You are on page 1of 22

Methodology for optimizing composite

towers for use on floating wind turbines


Cite as: J. Renewable Sustainable Energy 9, 033305 (2017); https://doi.org/10.1063/1.4984259
Submitted: 16 September 2016 • Accepted: 13 May 2017 • Published Online: 24 May 2017

Andrew C. Young, Andrew J. Goupee, Habib J. Dagher, et al.

ARTICLES YOU MAY BE INTERESTED IN

WindFloat: A floating foundation for offshore wind turbines


Journal of Renewable and Sustainable Energy 2, 033104 (2010); https://
doi.org/10.1063/1.3435339

Validation of a FAST semi-submersible floating wind turbine numerical model with


DeepCwind test data
Journal of Renewable and Sustainable Energy 5, 023116 (2013); https://
doi.org/10.1063/1.4796197

Comparisons of dynamical characteristics of a 5 MW floating wind turbine supported by a


spar-buoy and a semi-submersible using model testing methods
Journal of Renewable and Sustainable Energy 10, 053311 (2018); https://
doi.org/10.1063/1.5048384

J. Renewable Sustainable Energy 9, 033305 (2017); https://doi.org/10.1063/1.4984259 9, 033305

© 2017 Author(s).
JOURNAL OF RENEWABLE AND SUSTAINABLE ENERGY 9, 033305 (2017)

Methodology for optimizing composite towers for use


on floating wind turbines
Andrew C. Young,1 Andrew J. Goupee,2,a) Habib J. Dagher,1
and Anthony M. Viselli1
1
Advanced Structures and Composites Center, University of Maine, Orono,
Maine 04469-5793, USA
2
Department of Mechanical Engineering, University of Maine, Orono,
Maine 04469-5793, USA
(Received 16 September 2016; accepted 13 May 2017; published online 24 May 2017)

A methodology for the design and optimization of a composite wind turbine tower for
use on a floating offshore platform is presented. A composite turbine tower on a
floating offshore platform not only has the potential to reduce maintenance and
upkeep costs associated with the use of steel offshore but also has potential to reduce
the tower mass and subsequently the support platform mass. The optimization
problem is formulated to obtain a turbine tower that meets all strength and
serviceability criteria and minimizes the tower mass. The optimization and design
process link a number of dynamic analyses and finite element routines using a genetic
algorithm. This work documents the optimization and design software and illustrates
its use in various case studies for a 6 MW floating wind turbine system. Case studies
include the optimization of a steel tower as a comparison to a composite material
tower and the use of a cored, sandwich panel composite tower versus a solid
composite tower. The results demonstrate that significant reductions in the tower mass
are likely when comparing a steel tower with a composite tower. Also demonstrated
for the platform and turbine configuration considered in this work is that the
minimum mass tower design is driven towards a solid shell composite configuration,
rather than a sandwich panel tower shell, in the face of reasonable design constraints.
Published by AIP Publishing. [http://dx.doi.org/10.1063/1.4984259]

I. INTRODUCTION
The development of floating platforms to accommodate wind turbines has gained attention
recently with the deployment of demonstration projects such as Statoil’s Hywind,1 Principle
Power’s WindFloat,2 the University of Maine’s VolturnUS 1:8 project,3,4 Sway AS’s concept in
Norway,5 and the GOTO6 and Fukushima Forward7 projects in Japan. These projects move
from traditionally fixed based offshore turbine structures to floating structures that can be
deployed in greater water depths where the visual impact is reduced and wind resources can be
more favorable.8 Although a floating offshore wind turbine has access to greater wind resour-
ces, additional design and engineering challenges exist to adapt existing turbine and tower tech-
nologies to the floating offshore environment. Both the material and structural technologies
must be further developed to maximize the benefits of a floating offshore wind turbine system.
A significant part of developing the floating wind turbine technology is the development of the
wind turbine tower.
The turbine tower connects the wind turbine to the foundation of the structure (whether a
fixed or floating base). The tower transfers loading from the turbine to the foundation. In the
case of a floating platform, the inertial and gravity loadings of the tower are a larger consider-
ation as the motions of the system are much greater than with a fixed base tower.

a)
Author to whom correspondence should be addressed: agoupe91@maine.edu. Tel.: 207-581-3657.

1941-7012/2017/9(3)/033305/21/$30.00 9, 033305-1 Published by AIP Publishing.


033305-2 Young et al. J. Renewable Sustainable Energy 9, 033305 (2017)

Both land and offshore wind turbine towers are traditionally constructed from steel9
although the use of a composite material wind turbine tower has been studied by others.
Gutierrez et al.10 investigated the use of a wind turbine tower made using composite materials.
Their work was driven primarily by the potential to easily fabricate the composite structures
close to where they would be deployed, a benefit in mountainous or otherwise rugged terrain.
Both a solid composite material tower and a composite tower shell with a concrete core
were analyzed and tested at the 2 MW turbine scale. Polyzois et al.11 conducted numerical and
experimental investigations into the use of composite material wind turbine towers. The designs
explored consisted of multiple composite material filament wound cells that were joined
together with resin. The technology has been adapted from its successful use in similar struc-
tures such as electrical transmission towers. Lim et al.12 investigated the use of filament wound
composite wind turbine towers at the 2 MW scale. Load analyses and optimization were per-
formed on sandwich panel towers composed of E-glass and polyester or epoxy composite mate-
rial face sheets with a polyester core. Various winding angles were investigated.
Composite materials are an attractive option for wind turbine towers because of their high
specific strength, high specific stiffness, and high corrosion resistance.13 As the wind turbine
structure moves from a fixed, land-based design to a floating, offshore design, the mass and cor-
rosion resistant properties become even more attractive. Furthermore, a composite wind turbine
tower can be built and assembled close to the deployment site, reducing transportation costs
and making the use of large scale components possible.10
In light of the promise composite wind turbine towers hold for floating wind turbine appli-
cations, the University of Maine-led DeepCwind Consortium designed, fabricated, tested, and
deployed a prototype composite (E-glass fiber and polyester resin) tower atop the 1:8-scale
VolturnUS 1:8 floating wind turbine in 2013 and 2014.14 An image of the VolturnUS 1:8 is
shown in Fig. 1. Further development was carried out by the University of Maine and Ershigs,
Inc. on composite tower technology, which culminated in the successful mechanical testing of a
1:2-scale, 18.29 m-long tower sub-component which included a joint at the midspan.15 Both the
VolturnUS 1:8 deployment and the 1:2-scale laboratory work have helped to demonstrate the
feasibility of composite tower technology for floating wind turbine applications.

FIG. 1. Image of the VolturnUS 1:8 floating wind turbine with a composite tower.
033305-3 Young et al. J. Renewable Sustainable Energy 9, 033305 (2017)

In an offshore floating structure, the mass of the structural components becomes critical.
An increase in the mass will require an increase in the volume that the floating foundation
displaces in order to maintain buoyancy, potentially leading to a further increase in the mass
of the floating support structure. Furthermore, an increase in the tower mass which is well
above the center of gravity of the floating support structure will decrease the stability of the
floating turbine, leading to a further increase in the required size of the floating foundation.
Optimization of the wind turbine tower to minimize the mass while still maintaining the
required strength, fatigue, stability, and serviceability criteria of the system including the tur-
bine is therefore crucial for minimizing the cost of the system. Young16 estimated that almost
2 kg of hull steel and over 7 kg of hull ballast could be removed for every kg of tower mass
removed for a floating wind turbine using a generic spar-buoy platform. Composite materials
offer great promise for reducing the tower mass; however, a systematic method for achieving
the maximum tower mass reduction while still meeting all structural performance require-
ments is needed.
With this need in mind, a methodology to obtain the optimal, minimum mass design of a
composite wind turbine tower for use on a floating offshore platform is presented. There cur-
rently exists a large wealth of literature in the broad area of laminated composite optimization
to draw from (e.g., see Refs. 17–24). In addition, numerous optimization strategies have been
developed and refined for the purposes of structural optimization as discussed in the review
article by Sigmund and Maute.25 In this work, as has been utilized in many of the previously
noted laminated composite optimization problems, a genetic algorithm (GA) is employed in a
single objective, constrained tower optimization problem. As noted in the study by Sigmund
and Maute, evolutionary optimization approaches do have certain drawbacks including effi-
ciency, a lack of appropriate algorithm stopping criteria, and limited means for flexibly incorpo-
rating multiple constraints.25 However, the primary aim of this work is to summarize the novel
formulation of a single objective, constrained tower optimization problem including the calcula-
tion of the objective function and design constraints for floating wind turbine applications.
Certainly, several options exist for undertaking the optimization problem posed in this work
aside from a GA. That noted, the successful use of a GA in design optimization has been dem-
onstrated (e.g., see Refs. 26–29), and it is relatively simple to implement. This, combined with
the fact that this work is not focused on developing more efficient optimization strategies but
on identifying the design driving aspects of towers for floating wind turbines and showing the
promise of composite towers for these applications, makes a GA a reasonable optimization
strategy choice.
In addition to the minimal tower mass objective, a number of constraints are employed to
obtain a satisfactory and feasible tower design. The natural frequency of the installed tower is
considered using a beam finite element (FE) model in order to ensure that the tower frequency
does not conflict with the forcing frequencies of the turbine or the floating system. Also, the
critical buckling load as well as the stress in the tower shell is considered by coupling the opti-
mization routine to the FE package ANSYS.30 The fatigue design of the towers is not consid-
ered as a design constraint within this optimization routine, but fatigue is checked later both
analytically and experimentally once an optimized composite tower design is developed.14
Incorporating a fatigue analysis into the optimization routine would have significantly increased
the computation time, and previous studies had shown that fatigue was not expected to be a
design driver for a composite tower.11 Tower loading is obtained from a multi-body dynamic
analysis using the National Renewable Energy Laboratory’s (NREL) turbine modeling tool
FAST, version 7.31 Case studies of floating offshore wind turbine systems using FAST have
been conducted by others32 and partially validated with model testing data.33–35
To demonstrate the potential benefits of a composite wind turbine tower on a floating off-
shore platform, a number of case studies are documented for an early stage development 6 MW
VolturnUS floating wind turbine. The VolturnUS platform is being developed by a consortium
of industry partners led by the University of Maine’s Advanced Structures and Composites
Center.3,4 The VolturnUS hull considered here consists of a concrete floating base supporting
either a steel or composite wind turbine tower and a generic 6 MW wind turbine. The case
033305-4 Young et al. J. Renewable Sustainable Energy 9, 033305 (2017)

studies include the use of the optimization routine to analyze three tower designs on the
VolturnUS hull: (1) a steel tower, (2) a composite tower consisting of a sandwich panel shell,
with inner and outer structural layers over a foam core, and (3) a solid shell composite tower.

II. TOWER OPTIMIZATION AND ANALYSIS PROCEDURE


In this section, the tower optimization and analysis routine is documented. All the software
was developed using MATLAB.36 The routine consists of two discrete modules, the tower anal-
ysis routine and the single objective GA routine. The analysis portion of the routine includes a
number of objective and constraint functions that are used in the optimization process.
The entire tower geometry is parameterized with a select number of variables. The varia-
bles are defined at discrete stations along the tower length and interpolated over the length of
the tower. The elevation of a tower station and the diameter of the tower at the station are pre-
defined. Tower section variables include the thickness of layers of the material in the tower
shell.
Tower sections are defined between the tower stations so that the quantities defined at the
stations can be interpolated throughout the tower. The number of tower sections between the
tower stations is predefined. The analysis routine linearly interpolates between the quantities
defined at the tower stations to define the geometry of the tower in the sections. While the
diameter of the tower varies linearly between tower stations, the thicknesses of the shell layers
are constant within each tower section.
The interpolation scheme is illustrated in Fig. 2. This figure presents a tower with three sta-
tions, one at the bottom, one midway up the tower, and the last at the top. The value of the var-
iable (e.g., tower thickness) is defined at the tower stations, for example, normalized design val-
ues of 1.0 at the tower bottom, 0.2 in the middle, and 0.5 at the tower top. The number of
tower sections in this example is predefined to be 10, that is, 5 between each tower section.
The solid line represents the linear interpolation of the design variable, and the dashed lines
represent the interpolated design variable in each of the three sections. It is interpolated in a
piece-wise manner over the length of the tower such that the value is constant within each
tower section. In an actual composite tower, layers of materials will be eliminated, or dropped,
as they are no longer needed, yielding distinct tower sections with uniform properties within
each section. This is representative of the modeling scheme employed here.

FIG. 2. Illustration of the variable interpolation scheme for tower design variables.
033305-5 Young et al. J. Renewable Sustainable Energy 9, 033305 (2017)

A tower design is completely described by the station variables. The number of optimiza-
tion variables depends on what is being optimized and the number of tower stations that are
used in the analysis. Each tower station contains information on the configuration of the tower
at the station, including the elevation of the tower station (hi), the diameter at the station (Di),
and the thicknesses of the material layers, from one to j, which make up the tower shell (ti1, ti2,
ti3, …, tij) in the ith tower station.
The tower design is subjected to a number of objective and constraint functions. The
objective function simply calculates the mass of the tower based on the density of the
materials and tower geometry. Constraints on the tower design are required to ensure that
the final tower design is feasible. A frequency constraint is used to ensure that the turbine
does not excite the tower at a resonant frequency, while stress and buckling constraints
ensure that the tower has the capacity to withstand the expected loading. The constraint
functions include both the custom beam FE code written in the MATLAB environment and
the commercial shell FE code from the FE analysis package ANSYS. The beam FE code is
used in the modal analysis of the tower and is described in the study by Coulling et al.33
The ANSYS shell FE analysis routines are used in the buckling and stress analyses of the
tower.
For the GA portion of the analysis, the real-coded GA algorithm documented in Goupee
and Vel37 was utilized. This implementation is based on select algorithms proposed by Deb.26
The number of variables to be optimized and the population size are predefined. The GA cre-
ates an initial population of individuals within the bounds of the design space. Each individual
is represented by the value of the optimization variables that are determined randomly. The
objective and constraint functions are then evaluated for each individual. Depending on the
value of the objective function, and whether or not the constraints are violated, a fitness value
is assigned to each individual such that smaller numeric values are indicative of a more desir-
able solution. The fitness value is constructed from the objective and constraint functions using
the constraint handling method prescribed by Deb.38 If no constraints are violated, then the fit-
ness is set equal to the objective function. If any one of the constraints is violated, then the fit-
ness is set equal to the sum of the constraint violations plus the objective function value for the
worst feasible solution in the population. If no solutions are feasible in the population, then the
value for the worst feasible solution can be set to an arbitrary value (zero is used in this work).
It is noted that individuals with better fitness values become more likely to pass on their genes,
or optimization variables, to the next generation.
The population of the next generation is determined based on the previous generation.
First, a reproduction scheme is employed that favors individuals with better fitness values for
selection into the mating pool. This operator is analogous to Darwin’s survival of the fittest
concept. In this work, the reproduction operator utilized is a tournament selection scheme.39
Once in the mating pool, two more operations are executed, crossover and mutation. Crossover
is a form of reproduction. The optimization variables are split between two parent individuals
to form an offspring. The crossover operation, which is undertaken with simulated binary cross-
over,40 searches for more fit individuals by combining individuals that are known to have an
above average fitness. The mutation operation, which is performed using a polynomial mutation
operator,38 will alter one or more of the optimization variables in a random manner with a pre-
defined probability. The mutation operation is therefore not applied to all individuals in the
generation. The mutation operation serves to maintain diversity in the population and can help
to prevent convergence to a local minimum.
The objective and constraint functions are again evaluated using the individuals of the
new generation, and fitness values are assigned. Termination criteria can be defined as a popu-
lation minimum fitness value failing to change by more than a preset amount over a specified
number of generations or simply based on the number of generations. In this manner, the fit-
ness of the population becomes better and better, and a feasible, optimal solution is obtained.
A simplified flow chart outlining the tower optimization and design process is presented in
Fig. 3.41
033305-6 Young et al. J. Renewable Sustainable Energy 9, 033305 (2017)

FIG. 3. Simplified tower optimization procedure using a genetic algorithm.

III. FORMULATION OF THE OBJECTIVE AND CONSTRAINT FUNCTIONS


The tower optimization and analysis routine consists of a single objective function and a
number of constraint functions. The objective function estimates the mass of the wind turbine
tower, while the constraint functions determine the fundamental bending frequencies of the
operational tower, the buckling load of the tower, and the maximum stress utilization ratio, i.e.,
the controlling component of stress in the tower shell divided by the ultimate stress of the
material (von Mises yield criteria are used for the steel tower, while the maximum stress failure
criteria are used in the case of the composite towers). The optimization routine utilizes a rela-
tively simple objective function, while the constraint functions are significantly more involved
and computationally intensive to evaluate.

A. Objective function
The mass of the tower is a function of the optimization variables, in particular, the layer
thicknesses, defined at the tower stations. The total mass is found by summing the volume of
each layer of the material in each tower section and multiplying by the respective density of
the material. The mass, M, of a tower with n layers in a section and m sections is therefore
n X
X m
M¼ Vjk qjk ; (1)
j¼1 k¼1

where qjk is the density of the layer in the section. The volume of the jth layer in the kth sec-
tion is idealized as a hollow frustum and is calculated as
8 9
1 >
> D 2
þ D D þ D 2 >
Vjk ¼ pðhkþ1  hk Þ < " k k kþ1 kþ1 #>
=; (2)
12 2
ðDk  2tjk Þ þ ðDk  2tjk ÞðDkþ1  2tjk Þ
>
> >
>
: þðDkþ1  2tjk Þ2 ;

where hk is the height at the bottom of the section, hkþ1 is the height at the top of the section,
Dk is the outer diameter of the layer at the bottom of the section, Dkþ1 is the outer diameter of
the layer at the top of the section, and tjk is the thickness of the layer in the section.

B. Constraint functions
The first of the constraints is on the fundamental natural bending frequencies of the tower.
It is desirable that the fundamental natural bending frequencies of a wind turbine tower avoid
the once per revolution (1P) and three times per revolution (3P) turbine excitation ranges (with
a reasonable buffer) so that the turbine does not excite the tower at a resonant frequency. The
1P and 3P excitations are due to rotor imbalance and blade/tower aerodynamic interactions,
033305-7 Young et al. J. Renewable Sustainable Energy 9, 033305 (2017)

respectively. It should be noted that using modern control techniques, such as frequency hop-
ping (e.g., see Ref. 42), tower resonance issues can be avoided. Thus, if the frequency con-
straint is found to control the design of the tower, the frequency constraint can be eliminated,
and the optimization analysis is rerun to find a lighter tower design keeping in mind that fre-
quency hopping will need to be employed.
The natural frequencies of the tower were determined using a beam FE model that was
developed by Coulling et al.33 for determining the mode shapes and natural frequencies of a
wind turbine tower on a floating foundation. The distributed mass and stiffness properties of the
tower are considered at the Gaussian quadrature points of three dimensional beam elements
(e.g., see Ref. 43). The platform and turbine mass and inertias are lumped at the tower base
and tower top nodes, respectively, to account for the effect of the masses of those components
on the natural frequency of the system. Springs are affixed to the tower base node to account
for the hydrostatic and mooring restoring stiffness of the floating system and serve as the
boundary conditions in the FE model. The beam FE routine then solves the eigenvalue problem
which is of the form
 
½K  x2 ½M fD g ¼ 0; (3)

where ½K is the tower stiffness matrix, x the natural frequency, ½M  the mass matrix, and fD g
the mode shape vector (e.g., see Ref. 43).
In addition to placing constraints on the natural frequencies of the tower, it is also critical
that the ultimate capacity of the tower is not exceeded by the tower loading. Based on previous
experience with steel and composite tower designs15 of similar diameter to thickness ratios, it
is known that the buckling capacity can control the ultimate structural capacity of the tower
shell. As such, the second constraint in the tower optimization and analysis routine is a limit on
the minimum buckling capacity of the tower shell.
A linear buckling analysis is conducted using a shell FE model to estimate when tower
failure is likely to occur. The FE formulation of the buckling eigenvalue problem is
 
½K þ kcr ½Kr  fdDg ¼ 0; (4)

where kcr is the critical buckling factor, ½Kr  is the stress stiffness matrix, and fdDg is a vec-
tor of buckling displacements (e.g., see Ref. 43). In this work, the FE analysis is conducted
using the software package ANSYS. The eight-node quadrilateral layered shell element
SHELL281 is used. The tower is modeled as having a fixed base, and the tower top forces are
distributed to the tower top by means of a rigid cap. The effects of gravitational loading are
included.
The maximum horizontal tower top force and maximum moment obtained from NREL’s
FAST analysis of the floating system are applied to the FE model as well as a downward
force representative of the turbine mass. The tower mass is included by applying a vertical
acceleration of 9.8 m/s2 and modeling the tower shell with a representative mass. The horizon-
tal force and moment are a combination of thrust due to wind loading and nacelle inertial
forces due to the motion of the structure. The buckling factor, or by how much the maximum
horizontal force and moment can be factored before shell buckling occurs, is determined
through the solution of Eq. (4). Figure 4 illustrates a buckled shape on a linearly tapered
tower optimized for the VolturnUS platform. The buckled shape due to bending was found to
be similar to other studies on the buckling behavior of thin walled composite cylinders.44,45
The buckling factor was found to be 3.24 (i.e., loads can be factored 3.24 times before buck-
ling in the tower is predicted), with a refined mesh (1433 elements), as determined from a
convergence study of this tower.
The third constraint in the optimization and analysis routine limits the allowable stress in
the tower shell. The constraint is defined such that the stress utilization ratio (maximum stress
in the material over the allowable stress in the material) does not exceed a predefined threshold
limit.
033305-8 Young et al. J. Renewable Sustainable Energy 9, 033305 (2017)

FIG. 4. Buckled shape of a composite tower.

The stress utilization ratio analysis is conducted using a shell FE model to estimate when
tower material failure is likely to occur. First, the equilibrium of the tower subjected to the
external forces is obtained from the solution of

½KfDg ¼ fFg; (5)

where fDg is a vector of tower displacements and fFg a vector of externally applied forces as
well as inertial forces applied as body forces. The tower shell strains are obtained locally using
element nodal displacements and gradients of the element shape functions.43 The strains are
then employed in classical lamination constitutive relationships to obtain the in-plane stresses
in any layer of the composite material comprising the tower shell.46
In this work, the FE problem of Eq. (5) is solved using the software package ANSYS with
the four-node quadrilateral layered shell element SHELL181 selected. Using the eight-node ele-
ment that was used in the buckling analysis only served to extend the analysis time without
increasing the accuracy of the results. The tower is also modeled as having a fixed base, and
the tower top forces are distributed to the tower top by means of a rigid cap. The effects of
gravitational loading are included.
All six components of the extreme tower top forces and moments (forces at the tower top
along the x, y, and z axes and moments about the x, y, and z axes, Fig. 5) from NREL’s FAST
analyses are applied to the FE stress utilization model. The extreme tower top forces and
moments include the mass and inertial effects from the nacelle. In order to accurately model
the distributed inertial forces due to the acceleration of the tower body, the six components of
translational and rotational acceleration obtained from FAST at the same instant as the extreme
loads, plus acceleration due to gravity, are applied to the tower base. When the extreme forces
are applied to the tower top and the six components of acceleration are applied to the tower
base, the reaction solution from the FE model agrees with the tower base forces obtained from
FAST.
The loading at the tower top produces stress in the tower shell that is then checked to
ensure that it does not exceed the allowable stress utilization ratio of the material (S). In the
case of a composite tower, the components of stress were analyzed on a layer by layer basis
using the maximum stress failure criteria and predefined material strengths. The in-plane and
033305-9 Young et al. J. Renewable Sustainable Energy 9, 033305 (2017)

FIG. 5. Tower top coordinate system.

shear stresses in the principal material axes (r11 , r22 , and r12 ; respectively) were compared
with the failure stress in the material (F11 , F22 ; and F12 ; respectively). When checking the ten-
sion and compression stress along the principal material axis (r11 and r22 ), the strength was
taken as the tensile failure stress when r was greater than zero and the compressive failure
stress when r was less than zero. The maximum stress utilization ratio failure criteria for the
composite material are given in the following equation:
 
jr11 j jr22 j jr12 j
max ; ; < S: (6)
F11 F22 F12

The routine loops through each layer of each tower section to determine the maximum
stress utilization ratio. The nodes where an abrupt change in the shell thickness occurs are
ignored in order to avoid numerical singularities.
Alternatively, for a steel tower, the maximum stress utilization ratio is taken as the maxi-
mum biaxial von Mises stress state in the tower over the yield stress of the material. The von
Mises stress criteria are shown in Eq. (7), where r1 and r2 are the principal stresses and rY is
the yield stress of the material

 2    2
r1 r1 r2 r2
 þ < S: (7)
rY r2Y rY

The use of constraints is required to influence the design of the tower during the optimiza-
tion process and ensure that the optimization routine converges on a feasible solution. The opti-
mal design will therefore be strongly influenced by the constraints that are chosen. Another
influence on the optimal design of the tower is the tower loading. In this paper, the load case
that was expected to control the tower design was selected from the extreme event table gener-
ated from FAST analyses using a realistic tower and held constant throughout the optimization
process. The extreme event table was generated from a subset of the design load cases (DLCs)
that would ultimately be required for a comprehensive analysis. The subset was chosen to
obtain worst case loading scenarios. Using loading from just the controlling extreme event
increases the computational efficiency of the optimization routine and serves as a valid means
of comparison between different tower configurations. Once an optimum tower is generated, a
033305-10 Young et al. J. Renewable Sustainable Energy 9, 033305 (2017)

final tower design must consider all the load cases generated from a comprehensive set of
DLCs to ensure that the final, optimized tower design meets or exceeds the criteria for all the
load cases. Multiple load cases can be incorporated into the optimization routine as additional
buckling or stress utilization constraints although the addition of these constraints will increase
the analysis time. In addition, a radically lightened tower may slightly alter the controlling
loads on the tower. As such, further improvement of the tower design may be obtained by
rerunning the FAST analyses for the optimized tower and using the new extreme events
obtained in another iteration of the optimization process. For the sake of simplicity, this process
was not undertaken in the following parts of this paper.

IV. OPTIMIZATION CASE STUDIES


Three case studies were conducted using the tower optimization and design routine. The
analyses were conducted for a generic 6 MW wind turbine atop an early stage VolturnUS float-
ing foundation. The case studies include both the use of a steel tower and the use of a solid
composite material tower wall versus a sandwich wall design with a foam core. The case stud-
ies serve to illustrate the design drivers for the tower.
Before the case studies are documented, a description of the FAST model that was used to
generate the tower loading is presented. The wind turbine that was used for both the FAST model-
ing and in the beam FE model used to determine the tower mode shapes and natural frequencies
was based on the NREL 5 MW reference turbine47 but was modified to represent a commercial
6 MW offshore turbine. A summary of the wind turbine configuration is presented in Table I.
As noted previously, the VolturnUS floating structure consists of a concrete floating, semi-
submersible base. The gross properties used for modeling purposes are shown in Table II. The
platform properties are representative of an early stage, conceptual VolturnUS design for sup-
porting a 6 MW wind turbine.
Table III summarizes the mooring system’s gross properties. This system utilizes catenary
mooring lines to provide both station keeping functions and help to stabilize the platform.
A limited set of DLCs were analyzed. The DLCs that were analyzed represent the cases
that are expected to control the tower design. The load cases were chosen from the American
Bureau of Shipping (ABS) Guide for Building and Classing Floating Offshore Wind Turbine
Installations.48 The DLCs included power production, parked and extreme survival cases. The
DLCs are documented in Table IV with the environments obtained from Viselli et al.49
Although the subset of DLCs that were analyzed is limited, they represent the controlling
load cases, and the loading that was obtained still serves to provide a valid basis of comparison
between different towers. The wind conditions NTM and EWM in Table IV are a normal turbu-
lence model and an extreme wind model, respectively. The velocity of the wind at the hub is

TABLE I. Gross property overview of the generic 6 MW proxy wind turbine.

Property Value

Power rating 6 MW
Rotor orientation, configuration Upwind, 3 blades
Control Variable speed, collective pitch
Rotor, hub diameter 129.2 m, 5.2 m
Hub height 100 m
Cut-in, rated, cut-out wind speed 3 m/s, 13.2 m/s, 25.0 m/s
Cut-in, rated rotor speed 7 rpm, 12 rpm
Cut-in, rated 1P frequency 0.117 Hz, 0.2 Hz
Cut-in, rated 3P frequency 0.35 Hz, 0.6 Hz
Overhang, shaft tilt, precone 5.02 m, 5 deg, 2.5 deg
Rotor mass 153 000 kg
Nacelle mass 240 000 kg
033305-11 Young et al. J. Renewable Sustainable Energy 9, 033305 (2017)

TABLE II. Gross property overview of the early stage VolturnUS 6 MW floating platform.

Property Value

Total draft 20 m
Elevation to the platform top 15 m
Platform mass (includes water ballast) 15.2  106 kg
Center of mass location below the still water line 8.0 m
Roll/pitch radius of gyration 19.0 m
Yaw radius of gyration 25.5 m
Total displacement 15 594 m3

TABLE III. Gross property overview of the early stage VolturnUS 6 MW mooring system.

Property Value

Water depth 90 m
Number of mooring lines 3
Angle between adjacent lines 120 deg
Unstretched mooring-line length 773.7 m
Radius to anchors from the centerline 800 m
Radius to fairleads from the platform centerline 42.4 m
Line extensional stiffness 2.72 GN
Mooring line wet weight 286.7 N/m

TABLE IV. DLC wind and wave conditions.

Wind conditions Wave conditions

DLC Wind model Vhub (m/s) hwind (deg) Wave model Hs (m) Tp (s) JONSWAP c hwave (deg)

1.6 NTM 3 < Vhub < 25 0 SSS 10.2 14.1 2.75 0


6.1 EWM 35.48 352,0,8 ESS 10.2 14.1 2.75 330,0,30
6.3 EWM 27.13 340,0,20 ESS 6.5 11.6 2.0 330,0,30
Survival EWM 39.81 352,0,8 ESS 12.0 15.3 3.0 330,0,30

Vhub , and hwind is the yaw orientation of the wind inflow normal to the rotor plane. The wave
conditions SSS and ESS in Table IV are the severe sea state and the extreme sea state, respec-
tively. Hs is the significant wave height, Tp the peak spectral period, c the JONSWAP peak
shape parameter, and hwave the orientation of the wave propagation direction relative to the plat-
form x coordinate.
Table V presents the DLC simulation conditions and load factors. The normal (N) load fac-
tor is 1.35, and the abnormal (A) load factor is 1.1. The load factors were included in all
analyses.

TABLE V. DLC load factors.

DLC Simulation conditions ABS load factor

1.6 Normal operational N


6.1 Parked N
6.3 Parked N
Survival Parked A
033305-12 Young et al. J. Renewable Sustainable Energy 9, 033305 (2017)

A. Steel tower baseline


Steel is typically used for land-based wind turbine applications.9 The use of a composite
wind turbine tower is expected to reduce the required mass of the system, leading to a reduced
mass of the floating base. The steel tower was analyzed using the same tower optimization and
analysis tools that were developed for composite towers, except for the failure criteria.
As was mentioned previously, the tower optimization routine consists of a relatively simple
objective function and much more computationally intensive constraint functions. Included in
the steel tower optimization routine were constraints on the natural frequency of vibration, the
buckling factor, and the stress utilization ratio. The natural frequency constraint in the optimiza-
tion analysis specified that the fundamental bending frequency avoids the 1P and 3P (x1P and
x3P ) turbine excitation frequency ranges. The fundamental frequencies of the tower were fur-
ther required to be above 0.3 Hz in order to avoid excessive tower compliance. In the event that
a feasible solution cannot be obtained between the 0.3 Hz and 3P excitation range, the natural
frequency of the tower must be greater than the 3P excitation frequency at the rated wind
speed, plus a 5% buffer of 0.03 Hz. The additional buffer frequency was included to move the
fundamental bending frequency away from the 3P excitation range at the rated wind speed.
This buffer was not included at the cut-in end of the 3P excitation range since the wind speeds
are much lower at the cut-in end, and the aerodynamic coupling between the blade and the
tower will not likely produce significant harmonic forcing for this operational condition. The
stress utilization criteria specified that the maximum factored von Mises stress does not exceed
the yield stress. A factor of 1.67 was chosen based on the ABS Guide for Buckling and
Ultimate Strength Assessment for Offshore Structures.50 The formulation of the optimization
problem is presented in Eq. (8). The buckling constraint specified that the fundamental buckling
mode (kcr ) should be greater than 2.25. The 2.25 factor includes the 1.67 factor used in the
strength analysis and a 1.35 factor to account for geometric imperfections in the tower shell. In
the final design of a steel tower, extensive studies into the sensitivity of the buckling load to
geometric imperfections in the tower shell will be required. In real structures, the elastic buck-
ling load tends to over-predict the actual buckling load.51 For this analysis, a geometrically per-
fect structure was analyzed with a geometric imperfection factor of 1.35.
The mathematical form of the objective function FðTÞ and nonlinear constraints gk ðTÞ uti-
lized in the GA, where (T) are the optimization variables, is detailed in Eq. (8).

Find T
Minimize F ðT Þ ¼ M ðT Þ;
x1
Subject to g1 ðTÞ ¼  1  0;
0:3Hz
  
x1 x1
g 2 ðT Þ ¼ 1  1  0;
x3Pcut in 1:05  x3Prated
kcr
g 3 ðT Þ ¼  1  0;
2:25 "     2 #
r1 2 r1 r2 r2
g4 ðTÞ ¼ 1  1:67  2
þ  0;
rY rY rY
Ti  0; i ¼ 1; 2; 3: (8)

Three optimization variables were utilized. The thickness of the steel shell was optimized
at the tower base (station 1), the tower mid-height (station 2), and the tower top (station 3)
(T ¼ {t11 , t21 , t31 }). There were no limits placed on the maximum thickness of the shell; how-
ever, checking the final shell thickness after convergence is required to ensure that a practical
tower has been designed. The tower height was 82 m, and the tower base and top diameters
were 7.0 m and 3.5 m, respectively. Five tower sections were specified between each station for
a total of ten tower sections. The steel tower shell was assumed to have an elastic modulus of
210 GPa, a Poisson ratio of 0.3, a yield strength of 275 MPa, and a density of 7.85 g/cm3. The
033305-13 Young et al. J. Renewable Sustainable Energy 9, 033305 (2017)

density of the material was further increased by 10% in order to account for connection and
other details not explicitly modeled in this study. The tower top was loaded with a thrust force
(positive x direction, Fig. 5) of 3.3 MN, an axial compressive force (-z direction) of 3.8 MN
(representative of the turbine mass), and a moment about the positive y axis of 10.3 MN m.
The loading was obtained from the FAST analysis of the floating system and included all ABS
specified load factors.48 The controlling load case was an extreme force at the tower top in the
positive x direction (producing a maximum base bending moment). The extreme load case was
found in DLC 1.6 at a rated wind speed of 13.2 m/s, with a significant wave height of 10.2 m
and a wave period of 11.6 s.
A total of 50 individuals were specified for the GA optimization routine. A termination cri-
terion was set such that if the change in the tower mass was less than 100 kg over a span of ten
generations, the optimization analysis was halted.
An optimal steel tower was obtained. A total of 68 generations were required to achieve
the optimal solution. This required 3400 function evaluations and approximately half a day of
computing time on a standard desktop computer. The mass of the tower was optimized to be
608 266 kg. The thicknesses at tower stations 1, 2, and 3 were found to be 62.8 mm, 58.8 mm,
and 20.8 mm, respectively. As can be seen in Fig. 6, the natural frequency of the tower falls on
the upper edge of the 3P frequency constraint at 0.63 Hz. The buckling capacity of the tower
was found to be 12.8 (i.e., the loads that were applied could be multiplied by a factor of 12.8
before a buckling failure is predicted to occur) after the termination of the optimization routine.
The maximum stress utilization ratio was found to be 0.578.
It was not possible for the optimization routine to find a feasible solution between the 1P
and 3P turbine excitation ranges while meeting the stress utilization ratio criteria. This is due in
part to the increase in the tower’s natural frequency caused by the connection to a compliant,
instead of nearly rigid foundation (e.g., see Ref. 52). As a result, the frequency of the tower
must be pushed above the 3P turbine excitation range, giving the design significantly more
mass than it would have otherwise. The optimal tower design is driven by the frequency avoid-
ance criteria, whereas the optimized solution did not approach the buckling criteria or the stress
utilization criteria.
As noted previously, frequency avoidance criteria can also be satisfied through control
schemes such as frequency hopping as opposed to requiring the bending frequencies of the
tower to lie outside of the frequency avoidance bands. To quantify the impact of eliminating
the frequency constraint on the mass of the steel tower, the optimization problem posed in
Eq. (8) is rerun with constraint g2 removed. The GA population size and the termination crite-
rion were kept the same when rerunning the analysis.
A revised, optimal steel tower was obtained. A total of 44 generations were required to
achieve the optimal solution. This required 2200 function evaluations and a little less than half
a day of computation time on a standard desktop computer. The mass of the tower was opti-
mized to be 511 261 kg, a 16% reduction over the steel tower incorporating the frequency con-
straint in its design. The thicknesses at tower stations 1, 2, and 3 were found to be 48.7 mm,

FIG. 6. Optimized steel tower on the VolturnUS platform, natural frequencies, and avoidance ranges.
033305-14 Young et al. J. Renewable Sustainable Energy 9, 033305 (2017)

47.0 mm, and 29.9 mm, respectively. As can be seen in Fig. 7, the fundamental natural frequen-
cies of the tower for the fore-aft (0.59 Hz) and side-side (0.58 Hz) directions fall within the 3P
range and would require frequency hopping to avoid resonance issues. The buckling capacity of
the tower was found to be 10.0 after the termination of the optimization routine. The maximum
stress utilization ratio was found to be 0.60, which is the maximum allowable stress utilization
ratio. It is worth noting that the lower 8 of 10 tower sections possessed maximum stress utiliza-
tion ratios greater than 0.58. This indicates that the tower is approaching a fully stressed
configuration.

B. Comparison of solid and sandwich panel composite towers


Cored sandwich panels are often used in composite structural applications. The panels typi-
cally consist of relatively rigid composite faces with a relatively flexible foam or balsa core.
The panel efficiently resists bending by accommodating the tensile and compressive stresses in
the composite faces and shear stresses in the foam core, analogous to an I-beam.53
Of major concern in a composite tower is the buckling capacity. Buckling in the tower is a
local phenomenon (see Fig. 4), and as such, the buckling capacity can be increased by increas-
ing the local bending stiffness of the tower shell. An efficient way to increase the tower wall
bending stiffness is to use a sandwich panel construction. It is not unreasonable to assume that
an efficient tower shell design could utilize a sandwich panel configuration.
Competing with the buckling capacity of the tower is the tower stiffness. A composite
tower will have a reduced stiffness when compared to a steel tower. Although this is not neces-
sarily detrimental to the use of a composite tower, it is desired to increase a composite tower’s
stiffness. Rather than simply avoiding the 1P turbine excitation frequency at a rated wind speed
of 0.2 Hz, it was desired that the fundamental bending frequency be increased above 0.3 Hz to
prevent excessive tower top displacements while in service. This was also done for the previous
steel tower optimization although it did not ultimately influence the steel tower design.
Although a sandwich panel design is the most efficient method of resisting local buckling,
a solid composite tower shell also has a degree of local buckling capacity. On the other hand,
even though a sandwich panel wall will increase the local stiffness of the laminate, it does little
to increase the global stiffness of the tower. This is because increasing the thickness of the core
only moves the stiff inner skin closer to the neutral axis of the tower.
Two optimization and analysis routines were conducted in an attempt to resolve whether a
sandwich panel or solid composite tower shell is more beneficial in reducing the overall weight
of the floating wind turbine system studied here. The optimization studies were identical except
for the optimization variables. Both the towers were 82 m long and linearly tapered, as in the
case of the steel tower. The tower base diameter was specified to be 7 m, and the tower top
diameter was specified to be 3.5 m (also identical to the steel tower). Both the towers utilized
two tower stations, one at the base (station 1) and the other at the top (station 2). Ten tower
sections were specified between stations. A layer of biaxial material was used on the outside of

FIG. 7. Optimized steel tower eliminating the frequency constraint on the VolturnUS platform, natural frequencies, and
avoidance ranges.
033305-15 Young et al. J. Renewable Sustainable Energy 9, 033305 (2017)

TABLE VI. Properties of unidirectional and biaxial E-glass/epoxy lamina.13

Property (Units) Unidirectional Biaxial

Density (g/cm3) 1.97 1.90


Longitudinal modulus, E1 (GPa) 41 24.5
Transverse in-plane modulus, E2 (GPa) 10.4 23.8
In-plane shear modulus, G12 (GPa) 4.3 4.7
In-plane Poisson’s ratio,  12 (-) 0.28 0.11
Longitudinal tensile strength, F11t (MPa) 1140 433
Transverse tensile strength, F22t (MPa) 39 386
Longitudinal compressive strength, F11c (MPa) 620 377
Transverse compressive strength, F22c (MPa) 128 335
In-plane shear strength, F12 (MPa) 89 84

the tower shell for both the configurations to increase the torsional rigidity of the tower. The
outer biaxial material was not optimized in the current study but taken as a constant for both
the configurations.
Generic E-glass/epoxy lamina and foam core material properties were used in the analysis.
The in-plane properties are presented in Tables VI and VII with the longitudinal direction
aligned with the long axis of the tower shell in the case of the structural laminate. The proper-
ties that were used in the current study can be found in Ref. 13. As in the case of the steel
tower, the density of the material was increased by 10% to account for connection and other
details not explicitly modeled in the current study.
In the case of the sandwich panel tower, the thicknesses of the inner structural composite
layer (ti1 ), foam core layer (ti2 ), and outer structural composite layers (ti3 ) were optimized at
both the tower stations (i equals one and two). The optimization variables for the sandwich
panel tower (Tsp ) are therefore as follows:

Tsp ¼ ft11 ; t12 ; t13 ; t21 ; t22 ; t23 g: (9)

In the case of the solid shell tower, only the thickness of a structural skin (ti1 ) was opti-
mized at each tower station (i equals one and two). The optimization variables for the solid
shell composite tower (Tss ) are therefore as follows:

Tss ¼ ft11 ; t21 g: (10)

The tower top was loaded with a thrust force (positive x direction, Fig. 5) of 3.3 MN, an
axial compressive force of 3.8 MN (negative z direction, representing the topside mass), a

TABLE VII. Properties of the foam core.13

Property (Units) Value


3
Density (g/cm ) 0.25
In-plane modulus, E1 (MPa) 240
In-plane modulus, E2 (MPa) 230
In-plane shear modulus, G12 (MPa) 115
Longitudinal tensile strength, F11t (MPa) 7.2
Transverse tensile strength, F22t (MPa) 7.2
Longitudinal compressive strength, F11c (MPa) 4.6
Transverse compressive strength, F22c (MPa) 4.6
In-plane shear strength, F12 (MPa) 5.0
033305-16 Young et al. J. Renewable Sustainable Energy 9, 033305 (2017)

moment about the positive y axis of 10.3 MN m as well as the corresponding translational, rota-
tional, and gravitational accelerations. The loading was obtained from the FAST analysis of the
floating system. The load case is an extreme force at the tower top in the positive x direction
(producing a maximum base moment). The extreme was found in DLC 1.6 at a rated wind
speed of 13.2 m/s, with a significant wave height of 10.2 m and a wave period of 11.6 s.
Both the towers were subjected to the same constraints. The natural frequency constraint in
the optimization analysis specified that the fundamental bending frequency avoids the 3P (x3P )
turbine excitation range and is above 0.3 Hz.
A relevant DNV standard, Composite Components,54 conservatively specifies a simpli-
fied partial safety factor of 2.5 for a brittle material in a “high” safety class when the load
effect is known with a coefficient of variation (COV) less than 20% and a COV of material
strength between 12.5% and 15%. The simplified partial safety factor includes a partial load
factor and a partial resistance factor. This value corresponds to a resistance factor of 1.85
when accounting for the appropriate ABS load factor of 1.35. An additional resistance factor
of 1.2 was applied to account for the effects of long term loading and environmental effects
described in Ref. 54. The resistance factors correspond to a maximum stress utilization ratio
of 0.45.
In the case of the buckling analysis, the same resistance factors were applied as in the
stress utilization ratio analysis, with an additional factor of 1.35 to account for geometric imper-
fections. In the final design of a composite tower, extensive studies into the sensitivity of the
buckling load to geometric imperfections in the tower shell will be required. In real structures,
the elastic buckling load tends to over-predict the actual buckling load.51 The combined resis-
tance factors correspond to a total factor of 3.0 for buckling.
The formulation for both the sandwich panel and solid shell optimization problems is as
follows:

Find Tsp or Tss


Minimize FðTÞ ¼ MðTÞ;
x1
Subject to g1 ðTÞ ¼  1  0;
 Hz
0:3  
x1 x1
g 2 ðT Þ ¼ 1  1  0;
x3Pcut in 1:05  x3Prated
kcr
g 3 ðT Þ ¼  1  0;
3   
1 r11 r22 r12
g 4 ðD Þ ¼ 1  max ; ;  0;
0:45 F11 F22 F12
Ti  0; i ¼ 1; 2; …; n  m: (11)

A total of 40 individuals were specified for the solid shell tower optimization analysis,
whereas 50 individuals were specified for the sandwich panel tower configuration because there
were more optimization variables. A termination criterion was set such that if the change in the
tower mass was less than 100 kg over a span of ten generations, the optimization analysis was
terminated. The solid shell tower required 30 generations to meet the termination criterion,
whereas the sandwich panel configuration needed 86 generations. This resulted in 1200 function
evaluations for the solid shell tower optimization and 4300 function evaluations for the sand-
wich panel tower optimization. On a standard desktop computer, the solid shell case required
approximately a day of runtime, whereas the sandwich panel case required just below three
days of computer runtime.
The results of the analyses are shown in Table VIII. As can be seen in the table, the sand-
wich panel tower has essentially converged to a solid composite configuration with a negligible
foam core thickness, accounting for the stochastic nature of the GA and the increase in optimi-
zation variables for the sandwich panel configuration. The optimized thickness of the foam core
was less than 0.3 mm at the tower top and bottom.
033305-17 Young et al. J. Renewable Sustainable Energy 9, 033305 (2017)

TABLE VIII. Comparison of sandwich panel and solid shell composite tower optimization analyses.

Sandwich panel configuration

Optimization variable Optimized value

t11 Inner composite 1.28 mm


t12 Foam core 0.27 mm
t13 Outer composite 65.9 mm
t21 Inner composite 2.84 mm
t22 Foam core 0.12 mm
t23 Outer composite 45.7 mm

Solid shell configuration

Optimization variable Optimized value

t11 Structural composite 66.8 mm


t21 Structural composite 48.2 mm

The results of the objective and constraint function analyses are presented in Table IX. The
controlling constraints are the frequency avoidance criteria and the buckling criteria, whereas
both the constraints have converged to the minimum allowable value. The stress utilization ratio
is below the allowable value of 0.45 for both the configurations. For the sandwich panel config-
uration, it can be seen in Table IX that the stress utilization ratio is higher than that for the
solid shell configuration. The maximum stress utilization ratio occurs in the thin layer of the
core material for the sandwich panel configuration. The highest stress utilization ratio in the
sandwich panel configuration not including the foam core occurs in the same location as the
solid shell configuration and is also 0.22.
Since the total tower weight is practically the same for the sandwich and solid composite
tower options, the use of a solid shell composite tower is a more efficient overall construction
for the given constraints used in this design problem. Removing the foam core from the tower
reduces the complexity of the tower construction and therefore reduces manufacturing costs.
Changing the optimization constraints to increase the required buckling factor or decrease the
required tower stiffness will modify the optimal tower design. For example, replacing the
stiffness-driven minimum 0.3 Hz tower bending frequency constraint with a lower value dic-
tated by 1P turbine excitation avoidance would likely lead to a lower-mass, albeit softer,
design. In addition, increasing the number of optimization variable tower stations would also
likely lead to a lower-mass design. It is also possible that with certain modifications to the opti-
mization or constraint parameters, a foam core design might prove to be desirable. However,
for the reasonable set of optimization variables and constraints presented in this analysis, a solid
shell composite wind turbine tower appears to be the best design. The solid wall configuration
was significantly less computationally extensive to obtain because less optimization variables
are required to fully describe the tower configuration.
The two composite towers are 55% lighter than the lightest steel tower optimized in the
previous section (Sec. IV A; a summary of the steel and composite towers can be found in

TABLE IX. Results of analyses on optimized tower configurations.

Sandwich panel configuration Solid shell configuration

Tower mass 231 367 kg 231 512 kg


Fundamental bending frequency 0.30 Hz 0.30 Hz
Optimized buckling factor 3.0 3.0
Maximum stress utilization ratio 0.29 0.22
033305-18 Young et al. J. Renewable Sustainable Energy 9, 033305 (2017)

TABLE X. Comparison of steel and composite configurations.

Steel (no frequency constraint) Composite

Tower mass 511 261 kg 231 512 kg


Fundamental bending frequency 0.58 Hz 0.30 Hz
Optimized buckling factor 10.0 3.0
Maximum stress utilization ratio 0.60 0.22

FIG. 8. Comparison of the optimized tower shells.

Table X). These composite towers can successfully straddle the 1P and 3P turbine excitation
ranges. The steel tower is forced to either have a fundamental bending frequency greater than
the 3P turbine excitation frequency at the rated wind speed or rely on frequency hopping.
For illustrative purposes, the relative thicknesses of the optimized steel (no frequency con-
straint), sandwich panel, and solid shell tower shells are shown in Fig. 8. As can be seen in
Fig. 8, the thickness of the foam core is negligible. The steel tower thicknesses are significantly
smaller than those for the optimal composite tower; however, the much larger density of steel
compared to the E-glass/epoxy composite yields a total steel tower mass that is 2.21 times the
mass of the optimal composite tower. As shown in Ref. 16, this large decrease in the tower
mass could yield an even larger reduction in the hull structural and ballast mass, driving down
the overall cost of the floating offshore wind turbine system.

V. CONCLUSIONS
An overview of the tower analysis routine that was developed to design minimum-mass
steel and composite material towers as well as a summary of the GA that was utilized in the
optimization process was presented. A summary of the tower objective and constraint functions
was provided, and three case studies utilizing the tower optimization and design routine were
conducted. The case studies included optimization of a steel tower for a 6 MW floating turbine
033305-19 Young et al. J. Renewable Sustainable Energy 9, 033305 (2017)

system and two composite materials towers using sandwich panel wall construction and solid
panel wall construction.
A minimum mass steel tower for a scenario considering three tower stations was obtained
for use on the 6 MW floating wind turbine. The natural frequency of the tower could not be
forced between the 1P and 3P values without violating the maximum stress utilization ratio cri-
teria. As such, the tower frequency had to be increased above the 3P excitation range. The
resulting required increase in stiffness leads to a significant increase in the tower mass. The
optimized steel tower mass was found to be 608 266 kg. Eliminating the 3P constraint, which
could be performed in practice through altered turbine rotor speed ranges or via variable speed
generator frequency hopping, yielded a total tower mass of 511 261 kg. This value is 2.21 times
the mass of the optimized composite tower. The significant decrease in the mass obtained with
the use of an optimized composite tower over a steel tower may lead to sizeable reductions in
the floating wind turbine hull mass and cost.
The use of a sandwich panel tower configuration, consisting of composite material face
sheets with a foam core, was compared to that of a solid shell composite tower configuration.
For the given set of constraints considered here, the sandwich panel configuration converged to
a negligible core thickness and resulted in practically the same mass as the optimum solid wall
configuration. Although different constraints may yield a more optimal sandwich wall design,
an important consideration is the added complexity and cost of utilizing a foam core. The solid
wall composite tower configuration appears to be well suited to offshore floating wind turbine
systems due to the relative simplicity of construction and ability to meet reasonable combina-
tions of design constraints.

ACKNOWLEDGMENTS
The authors would like to acknowledge the financial support of the U.S. Department of Energy
Office of Energy Efficiency and Renewable Energy Grant Nos. DE-EE0002981 and DE-
EE0003278, the National Science Foundation (PFI) Grant No. IIP-0917974, the State of Maine
2010 State Bond, the Maine Technology Institute Grant Nos. CIP 111 and CIP 120, and the
University of Maine and the support of the members of the DeepCwind Consortium including
Ashland Inc., PPG, Cianbro and Ershigs.
1
B. Skaare, F. G. Nielsen, T. D. Hanson, R. Yttervik, O. Havmoller, and A. Rekdal, “Analysis of measurements and simu-
lations from the Hywind Demo floating wind turbine,” Wind Energy 18(6), 1105–1122 (2015).
2
C. Cermelli, D. Roddier, and A. Weinstein, “Implementation of a 2MW floating wind turbine prototype offshore
Portugal,” in Proceedings of the 2012 Offshore Technology Conference, Houston, Texas, 30 April–3 May 2012.
3
A. M. Viselli, A. J. Goupee, H. J. Dagher, and C. K. Allen, “Design and model confirmation of the intermediate scale
VolturnUS floating wind turbine subjected to its extreme design conditions offshore Maine,” Wind Energy 19(6),
1161–1177 (2016).
4
A. M. Viselli, A. J. Goupee, and H. J. Dagher, “Model test of a 1:8 scale floating wind turbine offshore in the Gulf of
Maine,” J. Offshore Mech. Arct. Eng. 137(4), 041901 (2015).
5
J. H. Koh, A. Robertson, J. Jonkman, R. Driscoll, and E. Yin Kwee Ng, “Validation of SWAY wind turbine response in
FAST, a focus on the influence of tower wind loads,” in Proceedings of the International Offshore and Polar Engineering
Conference (ISOPE 2015), Kona Hawaii, 21–26 June 2015.
6
T. Utsunomiya, I. Sato, O. Kobayashi, T. Shiraishi, and T. Harada, “Design and installation of a hybrid-spar floating
wind turbine platform,” in Proceedings of the 34th International Conference on Ocean, Offshore and Arctic Engineering,
St. John’s, Newfoundland, 31 May—5 June 2015.
7
Hitachi, “The wind from the future – Fukushima floating offshore wind farm demonstration project,” Hitachi Rev. 63(3),
12–17 (2014); available at http://www.hitachi.com/rev/pdf/2014/r2014_technology02.pdf.
8
W. Musial, S. Butterfield, and B. Ram, “Energy from offshore wind,” in Proceedings of the 2006 Offshore Technology
Conference, Houston, Texas, 1–4 May 2006.
9
T. Burton, D. Sharpe, N. Jenkins, and E. Bossanyi, Wind Energy Handbook, 1st ed. (John Wiley & Sons, West Sussex,
England, 2001).
10
E. Gutierrez, S. Primi, F. Taucer, P. Caperan, D. Tirelli, J. Mieres, I. Calvo, J. Rodriguez, F. Vallano, C. Galiotis, and D.
Mouzakis, “A wind turbine tower design based on the use of fibre-reinforced composites,” ELSA-JRC, Contract No.
ENK5-CT-2000-00328, 2003.
11
D. J. Polyzois, I. G. Raftoyiannis, and N. Ungkurapinan, “Static and dynamic characteristics of multi-cell jointed GFRP
wind turbine towers,” Compos. Struct. 90, 34–42 (2009).
12
S. Lim, C. Kong, and H. Park, “A study on optimal design of filament winding composite tower for 2 MW class horizon-
tal axis wind turbine system,” Int. J. Compos. Mater. 3, 15–23 (2013).
033305-20 Young et al. J. Renewable Sustainable Energy 9, 033305 (2017)

13
I. M. Daniel and O. Ishai, Engineering Mechanics of Composites Materials, 2nd ed. (Oxford University Press, Oxford,
United Kingdom, 2006).
14
A. C. Young, S. Hettick, H. J. Dagher, A. M. Viselli, and A. J. Goupee, “VolturnUS, 1:8-scale FRP floating wind turbine
tower: Analysis, design, testing and performance,” in Proceedings of the 33rd International Conference on Ocean,
Offshore and Arctic Engineering, San Francisco, California, 8–13 June 2014.
15
A. M. Viselli, H. J. Dagher, S. M. Tomlinson, A. C. Young, A. J. Goupee, and S. A. Hettick, “Design, fabrication, and
testing of a composite tower for floating offshore wind turbine,” in Proceedings of the Composites and Advanced
Materials Expo, Orlando, Florida, 13–16 October 2014.
16
A. C. Young, “Investigations into the use of a composite tower on floating offshore wind turbine platforms,” M.S. thesis
(University of Maine, 2013).
17
B. Liu, R. T. Haftka, M. A. Akgun, and A. Todoroki, “Permutation genetic algorithm for stacking sequence design of
composite laminates,” Comput. Methods Appl. Mech. Eng. 186(2–4), 357–372 (2000).
18
J. H. Park, J. H. Hwang, C. S. Lee, and W. Hwang, “Stacking sequence design of composite laminates for maximum
strength using genetic algorithms,” Compos. Struct. 52(2), 217–231 (2001).
19
E. Lund and J. Stegmann, “On structural optimization of composite shell structures using a discrete constitutive para-
metrization,” in Proceedings of the Conference on Science of Making Torque from Wind, Delft, Netherlands, 19–21
April 2004.
20
R. H. Lopez, M. A. Luersen, and E. S. Cursi, “Optimization of laminated composites considering different failure
criteria,” Compos. Part B: Eng. 40(8), 731–740 (2009).
21
S. N. Omkar, J. Senthilnath, R. Khandelwal, G. N. Naik, and S. Gopalakrishnan, “Artificial bee colony (ABC) for multi-
objective design optimization of composite structures,” Appl. Soft Comput. 11(1), 489–499 (2011).
22
A. E. Assie, A. M. Kabeel, and F. F. Mahmoud, “Optimum design of laminated composite plates under dynamic
excitation,” Appl. Math. Modell. 36(2), 668–682 (2012).
23
W. Hu, D. Park, and D. H. Choi, “Structural optimization procedure of a composite wind turbine blade for reducing both
material cost and blade weight,” Eng. Optim. 45(12), 1469–1487 (2013).
24
C. C. Antonio, “A memetic algorithm based on multiple learning procedures for global optimal design of composite
structures,” Memetic Comput. 6(2), 113–131 (2014).
25
O. Sigmund and K. Maute, “Topology optimization approaches,” Struct. Multidiscip. Optim. 48, 1031–1055 (2013).
26
K. Deb, Multi-Objective Optimization Using Evolutionary Algorithms (Wiley, Chichester, England, 2001).
27
J. D. Poirier, S. S. Vel, and V. Caccese, “Multi-objective optimization of laser-welded steel sandwich panels for static
loads using a genetic algorithm,” Eng. Struct. 49, 508–524 (2013).
28
R. Belevicius, D. Jatulis, and D. Sesok, “Optimization of tall guyed masts using genetic algorithms,” Eng. Struct. 56,
239–245 (2013).
29
J. Lohn, D. Linden, and G. Hornby, “Advanced antenna design for a NASA small satellite mission,” in Proceedings of
the 22nd AIAA/USU Conference on Small Satellites, Logan, Utah (2008).
30
ANSYS Mechanical APDL Structural Analysis Guide (ANSYS, Inc., Southpointe, PA, 2012) available at http://
148.204.81.206/Ansys/150/ANSYS%20Mechanical%20APDL%20Structural%20Analysis%20Guide.pdf.
31
J. M. Jonkman and M. L. Buhl, Jr., “FAST user’s guide,” NREL Technical Report No. NREL/EL-500-38230, 2005.
32
J. M. Jonkman, “Dynamics modeling and loads analysis of an offshore floating wind turbine,” NREL Technical Report
No. NREL/TP-500-41958, 2007.
33
A. J. Coulling, A. J. Goupee, A. N. Robertson, J. M. Jonkman, and H. J. Dagher, “Validation of a FAST semi-submersible
floating wind turbine numerical model with DeepCwind test data,” J. Renewable Sustainable Energy 5, 023116 (2013).
34
I. Prowell, A. Robertson, J. Jonkman, G. M. Stewart, and A. J. Goupee, “Numerical prediction of experimentally
observed scale-model behavior of an offshore wind turbine supported by a tension-leg platform,” in Proceedings of the
2013 Offshore Technology Conference Houston, Texas, 6–9 May 2013.
35
J. R. Browning, J. Jonkman, A. Robertson, and A. J. Goupee, “Calibration and validation of the FAST dynamic simula-
tion tool for a spar-type floating offshore wind turbine,” in Proceedings of the Science of Making Torque from Wind
Conference, Oldenburg, Germany, 9–11 October 2012.
36
MATLAB 8, The Mathworks, Inc., Natick, Massachusetts, United States.
37
A. J. Goupee and S. S. Vel, “Two-dimensional optimization of material composition of functionally graded materials
using meshless analyses and a genetic algorithm,” Comput. Methods Appl. Mech. Eng. 195, 5926–5948 (2006).
38
K. Deb, “An efficient constraint handling method for genetic algorithms,” Comput. Methods Appl. Mech. Eng. 186(2-4),
311–388 (2000).
39
D. E. Goldberg and K. Deb, “A comparative analysis of selection schemes used in genetic algorithms,” in Foundations of
Genetic Algorithms 1 (FOGA-1) (Morgan Kaufmann, San Mateo, California, United States, 1991), pp. 69–93.
40
K. Deb and R. B. Agrawal, “Simulated binary crossover for continuous search space,” Complex Syst. 9, 115–148 (1995);
available at http://www.complex-systems.com/issues/09-2.html.
41
A. J. Goupee, “Methodology for the thermomechanical simulation and optimization of functionally graded materials,”
M.S. thesis (University of Maine, 2005).
42
K. Johnson, L. J. Fingeresh, and A. Wright, “Controls advanced research turbine: Lessons learned during advanced con-
trols testing,” NREL Technical Report No. NREL/TP-500-38130, 2005.
43
R. D. Cook, D. S. Malkus, M. E. Plesha, and R. J. Witt, Concepts and Applications of Finite Element Analysis, 4th ed.
(John Wiley & Sons, Hoboken, New Jersey, 2002).
44
C. Bisagni and P. Cordisco, “An experimental investigation into the buckling and post-buckling of CFRP shells under
combined axial and torsion loading,” Compos. Struct. 60, 391–402 (2003).
45
C. Bisagni, “Numerical analysis and experimental correlation of composite shell buckling and post-buckling,”
Composites, Part B 31, 655–667 (2000).
46
J. N. Reddy, Mechanics of Laminated Composite Plates and Shells: Theory and Analysis, 2nd ed. (CRC Press, Boca
Raton, Florida, 2004).
47
J. M. Jonkman, S. Butterfield, W. Musial, and G. Scott, “Definition of a 5-MW reference wind turbine for offshore system
development,” NREL Technical Report No. NREL/TP-500-38060, 2009.
033305-21 Young et al. J. Renewable Sustainable Energy 9, 033305 (2017)

48
American Bureau of Shipping, Guide for Building and Classing: Floating Offshore Wind Turbine Installations
(American Bureau of Shipping, Houston, Texas, 2013); available at https://www.eagle.org/eagleExternalPortalWEB/
ShowProperty/BEA%20Repository/Rules&Guides/Current/195_FOWTI/Guide.
49
A. M. Viselli, G. Z. Forristall, B. Pearce, and H. J. Dagher, “Estimation of extreme wave and wind design parameters for
offshore wind turbines in the Gulf of Maine using a POT method,” Ocean Eng. 104, 649–658 (2015).
50
American Bureau of Shipping, Guide for Buckling and Ultimate Strength Assessment for Offshore Structures (American
Bureau of Shipping, Houston, Texas, 2004).
51
W. C. Young and R. G. Budyas, Roark’s Formulas for Stress and Strain, 7th ed. (McGraw-Hill, New York, New York,
2002).
52
B. J. Koo, A. J. Goupee, R. W. Kimball, and K. F. Lambrakos, “Model tests for a floating wind turbine on three different
floaters,” J. Offshore Mech. Arct. Eng. 136(2), 021904 (2014).
53
D. Zenkert, The Handbook of Sandwich Construction (Chameleon Press Ltd., London, England, 1997).
54
DNV Offshore Standard DNV-OS-C501, Composite Components (Det Norske Veritas AS, 2010); available at http://
rules.dnvgl.com/docs/pdf/DNV/codes/docs/2013-11/OS-C501.pdf.

You might also like