Energy Conversion and Management 248 (2021) 114771

Contents lists available at ScienceDirect

Energy Conversion and Management

journal homepage: www.elsevier.com/locate/enconman

A variable twist blade for horizontal axis wind turbines: Modeling

and analysis
Hamid Khakpour Nejadkhaki , Azadeh Sohrabi , Tejas Prasad Purandare , Francine Battaglia ,
John F. Hall *
Department of Mechanical and Aerospace Engineering, University at Buffalo – State University of New York, Buffalo, NY 14260, USA

A R T I C L E I N F O A B S T R A C T

Keyword: A method for analyzing the performance of a wind turbine blade design with an adaptive variable twist is
Active variable twist presented. The possibility of implementing morphing features is becoming increasingly possible with innovative
Adaptive blade design materials and manufacturing techniques. This motivates a blade concept consisting of multiple shell sections
Additive manufacturing
mounted on a rigid spar and covered by non-structural skin. The design allows the blade shells to adapt the twist
Blade element momentum
Computational fluid dynamics
angle distribution (TAD). The study is conducted using data acquired from the National Renewable Energy
Horizontal axis wind turbine Laboratory (NREL) 20 kW wind turbine. A design problem is formulated to find the TAD that maximizes the
aerodynamic efficiency for a discrete set of points that represent the Region 2 wind speed. The TAD is found for
each point using a heuristic search algorithm. This algorithm searches through performance data that is acquired
using the AeroDyn open source simulation software. Results from the model are validated with computational
fluid dynamics (CFD) simulations using the Reynolds-averaged Navier-Stokes equations with the k-ω shear stress
transport turbulence model. Through this first part of the analysis, it becomes possible to understand the required
amount of blade deformation by the TAD function. Given this understanding, a stress analysis can be performed
on the blade structure. The scenario considers the combined effect of twisting with aerodynamic loading. The
blade load is determined by combining CFD with finite element analysis to facilitate the one-way fluid-structural
interaction. The outcome of this work demonstrates how the proposed method can be used to determine the gain
in efficiency, required range of motion, and structural performance of the blade shells for a given wind turbine
blade design.

technology [1,6].
1. Introduction
1.1. Wind energy operation objectives
In 2019 wind energy exceeded a capacity of 651 GW as the number
of new installations increased by 19% over the previous year [27]. To Wind turbine operation is defined by four regions as shown in Fig. 1
support the increasing energy demand, it is necessary to improve the [28]. The wind conditions are not suitable for operation in Regions 1 and
efficiency and reliability of wind turbine systems [1,2,7]. Blade design is 4 where the system is shut down. In Region 2 the wind speed is sufficient
crucial to these goals and is a research topic across multiple scales of to produce power at a partial level. Therefore, research goals for Region
development [3–5]. A challenge in blade design involves trade-offs be­ 2 focus on the adjustment of the tip-speed ratio as a means of maxi­
tween blade performance objectives that maximize production and mizing wind capture, and thus production [29–31]. In Region 3, the
those that mitigate deleterious loads. The consequence of these trade- wind speed is at or above rated speed and full production is achieved.
offs occurs during operation as energy production is often curtailed to Pitch control is used to dissipate wind energy and maintain operation at
damp drivetrain loads and vibration that promote system failure. The the rated power [8]. Still, in Region 3, there are instances where full
scale of these problems has also increased as wind turbine and rotor sizes power drops off as the blades are controlled to mitigate loads and vi­
continue to increase [1,2,7]. Accordingly, new blade designs, bration. Similar events occur in Region 2 where maximum efficiency is
manufacturing techniques, and advanced rotor architecture with adap­ given up [9]. Trade-offs between energy production and load damping
tive elements are part of the long-term objectives for wind energy are necessary due to the limitations of the conventional blade design.

* Corresponding author.
E-mail address: [email protected] (J.F. Hall).

https://doi.org/10.1016/j.enconman.2021.114771
Received 27 January 2021; Accepted 11 September 2021
Available online 30 September 2021
0196-8904/© 2021 Published by Elsevier Ltd.
H.K. Nejadkhaki et al. Energy Conversion and Management 248 (2021) 114771

Nomenclature θ Relative flow angle,deg

ν Poisson’s ratio
a Axial induction factor ρ Density, kg/m3
a’ Angular induction factor σ’ Local solidity
Cpres Pressure coefficient ϕ Twist angle,deg
cp Power coefficient Ω Blade rotational speed, rad/s
cp,conv Power coefficient for conventional pitch control ω Specific dissipation rate
cp,TAD Power coefficient for TAD improvement
Abbreviations
d Distance from blade root, m
AEP Annual energy production
E Modulus of elasticity, Pa
AM Additive manufacturing
e Axial induction factor error
BC Boundary condition
e’ Angular induction factor error
BEM Blade element method
FTH Thrust force, N
C3D10 10-node quadratic tetrahedron
G Shear modulus, Pa
CFD Computational fluid dynamics
i Cross-section index in design space exploration algorithm
CFRTP Continuous fiber reinforced thermoplastic
k Turbulence kinetic energy
DOE Department of Energy
M Million
FEA Finite element analysis
N Number of cells
FSI Fluid-structure interaction
n BEM analysis elements index
GCI Grid convergence index
P Pressure, N/m2
HAWT Horizontal axis wind turbine
R Rotor (blade) radius, m
IEA International energy agency
r Radial distance, m
NREL National renewable energy laboratory
rf Refinement factor
RANS Reynolds-averaged Navier Stokes
TR Rotor torque, N⋅m
SST Shear stress transport
υw Wind speed, m/s
TAD Twist angle distribution
y+ Dimensionless wall distance
UAE Unsteady aerodynamics experiment
Greek
Δ Relative difference

1.2. Conventional blade structures distribution selections [35]. These parameters are fixed in for rigid
blades. Optimization of these parameters involves trade-offs that focus
Typical wind turbine blade designs are shown in Fig. 2 [32]. Ther­ on the performance for a limited range of wind speed.
mosetting composite blades include the monolithic-skin monocoque,
single-shear web, and double-shear web design configurations [33]. A 1.3. Adaptive blade Innovation
load-carrying girder spar may also be used to strengthen these designs.
Another design is the rib and bulkhead structure. Here the internal The shape of morphing blades is adapted to specific conditions to
structure provides shear stiffness and supports the surrounding aero­ improve aerodynamic performance [10,11]. One morphing type in­
dynamic shell [34]. For all four cases, shells provide structural support volves an adaptive TAD. Loth and Moriarty proposed a morphing TAD
to torsional and edgewise bending loads. One portion of the shell is concept in which blade segments are connected through a tension cable.
subjected to pressure and the other to suction. Aerodynamic loads During operation, centrifugal force pulls on the cable causing a change
depend on blade design and operating conditions. The latter is influ­ in the angle of attack [12]. Gili and Frulla worked on a variable TAD for
enced by the wind speed, tip-speed ratio, and Reynolds number. Blade micro wind turbines [13]. The authors used an actuator to set the local
design primarily concerns airfoil, chord length, and twist angle pitch angle of three internal ribs. In doing so, it was possible to adjust the

Fig. 1. Four regions of wind energy system operation.

2
H.K. Nejadkhaki et al. Energy Conversion and Management 248 (2021) 114771

Fig. 2. Wind turbine blades featuring (a) monolithic skin monocoque, (b) single-shear web, (c) double-shear web (box spar), and (d) rib and bulkhead design.

blade twist distribution. Sirigu et al. advanced the work in [13] through on the system [10–20]. Advancements in materials and manufacturing
the introduction of a control design. The controller broadened the range have the potential to actualize the proposed adaptive blade technology.
of wind speed for which the capability was able to boost wind capture Thus, it is hypothesized that an adaptive blade could be constructed to
[14]. Wang et al. used blade element method (BEM) theory to study a improve the performance of wind energy production and reliability, and
blade design with an adjustable TAD. The study was based on a linear this will be demonstrated in this study.
twist distribution to improve the aerodynamic efficiency of a fixed speed
system [15]. Previous work indicates that this capability can improve 1.5. Blade concept
annual energy production by up to 8.5% for some blade designs [36,37].
The TAD can adjust to maximize efficiency at start-up and to change and A box girder is incorporated into the rib and bulkhead design to
maintain rotor speed at higher wind speeds [12]. The capability enables create the proposed concept. A series of flexible blade segments placed
load reduction with much less loss of efficiency [16]. Bend-twist coupled around a central spar enable the TAD transformation. The external
blades can reduce fatigue damage in wind turbines by as much as 70% surface of the blade is formed by a non-structural skin covering the
[17,18]. Variable TAD can also provide benefits to other aerodynamic segments. A small-scale model, shown in Fig. 3a, illustrates the assembly
applications [19,20]. and components. Fig. 3b shows the schematic of this design with eight
Morphing structures require innovative structural features that blade segments. Actuators are located at the ends of every other
enable shape transformation. The transformation may be enabled by segment. The selection of these locations tuned stiffness of each segment
composite materials that exhibit directional behavior such as compli­ is crucial in achieving the desired TAD as described in the authors’ other
ance in one direction and stiffness in another [22]. Additive work [38,39].
manufacturing (AM) is an enabling technology for these purposes
[23,24]. Optimized directional behavior is obtained through the precise
1.6. Objectives for overarching research and Current study
placement of layered materials and the alignment of reinforcement fi­
bers. Precise placement has also led to significant improvements in the
The authors overarching study will look at Region 2 and 3 separately
strength-to-weight ratio. Li et al. fabricated a carbon fiber reinforced
since the operational objectives differ. Studying Region 2 performance is
polylactic acid composite using 3D printing [25]. The technique
the topic of this paper. The Region 3 study is in process. Preliminary
increased tensile and flexural strengths by 13.8 and 164% each over that
findings in Region 3 are consistent with other studies that suggest TAD
produced with conventional methods. Liu used 3D printing to build
capability can reduce loads by more than 70%. That study further in­
continuous-fiber-reinforced thermoplastic (CFRTP) lattice truss struc­
dicates the higher control capability of the active TAD can mitigate
tures [21]. CFRTP has a high potential for aerospace applications and is
losses in production that occur during transient events in Region 3.
also durable, lightweight, and recyclable. [7]. The size of AM products is
Collectively, the adaptive TAD can have a significant impact on the
also increasing. Oak Ridge National Laboratory [26] has produced AM
system through production gains, load reduction, stability, and a
molds for wind turbine blades. Moreover, printing speed has increased.
simplified structural design. The objectives of the work in this paper are
Printing performed by the Big Area Additive Manufacturing machine is
to (1) explore a blade concept for an active TAD, (2) devise a model to
1000 times faster than that of conventional machines.
study the TAD in Region 2, (3) ensure the model provides reasonable
results, (4) determine the range of TAD transformation, and (5) analyze
1.4. Impetus for adaptive twist capability
the structural response of blade shells subjected to torsional and aero­
dynamic loads.
Information gained from the literature, shown in Table 1, motivates
for the study for an adaptive twist capability. Innovation is needed to
2. Case study model
improve the reliability and production level of wind energy systems
[1–7]. This will become increasingly important as the dependency on
A case study is presented to facilitate an understanding of the pro­
wind energy grows. Work of the authors and others shows adaptive
posed blade concept and associated analysis. This study employs a
blades provide new capabilities that improve energy production while
simulation model based on the NREL Unsteady Aerodynamics Experi­
also mitigating loads and other phenomena that promote wear and tear
ment (UAE) Phase VI turbine [40]. The system characteristics are
summarized in Table 2. Detailed geometry and experimental procedures
Table 1 can be found in the NREL report [40]. This model is selected because it
Salient findings from published reports and other studies.
provides certified data that has also been used in other published studies
Findings Sources [15,41]. Wind data acquired from the NREL wind resource dataset is
Need for new rotor technology [1–7] used as an input for this model [42].
Current blade design and limitations [1,2,7–9] A 3D model of the blade is prepared as shown in Fig. 4. The structure
Benefits of blade with adaptive twist angle [10–20] is assumed to be made from ULTEM 9085 material using fused deposi­
Potential blade manufacturing processes and benefits [7,21–26]
tion modeling. This material has a high strength-to-weight ratio and is

3
H.K. Nejadkhaki et al. Energy Conversion and Management 248 (2021) 114771

Fig. 3. (a) Small-scale 3D printed blade and (b) modular blade concept using shells.

Region 2. The TAD is defined in Section 3.1 for use in the analysis.
Table 2
Section 3.2 gives a brief overview of the AeroDyn [46] software used to
NREL UAE Phase VI turbine characteristics.
acquire the aerodynamic performance. This model is combined with a
Parameter Value heuristic search algorithm described in Section 3.3. The algorithm finds
Rated power 20 kW the TAD that provides the greatest efficiency at a given wind speed. A
Cut-in speed 5 m/s CFD model is presented in Section 3.4. Model performance is verified
Rated speed 13.5 m/s and subsequently used to validate results obtained from the AeroDyn
Blade airfoils S809
Number of blades 2
software. The CFD model is also used to determine the distribution of
Hub Height 12.192 m pressure across the blade surface. This distribution is implemented in
Rotor Diameter 10.058 m Section 3.5 for the structural analysis. Finite element analysis (FEA) is
Blade tip pitch angle 2◦ -14◦ implemented to examine the structural response due to the fluid struc­
Inlet 5–25 m/s
ture interaction (FSU) and applied torsional load that transforms the
TAD. The last step examines the ability of this design to maintain the
intended airfoil shape and the impact this may have on performance.

3.1. Representation of the TAD

The relative flow angle, θ, over the wind turbine blade cross-section
varies in moving from the root to the tip. This is the reason blades are
manufactured with a twist angle that changes along the length. These
changes can be expressed in terms of the twist angle, ϕ, that occurs at
each distance, d, from the blade root as shown in Fig. 5. The TAD refers
to the set of points defined by these two variables. Conventional rigid
blades have a fixed TAD. Consequently, the TAD is optimized for a
specific wind speed in the fixed speed system. Accordingly, the aero­
Fig. 4. 3D model of the segment. dynamic analysis finds the appropriate TAD as a function of wind speed.

intended for aerospace applications [43,44]. The blade geometry is

3.2. AeroDyn simulation
based upon a conventional design [45] with wall and rib thicknesses
corresponding to 1% and 3% of the chord length. This model evaluates
AeroDyn is used to compute aerodynamic loads on the blade. Cal­
the cross-sectional changes. Material properties used in the study are
culations for modeling the wind turbine rotor are based on the quasi-
provided in Table 3. Since the material is anisotropic, the properties are
steady BEM, which requires an iterative nonlinear solution [47]. The
defined in three directions, noted as 1, 2, and 3 [43]. The blade is
process is conducted by dividing the blade into elements that are
assumed to be printed in direction 3. Therefore, the length of the blade
analyzed separately. BEM is known for its efficiency and provides
span is direction 3, perpendicular to the chord line in the airfoil plane is
reasonable results for the study [48]. The classic BEM needs corrections
direction 2, and the chord line of the middle cross-section is direction 1.
to obtain reliable results [48]. BEM doesn’t consider the effect of the
vortices shed into the induced velocity field. This occurs at both the
3. Methods
blade tip and hub. The effect near the blade tip has the greatest influence
on the aerodynamic efficiency. Aerodyn employs Prandtl’s loss correc­
A series of modeling steps explore the aerodynamic and structural
tion factors to compensate for the losses in both locations [47]. Cor­
performance of a blade with an adaptive TAD. Aerodynamic perfor­
rections are also needed when the turbine is operating with a high tip-
mance is evaluated by comparing the aerodynamic efficiency of the
speed ratio. A high induction factor means that the momentum theory
adaptive blade to that of the original blade that has a fixed TAD. This
breaks down [48]. Aerodyn recognizes this and applies the Glauert
comparison is made for a series of points that represent wind speed in
correction factor in these cases. Given that the classic BEM is for

Table 3
Stiffness properties of ULTEM 9085 [41].
Property Modulus of elasticity [MPa] Poisson’s ratio [-] Shear modulus [MPa]

E1 E2 E3 ν12 ν13 ν23 G12 G13 G23

Value 2539.4 2327.9 2159.6 0.46 0.39 0.40 635.5 635.5 582.82

4
H.K. Nejadkhaki et al. Energy Conversion and Management 248 (2021) 114771

Fig. 5. Definition of the twist angle, ϕ, at distance,d

axisymmetric flow, a skew wake correction is used based on additional forces. The Reynolds-averaged Navier-Stokes equations (RANS) equa­
relationships developed by Glauert [47]. The sections results are com­ tions are employed to solve conservation of mass and momentum, and
bined in the last step to provide the thrust force, FTH , the rotor torque, the turbulence kinetic energy (k) and specific rate of dissipation (ω) are
TR . Results acquired from the BEM model are compared to those represented in the k-ω shear stress transport (SST) turbulence model. A
determined by CFD to ensure reliability. The process is explained as part fully implicit coupled scheme was employed for pressure–velocity
of the CFD work in Section 3.4. coupling. In all cases, the second-order upwind discretization scheme
was adopted to discretize the momentum and turbulence equations.
The 3D geometry of the wind turbine was based on the experiments
3.3. TAD search algorithm
conducted by NREL for the S809 blade. For the work herein, the tower
was excluded from the simulations, but the location of the tower and
A heuristic algorithm is implemented to find the set of TAD shapes
wind tunnel size are consistent with NREL, as shown in Fig. 7a. The
that maximize wind capture for a discrete set of wind speed that rep­
computational domain (24.4 m × 36.6 m × 55.9 m) is the same size as
resents Region 2. The procedure [49] illustrated in Fig. 6 finds the TAD
the wind tunnel where the NREL tests were performed. For reference,
at each point. To find the TAD, one blade element is studied at a time by
the dimensions of the wind tunnel are shown in terms of the blade radius
making discrete changes to the twist angle. As this occurs, the twist
(R). The tower was excluded to make the numerical simulation simpler
angle of all the other elements remains constant. The twist angle with
and reduce the computational time; it will be shown that the simulations
the maximum cp is considered to be the new twist angle of that cross-
compare extremely well with the NREL experiments.
section. The search domain for the twist of each element, ϕ(di ), ranges
The upstream boundary assumes a horizontal uniform flow of ve­
between those twist angle used in the previous ϕ(di− 1 ) and next elements
locity υw . The inlet turbulent intensity is set to 0.5%, which matches the
ϕ(di+1 ). When analyzing the first and last cross-sections of the blade, the
inlet data provided by NREL. The sidewalls of the tunnel were modeled
search domain is between ϕ(di ) and ϕ(di+1 ), and between ϕ(di− 1 ) and
as no-slip surfaces and the downstream flow exits at ambient pressure.
ϕ(di ), respectively. The search domain is directly related to the updated
For satisfactory convergence, the residuals were set to 1e-5 to ensure
TAD. Consequently, after each maximization step, the search domain
that they reduced three orders of magnitude.
may change for the next cross-section. When all of the cross-sections are
Fig. 7b shows a sample of the mesh used to discretize the computa­
investigated, the maximization procedure for the blade is repeated,
tional domain. The mesh includes two overset blocks. The first block
starting with the first non-circular cross-section. The reason for
(cylindrical section) is converted from a tetrahedral into a polyhedral
repeating the process is that the new TAD might provide a new search
arrangement to considerably reduce the number of cells and better align
domain and improve results for some cross-sections. It is repeated until
the cell faces and flow [50]. The second block is the outer domain using
no change occurs in the twist nor the power coefficient, cp . It is notable
hexahedral cells. The overset meshing technique was chosen to retain
that other search tools, such as a genetic algorithm could have been used
mesh resolution and to create separate meshes associated with the
to find the TAD. However, the heuristic algorithm [49] provided a
different blocks. A dynamic mesh was used for the rotational model. The
reasonable solution for this study. The approach was not computation­
regions closer to the blade use very fine cells and coarser cells are used
ally expensive and gave the authors to a better understanding of TAD
further away from the blade. To resolve the boundary layer, prism layers
aerodynamics.
were attached to the blade surface to capture the flow details near the
wall. The height of the first prism layer is set to 1e− 5 m to ensure that
3.4. CFD analysis the dimensionless node distance y+ ≈ 1 for cells adjacent to the blade is
achieved.
The 3D horizontal axis wind turbine (HAWT) was modeled using the
finite volume solver Ansys Fluent [50] to simulate the aerodynamic

5
H.K. Nejadkhaki et al. Energy Conversion and Management 248 (2021) 114771

Fig. 7. The HAWT model for (a) the computational domain and (b) over­
set mesh.

Table 4
Discretization error using the grid convergence index
methodology.
Parameter Value

N1, N2, N3 6.2 M, 2.7 M, 1.2 M

rf21 1.32
rf32 1.32
thrust1 713
thrust2 711
thrust3 694
GCI21 0.05%
Fig. 6. The design space exploration algorithm. GCI32 0.40%

3.4.1. Verification study

Therefore, the medium mesh will be used for the remainder of the
To determine if the numerical solution is sensitive to the cell size, the
simulations to reduce the computational time for the simulations.
grid convergence index (GCI) [51], which is based on the Richardson
extrapolation, was used to estimate the numerical error. Three meshes
3.4.2. Validation study
with a refinement factor (rf) of 1.32 were examined. The grid refinement
The NREL UAE Phase VI tests quantified the full-scale two-bladed
factor is the ratio of average cell height of two different meshes, which
HAWT and provided extensive experimental data. In this study, the CFD
should be greater than 1.3 as suggested by Celik et al. [51]. The coarse
model was validated using the NREL experimental data from the Phase
(3), medium (2), and fine (1) meshes are summarized in Table 4, where
VI sequence S report with a 3◦ pitch angle and wind speed υw = 5 m/s.
N is the number of cells. Subscripts refer to the specific mesh resolution
(Note: The NREL UAE report uses a letter or number to identify each
as noted parenthetically, and pairs of subscript numbers refer to com­
sequence test; S stands for a special test setting. The details of the setting
parisons between two meshes. The thrust produced on the S809 blades
can be found on page 21 of the UAE Phase VI report [40].) The thrust
when υw = 5 m/s, Ω = 72 rpm and 3◦ pitch angle was calculated from the
and torque for the NREL experiments are shown in Table 5 and
simulations and compared for each mesh, as shown in Table 4. The GCI
compared with the CFD predictions for each mesh. The medium and fine
comparing the thrust predicted using the three consecutive meshes are
mesh predictions are in close agreement with the experiment, and the
shown in Table 4 and indicate that with confidence, the solutions are
relative error between the experiment and fine mesh is 2.6% for the
mesh-independent with errors less than 0.4%. Further evidence is shown
thrust and 0.23% for the torque. Fig. 8 includes the experimental data to
in Fig. 8 using the pressure coefficient Cpres along the chord length at
directly compare with the CFD simulations as a further demonstration of
46.6% and 80% of the blade surface. The pressure coefficient is a
the good agreement.
normalized pressure, where Cpres = P/{½ρ[υw2 + (RΩ)2]}. The differ­
ences for each mesh are very minimal and the curves closely overlap.

6
H.K. Nejadkhaki et al. Energy Conversion and Management 248 (2021) 114771

Fig. 8. Measurement sections for pressure coefficients at different locations (a) in % blade radius, pressure coefficient at (b) 46.6% and c) 80% of the blade radius for
different grid resolution.

that occurs when the blade is contorted to its maximum displacement.

Table 5
The spar is not part of this analysis. The spar is based on conventional
Thrust and torque predictions for different CFD meshes and the NREL
blade design techniques. It is assumed to be rigid and capable of carrying
experiment.
the load. It is assumed that no load is transferred between consecutive
FTH (N) TR (N⋅m) blade segments. In this study, the segments facilitate torsional compli­
NREL experiment 732 433 ance and still maintain the required airfoil shape when deformed or
CFD Coarse 694 309 exposed to aerodynamic forces.
Medium 711 416
Fine 713 432
3.5.1. Fluid-Structure interaction
The aerodynamic loads on the blade surface can potentially deform
3.5. Structural analysis the cross-sectional geometry. A large deviation in the airfoil shape will
have an adverse impact on the aerodynamic efficiency of the turbine.
The aerodynamic design impacts the efficiency of energy captured Hence, the FSI analysis is used to study the effects of aerodynamic forces
from the wind as it moves across the blade [34]. For that step, it is on the deformation of a 3D model. FSI analyzes the coupled dynamics of
assumed that the shape of the cross-section would not be significantly structures in contact with the fluid, as shown in Fig. 9. The structural
deformed as the TAD changed. This step of the design procedure ensures deformation occurs in response to pressure developed by the fluid. In
that assumption. This is accomplished through an analysis of an extreme return, the structure imparts disturbances on the adjoining fluid, and
loading event on the portion of the blade that has the greatest amount of thus alters the fluid flow. However, the effect that the structure has on
twist. The first part of this process analyzes the blade as it is subjected to fluid flow can be neglected when the amount of deformation is small.
aerodynamic pressure. Torsion is then applied to simulate the behavior For the proposed design, the amount of deformation is minuscule in

7
H.K. Nejadkhaki et al. Energy Conversion and Management 248 (2021) 114771

Fig. 9. Schematic for a one-way FSI simulation.

comparison to the blade dimension. Therefore, the one-way FSI 3.5.2. Torsional loading
assumption can be implemented [52]. CFD is used to model the pressure The torsional deformation that the segment experiences, is based on
distribution over the turbine. FEA is then applied to determine the the maximum twist gradient. This gradient is determined by the aero­
amount of deformation. dynamic analysis. The amount of torque required to achieve this posi­
The objective of executing an FSI for design validation is to estimate tion is found using FEA simulation. The properties of the ULTEM 9085
the behavior of the blade by calculating deformations and stresses when material are used for the analysis. It is conservatively based on an
acted upon by pressure distribution of airflow. This estimation shall aid extreme-case scenario. There is a rib at both ends of the blade element.
in confirming the material selection and design robustness. Based upon One of these ribs is fixed, while torque is applied through on the other
the results a decision can be made if further design alteration is neces­ rib. The cross-section at the center of the segment has the least amount
sary. At present, the objective does not demand a high-fidelity FSI of support, being located away from the ends. This section is the most
simulation. This allows for simplifications in the simulation to save susceptible to failure. The geometry at this cross-section is compared
computational time. Simplifications are done in terms of geometry and before and after the torsional load is applied. This is done to ensure there
flow characteristics as discussed in the analysis and results section of the is not a significant change in the cross-sectional shape at this location.
paper. Full-scale FSI studies can be taken up at a later stage of design XFOIL [55] is open-source software that can calculate the lift and drag
development [53,54]. coefficients for a given airfoil shape. It is used here to determine any
The output of the CFD simulation determines the aerodynamic changes to these coefficients before and after deformation. The variation
pressure acting on the blade segment. The aerodynamic pressure is in this coefficient provides a quantitative sense of the cross-sectional
applied to the structural model as shown in Fig. 10. The distribution of change. For this type of loading the amount of deformation is negli­
this pressure is a function of the airflow and the angle of attack. The FEA- gible as shown in the results in Section 4.
based solver calculates the structural deformation and stress in the blade
segment. For the FEA analysis, the blade segments are assumed to be 4. Results
fixed to the ribs (that attach to the spar). The contact surfaces are
considered to be fixed as boundary conditions. The NREL UAE Phase VI turbine was used for the case study. The

Fig. 10. Pressure distribution (in Pa) over the blade segment.

8
H.K. Nejadkhaki et al. Energy Conversion and Management 248 (2021) 114771

TAD shape associated with maximum efficiency was found for a discrete
set of points of wind speed in Region 2. The region was evaluated at 1 m/
s increments from the cut-in to rated speed. To find the gain in the ef­
ficiency, the maximum power coefficient, cp , is obtained using the
procedure described in Section 3.2. The results are compared to the
maximum efficiency that can be found using a rigid blade with con­
ventional pitch control, noted as cp,conv, and the improvement using
TAD, noted as cp,TAD, shown in Table 6. The power coefficient relative
difference Δcp between the conventional pitch control and TAD is also
provided to demonstrate improvements using TAD. It is notable that at 9
m/s there is virtually no increase in Δcp, which is less than 0.2%, sug­
gesting that TAD is optimal near this wind speed. In moving away from
this speed, the improvement in cp by TAD generally increases. It reaches
3.7 % at 5 m/s which demonstrates the usefulness of the proposed design
for low wind speed. The information in Table 6 was considered with
wind data to estimate the gain in annual energy production (AEP). In­
formation about the NREL Western Datasets used in this calculation can
be acquired from [42]. The results suggest that the contribution from
Region 2 alone would be in the range of 1.7–2.0% would increase AEP
for this particular blade design. These results do not include additional
gains in production that can be obtained due to the improved transient
response of the TAD. These events are being studies and will be reported
on separately. The optimum twist angles, ϕp and ϕTAD , were calculated
for the pitch controlled and variable TAD blade, respectively. The op­
timum TAD, ϕTAD and the difference, ϕTAD − ϕp , between the angles are
shown in Fig. 11, as functions of wind speed and distance from the blade
root. The general trend shows an increase in the optimum twist angle as
the wind speed increases. Also, in moving towards the tip, the twist
angle reduces. As it is seen in Fig. 11b, the twist angle in TAD and a
conventional pitch-controlled system was very similar near 9 m/s,
which is consistent with the results in Table 6. This again suggests that
the blade used in this study has a TAD that is optimal at this wind speed.
Towards the cut-in speed, the differences in the TAD also increase be­
tween the two methods. The difference is most noticeable at cut-in
speed. Moreover, it is most pronounced near the root, where the twist
angle is greater for pitch control. At this same speed, it is observable that
the twist angles become more similar toward the tip. At the rated speed,
the TAD is also considerably different between the two methods. How­
ever, the amount of difference is less than that occurring at cut-in. Also,
the active TAD would have a higher degree of twist than that acquired
through pitch control. This is in contrast to the active TAD at cut-in,
which has a lesser degree. However, at both the cut-in and rated
speeds, there is little variation between the twist angle moving towards Fig. 11. (a) Optimum twist angle, ϕTAD as a function of wind speed and dis­
the tip. tance from the blade root and, (b) difference between optimum twist angle in
Fig. 12 shows how the angle of twist varies with wind speed at each TAD and pitch-controlled systems,ϕTAD − ϕp
distance as measured from the root. A linear estimation (shown by the
dotted line) of the twist versus wind speed was provided for each point These are based on the Region 2 simulation results. To investigate the
in Table 7. The slope of these fitted lines is higher for the points close to cross-sectional change, it is necessary to consider the “free shape” of the
the root and reduces as towards the tip. The changes in the slope are also blade. This is analogous to the free position of a spring, in which there is
greater near the root and moderate towards the tip. For the pitch- no loading or deformation. The free shape is established as the mean
controlled system, the slope is constant for the whole blade, since the position that is shown in Fig. 13. The amount of variation from this free
entire blade is being pitched. In the TAD blade, one may conclude that position is important in determining the required actuation and defor­
the twist angle varies closest to the root. This suggests that less move­ mation. Specifically, the latter affects the structural design of the blade.
ment, and thus less actuation, is required near the outer portions of the Hence, the maximum twist angle per unit of length is extracted from this
blade. This could be beneficial inasmuch as blade loads are concerned. figure for consideration in the structural analysis that follows. The
Fig. 13 shows the minimum and maximum twist configurations.

Table 6
Gain in the maximum cp by twist modification for wind speeds 5–13 m/s.
υw [m/s]

parameter 5 6 7 8 9 10 11 12 13

cp,conv [–] 0.4465 0.4839 0.4346 0.3702 0.3144 0.268 0.2307 0.2002 0.1742
cp,TAD [–] 0.4631 0.4889 0.4392 0.3715 0.3148 0.2693 0.2325 0.2033 0.1787
Δcp [%] 3.718 1.033 1.058 0.351 0.127 0.485 0.780 1.548 2.583

9
H.K. Nejadkhaki et al. Energy Conversion and Management 248 (2021) 114771

maximum twist angle per unit of length occurs in the blade segment that
is situated between 1.8 and 2.16 m as measured from the blade root. The
amount of chan ge in the twist angle as measured between the chord
lines at each end is 1.96◦ . This scenario is used to analyze the structural
response to deformation and aerodynamic loading.
CFD workflow was used to model both the optimized blades and the
S809 blades for υw = 5 m/s with a 2◦ pitch and υw = 13 m/s with a 14◦
pitch. Using the simulated flow, torque and thrust were calculated as
shown in Table 8 and compared with BEM. The relative error is the
difference between the BEM and CFD results relative to the CFD results.
In general, the relative error was between 9 and 17% but it is important
to bear in mind that the CFD predictions are for a full three-dimensional
flow field whereas BEM uses correction factors. Table 9 shows the thrust
and torque improvements by using the optimal blades versus the S809
blades of the same blade pitch angle and wind speed. The CFD pre­
dictions compared to BEM are in the same order of magnitude and
demonstrate that BEM can be used to assess the optimization of a blade.
Furthermore, the CFD results show that using the optimal blade im­
proves the torque by 5%, which is very substantial for wind turbine
operations.
An FSI analysis was conducted on the blade segment. This was done
Fig. 12. Angle of twist variation with wind speed at each distance as measured
to ensure that the blade segment maintains its geometry under the
from the root.
applied loads. Executing a full-scale FSI over the entire span of the blade
shall require high computational effort. As a preliminary study, the
Table 7 blade segment, described in Section 2 is used for the simulation. Hence
Linear estimation of the twist versus wind speed at the geometry is simplified from a full wind turbine blade to almost a
each distance from the root. twisted box wing shape. Ansys Workbench was used for the simulation.
d (m) ϕ (◦ ) The pressure distribution from the CFD simulation is then applied as
a load onto the blade segment. The blade segment has properties of
0.800 2.857υw − 2.168
ULTEM 9085. It is observed that the deformations are of the magnitude
1.278 2.708υw − 6.438
of 0.004 mm. This amount is negligible when compared to the size of the
1.714 2.320υw − 7.214
blade segment. Also, the maximum Von-Mises stress is less than 1 MPa.
2.116 1.945υw − 6.872
The design is acceptable considering that the ULTEM 9085 has a
2.552 1.611υw − 5.758 strength of more than 30 MPa [56]. This provides a high factor of safety
2.954 1.587υw − 6.486 from the pressure loads. Considering the magnitude of this value, further
3.390 1.590υw − 7.219 improvements in the simulation settings were not necessary. The anal­
3.792 1.614υw − 8.146 ysis results have been reported in Table 10. Cut-out speed, 25 m/s, is the
4.144 1.587υw − 8.284 greatest operating wind speed possible and is used here.
4.522 1.587υw − 8.881 The segment deformation is subsequently analyzed in the extreme
twist configuration. The middle cross-section is considered in this case as
it is shown in Fig. 14. The prepared 3D model was imported into ABA­
QUS software [57] for finite element analysis. To apply the torsion, one
end of the segment is fixed. Torque is applied to the opposing end
through the rigid body constraint. This emulated the ribs in the structure
that would be supporting the flexible shell. The applied load flexed the
segment by 1.96◦ . This was found by calculating the angle between the
chord lines at the ends. A 10-node quadratic tetrahedron (C3D10) was
used for meshing. Since the computational time for this case was not so
high, the mesh size is refined to be one-third of the default mesh size
suggested by ABAQUS. Fig. 14 shows the applied torque and boundary
condition.
In Fig. 15a, the middle cross-section of the free position is compared
to that of the twisted position. The chords of the twisted and free-
position sections are aligned to make a comparison. Fig. 15b illus­
trates how closely the two sections are matched to one another. The
results are virtually the same when considering the twist that occurs in
the opposite direction. In both cases, the amount of variation was
insignificant. Moreover, the Von-Mises stress resulting from this defor­
mation was less than 1 MPa. An XFOIL analysis of the twisted cross-
section determined any change in the aerodynamic performance. The
aerodynamic performance coefficients used for this segment in the
AeroDyn, have been reported for Reynolds number 7.5e5. This is also
the value in the analysis of the middle cross-section by XFOIL. The angle
Fig. 13. The minimum, maximum and free position of the twist angle as they
change along the flexible blade, as measured from the root.
of attack was set to 7.06◦ , as it was used in the FSI analysis. For this
setting, the lift coefficient was found to be 1.2979 and 1.2966 before and
after twisting the element, respectively which is very small.

10
H.K. Nejadkhaki et al. Energy Conversion and Management 248 (2021) 114771

Table 8
Thrust, torque and relative error for different blade configurations.
Blade type Pitch angle ( ) υw (m/s) FTH (N) TR (N⋅m)

CFD BEM Relative error CFD BEM Relative error

S809 2 5 920 1027 10.42% 420 367 14.44%

Optimal twist 2 5 948 1053 9.97% 444 381 16.54%
S809 14 13 1554 1782 12.79% 2778 2514 10.50%
Optimal twist 14 13 1575 1802 12.60% 2929 2598 12.74%

the aerodynamic and structural performance of the adaptive blade. The

Table 9 technique can be used to determine the TAD that maximizes the power
Improvement of optimal twist blades.
coefficient at a given wind speed in Region 2. The technique defines the
Pitch angle ( ) υw (m/s) FTH (N) TR (N⋅m) required geometry and range of travel for the adaptive TAD. Given the
CFD BEM CFD BEM aerodynamic loads and required torsional deformation enabled a
structural analysis of the flexible shells. A case study was conducted
2 5 3.04% 2.53% 5.71% 3.81%
14 13 1.35% 1.12% 5.44% 3.34%
using the NREL Phase VI wind turbine data. This model was selected
because it is known to have reliable data. Simulation results from this
study suggest the adaptive TAD can improve the efficiency across Region
2. Using this capability improved the power coefficient. An increase of
Table 10
3.7% is realized at cut-in speed while a 2.58% gain occurs at the rated
Characteristics of FSI analysis.
speed. CFD results were obtained at the cut-in and rated speed and
Wind speed Angle of Maximum Maximum Von-Mises
showed similar gains to those obtained through the AeroDyn software.
(m/s) attack (◦ ) Deformation (mm) Stress (Pa)
The model in this paper is also being studied using tunnel experiments.
25 7.06 0.004 3.28e5 The method in this paper provides a foundation for additional studies.
Work on the AWT-27 turbine [58] shows a higher increase in efficiency
that will be reported in the future. Future work will analyze the effect
the active TAD has on load and vibration mitigation during transient
events. The subject of this study will be the DTU 10 MW model [59].
AM is recognized as an enabling technology for manufacturing the
proposed blade. The loading was also analyzed through a structural
simulation of a 3D printed blade segment made of ULTEM 9085. The
structural analysis indicated that this material could readily withstand
stresses induced by the aerodynamic and torsional loads. Moreover, the
proposed design can be deformed without negative consequences on the
required airfoil geometry. The ongoing research will focus on AM pro­
cesses, materials, and internal structures that enable the blade TAD to
create the required range of transformation.
Fig. 14. Load and boundary conditions on the analyzed model.
CRediT authorship contribution statement
5. Conclusion
Hamid Khakpour Nejadkhaki: Data curation, Formal Analysis,
A blade concept for an adaptive TAD has been presented. The blade Methodology, Writing - original draft. Azadeh Sohrabi: Formal Anal­
includes a rigid spar, segments with torsional compliance, and non- ysis, Methodology, Validation, Visualization, Software, Writing – orig­
structural skin. A modeling approach was also presented for analyzing inal draft. Tejas Prasad Purandare: Software, Writing – original draft.

Fig. 15. The free-shape cross-section (-.-) compared to the twisted section (×), (a) as it occurs, (b) with the chord lines aligned.

11
H.K. Nejadkhaki et al. Energy Conversion and Management 248 (2021) 114771

Francine Battaglia: Methodology, Resources, Supervision, Writing - [26] Nuttall D, Elliott A, Post B, Love L. Advanced Infusion Techniques with 3-D Printed
Tooling. Oak Ridge, TN (United States): Oak Ridge National Laboratory (ORNL);
review & editing. John F. Hall: Conceptualization, Data curation, Re­
2016. ORNL/TM-2016/19.
sources, Supervision, Writing – original draft, Writing - review & [27] Lee J, et al. Global Wind Report 2019. Brussels: Global Wind Energy Council
editing. (GWEC); 2020.
[28] Hamid Khakpour Nejadkhaki SC, Hall John F. A design methodology for selecting
ratios for a variable ratio gearbox used in a wind turbine with active blades.
Declaration of Competing Interest Renewable Energy 2018;118:1041–51. https://doi.org/10.1016/j.
renene.2017.10.072.
[29] Johnson KE. Adaptive torque control of variable speed wind turbines. Technical
The authors declare that they have no known competing financial Report No. NREL/TP-500-36265. Golden, CO: National Renewable Energy
interests or personal relationships that could have appeared to influence Laboratory; 2004.
[30] Rex A, Johnson K. Methods for controlling a wind turbine system with a
the work reported in this paper.
continuously variable transmission in region 2. J Sol Energy Eng 2009;131(3):
2009.
References [31] Jeong HG, Seung RH, Lee KB. An improved maximum power point tracking
method for wind power systems. Energies 2012;5(5). https://doi.org/10.3390/
en5051339.
[1] Wind Vision Detailed Roadmap Actions: 2017 Update; EERE Publication and
[32] de Goeij WC, van Tooren MJL, Beukers A. Implementation of bending-torsion
Product Library, DOE/GO-102018-5056; Other: 7985 United States 10.2172/
coupling in the design of a wind-turbine rotor-blade. Appl Energy 1999,;63(3):
1497756 Other: 7985 EE-LIBRARY English, 2019. [Online]. Available: <htt
191–207. https://doi.org/10.1016/S0306-2619(99)00016-1.
ps://www.osti.gov/servlets/purl/1497756>.
[33] Forcier L-C, Joncas S. New structural design concepts for large thermoplastic wind
[2] N.A., World Energy Outlook 2020, International Energy Agency (IEA), October
turbine blades using structural optimization techniques. The 51st AIAA/ASME/
2020 2020. [Online]. Available: <https://www.iea.org/reports/world-energy-out
ASCE/AHS/ASC Structures, Structural Dynamics, and Materials Conference,
look-2020>.
Orlando, FL, United states. 2010.
[3] Wang Z, Suiker AS, Hofmeyer H, van Hooff T, Blocken B. Coupled aerostructural
[34] Mølholt Jensen F, Branner K. 1 - Introduction to wind turbine blade design. In:
shape and topology optimization of horizontal-axis wind turbine rotor blades.
Brøndsted P, Nijssen RPL, editors. Advances in Wind Turbine Blade Design and
Energy Convers Manage 2020;212:112621.
Materials. Woodhead Publishing; 2013. p. 3–28.
[4] Saleem A, Kim M-H. Effect of rotor tip clearance on the aerodynamic performance
[35] Manwell JF, McGowan JG, Rogers AL. Aerodynamics of wind turbines. John Wiley
of an aerofoil-based ducted wind turbine. Energy Convers Manage 2019;201:
& Sons Ltd 2009:91–155.
112186.
[36] Liu X, Wang L, Tang X. Optimized linearization of chord and twist angle profiles for
[5] Shen X, Yang H, Chen J, Zhu X, Du Z. Aerodynamic shape optimization of non-
fixed-pitch fixed-speed wind turbine blades. Renewable Energy 2013;57:111–9.
straight small wind turbine blades. Energy Convers Manage 2016;119:266–78.
https://doi.org/10.1016/j.renene.2013.01.036.
https://doi.org/10.1016/j.enconman.2016.04.008.
[37] Hassanzadeh A, Hassanzadeh Hassanabad A, Dadvand A. Aerodynamic shape
[6] Hamid Khakpour Nejadkhaki JFH. Modeling and design method for an adaptive
optimization and analysis of small wind turbine blades employing the Viterna
wind turbine blade with out-of-plane twist. J Sol Energy Eng 2018;140(5):9.
approach for post-stall region. Alexandria Eng J 2016,;55(3):2035–43. https://doi.
https://doi.org/10.1115/1.4040104.
org/10.1016/j.aej.2016.07.008.
[7] Philibert C, Holttinen H. Technology Roadmap: Wind Energy. International Energy
[38] Khakpour Nejadkhaki H, Hall JF. Control framework and integrative design
Agency France; 2013.
method for an adaptive wind turbine blade. J Dyn Syst, Measur, Control 2020;142
[8] Johnson KE, Fingersh LJ. Adaptive pitch control of variable-speed wind turbines.
(10). https://doi.org/10.1115/1.4047098.
J Sol Energy Eng 2008;130(3). https://doi.org/10.1115/1.2931505.
[39] Mou F, Nejadkhaki HK, Estes A, Hall JF. Weighted least squares approach for an
[9] Rahimi M. Improvement of energy conversion efficiency and damping of wind
adaptive aerodynamic engineered structure with twist transformation. J Energy
turbine response in grid connected DFIG based wind turbines. Int J Electr Power
Resour Technol 2019;141(5). https://doi.org/10.1115/1.4042642.
Energy Syst 2018;95:11–25.
[40] Hand MM, et al. Unsteady Aerodynamics Experiment Phase VI: Wind Tunnel Test
[10] Ponta FL, Otero AD, Rajan A, Lago LI. The adaptive-blade concept in wind-power
Configurations and Available Data Campaigns. technical report. National
applications. Energy Sustainable Dev 2014;22:3–12. https://doi.org/10.1016/j.
Renewable Energy Laboratory; 2001.
esd.2014.04.004.
[41] Liu X, Lu C, Liang S, Godbole A, Chen Y. Vibration-induced aerodynamic loads on
[11] Xie W, Zeng P, Lei L. A novel folding blade of wind turbine rotor for effective power
large horizontal axis wind turbine blades. Appl Energy 2017;185(Part 2):1109–19.
control. Energy Convers Manage 2015;101(Supplement C):52–65. https://doi.org/
https://doi.org/10.1016/j.apenergy.2015.11.080.
10.1016/j.enconman.2015.05.037.
[42] Nejadkhaki HK, Lall A, Hall JF. A methodology to synthesize gearbox and control
[12] Eric Loth MS, Moriarty Patrick. Morphing segmented wind turbine concept. The 28
design for increased power production and blade root stress mitigation in a small
th AIAA Applied Aerodynamics Conference, Chicago, Illinois. 2010.
wind turbine. J Mech Des 2017;139(8). https://doi.org/10.1115/1.4036998.
[13] Gili P, Frulla G. A variable twist blade concept for more effective wind generation:
[43] El-Gizawy AS, Corl S, Graybill B. Process-induced properties of fdm products.
design and realization. Smart Sci. 2016;4(2):78–86. https://doi.org/10.1080/
Proceedings of The ICMET. 2011.
23080477.2016.1191002.
[44] Bagsik A, Schöppner V, Klemp E. FDM part quality manufactured with Ultem 9085.
[14] Sirigu G, Cassaro M, Battipede M, Gili P, Frulla G. Wind generator innovative blade
In: 14th International Scientific Conference on Polymeric Materials; 2010.
design: variable twist and start-up control. Int. J. Mech. 2016;10:53–61.
p. 307–15.
[15] Wang W, Caro S, Bennis F, Salinas Mejia OR. A simplified morphing blade for
[45] Burton T. Wind Energy: Handbook. Chichester; New York: J. Wiley; 2001.
horizontal axis wind turbines. J Sol Energy Eng 2013;136(1). https://doi.org/
[46] Hayman GJ, Jonkman JM, Jonkman BJ, Damiani RR. AeroDyn v15 User’s Guide
10.1115/1.4025970.
and Theory Manual [Online]. National Renewable Energy Laboratory (NREL);
[16] Fuglsang P, Thomsen K. Site-specific design optimization of 1.5–2.0 MW wind
2016.
turbines. J Sol Energy Eng 2001;123(4):296–303. https://doi.org/10.1115/
[47] Moriarty PJ, Hansen AC. AeroDyn Theory Manual. Golden, CO (US): National
1.1404433.
Renewable Energy Lab.; 2005.
[17] Van Dam C, Berg DE, Johnson SJ. Active Load Control Techniques for Wind
[48] Hansen MOL. In: Aerodynamics of Wind Turbines. UK and USA: Earthscan; 2008.
Turbines. Sandia National Laboratories; 2008.
p. 192.
[18] Lobitz DW, Veers PS, Eisler GR, Laino DJ, Migliore PG, Bir G. The Use of Twist-
[49] Eiselt HA, Sandblom C-L. Integer Programming and Network Models. Springer
coupled Blades to Enhance the Performance of Horizontal Axis Wind Turbines,
Science & Business Media; 2013.
SAND2001-1003. Albuquerque, NM: Sandia National Laboratories; 2001.
[50] N.A.. ANSYS Fluent Theory Guide, Release 19. Canonsburg, PA: ANSYS, Inc.; 2019.
[19] Runge J-B, Osmont D, Ohayon R. Twist control of aerodynamic profiles by a
[51] Celik IB, Ghia U, Roache PJ, Freitas CJ. Procedure for estimation and reporting of
reactive method (experimental results). J Intell Mater Syst Struct 2012;24(8):
uncertainty due to discretization in CFD applications. J Fluids Eng-Trans ASME
908–23. https://doi.org/10.1177/1045389X12437884.
2008;130(7).
[20] Jae-Sang P, Seong-Hwan K, Sung Nam J, Myeong-Kyu L. Design and analysis of
[52] Benra F-K, Josef Dohmen H, Pei J, Schuster S, Wan B. A comparison of one-way and
variable-twist tiltrotor blades using shape memory alloy hybrid composites
two-way coupling methods for numerical analysis of fluid-structure interactions.
[Online] Smart Mater Struct 2011;20(1).
J Appl Math 2011. https://doi.org/10.1155/2011/853560.
[21] Liu S, Li Y, Li N. A novel free-hanging 3D printing method for continuous carbon
[53] Lee K, Huque Z, Kommalapati R, Han S-E. Fluid-structure interaction analysis of
fiber reinforced thermoplastic lattice truss core structures. Mater Des 2018,;137
NREL phase VI wind turbine: aerodynamic force evaluation and structural analysis
(Supplement C):235–44. https://doi.org/10.1016/j.matdes.2017.10.007.
using FSI analysis. Renewable Energy 2017,;113:512–31. https://doi.org/
[22] Bagherpour T, Li X, Manolas D, Riziotis V. Modeling of material bend-twist
10.1016/j.renene.2017.02.071.
coupling on wind turbine blades. Compos Struct 2018;193:237–46.
[54] Rezaeiha A, Montazeri H, Blocken B. Towards accurate CFD simulations of vertical
[23] Yang S, Zhao Y. Additive manufacturing-enabled design theory and methodology: a
axis wind turbines at different tip speed ratios and solidities: Guidelines for
critical review. Int J Adv Manuf Technol 2015;80.
azimuthal increment, domain size and convergence. Energy Convers Manage
[24] Fasel U, Keidel D, Baumann L, Cavolina G, Eichenhofer M, Ermanni P. Composite
2018,;156:301–16. https://doi.org/10.1016/j.enconman.2017.11.026.
additive manufacturing of morphing aerospace structures. Manuf Lett 2020;23:
[55] Drela M, Youngren H. XFOIL 6.94 User Guide, 2001. Cambridge, MA, USA:
85–8.
Massachusetts Institute of Technology; 2019.
[25] Li N, Li Y, Liu S. Rapid prototyping of continuous carbon fiber reinforced polylactic
[56] V. Schöppner and K. P. KTP, Mechanical properties of fused deposition modeling
acid composites by 3D printing. J Mater Process Technol 2016,;238(Supplement
parts manufactured with Ultem* 9085, in: Proceedings of 69th Annual Technical
C):218–25. https://doi.org/10.1016/j.jmatprotec.2016.07.025.

12
H.K. Nejadkhaki et al. Energy Conversion and Management 248 (2021) 114771

Conference of the Society of Plastics Engineers (ANTEC’11), 2011, vol. 2, pp. 1294- [58] Jonkman JM, Buhl Jr ML. FAST User’s Guide-Updated August 2005. Golden, CO:
1298. National Renewable Energy Laboratory (NREL); 2005.
[57] N.A.. Abaqus Analysis User’s Guide (Solid Continuum Elements). Providence, RI, [59] Bak C, et al. The DTU 10-MW reference wind turbine. Danish Wind Power Research
USA: Dassault Systèmes Simulia Corp.; 2014. 2013. 2013.

13