effect of responses of the platform. A tower and a turbine are mounted on
wave loading on platform motions, and wind speed is assumed top of the platform and the blades are rotating, thus causing a gy-
constant. Other aerodynamic loading and aeroelastic effects, such roscopic effect. Surge, sway, and heave for translational and roll,
as wind loads on the elastic blades and wind loads on the turbine pitch, and yaw for rotational motion is used when describing the
tower, are not considered here. Preliminary results have been floating wind turbine motion, as shown in Fig. 1.
reported in Nematbakhsh et al. [26], where an overview of the The response of the wind turbine to wind and waves is studied,
modeling strategy was given and results for a scale model of a using a three-dimensional rectangular domain that encompasses
TLP platform were shown. Nematbakhsh et al. [27,28] showed the water, the water-air interface, and part of the air above the
how to extend the model to a ballasted platform and validated the water. The density ratio of water to air is set to 1000. The domain
numerical results against a 100:1 scale model experiment is discretized using a regular structured grid, consisting of straight
described in Olinger et al. [29]. The present paper expands signifi- but unevenly spaced grid lines. Figure 2 shows the grid and the
cantly on the work described in Refs. [26–29] by giving a more wind turbine from different view angles. The grid lines are spaced
complete description of the model, adding rotor gyroscopic in such a way to resolve relatively well the region around the plat-
effects, presenting further validation of the model, and presenting form. The platform is located one wavelength from the wave gen-
results for a full-scale TLP. erator to minimize effects of waves reflected from the platform
The rest of the paper is organized as follows. In the next sec- and to let the waves develop fully before interacting with the plat-
tion, the computational model is described, focusing on the IB form. The computational domain is 1600  100  175 m in length,
method and the auxiliary forces. The free-surface tracking method width, and height, respectively. For the baseline simulation,
(level set method) is validated by comparing the results with 150  64  64 grid points are used to resolve the domain and the
available analytical solution. The IB method for solid-fluid inter- tank is resolved by 9000 control volumes.
action is validated by comparing the results with both experimen- At the upstream end, waves are generated by specifying the inlet/
tal and numerical data, and the combination of both free surface outlet velocities using an analytical solution for a second-order, free
and solid-fluid interaction problem is verified and validated by surface Stokes wave [31]. The top boundary conditions allow in and
performing grid study and comparison with experimental and nu- out flow to match the velocity specified at the wave maker end, thus
merical data. In the results section, the response of the wind tur- ensuring that mass conservation is satisfied over the entire numerical
bine to a single amplitude and frequency wave is studied. The domain at every time step. Full slip, no-through flow boundary con-
flow field around the platform is examined, the validity of the lin- ditions are imposed on all other sides of the flume. No-slip boundary
ear assumption for the platform response to different amplitude conditions are imposed on the exterior surfaces of the buoyant tank,
waves is investigated, a simple design study is performed, and but boundary layers are, however, not fully resolved, since the grid
finally the interaction of the wind turbine with a large amplitude resolution at the tank surface is not fine enough. The various parame-
nonlinear wave is presented. ters for the floating wind turbine are listed in Table 1.

121103-2 / Vol. 135, DECEMBER 2013 Transactions of the ASME

 loaded from http://asmedigitalcollection.asme.org/fluidsengineering/article-pdf/135/12/121103/6189147/fe_135_12_121103.pdf?casa_token=-W2teFbAAG0AAAAA:uUHxG5Ln8NMvKu-N2lCb3NyN5AwfOYEhG4rLiTHdnFoh6EoNrqd9JcKRP3V64CQou0AkbB_aRQ by University Of Tasmania user on 02 Apri
Fig. 1 The baseline tension leg platform with a 5-MW wind turbine. Notation:
1-rotor, 2-nacelle, 3-tower, 4-platform tank, 5-tank ballast section, 6-spoke, 7-tether.

Fig. 2 Three different views of the upstream part of the computational grid. The grid consists
of straight lines that are unevenly spaced to give a fine resolution around the platform.

In the limit of a very fine grid size, the boundary layer would be turbines, the Reynolds number is high enough so that skin friction
fully captured by the current method, since the no-slip boundary forces can be neglected in comparison with inviscid inertial force
condition is imposed on the solid-fluid interface. Due to the lim- and pressure forces due to the vortex shedding [33].
ited resolution in three-dimensional modeling, the boundary layer The fluid flow is described by the one-fluid Navier–Stokes
is thicker than it should be (see Yang and Stern [32]). Therefore, equations, where one set of equations is used for the whole do-
the skin friction forces on the platform are predicted less accu- main and the different fluids are identified by their different mate-
rately. However, the vortex shedding should be captured relatively rial properties. The momentum equation, with extra forces added
well, as will be shown in Sec. 3 for a flow behind a cylinder. For to couple in the effect of the tower, rotor, nacelle, tethers, and the
large offshore structures in the ocean, such as floating wind wind is

Journal of Fluids Engineering DECEMBER 2013, Vol. 135 / 121103-3

 loaded from http://asmedigitalcollection.asme.org/fluidsengineering/article-pdf/135/12/121103/6189147/fe_135_12_121103.pdf?casa_token=-W2teFbAAG0AAAAA:uUHxG5Ln8NMvKu-N2lCb3NyN5AwfOYEhG4rLiTHdnFoh6EoNrqd9JcKRP3V64CQou0AkbB_aRQ by University Of Tasmania user on 02 Apri
Table 1 The parameters used for the computations of the float- and angular momentums are found by integrating the velocity
ing wind turbine obtained from the solution of the Navier–Stokes equations over
the part of the domain occupied by the platform tank,
Tank mass 4.97  106 kg
Ballast mass 4.67  106 kg ð
Wind turbine reserve 7.25  103 m3 ms uscg ¼ CuqdX
buoyancy (including spokes) ðX (4)
Tethers stiffness 1.8  106 KN
Pretension of tethers 22% of the buoyancy Iscg xs ¼ Cðr  uÞqdX
X
Tower mass 0.347  106 kg
Nacelle þ rotor mass 0.350  106 kg
Rotor second moment of inertia 0.0047  1010 kgm2 The components of the moment of inertia tensor with respect to
Rotor speed 12.1 rpm center of gravity,
Wave period 10 s 2 3
Wave height 5.3 m Ixx Ixy Ixz
Wave gage position in upstream of TLP k/2 ¼ 78 m 6 7
Wave generator distance from TLP k ¼ 156 m Iscg ¼ 4 Ixy Iyy Iyz 5 (5)
Wind thrust force 600 KN (correspond to Ixz Iyz Izz
14 m/s [17])
Center of gravity in Z direction 2.75 m (respect to water
free surface) are computed at every time step by integrating over the part of the
domain occupied by the platform tank. After the velocity at the
center of gravity and the angular velocity have been found, the ve-
locity inside the platform tank is corrected by
 
@u
q þ r  ðuuÞ ¼ rp þ qg þ r  lðru þ ruT Þ uCorr ¼ u þ Cðuscg þ ðxs  rÞ  uÞ (6)
@t
þ FEqF þ FEqM þ FTethers þ FWind (1) and the rigid body constraint in the solid part will be satisfied.
Since the original velocity field and the velocity given by Eq. (6)
The fluids are taken to be incompressible, so the continuity equa- are both divergence free, the new velocity field should be diver-
tion is gence free everywhere. Sometimes this is not exactly true at the
solid-fluid interface; in this case, a few iterations are used. Yabe
ru¼0 (2) et al. [24] and Patankar and Sharma [25] give further details on
the immersed boundary method used.
The last four terms on the right-hand side of Eq. (1) will be While the interaction of the floating tank with the incoming
described later. The governing equations are solved by an explicit waves is captured by solving the Navier–Stokes equations, the
second-order predictor-corrector method on a staggered grid. The effects of the wind and the dynamic effects of other parts of the
advection terms are discretized using a second-order ENO method platform (tower, nacelle, and rotor) are included, using reduced
and a simple second-order, centered difference approximation is models that result in forces and/or moments that act on the plat-
used for the viscous terms. The pressure equation is solved using form. The weight and inertia of the tower, rotor, and nacelle are
a semicoarsing multigrid method [34]. modeled by placing an equivalent point mass at the proper dis-
To track the free surface, a level set method has been used tance from the center of gravity of the platform tank,
[35–37] and advected using a second-order ENO method and a
second-order predictor-corrector method for marching in time. FEqF1 ¼ mEq g þ mEq ðascg þ as  rÞ
The reinitialization equation is iterated a few times in fictitious ¼ mEq ðg þ ðascg þ as  rÞÞ (7)
time to reset the level set function to a distance function [37].
Once the level set function has been updated, the density and vis-
cosity are adjusted to match the new position of the free surface. This force also generates a moment equal to MEq1 ¼ r  FEqF1
The platform consists of a tank that is ballasted with a higher that is added to the center of gravity of the tank. The wind thrust
density material at the bottom, and to track its motion, an force on the rotor is treated in the same way as FEqF1 , since it is
immersed boundary method has been used [24,25]. A marker modeled as a concentrated thrust force (Table 1).
function C, defined such that C ¼ 1 inside the solid and zero Gyroscopic effects, due to rotation of the rotor, can be impor-
everywhere else, is used to identify the solid. Determining tant when the wind turbine is in operational mode and the rotor is
whether a point is inside or outside a given region is, in general, a rotating. Their effect on the motion should, however, be small
complex problem [38], but here, platform geometry is relatively under operating conditions. Therefore, using the dynamic equa-
simple (cylinder). Thus, a point is inside the cylinder if it is less tions of the motion for the rotor [39,40], it can be shown that the
than a radius away from the centerline and less than half the cylin- dominant gyroscopic moments are pitch-induced yaw motion and
der height from the midsection plane. The marker function needs yaw-induced roll motion. Those can be written in the following
to transit smoothly from one value to the other, with the width of form:
the transition being of the order of the grid spacing. Therefore, a
marker function given in Eq. (3) is used. MYaw ¼ ðIyy  Ixx ÞcxPitch
(8)
3
MPitch ¼ ðIxx  Izz ÞcxYaw
2
ðri  xi Þ þ 1:5e ðri  xi Þ
ci ¼ 0:5 þ 0:5 ; i ¼ 1; 2; C ¼ c 1 c2
ððri  xi Þ2 þ e2 Þ1:5 These moments are transferred from the rotor to the floater tank
(3) center of gravity (MEq2 ), since the rotor cannot have independent
roll or pitch motion with respect to the platform (except for tur-
where e can be adjusted to control the thickness of the transition bine alignment with wind direction, which happens slowly enough
zone. The Navier–Stokes equations are solved in the whole do- so that its gyroscopic effects can be neglected [41]).
main, including in the region occupied by the platform tank, To model the four pretensioned vertical tethers that attach the
resulting in a velocity field that generally does not satisfy the rigid platform to the ocean floor, a simple Hooke’s law for the tether
body motion constraint. To correct the velocity, first the linear tension is used in Eq. (1).

121103-4 / Vol. 135, DECEMBER 2013 Transactions of the ASME

 loaded from http://asmedigitalcollection.asme.org/fluidsengineering/article-pdf/135/12/121103/6189147/fe_135_12_121103.pdf?casa_token=-W2teFbAAG0AAAAA:uUHxG5Ln8NMvKu-N2lCb3NyN5AwfOYEhG4rLiTHdnFoh6EoNrqd9JcKRP3V64CQou0AkbB_aRQ by University Of Tasmania user on 02 Apri
Table 2 Modeling strategies for the different components of the floating wind turbine platform

Components Effects considered Modeling method Resultant force and moment on the floater tank

Tank Fully captured IB method –

Tower Weight and inertia Lumped mass FEqF1 and MEq1
Nacelle Weight and inertia Lumped mass FEqF1 and MEq1
Rotor Weight and inertia Gyroscopic Rotating lumped mass FEqF1 and MEq1 MEq2
Tethers Force and moment on the platform Hook’s law Eq. (9) and MEq3
Spokes Volume Hydrostatic FEqF2

FTethers ¼ kðmaxðli  l0 Þ  l0 Þei (9)

Equation (9) shows that, if the tension becomes zero, the tethers
will go slack. The vertical tethers are attached to the end of hori-
zontal spokes extending out from the bottom of the tank to keep
the platform stable. The forces from the tethers are added at the
surface of the tank, but the additional moment from the spokes
(MEq3 ) are included in the model separately. The spokes are not
modeled using an immersed boundary method, but their buoyancy
effects are included as a force FEqF2 . The drag force on the
spokes is assumed to be small and is neglected. The associated
moment due to the drag force on spokes can also be ignored in the
limit of linear pitch and roll motion of TLP. This is a reasonable
assumption for TLP, due to the high stiffness of the tethers.
The sum of all the forces given by the models of the tower, na-
celle, rotor, and spokes results in a total body force on the floater
tank (fourth term on the right hand side of Eq. (1)), which is added
to the center of the ballasted volume of the tank, distributed over
six grid points in each direction,

FEqF ¼ FEqF1 þ FEqF2 (10)

A summation of the moments from the models for the tower, na-
celle, rotor, and spokes gives a total moment,

MEq ¼ MEq1 þ MEq2 þ MEq3 (11)

The moment is represented as three force couples (FEqM in Eq.

(1)) in the pitch, roll, and yaw directions and smoothed over the
ballasted volume of the tank, using a truncated 3D Gaussian to
distribute the forces. Table 2 shows how the different parts of
floating wind turbine are treated.
The overall numerical algorithm is shown in Fig. 3. The outer
loop is the time integration, and the second loop is the iteration
where the tower inertia is coupled with the Navier–Stokes solver.
Each iteration consists of predicting the platform motion with the
Navier–Stokes equations and correcting the prediction by linear
and angular momentum conservation (Eq. (4)). In step 1, the val-
ues from the previous time step, including position, velocity, and
acceleration, are assumed as the initial guess. The same is done
for the marker function and the velocity in the whole domain. The
density and viscosity at every grid point in the whole domain,
including inside the platform tank, are then assigned, based on the
free surface and the platform position (step 2). Equation (1) is
solved in step 3, and the velocity in the whole domain at the new
time level is found. In step 4, the platform position is updated and
used in step 5 to update the marker function identifying the solid.
In step 6, the platform position from step 4, along with the marker
function found in step 5, is used to calculate the linear and angular
velocity of the platform based on the conservation of linear and
angular momentum. These two values are used to correct the ve-
locity found by solving the Navier–Stokes equations in step 7 and
then used for tracking the interface by solving the level set and
the reinitialization equation. Convergence is reached when the Fig. 3 A flow chart of the algorithm for numerical modeling of
values calculated in steps 4–7 do not change. Usually, the angular a floating wind turbine, based on the Navier–Stokes equations.
acceleration of the platform (due to the height of the tower) con- xsi : points required for tracking the wind turbine; /: distance
verges more slowly than other quantities. Once the solution has function for tracking the free surface.

Journal of Fluids Engineering DECEMBER 2013, Vol. 135 / 121103-5

 loaded from http://asmedigitalcollection.asme.org/fluidsengineering/article-pdf/135/12/121103/6189147/fe_135_12_121103.pdf?casa_token=-W2teFbAAG0AAAAA:uUHxG5Ln8NMvKu-N2lCb3NyN5AwfOYEhG4rLiTHdnFoh6EoNrqd9JcKRP3V64CQou0AkbB_aRQ by University Of Tasmania user on 02 Apri
Table 3 Comparison of shedding frequency and lift and drag
coefficients of a cylinder at Re 5 150 under uniform flow

Strouhal Max. and min. Mean drag

number lift coefficient coefficient

Current method (2D and 0.178 60.62 1.50

infinite 3D cylinder)
Lai and Peskin first order 0.156 60.4 1.60
IB method [43]
Lai and Peskin formally second 0.183 60.58 1.45
order IB method [43]
He and Doolen (lattice 0.179 60.49 1.261
Boltzmann method) [44]
Henderson [45] — — 1.33
Williamson (exp.) [46] 0.183 — —
Hammache and Gharib (exp.) [47] 0.176 — —

Fig. 4 Comparison of the numerical results with the analytical inclusion of the advection terms, which are ignored in the analyti-
solution [22] for the amplitude of a decaying wave in a closed
flume
cal solution.
For the second test, since a floating wind turbine tank that is cir-
cular in shape will be considered, the flow field behind a circular
converged, the other quantities listed in the box are updated in cylinder in two dimensions is studied. The problem is solved in a
step 9; otherwise, these values are assigned as the new guess to 25Dc  6.25Dc domain with a Reynolds number equal to 150. The
start the procedure again. After convergence, the variables shown cylinder diameter and the far field velocity is defined as character-
in step 9 are updated and are used in step 10 for updating the free istic length and velocity of the problem. The cylinder diameter is
surface position (solving the level set and the reinitialization equa- resolved by 20 elements. Figure 5 shows the vortices behind the
tion). Then, the next time step can be started. As implemented, the cylinder, and it is clear that the Von Karman street is captured
numerical model is second order in time, but for simplicity, only well behind the cylinder. Table 3 compares the Strouhal number,
the first order version is shown in Fig. 3. For the initial time steps, the mean drag force, and oscillation of the lift coefficient (peak to
usually around ten iterations are required for convergence, but af- peak) with experimental and numerical data. The vortex shedding
ter that, one or two iterations are usually sufficient. For tall towers, frequency agrees very well with earlier results [43–47]. For the
fluctuations of the angular acceleration are large and underrelaxa- lift and drag coefficients, the results are close to data reported by
tion is used after each iteration (step 8). Lai and Peskin [43], which, like the currently used approach, is
based on the IB method. The reason for the difference in the lift
and drag coefficients compared with the results from Henderson
3 Model Validation [45] and He and Doolen [44] is possibly due to the differences in
To verify the numerical method and to assess the grid resolu- the domain size between the studies. In addition to the two-
tion requirements for the problems considered here, a number of dimensional study shown here, the vortex shedding behind a cir-
comparisons with existing results and grid refinement studies have cular cylinder with infinite length has also been simulated, and the
been performed. Here, four examples are discussed. Strouhal number is compared with the experimental results shown
The accuracy of the free surface tracking is examined by com- in Table 3 to guarantee the accuracy of the method in three dimen-
paring the numerical results with the analytical solution for the sions as well.
free surface oscillations of a viscous liquid in a flume with full In the third test, the presented numerical method predictions of
slip walls by Wu et al. [42], for a long, small amplitude wave and the pitch motion of a rectangular free-floating block is compared
a Reynolds number equal to 200. Here, pffiffiffiffiffiffiffiffithe characteristics length with experimental results from Jung et al. [48]. The block is made
and velocity are defined as Dl and gDl , accordingly. The flume of acrylic with a uniform density of 1048 kg/m3 and dimensions
is taken to be 2 m long and the mean depth of the liquid to be 1 m of 0.1 m height, 0.3 m length, and 0.9 m width, which floats in a
and start with a single wave with an amplitude of 10 cm. The sim- flume that is 0.9 m wide, 1.2 m deep, and with a mean water depth
ulation is done using a flume that is 1.2 m high, using a mesh with of 0.9 m. The block width is equal to the width of the flume, so
240  120  4 grid points in the length, depth, and width direc- the problem can be taken to be two dimensional [48]. The block
tion. The density ratio of fluid to air is 1000, so the effects of the can pitch freely, but heave and sway motions are prevented. The
air on the free surface motion is negligible. Figure 4 shows the ini- block is initially tilted at a 15-degree angle in pitch direction
tial condition and the amplitude of the free surface at the center of about its center of gravity and is then allowed to pitch freely until
flume versus time as predicted by both the simulation and the ana- its motion has decayed. Since the decay is affected more by radia-
lytical solution. The agreement is very good, and the slight over- tion of energy by the waves rather than viscous effects, a rela-
prediction of the amplitude by the numerical method is due to the tively long computational domain of 10 m is used and the

Fig. 5 Vortex shedding behind a two-dimensional cylinder for Reynolds number

equal to 150. Dashed and solid lines represent negative and positive vorticity.

121103-6 / Vol. 135, DECEMBER 2013 Transactions of the ASME

 loaded from http://asmedigitalcollection.asme.org/fluidsengineering/article-pdf/135/12/121103/6189147/fe_135_12_121103.pdf?casa_token=-W2teFbAAG0AAAAA:uUHxG5Ln8NMvKu-N2lCb3NyN5AwfOYEhG4rLiTHdnFoh6EoNrqd9JcKRP3V64CQou0AkbB_aRQ by University Of Tasmania user on 02 Apri
Fig. 6 (a) Results from the grid resolution study for the pitch motion of a block. (b) A compar-
ison of the pitch motion results from the finest grid with the experimental results of Ref. [48].

flow and experimental data. It is noted that flow field results, not
shown here, for the long waves (low frequencies) show vortices
on both sides of the block, in agreement with observations
reported by Ref. [48]. These vortices are due to the interaction of
the large amplitude pitch motion of the block with the incoming
waves. If the pitch motion of the block is small, there are no vorti-
ces and linear potential theory can be accurate enough.

4 Results
In this section, the method described and validated above is
used to study the response of the described floating wind turbine
(Fig. 1 and Table 1) to regular periodic waves of different ampli-
tude. Then, briefly, relatively simple design variations are exam-
ined by changing the tank aspect ratio. Finally, the nonlinear
response of the platform to a large amplitude nonlinear wave is
studied.

Fig. 7 Comparison of the response amplitude operators (RAO)

4.1 Response Under Periodic Wave. The floating wind tur-
given by current developed numerical model with the experi- bine and its response to a single frequency and amplitude incom-
mental and theoretical results of Ref. [48] ing waves is shown at two times (separated by a half wave period)
in Fig. 8. The incoming wave and the wind force on the rotor push
the platform downstream, but the prestressed tethers keep the plat-
calculation is stopped before the waves reflect from the bounda- form nearly horizontal, and the major response is an oscillatory
ries. Figure 6(a) shows a grid refinement study, where it is seen surge motion. In Fig. 9(a), the amplitude of the wave, measured
that the results are nearly converged on the 640  240 grid. The halfway between the wave maker and the platform, is shown. The
pitch frequency is relatively well captured by all but the coarsest wave height is slightly lower than the nominal height at the wave
grids, but the damping, which depends on the radiation of energy, maker, due to the presence of the platform. In the absence of the
needs relatively fine grids. A comparison between the results from platform, the waves propagate essentially unchanged downstream
the finest grid with the experimental results of Ref. [48] is shown until they reach the coarsely resolved part of the domain and
in Fig. 6(b). The agreement for the frequency is relatively good, damp out. The nonzero mean water level in Fig. 9(a) is due to
but the experimental results decay faster, possibly due to the asymmetry of the Stokes wave at the inlet. The platform has six
three-dimensional and surfactant effects. degrees of freedom (translations and angles), but the main
In the last test reported here, the pitch response of the same responses are surge, pitch, and heave. Yaw and roll motion may
block described in the third test will be further studied for various happen due to gyroscopic effects (to be discussed later), and there
incoming wave frequencies, and the results are compared with the are essentially no net forces in the sway direction.
experimental data of Ref. [48] and linear potential flow theory. Figures 9(b) and 9(c) show the response of the TLP to the
The block is initially horizontal, and the wave generator is located incoming waves. The dominant response is the surge, as seen in
1.7 wavelengths away from the block. In Fig. 7, the response am- Fig. 9(b), which is restricted by the pretension of the tethers. The
plitude operators (RAO) are shown. The results are in relatively surge motion shows a transient response for about 100 seconds af-
good agreement, especially at higher wave frequencies, where dif- ter the start of the simulation and then an approximately steady
fraction of the waves is important. The numerical results show a state. This evolution is similar to the surge response of a freely
slight shift (about 2%) in the natural frequency toward higher fre- floating structure reported by Koo and Kim [49], and as they also
quencies. Apart from this shift, the RAO responses on either side point out, the frequency of the transient response is a combination
of the resonant peak are in good agreement with both potential of the natural frequency of the TLP in surge and the incoming

Journal of Fluids Engineering DECEMBER 2013, Vol. 135 / 121103-7

 loaded from http://asmedigitalcollection.asme.org/fluidsengineering/article-pdf/135/12/121103/6189147/fe_135_12_121103.pdf?casa_token=-W2teFbAAG0AAAAA:uUHxG5Ln8NMvKu-N2lCb3NyN5AwfOYEhG4rLiTHdnFoh6EoNrqd9JcKRP3V64CQou0AkbB_aRQ by University Of Tasmania user on 02 Apri
Fig. 8 Two frames showing the motion of the floating platform (a) at t 5 213 s and (b) at 218 s (when the wave has propagated
half its wavelength) for the baseline run

wave frequency. The steady state response has, however, nearly yaw and pitch augment each other and can potentially lead to an
the same frequency as the incoming wave, since radiation and vis- instability of the whole platform.
cous damping are not large enough to greatly affect the response The tether tension forces are shown in Fig. 9(d) for the
frequency. In steady state conditions, the TLP response is approxi- upstream and downstream tethers. Because yaw, roll, and sway
mately single frequency at constant amplitude. The slight varia- motions of the platforms are small, the tension in the two
tion in the surge amplitude response, even in steady state, is due upstream tethers are nearly the same. The same is true for the
to the imperfection of incoming wave amplitude described earlier. downstream tethers; therefore, only one upstream and one down-
The amplitude of the heave motion is small compared to the stream tether force are plotted. The figure shows that the tension
surge, because of the high stiffness of the tethers, as seen in in the upstream tethers is higher, because of the mean pitch angle
Fig. 9(b), and like the surge response, the frequency of the heave at of the platform. The transient part of the tether response includes
steady state is nearly the same as the incoming wave. Since the lin- the initial motion, where the platform is trying to find a stable
ear heave response is small and the natural frequency of the heave position, and the later motion, where the effects of the platform
motion of the platform is usually far from the typical peak ocean natural frequency have not yet disappeared. The highest value for
waves frequencies, heave is sometimes neglected in initial design the tension is seen in the very beginning of the simulations, when
studies, such as by Wayman et al. [50], but it might be important in the wind force is applied impulsively.
nonlinear heave response, such as TLP ringing phenomena. The computational model described here can capture details of
The pitch response is very important, since a rocking motion of the flow field around the floating wind turbine tank usually not
the turbine can adversely affect the efficiency and operational available from currently used time domain and linear frequency do-
conditions of the rotor. The pitch response for a TLP floating main methods. As summarized by Sarpkaya [51], oscillatory flow
wind turbine is significantly larger than for traditional TLP struc- around a smooth circular cylinder is governed by the
tures in the oil industry, since the structure is lighter, the center of Keulegan–Carpenter number (KC) and the Reynolds number (Re).
gravity is higher, and there is a moment due the wind force on the At small KC numbers, the effects of the Re number are small. For
rotor. Thus, design concepts developed in the oil industry are the TLP floating wind turbine tank, Um is a function of both wave
unlikely to be applicable for a TLP for a wind turbine. Figure 9(c) and buoyant tank velocity. Knowing that the oscillation frequency
shows the pitch response of the TLP. The frequency is determined of the floating wind turbine tank is nearly the same as the incoming
by the incoming wave, as before, except at the very early time wave frequency, the relative maximum velocity of the oscillatory
when the wind force tilts the tower but the waves have not yet flow with respect to the tank is the summation of the flow and tank
reached the platform. The natural frequency of the pitch motion is velocity, which leads to a maximum of KC  1.25. According to
significantly higher than the frequency of the incoming wave, due Guilmineau and Queutey [52], separation of the flow behind the
to the stiffness of the tethers and the relatively large length of the cylinder will start from KC numbers between 1 and 2 and the wake
spokes. Therefore, its influence on the pitch response is small. will remain closed until around KC ¼ 4. But the study performed
Because of the high stiffness of the tethers, the restoring forces of by Guilmineau and Queutey is in the absence of a free surface. Yu
the TLP in the pitch direction is mainly because of the tethers et al. [53] claimed that presence of the free surface may inhibit vor-
rather than hydrodynamic effects. At steady state, when the struc- tex generation near the wake of a circular cylinder. Thus, for the
ture responds with a single amplitude and frequency, the mean floating wind turbine, generation of the vortices is expected for
pitch response is due to the difference of the mean moment from higher KC values than the values reported in Ref. [52].
the wind force and the mean moment due to the wave force. In steady state condition, vortices have not been seen behind
No noticeable difference in the yaw and pitch response is seen the cylinder, but for the transient region, weak vortices have been
when the gyroscopic effect is added, and an order of magnitude observed, as shown in Fig. 10. The reason for this can be seen in
analysis for the moments supports the numerical results, which Fig. 9(b), where in the very beginning of the transient region,
are also in qualitative agreement with the study done by Jensen there are two large surge motions (before and after the first peak).
[39]. The small values for the yaw and pitch gyroscopic moments The large surge motion gives the flow enough time for the forma-
are due to the platform large second moment of inertia in the yaw tion of vortices behind the cylinder due to a temporary higher KC
direction and restoring moment of the tethers. Although gyro- number. As steady state is approached, the amplitude of the surge
scopic effects are negligible in the present work, they may be motion is reduced and the flow field does not have enough time
important in other design studies, since gyroscopic moments in for the formation of vortices. This result has been confirmed by

121103-8 / Vol. 135, DECEMBER 2013 Transactions of the ASME

 loaded from http://asmedigitalcollection.asme.org/fluidsengineering/article-pdf/135/12/121103/6189147/fe_135_12_121103.pdf?casa_token=-W2teFbAAG0AAAAA:uUHxG5Ln8NMvKu-N2lCb3NyN5AwfOYEhG4rLiTHdnFoh6EoNrqd9JcKRP3V64CQou0AkbB_aRQ by University Of Tasmania user on 02 Apri
Fig. 9 (a) The wave height, measured half a wavelength upstream of tension leg platform; (b) the surge and heave
response; (c) the pitch response; (d) the upstream and downstream tether forces

conducting 2D simulations with different grid resolutions (not Figure 11(a) shows the peak-to-peak surge response to different
presented in the paper) on an oscillating cylinder with the same incoming wave heights for a single frequency wave. It can be seen
Re and KC numbers as for the baseline 3D simulation. that the linear assumption is verified by the current approach up to
Note that, even for the highest wave heights, which 5-MW TLP about 10 m wave height, but above that the results show a nonlin-
floating wind turbines are usually designed for, the KC number is ear trend. The nonlinearity is toward the safe side and the linear
not high enough for the creation of vortex shedding behind the assumption leads to an overprediction of the wind turbine
floating tank. response. Linear curve fitting of the TLP surge response for the
After studying the surge response of the wind turbine to moder- moderate and low amplitude wave heights is also shown in
ate amplitude waves, the effects of changing the wave height, Fig. 11(a), which in the limit of zero wave height yields a zero
while keeping the wave frequency the same, is examined. As sum- surge response, as is expected. One probable reason for the nonli-
marized in Sec. 1, many previous studies of the effects of wave nearity in the larger amplitude waves can be due to the limited
amplitude on floating wind turbines have been conducted in the height of the platform tank outside the water. This will limit the
frequency domain, with a linear assumption for the response of platform area on which the waves can apply forces and will
the platform to different incoming wave heights [16]. On the other reduce the platform response to the wave. The significantly higher
hand, time domain results are mostly based on experimental data. wave disturbance around the platform in Fig. 11(b) compared

Journal of Fluids Engineering DECEMBER 2013, Vol. 135 / 121103-9

 loaded from http://asmedigitalcollection.asme.org/fluidsengineering/article-pdf/135/12/121103/6189147/fe_135_12_121103.pdf?casa_token=-W2teFbAAG0AAAAA:uUHxG5Ln8NMvKu-N2lCb3NyN5AwfOYEhG4rLiTHdnFoh6EoNrqd9JcKRP3V64CQou0AkbB_aRQ by University Of Tasmania user on 02 Apri
Fig. 12 A comparison of the peak-to-peak pitch responses and
the tether forces at steady state for different designs of the
floating wind turbine tank. R shows the radius of three different
platform tanks, which is normalized based on the radius of the
standard tank.

Fig. 10 Top view of the velocity vectors at the midsection of as the radius of the floater tank is increased. This decrease in the
the floating wind turbine tank in the very beginning of the tran- pitch angle is probably because of the increase in the stabilizing
sient region (t 5 23 s). The solid line shows the floating wind moment arm of the tether forces (floater tank radius plus spoke
turbine tank border. Note that slight deviation of the velocity length). On the other hand, the tether peak-to-peak response has a
field on the solid border grids with respect to the inside solid is nonlinear trend. The decrease in tether forces as the radius
due to transition region from solid to fluid. increase is due to the same reason as for the pitch angle, but the
nonlinear trend is likely due to change in the tank shape that
with Fig. 8, which is for incoming wave heights 13.5 m and 5.3 m, affects the force from the incoming wave. The changes here are
accordingly, can be related to this effect. unlikely to greatly affect the hydrodynamic restoring force
because of the high stiffness of the tethers.
4.2 Effects of Tank Diameter and Response to Large While changes in the response of the platform under operating
Wave. An extensive design study would include changing the conditions are obviously very important, these dynamics are
various parameters for the rotor, tower, horizontal spokes, and likely to be captured reasonably well using linear models. The
tethers. Here, however, the focus is on varying the buoyant tank fully nonlinear model presented here is, however, valid for arbi-
geometry. The floater displacement, mass, and mooring system trary surface waves and platform motion and should be particu-
are kept the same, but the buoyant tank diameter is varied larly useful in computing what happens in off-design and
by 610% and the responses of different designs are studied and extreme conditions. Figure 13 shows results from one simulation
compared for moderate amplitude periodic waves described in where the simulation is started by releasing a large mass of water
Subsection 4.1. In both cases, the height of the floater tank is at one end of the domain; therefore, considerable platform
changed so that the volume is the same. Increasing the radius motion is expected. Here, the change in the wind thrust as the
leads to a more barge-like shape, whereas reducing the radius turbine rotor moves in surge direction may not be negligible any-
makes the tank more spar-buoy like. A comparison of the steady more, while the pitch motion is still small enough (less than 1
state peak-to-peak pitch angle and the tether forces is shown in deg) to be neglected. Thus, instead of applying a constant thrust,
Fig. 12. There is a linear decrease in the peak-to-peak pitch angle the relative velocity between the wind and rotor surge motion is

Fig. 11 (a) Surge response of the TLP to different incoming wave heights. At a wave period of 10 s, a linear trend is observed
for low and moderate amplitude waves but not for very large waves. (b) Wind turbine interacting with a 13.5 -m wave. Waves radi-
ated from the floater tank can be seen around the platform.

121103-10 / Vol. 135, DECEMBER 2013 Transactions of the ASME

 loaded from http://asmedigitalcollection.asme.org/fluidsengineering/article-pdf/135/12/121103/6189147/fe_135_12_121103.pdf?casa_token=-W2teFbAAG0AAAAA:uUHxG5Ln8NMvKu-N2lCb3NyN5AwfOYEhG4rLiTHdnFoh6EoNrqd9JcKRP3V64CQou0AkbB_aRQ by University Of Tasmania user on 02 Apri
Fig. 13 The response of the TLP to a large nonlinear wave. (a) The very beginning of the simulation; (b) after 6.25 s. Splashing
of the water can be seen behind the tank.

Fig. 14 (a) The pitch and surge response of the TLP for large nonlinear wave versus time; (b)
the tether forces versus time

computed and used to calculate wind thrust. Following Roddier After the first wave passes the wind turbine, due to reduction of
et al. [17], an appropriate drag coefficient of 0.4 is assumed for the wave forces on the platform, the surge response decreases and
the turbine rotor. The wind speed is the same as for the earlier the pitch angle and the tether forces increase to let the wind tur-
simulations. bine move toward its static stability position. The second wave is
The wave maker is turned off and the elevation of the water in not as strong as the first one, but nearly the same trend can be
the first 31.2 m of the domain is raised by 20 m. At the very begin- seen in the surge, pitch, and tether forces. The second wave is
ning, the water starts to slump down, sending a large wave toward present, since the increase of the water height from the first large
the platform (Fig. 13). Pitch angle and surge response along with wave leads to a decrease in the water height behind it. This
tether forces are shown in Fig. 14. Note that the initial increase in decrease gives enough potential energy to the water for the gener-
pitch, surge, and tether forces on the platform are due to the ation of a second wave. Gradually, the waves become smaller and
effects of the wind and initial oscillation of the platform to satisfy the wind turbine returns back to the static equilibrium position,
the static stability. As the water hits the platform, the surge motion although the wind force is still pushing the wind turbine. Overall,
is still increasing, but the pitch response starts decreasing. The this simulation shows the ability of the method to handle com-
decrease of the pitch response is because of the elevation of plat- pletely nonlinear and nonperiodic conditions for floating wind tur-
form center of gravity in comparison with the resultant hydrody- bine interacting with large waves.
namic force. Both upstream and downstream tethers have
continuously decreasing tension. This decrease is higher in
upstream tethers due to the decrease in the pitch angle. The con- 5 Conclusions
tinuous reduction in the tether forces is likely due to the down- A computational model is developed to predict the response of
ward vertical load mentioned by Bea et al. [54] on the platform a floating wind turbine to large amplitude waves. The method is
and reduction in reserve buoyancy. based on solving the unsteady Navier–Stokes equations with a

Journal of Fluids Engineering DECEMBER 2013, Vol. 135 / 121103-11

 loaded from http://asmedigitalcollection.asme.org/fluidsengineering/article-pdf/135/12/121103/6189147/fe_135_12_121103.pdf?casa_token=-W2teFbAAG0AAAAA:uUHxG5Ln8NMvKu-N2lCb3NyN5AwfOYEhG4rLiTHdnFoh6EoNrqd9JcKRP3V64CQou0AkbB_aRQ by University Of Tasmania user on 02 Apri
level set method to predict the free surface motion and an Um ¼ maximum relative oscillation velocity
immersed boundary method for tracking the floating wind turbine. x1 ¼ distance from an arbitrary grid point to the platform
The tethers, tower, nacelle, and rotor are included using reduced tank axis
order models, leading to a reasonably efficient computational x2 ¼ distance from an arbitrary grid point to the platform
approach. Wind speed is assumed constant. Various comparisons tank midsection plane
with analytical, numerical, and experimental data reveal the accu- as ¼ angular acceleration of the platform tank
racy of the proposed method for nonlinear modeling of floating c ¼ rotor speed
wind turbine. Results are presented for a full-scale turbine C ¼ volume occupied by the platform tank
mounted on a tension leg platform. Pitch, heave, and surge e ¼ smoothing parameter for the platform tank/fluid
response, as well as the tension in the tethers, are shown as a func- interface
tion of time for a periodic wave with a given frequency and ampli- k ¼ wave length
tude. Studying the flow field around the floating wind turbine l ¼ dynamic viscosity
shows that the chance of strong vortex shedding behind the TLP is  ¼ kinematic viscosity
small, due to the large diameter of the TLP tank and the resulting q ¼ density
small Keulegan–Carpenter (KC) number. Thus, the flow remains xs ¼ angular velocity of the platform tank
attached to the floater for regular incoming ocean waves. The X ¼ total volume of the computational domain
effects of varying wave amplitude on the surge response of wind
turbine are studied. The results show that the linear assumption References
for the response of floating wind turbines is accurate for a wide [1] Heronemus, W. E., 1972, “Power From Offshore Winds,” Proceedings of the
8th Annual Conference and Exposition on Applications of Marine Technology
range of wave heights but leads to overprediction for large waves to Human Needs, Washington, DC, pp. 435–466.
(for this case, over 10 m wave height). Also, the difference in the [2] Tong, K., 1998, “Technical and Economical Aspects of a Floating Offshore
platform motion and the tether tension is examined when the Windfarm,” J. Wind Eng. Ind. Aerodyn., 74, pp. 399–410.
shape of the buoyant tank is changed. The present method also [3] Henderson, A. R., and Patel, M. H., 2003, “On the Modeling of a Floating Off-
shore Wind Turbine,” Wind Eng., 6(1), pp. 53–86.
allows simulations of large amplitude waves and fully nonlinear [4] Ushiyama, I., Seki, K., and Miura, H., 2004, “A Feasibility Study for Floating
motion of the platform. One such result is presented here, for Offshore Windfarms in Japanese Waters,” Wind Eng., 28(4), pp. 383–397.
which the tension in the tethers becomes zero. [5] Butterfield, S., Musial, W., Jonkman, J., and Sclavounos, P., 2005,
“Engineering Challenges for Floating Offshore Wind Turbines,” Copenhagen
Offshore Wind Conference, Copenhagen, Denmark.
Acknowledgment [6] Musial, W., Butterfield, S., and Boone, A., 2004, “Feasibility of Floating Plat-
form Systems for Wind Turbines,” 23rd ASME Wind Energy Symposium,
This work was funded by the Energy for Sustainability Program Reno, NV.
of the National Science Foundation (Grant CBET 0853579). [7] Sclavounos, P., 2009, “Floating Offshore Wind Turbines,” Mar. Technol. Soc.
J., 42(2), pp. 39–43.
[8] Utsunomiya, T., Nishida, E., and Sato, I., 2009, “Wave Response Experiment
on SPAR-Type Floating Bodies for Offshore Wind Turbine,” Proceedings of
Nomenclature the Nineteenth International Offshore and Polar Engineering Conference,
ascg ¼ center of gravity acceleration of the platform tank Osaka, Japan.
[9] Phuc, P. V., and Ishihara, T., 2007, “A Study on the Dynamic Response of a
C¼ marker function Semi-Submersible Floating Offshore Wind Turbine System Part 2: Numerical
D¼ floater tank diameter Simulation,” Twelfth International Conference on Wind Engineering, Cairns,
Dc ¼ cylinder diameter Australia.
Dl ¼ half length of the flume [10] Karimirad, M., Meissonnier, Q., Gao, Z., and Moan, T., 2011, “Hydroelastic
Code-To-Code Comparison for a Tension Leg Spar-Type Floating Wind
ei ¼ unit vector along the tether Turbine,” Mar. Struct., 24(4), pp. 412–435.
fe ¼ wave/tank oscillation frequency [11] Larsen, T. J., 2009, HAWC2. The User’s Manual, version 3-8, Riso DTU
fq ¼ vortex shedding frequency National Laboratory for Sustainable Energy, Roskilde, Denmark.
FEqF ¼ equivalent force for using reduced order models for the [12] Hansen, M. O., Aas Jakobsen, K., Holmas, T., and Amdahl, J., 2009,
“VPONE–a New FEM Based Servo Hydro-and Aeroelastic Code for Wind
tower, the rotor, the nacelle, and the spokes Turbines,” Proceedings of European Offshore Wind, Stockholm, Sweden.
FEqM ¼ equivalent moment for using reduced order models for [13] Karimirad, M., and Moan, T., 2011, “Extreme Dynamic Structural Response
the tower, the rotor, the nacelle, and the spokes Analysis of Catenary Moored Spar Wind Turbine in Harsh Environmental Con-
FTethers ¼ tether forces ditions,” ASME J. Offshore Mech. Arct. Eng., 133, p. 041103.
[14] Sclavounos, P. D., Lee, S., DiPietro, J., Potenza, G., Caramuscio, P., and De
FWind ¼ wind force Michele, G., 2010, “Floating Offshore Wind Turbine: Tension Leg Platform
g¼ gravity acceleration and Taught Leg Buoy Concepts Supporting 3-5 Mw Wind Turbines,” European
H¼ heave amplitude Wind Energy Conference and Exhibition, Warsaw, Poland.
Is ¼ second moment of inertia tensor for the platform tank [15] Shim, S., and Kim, M. H., 2008, “Rotor-Floater-Tether Coupled Dynamic Anal-
ysis of Offshore Floating Wind Turbines,” Proceedings of the Eighteenth Inter-
ki ¼ restoring coefficient of the tethers national Offshore and Polar Engineering Conference, Vancouver, BC.
KC ¼ Keulegan–Carpenter number ¼ Um/feD [16] Jonkman, J. M., 2009, “Dynamics of Offshore Floating Wind Turbines. Model
l0 ¼ neutral length of the tethers Development and Verification,” Wind Eng., 12(5), pp. 459–492.
li ¼ tethers length in tension [17] Roddier, D., Cermelli, C., Aubault, A., and Weinstein, A., 2010, “WindFloat: A
Floating Foundation for Offshore Wind Turbine,” J. Renewable Sustainable
mEq ¼ total mass of the tower, the rotor, and the nacelle Energy, 2(3), p. 033104.
ms ¼ mass of platform tank [18] Nielsen, F. G., Hanson, T. D., and Skaare, B., 2006, “Integrated Dynamic Anal-
MPitch ¼ gyroscopic moment in pitch direction ysis of Floating Offshore Wind Turbines,” 25th International Conference on
MYaw ¼ gyroscopic moment in yaw direction Offshore Mechanics and Arctic Engineering, Hamburg, Germany.
[19] Lee, C. H., and Newman, J. N., 2006, WAMIT User Manual, versions 6.3,
p¼ pressure 6.3PC, 6.3S, 6.3S PC., WAMIT, Inc., Chestnut Hill, MA.
r1 ¼ platform tank radius [20] Jonkman, J. M., and Sclavonous, P. D., 2006, “Development of Fully Coupled
r2 ¼ half of the platform tank height Aeroelastic and Hydrodynamic Models for Offshore Wind Turbines,” Proceed-
Re ¼ Reynolds number ¼ uD/ ings of the 44th AIAA Aerospace Sciences Meeting, Reno, NV.
[21] Jonkman, J. M., and Buhl, M. L., 2007, “Loads Analysis of a Floating Offshore
S¼ surge amplitude Wind Turbine Using Fully Coupled Simulation,” Wind Power Conference and
St ¼ Strouhal number ¼ fqDc/u1 Exhibition, Los Angeles, CA.
t¼ time [22] Karimirad, M., 2010, “Dynamic Response of Floating Wind Turbines,” Sci.
u¼ velocity Iran., 17(2B), pp. 146–156.
[23] Stewart, G. M., Lackner, M. A., Robertson, A. N., Jonkman, J. M., and Goupee,
uCorr ¼ corrected velocity by the immersed boundary method A. J., 2006, “Calibration and Validation of a FAST Floating Wind Turbine
uscg ¼ center of gravity velocity of the platform tank Model of the DeepCwind Scaled Tension-Leg Platform,” 22nd International
u1 ¼ far field velocity Offshore and Polar Engineering Conference, Rhodes, Greece.

121103-12 / Vol. 135, DECEMBER 2013 Transactions of the ASME

 loaded from http://asmedigitalcollection.asme.org/fluidsengineering/article-pdf/135/12/121103/6189147/fe_135_12_121103.pdf?casa_token=-W2teFbAAG0AAAAA:uUHxG5Ln8NMvKu-N2lCb3NyN5AwfOYEhG4rLiTHdnFoh6EoNrqd9JcKRP3V64CQou0AkbB_aRQ by University Of Tasmania user on 02 Apri
[24] Yabe, T., Xiao, F., and Utsumi, T., 2001, “The Constrained Interpolation Profile [38] Ogayar, C. J., Segura, R. J., and Feito, F. R., 2005, “Points in Solid Strategies,”
Method for Multiphase Analysis,” J. Comput. Phys., 169(2), pp. 556–593. Comput. Graphics, 29, pp. 616–624.
[25] Patankar, N. A., and Sharma, N., 2005, “A Fast Projection Scheme for the [39] Jensen, C., 2011, “Numerical Simulation of Gyroscopic Effects in Ansys,” M.S.
Direct Numerical Simulation Of Rigid Particulate Flows,” Commun. Numer. thesis, Aalborg University, Aalborg, Denmark.
Methods Eng., 21(8), pp. 419–432. [40] Arnold, R. N., and Maunder, L., 2006, GyroDynamics and Its Engineering
[26] Nematbakhsh, A., Olinger, D. J., and Tryggvason, G., 2011, “A Computational Application, Academic, New York, pp. 83–93.
Simulation of the Motion of Floating Wind Turbine Platforms,” WIT Trans. [41] Pao, L. Y., and Johnson, K. E., 2009, “A Tutorial on the Dynamics and Control
Built Environ., 115, pp. 181–191. of Wind Turbines and Wind Farms,” American Control Conference, St. Louis,
[27] Nematbakhsh, A., Olinger, D. J., and Tryggvason, G., 2012, “Development and MO.
Validation of a Computational Model for Floating Wind Turbine Platforms,” [42] Wu, G. X., Taylor, R. E., and Greaves, D. M., 2001, “The Effect of Viscosity
50th AIAA Aerospace Sciences Meeting, Nashville, TN, AIAA Paper No. on the Transient Free-Surface Waves in a Two-Dimensional Tank,” J. Eng.
2012-0373. Math., 40, pp. 77–90.
[28] Nematbakhsh, A., Olinger, D. J., and Tryggvason, G., 2012, “A Nonlinear [43] Lai, M., and Peskin, C. S., 2000, “An Immersed Boundary Method With Formal
Computational Model for Floating Wind Turbines,” 5th Symposium on Trans- Second-Order Accuracy and Reduced Numerical Viscosity,” J. Comput. Phys.,
port Phenomena in Energy Conversion From Clean and Sustainable Resources, 160(2), pp. 705–719.
Rio Grande, PR. [44] He, X., and Doolen, D., 1997, “Lattice Boltzmann Method on a Curvilinear
[29] Olinger, D. J., DeStefano, E., Murphy, E., Naqvi, S., and Tryggvason, G., 2012, Coordinate System: Vortex Shedding Behind a Circular Cylinder,” Phys. Rev.
“Scale-Model Experiments on Floating Wind Turbine Platforms,” 50th AIAA E, 56(1), pp. 434–440.
Aerospace Sciences Meeting, Nashville, TN, AIAA Paper No. 2012-0375. [45] Henderson, R. D., 1995, “Details of the Drag Curve Near the Onset of Vortex
[30] Jonkman, J. S., Butterfield, S., Musial, W., and Scott, G., 2009, “Definition of a Shedding,” Phys. Fluids, 7, pp. 2102–2104.
5-MW Reference Wind Turbine for Offshore System Development,” Technical [46] Williamson, C., 1988, “The Existence of Two Stages in the Transition to
Report No. NREL/TP-500-38060. Three-Dimensionality of a Cylinder Wake,” Phys. Fluids, 31, pp. 3165–3168.
[31] Sorensen, R. M., 2006, Basic Coastal Engineering, Springer, New York. [47] Hammache, M., and Gharib, M., 1989, “A Novel Method to Promote Parallel
[32] Yang, J., and Stern, F., 2009, “Sharp Interface Immersed-Boundary/Level-Set Vortex Shedding in the Wake of Circular Cylinders,” Phys. Fluids, 1(10), pp.
Method for Wave–Body Interactions,” J. Comput. Phys., 228(17), pp. 1611–1614.
6590–6616. [48] Jung, K. H., Chang, K. A., and Jo, H. J., 2006, “Viscous Effect on the Roll
[33] Bearman, P. W., Downie, M. J., Graham, J. M., and Obasaju, E. D., 1985, Motion of a Rectangular Structure,” J. Eng. Mech., 132(2), pp. 190–200.
“Forces on Cylinders in Viscous Oscillatory Flow at Low Keulegan-Carpenter [49] Koo, W., and Kim, H., 2004, “Freely Floating-Body Simulation by a 2D Fully
Numbers,” J. Fluids Mech., 154(1), pp. 337–356. Nonlinear Numerical Wave Tank,” J. Ocean Eng., 31(16), pp. 2011–2046.
[34] Falgout, R., Cleary, A., Jones, J., Chow, E., Henson, V., Baldwin, C., Brown, [50] Wayman, E. N., Sclavounos, P. D., Jonkman, J., and Musial, W., 2006,
P., Vassilevski, P., and Yang, U. M., “Hypre User’s Manual, Version 2.7,” Cen- “Coupled Dynamic Modeling of Floating Wind Turbine Systems,” Offshore
ter for Applied Scientific Computing (CASC), Lawrence Livermore National Technology Conference, Houston, TX.
Laboratory, Livermore, CA. [51] Sarpkaya, T., 1986, “Force on a Circular Cylinder in Viscous Oscillatory Flow
[35] Osher, S. J., and Sethian, J. A., 1988, “Fronts Propagating With Curvature- at Low Keulegan-Carpenter Numbers,” J. Fluid Mech., 165, pp. 61–71.
Dependent Speed: Algorithms Based on Hamilton-Jacobi Formulations,” J. [52] Guilmineau, E., and Queutey, P., 2002, “A Numerical Simulation of Vortex
Comput. Phys., 79, pp. 12–49. Shedding From and Oscillating Circular Cylinder,” J. Fluids Struct., 16(6), pp.
[36] Fedkiw, R. P., Aslam, T., Merriman, B., and Osher, S., 1999, “A Non- 773–794.
Oscillatory Eulerian Approach to Interfaces in Multimaterial Flows (The Ghost [53] Yu, G., Avital, E. J., and Williams, J., 2008, “Large Eddy Simulation of Flow
Fluid Method),” J. Comput. Phys., 152(2), pp. 457–492. Past Free Surface Piercing Circular Cylinders,” ASME J. Fluids Eng., 130(10),
[37] Sussman, M., and Fatemi, E., 1999, “An Efficient Interface-Preserving Level p. 101304.
Set Redistancing Algorithm and Its Application to Interfacial Incompressible [54] Bea, R., Xu, T., Stear, J., and Ramos, R., 1999, “Wave Forces on Decks of Off-
Fluid Flow,” SIAM J. Sci. Comput., 20(4), pp. 1165–1191. shore Platforms,” J. Waterway, Port, Coastal, Ocean Eng., 125(3), pp. 136–144.

Journal of Fluids Engineering DECEMBER 2013, Vol. 135 / 121103-13