You are on page 1of 20

Renewable Energy 113 (2017) 512e531

Contents lists available at ScienceDirect

Renewable Energy
journal homepage: www.elsevier.com/locate/renene

Fluid-structure interaction analysis of NREL phase VI wind turbine:


Aerodynamic force evaluation and structural analysis using FSI
analysis
Kyoungsoo Lee a, *, Ziaul Huque a, b, Raghava Kommalapati a, c, Sang-Eul Han d
a
Center for Energy and Environmental Sustainability (CEES), Prairie View A&M University, Prairie View, TX, 77446, USA
b
Department of Mechanical Engineering, Prairie View A&M University, Prairie View, TX, 77446, USA
c
Department of Civil & Environmental Engineering, Prairie View A&M University, Prairie View, TX, 77446, USA
d
School of Architecture, Department of Architectural Engineering, Inha University, Inchoen, 402-751, South Korea

a r t i c l e i n f o a b s t r a c t

Article history: The fluid-structure interaction (FSI) analyses were performed to obtain the structural responses of 3D
Received 13 May 2015 wind turbine research on the National Renewable Energy Laboratory (NREL) Phase VI wind turbine using
Received in revised form the commercial program ANSYS. The surface pressure information was imported from the results of
29 December 2016
computation fluid dynamics (CFD) which were studied in advance. To perform the structural analyses for
Accepted 22 February 2017
Available online 12 July 2017
both of FSI and blade element momentum (BEM) method, the structural model of full NREL Phase VI
wind turbine were developed which is assembled by a rotor, nacelle, tower and blades. The aerodynamic
forces also can be calculated from structural model by the value of reaction on rotor shaft. From those
Keywords:
Fluidestructure interaction
new aerodynamic evaluating process, the accuracy of FSI process which utilizing CFD results were
Computational fluid dynamics described. And the advantages and characteristics of combination of CFD/FSI which were demonstrated
National Renewable Energy Laboratory by investigating the structural response of NREL Phase VI turbine for 7 wind speed cases. From the
Phase VI wind turbine blade studies of structural analysis, the wind turbine can be said that it is not governed by torque force, which
Aerodynamic force is the main interest in power efficiency of wind turbine, but the thrust force. Finally, the practical
Structural analysis applicability of this methodology is discussed.
Wind turbine © 2017 Published by Elsevier Ltd.

1. Introduction (BEM) method to evaluate these aerodynamic characteristics using


wind tunnel test data. The validity and applicability of the BEM
In the case of a 3D wind turbine blade, airfoil sections are used to method have already been verified in earlier studies [4e8]. How-
maximize its lift force [1e3] to increase the power output. The ever this is related to the solution of a few applications peculiar to
unstable phenomena such as separation, transition, and reattach- these mathematical models, and the limitations of these numerical
ment occur on the blade surface. As a result, it encounters stall codes are well known in scientific literature [5].
phenomena, where the lift and drag force changes suddenly with Wind tunnel test directly measures the surface pressure at
an increase in the wind speed. The dominant aerodynamic char- placed pressure tabs, and can be said as the best choice, if it is
acteristics of the blade surface pressure of a wind turbine have available. However, it needs high quality experimental facility and
important affect on the structural design process. Various aero- other supports to perform, and getting accurate aerodynamic
dynamic forces on the sections must be defined in order to design a characteristics of full scale wind turbine blade like NREL Phase VI, is
wind turbine in local and global coordinate systems. fundamentally dependent to the topology of surface pressure tabs.
The design codes generally use the blade element momentum There are little cases of public assessable full scale wind tunnel test
data. Therefore, studies are currently being conducted to develop
alternatives to make improvements because of the commonly
known difficulties of wind tunnel test.
* Corresponding author.
One alternative is the combination of a fluidestructure inter-
E-mail addresses: kylee@pvamu.edu (K. Lee), zihuque@pvamu.edu (Z. Huque),
rrkommalapati@pvamu.edu (R. Kommalapati), hsang@inha.ac.kr (S.-E. Han).
action (FSI) [9,10] and a computational fluid dynamics (CFD)

http://dx.doi.org/10.1016/j.renene.2017.02.071
0960-1481/© 2017 Published by Elsevier Ltd.
K. Lee et al. / Renewable Energy 113 (2017) 512e531 513

[11e19] in evaluating of structural and aerodynamic behavior during a 2-way FSI simulation, especially in the interface and fluid
characteristics respectively. The CFD is usually used to obtain domain, rather than the structural. In addition, the averaged pres-
aerodynamic coefficient information [11e19] accurately. However, sure and velocity distributions in the fluid domain from an un-
the commonly known representative difficulties of CFD are that the steady CFD simulation that adopted in the 2-way FSI eventually
analysis results are highly dependent on the grid and turbulence might be damped to converged steady state, if the wind turbine’s
models. In addition, huge amount of computing efforts are neces- dynamic vibrations are not the main concern.
sary for this method. And it is difficult to verify the accuracy of the If the 1-way FSI is used, the aerodynamic and flow character-
analysis results. Nowadays, many studies are being conducted to istics can easily be obtained and verified compared with the 2-way
enhance the turbulence model CFD accuracy [20e29], and the FSI, and the researcher can have more confidence in the CFD results.
validity of using CFD for a wind turbine has been reported [11e19]. A steady state RANS simulation can provide accurate results with
If the CFD accuracy is verified, it has advantages in comparison with the proper turbulence and grid models, and there are few potential
wind tunnel test and BEM. The 2D aerodynamic coefficients, which numerical difficulties with the interface when transferring the
were obtained from wind tunnel test for BEM process, can be ob- surface pressure data.
tained from CFD analysis easily. The difficulties of computing time To obtain valid CFD analysis results, which will be used for a FSI
and efforts of CFD can be overcome by accurate analysis results and analysis, various parametric CFD studies were performed previ-
its applicability for another design process of FSI. Even though the ously. The flow results when using the SST-Gamma theta model
many advantages of CFD over wind tunnel test and BEM, this [20e27] were studied from various parameters for the domain size
method can not perform the structural analysis for structural and grid resolution models with an automatic near-wall treatment
design fundamentally. Obtained accurate blade surface pressure method. The resulting determined domain and grid model com-
information by consider the millions of grid in CFD analysis must be bination was used in the Reynolds-averaged NaviereStokes (RANS)
simply converted for conventional code based BEM, if we don’t turbulence model studies, and the accuracy of the SST-Gamma
have any other options. Fundamentally, FSI can replace the BEM in theta turbulence-transition model was studied.
wind turbine structure design by importing the CFD pressure In this paper, to perform the FSI analysis, the structural prop-
result. It can be adopted in design a wind turbine to improve the erties of the NREL Phase VI wind turbine are described, and the
accuracy of the aerodynamic characteristics by utilizing the valid aerodynamic forces analyzed using the 1-way FSI process is verified
fluid results that were obtained and verified in a CFD analysis. by comparing them with the results of the CFD and BEM methods.
Naturally, the FSI accuracy is highly dependent on the CFD results, The accuracy and applicability of FSI analysis will be discussed by
and it can be understood as an extension of CFD that import the investigating the structural response behaviors. The linear static
surface pressure information from the fluid domain for structural analyses for the NREL Phase VI wind turbine was performed using
loading. the commercial finite element program ANSYS. Through this ac-
In the FSI analysis process, the surface pressure information is curacy verification process, the practical applicability of proposed
directly transferred to the structural surface pressure loading with methodology will be discussed like structural response in pitch
the help of topological interfaces, and then a linear or nonlinear angle control or fatigue analysis.
structural finite element (FE) analysis is performed using the
developed FE module. Therefore, the main idea behind it is to use 2. Evaluation of aerodynamic characteristics using CFD
the numerical results that were obtained from a CFD analysis for analysis
the interface for the FE structural module. The resulting aero-
dynamic forces and structural responses can be calculated from In general, to perform the structural design of a wind turbine
the structural analysis. In addition, FSI is an analytical technique system using the BEM analysis process, the torque and thrust
or process that combines the fluid and structural domains inter- forces need to be calculated in global coordinate system from
actively. Although the basic characteristics of an FSI analysis are its local aerodynamic force quantities (drag, lift, tangent, or normal
simplicity and ease of use as a structural analysis method, the force), which are obtained from 2D [1] or 3D [2,3] wind tunnel
more important concept of FSI is that it can easily evaluate the test. These global aerodynamic loading values are applied to the
flow and aerodynamic characteristics of the fluid domain for a nodal loading condition for the structural analysis. However,
wind turbine. there are few cases in which the CFD results are used for the BEM
As mentioned by Mentor [20], with the increases in hardware calculation, in place of wind tunnel test results, because the CFD
and computing efficiency, CFD engineers often attempt to solve analysis results are highly affected by the grid and turbulent
problems in new areas with 3D complex geometries and finer grids, model, and the computational efforts required for verified results
which have previously been impossible. The time consuming and are enormous. The reliability and accuracy of CFD analysis is still
low efficiency in computations for DNS or unsteady transient uncertain for the researcher and engineer who are familiar with
simulations (LES, DES, URANS) make it difficult to use these for the BEM process. Compared with CFD, the BEM method, which
complicated problems. Instead, they might prefer to use efficient allows a simple process and gives values in the evaluation of
steady state RANS turbulence models, through which valid results aerodynamic characteristics, has advantages in efficiency and
can easily be obtained. Likewise, new studies are being conducted accuracy as a design method for wind turbines within an
to integrate and combine the multi-physics of the fluid and struc- acceptable range.
tural domains through a FSI analysis within a reasonable Nevertheless, studies to develop advanced computational
computing cost. As previously explained, this tendency can be said methods are continuing in an effort to overcome the fundamental
to be the continuation of an effort to solve new problems that could difficulties and limitations of the BEM method, which is normally
not be solved previously. dependent on 2D wind tunnel test results, it is necessary to artifi-
The easier 1-way FSI has more advantages or may be the only cially modify the aerodynamic force coefficients in order to apply
available choice compared with the more complex unsteady 2-way the 2D wind tunnel test results for a 2D airfoil to the 3D NREL Phase
FSI analysis, which use the steady RANS and unsteady turbulence VI wind turbine [4,5]. In addition, 3D wind tunnel test results are
models, respectively. If the 2-way FSI analysis is used, the possi- essential to eventually verify the modified aerodynamic force co-
bilities of obtaining convergence and accurate results are not efficients. Moreover, to enhance the power generating efficiency
guaranteed. There can be many potential numerical singularities and economic structural design of wind turbines, an accurate
514 K. Lee et al. / Renewable Energy 113 (2017) 512e531

aerodynamic force evaluation must be performed. sections used in the 5-sections calculation. As a result, when the
On the assumption that its accuracy can be secured, CFD has integration along the span is performed, the 5-sections calculation
many advantages in comparison with the BEM method, even yields higher values.
though it requires a huge computational effort. Thus, much effort The aerodynamic force coefficients are frequently used to verify
has been made to enhance and verify the accuracy of CFD for a the accuracy of the results in wind tunnel test and CFD analyses. In
wind turbine [11e19]. However, these studies mainly focused on some cases, however, these normalized aerodynamic force co-
verifying the accuracy of the applied models by comparing their efficients may distort the real values of the aerodynamic force
results with the existing results, like those from wind tunnel test. characteristics. Sometimes or many of time, CFD analysis are
There are not yet a sufficient number of reported results for the focused on calculating and evaluating the aerodynamic character-
application of CFD analysis to the design of wind turbines. In order istics of airfoil and blade especially various aerodynamic co-
to enhance the reliability and applicability of CFD in the design of efficients. But those are normalized quantities which are valid for
wind turbines, the accuracy and comparative advantages of CFD 2D case basically. In 3D case, those coefficients are not proportional
need to be evaluated in relation to the aerodynamic characteris- or little related to actual values along the span wise direction.
tics. Therefore, in this study, the aerodynamic forces were evalu- Usually the most of aerodynamic coefficients near the hub
ated using the CFD analysis in the BEM process for determining (inboard) are higher than tip area (outboard). However the actual
the structural response. The accuracy, applicability, and reliability aerodynamic forces do not match with those coefficients. More-
were investigated by comparing them with those of NREL’s wind over, the effects of the number of sections division could not be
tunnel test. seen, if normalized coefficient is used.
In NREL’s wind tunnel test, pressure tabs were placed at five- The aerodynamic force coefficients at the 30.0% section in the
sections (30.0%, 46.7%, 63.3%, 80.0%, 95.0%), which could be an deep-stall wind-speed region (20 m/s, 25 m/s) had relatively smaller
insufficient number of sections on the span. In order to enhance the values than at lower wind speeds, as shown in Fig. 3. However,
validity of the evaluated aerodynamic characteristics for the blade, considering the actual aerodynamic force values in Fig. 4, the deep-
it is necessary to divide the blade into a larger number of elements. stall region’s aerodynamic forces are significantly higher than the
It is generally known that 10 or more elements need to be used to others. Similarly, the torque force, which is the main interest when
ensure the empirical accuracy of the BEM method. In this study, 5- designing a wind turbine to increase the power generating effi-
and 16-section division cases were considered to calculate the ciency, has a much lower force value than the normal force, as
aerodynamic characteristics, which were obtained from CFD explained in Fig. 4. So, it can be said that the real aerodynamic forces
analysis. and resulting structural responses in the global coordinate system
Figs. 1 and 2 show the evaluated aerodynamic characteristics for are dependent not on the torque but on the thrust force. The
the local and global coordinate systems when the SST-Gamma theta sectional thrust and torque coefficients (Fig. 3 (c), (d)) in global
model was used. As shown in Fig. 1, the effects of separation and coordinate system are naturally non-unit. Those do not include the
stall are not clear at 10 m/s, which is the stall-onset wind speed. As division length (m) because of the sectional quantity. After consid-
described in Figs. 5 and 2, it was difficult to simulate the exact ering (multiply) the divisional length (m), those resulted in Thrust
aerodynamic characteristics because of the stall delay where the and Torque force (N). The rotor shaft Thrust and Torque (N$m) can
20 angle of attack (AOA) occurs relatively close to the hub for the be calculated by integration or summation of those sectional values.
10 m/s wind speed case. As shown in Fig. 1, compared to the normal Another method is to use the Normal and Tangent force coefficients
force, the tangent force coefficients changed rapidly on the blade to calculate the Thrust and Torque coefficients.
span under the influence of separation and stall at 13 and 15 m/s, The actual sectional aerodynamic force (Thrust and Torque
but became smooth in the deep-stall region. There are few changes force) from smaller number of section (5sections) must be higher
in the normal force coefficient along the span even in the dynamic- than more number of sections (16sections). The torque is not force
stall region, and it was directly connected with the thrust force (N) but vector moment (N$m). And it is difficult to be used as the
results in Fig. 1(b). nodal loads. One of the practical purposes of evaluating aero-
Fig. 2 represents the global aerodynamic force characteristics of dynamic force coefficient can be used for the blade structural
the torque and thrust force in the Y and X directions, respectively. design. In this aspect, we have to evaluate the nodal loading values
Figs. 3 and 4 show the evaluated aerodynamic coefficient and force at each section. More number of sections can produce smooth
in the local and global coordinate systems, respectively, for the 16- distribution of nodal loading than smaller.
sections only. However, such a difference between the section division
In the case of the aerodynamic force coefficients for the 16- methods can not be seen in aerodynamic coefficient, because the
sections in Figs. 1 and 2(a) in the local and global coordinates, aerodynamic coefficients should be identical for any numbers of
respectively, the CFD results of this study agreed well with NREL, section divisions. And the graphs of coefficient, sectional force per
and the aerodynamic force results obtained from 5-sections agreed unit length (N/m) and Aerodynamic force (N) are quite different
well with NREL’s 5-section results, as shown in Fig. 2(b) and (c). when compared with each other, even though these terms are
However, the results from 16-sections showed underestimated derived from sectional pressure coefficient. In other words,
curves compared to those for 5-sections and could not be compared frequently using pressure coefficient and resulting aerodynamic
with the 5-section results because of the different number of force coefficients are convenient but sometime make a distorted
sections. view about real aerodynamic force value in global coordinate sys-
As shown in Fig. 2, the 5-sections torque force (N) values ob- tem. The evaluating processes of the torque (N*m) and Thrust (N)
tained in this study are in good agreement with NREL experimental for the 16 and 5 sections are derived from BEM equations. So the
results for low wind speeds, although the values obtained from 16- numbers of section on blade span affect important contribution to
sections are less than 5-sections, but they are smoother and more the accuracy of analysis. In this study, the effects of 16-sections in
accurate. Because using 16-sections provides more data from evaluating aerodynamic characteristics were demonstrated
consecutive sections, the 16-sections results can be considered compared with 5-sections.
more accurate. In fact, the 16-sections calculations must yield lower The pressure characteristics on the blade span were well re-
values than the 5-sections calculations because the section length ported by NREL for 5-sections only. However, there is limited
(sections) used in the 16-sections calculation are shorter than the pressure information for further evaluation. This is one of the
K. Lee et al. / Renewable Energy 113 (2017) 512e531 515

Tangent force c oefficient (7m/s) Nor mal force coefficient (7m/s)


0.40 3.0
Tangent force coefficient : present study, CFD (7m/s) Normal force coefficient : present study, CFD (7m/s)
Tangent force coefficient : NREL(3D experiment, 7m/s) Normal force coefficient : NREL(3D experiment, 7m/s)
2.5

0.30
2.0

0.20 1.5

1.0

0.10
0.5

Span ratio (r/R) Span ratio (r/R)


0.00 0.0
0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0

Tangent force coefficient (10m/s) Normal force coefficient (10m/s)


0.40
3.0
Normal force coefficient : present study, CFD (10m/s)
Tangent force coefficient : present study, CFD (10m/s)
Normal force coefficient : NREL(3D experiment, 10m/s)
Normal force coefficient : NREL(3D experiment, 10m/s) 2.5
0.30

2.0

0.20 1.5

1.0
0.10

NREL (3D experiment) 0.5

Span ratio (r/R) Span ratio (r/R)


0.00 0.0
0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0

Tangent force coefficient (13m/s) Normal force coefficient (13m/s)


3.0
0.40
Tangent force coefficient : present study, CFD (13m/s) Normal force coefficient : present study, CFD (13m/s)
Tangent force coefficient : NREL(3D experiment, 13m/s) Normal force coefficient : NREL(3D experiment, 13m/s)
2.5
0.30

2.0

0.20
1.5

0.10 1.0

0.5
0.00
0.0 0.2 0.4 0.6 0.8 1.0 Span ratio (r/R)
0.0
Span ratio (r/R)
-0.10 0.0 0.2 0.4 0.6 0.8 1.0

Tangent force coefficient (15m/s) Normal force coefficient (15m/s)


0.40 3.0
Tangent force coefficient : present study, CFD (15m/s)
Tangent force coefficient : NREL(3D experiment, 15m/s)
2.5
0.30

2.0

0.20
1.5

0.10 1.0

0.5 Normal force coefficient : present study, CFD (15m/s)


0.00 Normal force coefficient : NREL(3D experiment, 15m/s)
0.0 0.2 0.4 0.6 0.8 1.0
Span ratio (r/R)
0.0
Span ratio (r/R) 0.0 0.2 0.4 0.6 0.8 1.0
-0.10
Tangent force coefficient (25m/s) Normal force coefficient (25m/s)
0.40 3.0
Tangent force coefficient : present study, CFD (25m/s)
Tangent force coefficient : NREL(3D experiment, 25m/s)
2.5
0.30
2.0

0.20
1.5

0.10 1.0

Normal force coefficient : present study, CFD (25m/s)


0.5
0.00 Normal force coefficient : NREL(3D experiment, 25m/s)
0.0 0.2 0.4 0.6 0.8 1.0 Span ratio (r/R)
0.0
Span ratio (r/R) 0.0 0.2 0.4 0.6 0.8 1.0
-0.10

Fig. 1. Sectional tangent and normal force coefficient for various wind speed cases.
516 K. Lee et al. / Renewable Energy 113 (2017) 512e531

Torque force (N, 10m/s) Thrust force (N, 10m/s)


400.0
100.0

Thrust force : present study, CFD (10m/s) : 16sections


Torque force : present study, CFD (10m/s) : 16sections
Torque force : present study, CFD (10m/s) : 5sections Thrust force : present study, CFD (10m/s) : 5sections
80.0 Torque force : NREL(3D experiment, 10m/s) : 5sections Thrust force : NREL(3D experiment, 10m/s) : 5sections
300.0

5sections
60.0

200.0

5sections
40.0 NREL (3D experi.)

16sections
100.0
20.0 16sections

Span (m) Span (m)


0.0 0.0
0.0 1.0 2.0 3.0 4.0 5.0 0.0 1.0 2.0 3.0 4.0 5.0

Torque force (N, 13m/s) Thrust force (N, 13m/s)


150.0 400.0

T orque force : present study, CFD (13m/s) : 16sections T hrust force : present study, CFD (13m/s) : 16sections
T orque force : present study, CFD (13m/s) : 5sections T hrust force : present study, CFD (13m/s) : 5sections
T orque force : NREL(3D experiment, 13m/s) : 5sections T hrust force : NREL(3D experiment, 13m/s) : 5sections
300.0

100.0 5sections

200.0
5sections

50.0 16sections 16sections


100.0

Span (m) Span (m)


0.0 0.0
0.0 1.0 2.0 3.0 4.0 5.0 0.0 1.0 2.0 3.0 4.0 5.0

Torque force (N, 15m/s) Thrust force (N, 15m/s)


150.0 Torque force : present study, CFD (15m/s) : 16sections 400.0
Torque force : present study, CFD (15m/s) : 5sections Thrust force : present study, CFD (15m/s) : 16sections
Torque force : NREL(3D experiment, 15m/s) : 5sections Thrust force : present study, CFD (15m/s) : 5sections
Thrust force : NREL(3D experiment, 15m/s) : 5sections

300.0

100.0 5sections

200.0

16sections 5sections
50.0
16sections
100.0

Span (m) Span (m)


0.0 0.0
0.0 1.0 2.0 3.0 4.0 5.0 0.0 1.0 2.0 3.0 4.0 5.0
Torque force (N, 25m/s) Thrust force (N, 25m/s)
250.0 800.0
Torque force : present study, CFD (25m/s) : 16sections
Torque force : present study, CFD (25m/s) : 5sections T hrust force : present study, CFD (25m/s) : 16sections
Torque force : NREL(3D experiment, 25m/s) : 5sections T hrust force : present study, CFD (25m/s) : 5sections
200.0 T hrust force : NREL(3D experiment, 25m/s) : 5sections
600.0

150.0

400.0
16sections
100.0 5sections
16sections 5sections
200.0
50.0

Span (m)
0.0
Span (m)
0.0
0.0 1.0 2.0 3.0 4.0 5.0 0.0 1.0 2.0 3.0 4.0 5.0

Fig. 2. Aerodynamic force for various wind speed cases and section division cases.

important limitations on wind tunnel and BEM studies. The aero- evaluation of a larger number of sections should be considered,
dynamic force coefficients in the local and global coordinate sys- leading to smoother sectional aerodynamic force quantities. In the
tems were calculated for 5-and 16-sections using the CFD results. If following sections, the aerodynamic force characteristics for 5- and
valid and verified CFD results are available and can be used, the 16-sections are applied to the nodal loading values on the blade in a
K. Lee et al. / Renewable Energy 113 (2017) 512e531 517

Tangent force coefficient Normal force coefficient


0.50 3.0
Tangent force coefficient : present study, CFD (25m/s) : y+<1.0, SST(Gamma theta) Normal force coefficient : present study, CFD (25m/s) : y+<1.0, SST(Gamma theta)
Tangent force coefficient : present study, CFD (20m/s) : y+<1.0, SST(Gamma theta) Normal force coefficient : present study, CFD (20m/s) : y+<1.0, SST(Gamma theta)
Tangent force coefficient : present study, CFD (15m/s) : y+<1.0, SST(Gamma theta) Normal force coefficient : present study, CFD (15m/s) : y+<1.0, SST(Gamma theta)
Tangent force coefficient : present study, CFD (13m/s) : y+<1.0, SST(Gamma theta) Normal force coefficient : present study, CFD (13m/s) : y+<1.0, SST(Gamma theta)
0.40 Tangent force coefficient : present study, CFD (10m/s) : y+<1.0, SST(Gamma theta) Normal force coefficient : present study, CFD (10m/s) : y+<1.0, SST(Gamma theta)
Tangent force coefficient : present study, CFD (7m/s) : y+<1.0, SST(Gamma theta) Normal force coefficient : present study, CFD (7m/s) : y+<1.0, SST(Gamma theta)
Tangent force coefficient : present study, CFD (5m/s) : y+<1.0, SST(Gamma theta) Normal force coefficient : present study, CFD (5m/s) : y+<1.0, SST(Gamma theta)
Normal force coefficient : NREL(3D experiment, 10m/s) Normal force coefficient : NREL(3D experiment, 10m/s)
Tangent force coefficient : NREL(3D experiment, 7m/s) Normal force coefficient : NREL(3D experiment, 7m/s)
0.30 Tangent force coefficient : NREL(3D experiment, 5m/s) 2.0 Normal force coefficient : NREL(3D experiment, 5m/s)

10m/s 13m/s
0.20 10m/s

7m/s 1.0 7m/s


0.10

13m/s
5m/s
0.00
0. 2 0.4 0.6 0.8 1. 0

Span ratio (r/R) Span ratio (r/R)


0.0
-0.10 0.2 0.4 0.6 0.8 1.0

(a) Tangent force coefficient (b) Normal force coefficient

Thrust coefficient
Torque coefficient 3.0 Torque coefficient : present study, CFD (25m/s) : 16sections
1.5 Thrust coefficient : present study, CFD (25m/s) : 16sections
Thrust coefficient : present study, CFD (25m/s) : 16sections
Torque coefficient : present study, CFD (25m/s) : 16sections Thrust coefficient : present study, CFD (20m/s) : 16sections
Torque coefficient : present study, CFD (20m/s) : 16sections Thrust coefficient : present study, CFD (15m/s) : 16sections
Torque coefficient : present study, CFD (15m/s) : 16sections 2.5 Thrust coefficient : present study, CFD (13m/s) : 16sections
Torque coefficient : present study, CFD (13m/s) : 16sections Thrust coefficient : present study, CFD (10m/s) : 16sections
1.2 Torque coefficient : present study, CFD (10m/s) : 16sections Thrust coefficient : present study, CFD (7m/s) : 16sections
Torque coefficient : present study, CFD (7m/s) : 16sections Thrust coefficient : present study, CFD (5m/s) : 16sections
Torque coefficient : present study, CFD (5m/s) : 16sections 2.0
20m/s 15m/s 20m/s
0.9 25m/s 25m/s
13m/s Thrust coefficient : 25m/s 1.5 15m/s
13m/s
0.6 10m/s 10m/s
1.0

7m/s 7m/s
0.3 0.5

5m/s Torque coefficient : 25m/s


5m/s
0.0
0.0 0.0 0.2 0.4 0.6 0.8 1.0
0.0 0.2 0.4 0.6 0.8 1.0
Span ratio (r/R)
Span ratio (r/R)

(c) Torque coefficient (d) Thrust coefficient


Fig. 3. Local aerodynamic force coefficient of 16sections case for various wind speed cases.

Torque force (N)


100.0
Thrust force : present study, CFD (25m/s) : 16sections Thrust force (N) Torque force : present study, CFD (25m/s) : 16sections
Torque force : present study, CFD (25m/s) : 16sections 200.0 Thrust force : present study, CFD (25m/s) : 16sections
Torque force : present study, CFD (20m/s) : 16sections Thrust force : present study, CFD (20m/s) : 16sections
Torque force : present study, CFD (15m/s) : 16sections Thrust force : present study, CFD (15m/s) : 16sections
Torque force : present study, CFD (13m/s) : 16sections Thrust force : present study, CFD (13m/s) : 16sections
Torque force : present study, CFD (10m/s) : 16sections Thrust force : present study, CFD (10m/s) : 16sections
80.0 Thrust force : present study, CFD (7m/s) : 16sections
Torque force : present study, CFD (7m/s) : 16sections
25m/s Torque force : present study, CFD (5m/s) : 16sections Thrust force : present study, CFD (5m/s) : 16sections
150.0
25m/s

60.0 20m/s
Thrust : 25m/s
20m/s
100.0
Torque force : 25m/s
40.0
15m/s

15m/s
13m/s 50.0
20.0
13m/s
10m/s
10m/s
7m/s 7m/s
0.0 0.0
0.0 1.0 2.0 3.0 4.0 5.0 0.0 1.0 2.0 3.0 4.0 5.0
Span (m) Span (m)

Fig. 4. Global aerodynamic force of 16sections case for various wind speed cases.

structural analysis based on BEM, and the structural response re- 3. FE model development for FSI analysis
sults are examined in comparison with those obtained by the FSI
analysis. The near-wall grid resolution and its modeling technique (low-
518 K. Lee et al. / Renewable Energy 113 (2017) 512e531

Re. [20e27] and wall function methods [28,29]) have direct effects
on the performance of the RANS turbulence model in the CFD
analysis. Therefore, in order to successfully perform a FSI analysis, it
is necessary to prepare valid grid models for CFD. Likewise,
developing and verifying the structural FE model for the whole
wind turbine system, including the tower, rotor, nacelle, and blades,
are important and essential processes for the FSI analysis.
In the FSI analysis, the information on the blade surface ob-
tained by the CFD analysis is transferred directly as a pressure load
for the blade structure. The interface boundary condition provides
topological information between the two different domains, and
the pressure of the fluid domain is transferred as a load for the
structure through the interface. In general, the fluid grid and FE
mesh for the CFD and structural analyses, respectively, differ from
each other in shape and characteristics. In addition, the FE element
for structural analysis can be much coarser than the grid model for
fluid analysis, it is necessary to use relatively more elements for
the FE meshes in the FSI analysis than in a normal FE analysis to
import the surface pressure information of the fluid domain pre-
cisely into a pressure load for the structure. The blade modeling in
the BEM method can be divided into two cases: using one-
dimensional (1D) beam elements [8] and using the FE model
[6,7]. The accuracy and quality of the geometry and mesh of the
wind turbine blade are not a critical part in the structural analysis
for using BEM. In this case, the structural model is generated
under the conditions where its valid structural performance is
obtained.
However, in the FSI analysis, the geometries of the structural
and fluid domains should be matched with each other as exactly as
possible, and finer FE meshes are generated to minimize any
possible error when importing the pressure information of the fluid
domain. Thus, the geometry and FE mesh modeling processes in FSI
are more difficult than usual for BEM. In fact, numerical instability
may occur in the grids and meshes between two different domains
during the FSI analysis. In recent studies on FSI, the arbitrary
Lagrangian-Eulerian (ALE) technique was adopted to successfully
overcome the grid compatibility problem between the fluid and
structural domains [9,10].

4. Wind turbine structural model

The wind blade has a complex 3D curved surface with a


sectional airfoil shape and flexible nature, and is manufactured
using composite materials. Thus, its aerodynamic characteristics
are very sensitive to the curved shape. Determining the properties
of the blade is more difficult than that determining the properties
of the other parts in the wind turbine system. Thus, many studies
on the structural characteristics of a wind turbine have focused on
the blade [6e10].
In contrast, the tower is the main structural part that supports
the entire wind turbine. Its behavior and design process are
easily predictable. Therefore, the tower could often be omitted in
many cases, despite its importance. In order to consider the
tower in the structural model, the mechanical parts such as the
rotor, shaft, nacelle, and hub, which connect the blade and the
tower, must also be included in the FE model, but there are many
uncertainties. As mentioned in their report, NREL did not
perform a FE analysis because of the difficulties in determining
the uncertainties when developing a structural model [2]. As
previously explained, even though the blade is the most complex
part in a wind turbine system, the most important structural part
of a wind turbine is not the blade but the tower, and the struc-
Fig. 5. Blade surface mesh for fluid and structure (10 m/s). tural design of a wind turbine must involve the tower. Therefore,
the structural model should include the tower for a more accu-
rate FSI analysis.
K. Lee et al. / Renewable Energy 113 (2017) 512e531 519

.
Fig. 6. NREL Phase VI wind turbine structural FEM model: stiffness and mass definitions.
520 K. Lee et al. / Renewable Energy 113 (2017) 512e531

Table 1
Material properties: stiffness [30].

Components Material (Pa) Density (kg/m3) Description

Blade E ¼ 1.560E þ 10 1035 Effective stiffness and density


v ¼ 0.42
Nacelle, Hub, Shaft, etc Steel e Point mass (Table 2)
(Blade: E ¼ 1.560E þ 10
v ¼ 0.42)
Rigid link Rigid e Point mass (Table 2)
Tower Steel (ASTM A106) 10,350 Effective density

Table 2 6. Distribution of wind pressure characteristics on blade


Mass definitions for parts [2,3]. surface
Components Mass (kg) Description

Pitch shaft, bull gear, instrumentation, 38.6 Point mass


Fig. 8 shows the wind pressure distributions on the blade sur-
bearings, nut and spacers face obtained through a CFD analysis in comparison with sectional
Hub 237.8 Point mass wind pressure information for seven wind-speed cases. The
Boom, instrumentation enclosures and camera 141.9 Point mass sectional wind pressures shown in Fig. 8 have a different meaning
Nacelle 1256.0 Point mass
from that of the pressure coefficients evaluated in Fig. 1 or in the
Blade1 mass (standard tip) 60.2 Effective mass
Blade3 mass (standard tip) 60.2 Effective mass wind tunnel test. In other words, the amount of pressure distri-
Total rotor mass 577.3 e bution is a main concern in the FSI analysis, and thus it is not
Nacelle, hub, pitch shafts, and boom mass 1712.0 Point mass necessary to convert this pressure distribution information into a
Tower mass 3300.0 Effective mass
non-dimensional pressure coefficient. In fact, in the pressure co-
Pipe-609.6  17.5 (base) (3317.0) (mass of NREL)
Pipe-406.4  21.4 (top)
efficient graph of Fig. 1 in section 2, the pressure coefficient de-
creases as it moves closer to the tip and becomes the largest at the
section near the hub in the 5e15 m/s wind speed region. However,
the real wind pressure distribution was contrary to these, and was
Fig. 5 illustrates the fluid grid and FE mesh model for the CFD found to be the largest and smallest at the tip and hub, respectively.
and structural analyses with the transferred pressure information. In the deep-stall region (20 m/s, 25 m/s), the wind pressure dis-
Fig. 6, Tables 1 and 2 briefly explain the definitions of the stiffness tributions show a similar tendency as a whole.
and mass of the structural FE model of the NREL Phase VI wind The wind pressure distribution characteristics on the blade
turbine that was used in this study which is based on NREL’s surface are greatly affected by the wind speed and resulting stall.
report information [2,3]. Unlike previous studies [6e10], the entire In the pre-stall region (5 m/s, 7 m/s) and at the stall onset wind
wind turbine system composed of the blade, tower, and me- speed (10 m/s), a strong suction pressure occurs at the blade’s
chanical parts (nacelle, rotor, hub, etc.) was considered in the FSI leading edge near the tip, and relatively regular pressure distri-
and BEM analyses, as fully described in Lee et al. [30]. Further butions appear at the suction side. Afterward, in the dynamic-stall
information is provided elsewhere [30]. Thus, the development region (13 m/s, 15 m/s), the leading edge area, where the suction
and verification process for the FE model is omitted in this occurs, is decreased, and irregular pressure distributions at the
manuscript. blade’s suction side near the hub occur as a result of the unstable
separation phenomenon of the flow in the dynamic-stall region.
5. Aerodynamic force evaluation using FSI analysis Because of this, the tangent force coefficient and torque force
decrease. In the deep-stall region, with a 20e25 m/s wind speed
Fig. 7 describes the structural analysis process for the FSI and range, the effects of the leading-edge suction become insignificant,
BEM using the CFD results. In the FSI analysis, the CFD results were and strong suction occurs at the isolated small area of the leading
directly transferred to the surface pressure load. As described in edge near the hub, which has little influence on the aerodynamic
Fig. 2, the evaluated aerodynamic force in the global coordinates force.
(torque and thrust force) were used as the nodal load for each
section division type in the BEM process. The analytical results 7. Evaluation of aerodynamic force characteristics using FSI
according to the FSI and BEM methods are described in the
following sections by evaluating the aerodynamic force and struc- Fig. 9 shows the pressure loads introduced on the blade surface
tural response, respectively. The effects of the number of blade el- through the FSI process with regard to each wind speed, and il-
ements will also be discussed. lustrates the clear and direct pressure load information in com-
As shown in Fig. 7, both the BEM and FSI methods for the parison with that of the CFD results (Fig. 8). The tendencies of the
structural response use the CFD results for their loading informa- pressure distribution introduced on the structural blade surface
tion. However, if the FSI method is directly used for the structural through the FSI analyses were clearly distinguished according to
analysis, the analysis processes are drastically reduced and the wind speed and stall effects. As explained in Fig. 9, the pressure
simplified compared to BEM by omitting the evaluation process for distributions on the blade surface can be classified into the pre-stall
all of the aerodynamic force quantities. The accuracy of the FSI can (5 m/s, 7 m/s), dynamic-stall (10 m/s, 13 m/s, 15 m/s), and deep-stall
be fundamentally dependent on valid CFD results for the pressure (20 m/s, 25 m/s) regions, and these pressure distribution ten-
value on the blade surface, which is defined as the wall boundary in dencies are applied as pressure loads for the blade through the FSI
the CFD analysis. The FSI can prevent any errors that may occur analysis. As suggested in NREL’s wind tunnel test and earlier
during the calculation process in evaluating the aerodynamic force, research, the power generation efficiency of the wind turbine is not
where the surface pressure is converted to a nodal quantity for the proportional to the wind speed. In contrast, a region occurs where
BEM analysis. the power generation decreases or is constant as the wind speed
K. Lee et al. / Renewable Energy 113 (2017) 512e531 521

Fig. 7. 1-way FSI and BEM process using CFD.

increases according to the design and control methodology. The accuracy is highly dependent on the number and locations of the
stall-regulated NREL Phase VI wind turbine experienced the post- pressure tabs, through a dedicated and high dense placement of
stall region, and afterward, its power generation gradually in- pressure tabs. Thus, the discrepancies between the aerodynamic
creases again. forces in the NREL measured results and those evaluated using the
Fig. 10 illustrates the torque and thrust forces that were deter- 5-section pressure coefficients were calculated using the BEM
mined by the resultant reaction force about the primary axis in the process. The results shown in Fig. 10 can explain the accuracies and
structural wind turbine FE model through the FSI and BEM, validities of the 3D measured resultant aerodynamic forces. The
respectively. The CFD results were used to evaluate both the pres- BEM results from the structural wind turbine FE model and eval-
sure on the surface and aerodynamic forces on the nodal points, uated 5-section nodal aerodynamic loading can be different. To
which were applied as the loading information for each analysis, as improve the accuracies of the BEM calculations, a larger number of
illustrated in Fig. 7. blade elements or pressure tabs should be installed and considered
Even though the FSI process imported the pressure information in a wind tunnel test and BEM, respectively, but this has limitations.
from the CFD, its resulting aerodynamic torque force showed good The BEM calculations for the 5- and 16-sections in this study, using
agreement with the original CFD results, with a negligible the CFD results, show the importance and effect of the number of
discrepancy. However, in the BEM approach, the number of blade blade element divisions. A larger number of sections for the BEM
elements was very important to ensure the accuracy of the 3D produced more reasonable results that are close to those of the FSI
aerodynamic forces. analysis. As explained before, if the FSI analysis is performed, the
The NREL’s reported aerodynamic forces for the torque and entire evaluating process for the aerodynamic forces in the local
thrust force were measured using instruments. Strain gauges and and global coordinates was omitted. The consideration of the
load cells were installed at the low-speed shaft, teeter damper, and sectional divisions was also unnecessary, but the fundamental
link. Thus, the aerodynamic forces could be considered to be more surface pressure on the CFD domain needed to be transferred from
accurate than the aerodynamic characteristics evaluated from the the interface. The verified RANS CFD analysis results can give ac-
pressure coefficient, which were measured using pressure tabs. To curate results for the 1-way FSI analysis.
evaluate the torque and thrust force from the pressure coefficients, The imported pressure load distribution characteristics for the
the conventional BEM calculation processes such as line integral structural surface in Fig. 9 are important information, which can
and coordinate transformations must be performed, and the explain the 3D aerodynamic forces of the torque and thrust force
522 K. Lee et al. / Renewable Energy 113 (2017) 512e531

Fig. 8. Blade surface pressure for 7-wind speed cases from CFD.
K. Lee et al. / Renewable Energy 113 (2017) 512e531 523

Fig. 9. Imported pressure load from CFD through FSI process.


524 K. Lee et al. / Renewable Energy 113 (2017) 512e531

Torque (J) Thrust force (N)


4000 4000
Torque : CFD, SST(Gamma theta)
Torque : FSI using CFD
Torque : BEM, 16sections using CFD Thrust force :
Torque : BEM, 5sections using CFD NREL (3D experiment)/BEM (NREL’s 5-section)
Torque : NREL(3D experiment)
Torque : BEM, 5sections using NREL's
------------------------------------------
Thrust : CFD, SST(Gamma theta)
3000 Thrust : FSI using CFD 3000
Thrust : BEM, 16sections using CFD
Thrust : BEM, 5sections using CFD
Thrust : NREL(3D experiment)
Thrust : BEM, 5sections using NREL's
------------------------------------------

2000 2000

CFD/FSI/BEM(16sectins)

1000 1000
Torque :
NREL (3D experiment)
BEM (5-section)
Torque :
BEM (NREL’s 5-section)

Wind speed (m/s)


0 0
5 10 15 20 25

Fig. 10. Evaluation of torque and thrust from FSI and BEM (5 and 16sections).

(Fig. 10). In the wind speed range of 7e13 m/s, a negative suction structural analysis was performed with the BEM method using
pressure is mainly distributed at the leading edge near the tip, the CFD results, the accuracy of the analysis results could be
which can cause the blade to rotate. enhanced by considering a larger number of elements. Moreover,
As shown in Fig. 9, suction surface is covered by a wider that in the direct FSI analysis, the pressure information for the blade
region area at a wind speed of 7 m/s, and the maximum rotational surface could be precisely applied as a pressure load for the blade
force of the torque occurred at a wind speed of 10 m/s. With an surface using the interaction process without the aerodynamic
increase in the wind speed, the pressure distribution changes to- conversion process. As a result, the FSI analysis could reflect the
ward the mid-surface of the blade, and the power generating effi- aerodynamic characteristics more accurately than the BEM
ciency of the wind turbine gradually decreases. Therefore, if the FSI method. Therefore, the FSI method of this study can be consid-
process is used for the pressure load distribution, it is possible to ered to be an analysis method that can consider more advanced
recognize the aerodynamic characteristics more clearly in com- aerodynamic characteristics in the structural design of a wind
parison with the sectional aerodynamic forces (Fig. 2), which are turbine.
based on the pressure coefficients (Fig. 1).
Based on the torque analysis results in Fig. 10, it could be un- 8. Evaluation of structural behavior characteristics using FSI
derstood that the NREL Phase VI wind turbine displays its highest analysis
power generating performance at a wind speed of 10 m/s. The ac-
curacy and efficiency of the FSI method were also described and Fig. 11 shows the deformed geometries and deformation ob-
discussed. Moreover, in the case of the thrust force, it can be seen tained by the FSI analysis and BEM method at the 10 m/s speed case
that the effects of stall on it are relatively small in comparison with respectively. The deformed geometries are shown at a magnifica-
the rotational torque force. The thrust force has a disposition to tion of 50. In the case of the BEM method, analyses were performed
increase gradually and consistently as the wind speed increases. on 5- and 16-sections using the aerodynamic characteristics eval-
Compared with the torque force, the thrust force has little effect on uated in section 2 which were applied as a nodal load. Fig. 12 shows
the power generation of the wind turbine. If the thrust force in- the deformed configurations of a wind turbine obtained by the FSI
creases constantly, it will have severe structural safety effects on analysis according to the wind speeds for the whole (Fig. 12(a)) and
the blade and turbine. Thus, it is necessary to decrease the thrust blade alone (Fig. 12(b)) models. For the blade alone model, the
force to be as low as possible. tower and nacelle’s boundaries were constrained to obtain just the
In this section, the FSI method was used to evaluate the blade behavior.
aerodynamic force characteristics of the NREL Phase VI wind As can be seen in Fig. 12, the deformation increased in pro-
turbine, which were compared with those obtained by the BEM portion to the wind speed in the pre-stall region (5 m/s, 7 m/s) and
method. In order to obtain accurate and valid BEM calculations, a at the stall-onset velocity (10 m/s). After this, the deformation
larger number of blade element divisions was required. When remained similar in the dynamic-stall region, even though the
K. Lee et al. / Renewable Energy 113 (2017) 512e531 525

(a) FSI (b) BEM: 5sections (c) BEM: 16sections


Fig. 11. Results of structural analysis through FSI and BEM: 10 m/s wind speed case (scale: 50).

wind speed was higher. In the deep-stall region (20 m/s, 25 m/s), Fig. 14(a) and (b) show the comparative displacement results
as the wind speed increased again, the deformation also obtained by the FSI analysis only, according to the wind speeds,
increased. These structural responses can be highly related to the which were derived from Fig. 13 for the X and Y axes. It can be
pressure loading and resulting aerodynamic force on the blade seen in Fig. 14 that the deformation along the Y axis is small in
surface. comparison with that along the X axis. In the displacement re-
The graph of Fig. 13 describes various displacement results that sults along the X axis in Fig. 14(a), the displacement, at a wind
were obtained by using FSI and BEM at 16 locations in the blade’s speed of 13 m/s or 15 ms in the dynamic-stall region, showed
span-wise direction for each wind speed case. The global X and Y results that were almost similar to those at a velocity of 10 m/s,
directional displacements are plotted together. The global X and Y which is the stall-onset wind speed. In the aerodynamic char-
directions are parallel to the thrust and torque forces, respectively. acteristics (thrust force) explained in Fig. 10, the thrust force in
As shown in Fig. 13, the blade deformation and structural response the wind speed range of 13e15 m/s has a larger value than the
are governed by the X direction displacement, which is the wind speed of 10 m/s. Nevertheless, the reason why the
downstream direction on the global coordinate system. In com- respective deformations along the X axis at these three wind
parison, the blade deformation along the Y axis, which is the speeds almost agree with each other is that, as explained in
rotational direction of the blade, had a relatively small value in Fig. 4(b), the thrust force component pressure at a wind speed of
comparison with that along the X axis. 10 m/s has a larger distribution near the tip region than at
The wind blade’s sectional local weak and strong axes are usu- 13e15 m/s. As a result, although all of the thrust forces at wind
ally oriented in the global X and Y directions, respectively, to in- speeds of 13 m/s and 15 ms were larger than those at 10 m/s, the
crease the power generation efficiency. In order to maximize the relative displacement for the thrust force was found to be larger
rotational force of the blades and rotor, the direction of rotation at a wind speed of 10 m/s. It could also be seen that a large
becomes the strong axis on the local blade section, which is highly deformation occurs along the X axis in the deep-stall region
related to the airfoil properties. In contrast, the downstream di- (20 m/s, 25 m/s).
rection becomes the weak axis, and thus a larger deformation may Fig. 14(b) shows the displacement analysis results for the Y axis.
occur in this direction. The displacement was the largest at a wind speed of 25 m/s. In the
As a result of the structural analysis using FSI and BEM, the 10e20 m/s wind speed range, the deformation along the Y axis was
deformation due to NREL’s aerodynamic characteristics was found almost similar, and it could be seen that this is related to the
to be largest. The analysis results obtained from 16-sections were rotational torque distribution of Figs. 4(a) and 10. Moreover,
closer to the FSI results than the results obtained from 5-sections. considering the displacement along the X axis and the section
The overall structural responses were highly related to the evalu- stiffness, the displacements along the Y axis are relatively very
ated aerodynamic characteristics, which are described in Fig. 10. small and are almost ignorable.
526 K. Lee et al. / Renewable Energy 113 (2017) 512e531

Fig. 12. Displacement for whole turbine and blade alone model using FSI analysis (scale: 50).

CFD and wind tunnel test analyses for wind turbines mainly predict, not only results for the blade alone, but also for the tower
focus on the contents related to the enhancement of their power deformation.
generation. Thus, the geometry optimization of wind turbine As shown in Figs. 12e15, the deflections of blade at the tip vary
blades is pursued in order to maximize it. However, as can be from 15 cm (5 m/s) to 60 cm (25 m/s), when the whole turbine
seen in Fig. 14, the structural responses of blades and towers are system of blade, rotor, nacelle and tower are considered. And the
mainly governed by the thrust force, not the torque force. In ratio of deflection/span can be sail to be 0.03 to 0.12, in which can
comparison, the contribution of the rotational force to their be understood as large deflections.
structural deformation characteristics is almost insignificant. The main focus of this studies is to show the structural response
Thus, when designing the structure of a blade, the deformation of displacement of blade which connected with rotor and tower.
along the Y axis caused by its torque force may not be a matter of The CFD results which were used to evaluate the aerodynamic force
consideration. in 1-way FSI process, are sufficient and enough to perform the large
Fig. 15 shows the FSI analysis results obtained by boundary deflection analysis of rotor-blade. Actually, the way of our CFD grid
constraining the tower and nacelle to obtain the blade alone re- is simply defined to get accurate fluid behaviors. The near-wall yþ,
sponses, and it can be seen that the overall structural responses are domain size and turbulence model are considered to define it. The
similar to the results of Fig. 14, where the tower and nacelle were way of CFD grid to handle the large deflection of rotor-blade can be
also considered. If the tower deformation component is removed a critical point that is worth more detailed explanation which were
from Fig. 14, the results of Fig. 14 become similar to those obtained introduced in Part1 of this study.
by considering the blade alone, as shown in Fig. 15. It can be un- Fig. 16 describes the redefined displacement results for the tip
derstood that the analysis method of this study can efficiently region, which were already illustrated in Fig. 13 for wind speed
K. Lee et al. / Renewable Energy 113 (2017) 512e531 527

Displacement (5m/s, mm) Displacement (15m/s, mm)


30.0 60.0
X-dir. Displacement : present study (FSI, 5m/s) X-dir. Displacement : present study (FSI, 15m/s)
X-dir. Displacement : present study (16sections aerodynamic force, 5m/s) X-dir. Displacement : present study (16sections aerodynamic force, 15m/s)
X-dir. Displacement : present study (5sections aerodynamic force, 5m/s) X-dir. Displacement : present study (5sections aerodynamic force, 15m/s)
25.0 X-dir. Displacement : NREL (3D expriment, aerodynamic force, 5m/s) 50.0 X-dir. Displacement : NREL (3D expriment, aerodynamic force, 15m/s)
---------------------------------------------------------------------------------------------
---------------------------------------------------------------------------------------------
Y-dir. Displacement : present study (FSI, 15m/s)
Y-dir. Displacement : present study (FSI, 5m/s)
Y-dir. Displacement : present study (16sections aerodynamic force, 15m/s) NREL(5sections)
Y-dir. Displacement : present study (16sections aerodynamic force, 5m/s) Y-dir. Displacement : present study (5sections aerodynamic force, 15m/s)
20.0 Y-dir. Displacement : present study (5sections aerodynamic force, 5m/s) 40.0 Y-dir. Displacement : NREL (3D expriment, aerodynamic force, 15m/s)
Y-dir. Displacement : NREL (3D expriment, aerodynamic force, 5m/s)

FSI
15.0 NREL(5sections) 30.0 16sections
FSI 5sections
16sections
10.0 5sections 20.0

5.0 10.0 X-dir Disp.


X-dir Disp.
Y-dir Disp. Span (m) Y-dir Disp. Span (m)
0.0 0.0
0.0 1.0 2.0 3.0 4.0 5.0 6.0 0.0 1.0 2.0 3.0 4.0 5.0 6.0

Displacement (7m/s, mm) Displacement (20m/s, mm)


30.0 80.0
X-dir. Displacement : present study (FSI, 7m/s) X-dir. Displacement : present study (FSI, 20m/s)
X-dir. Displacement : present study (16sections aerodynamic force, 7m/s) X-dir. Displacement : present study (16sections aerodynamic force, 20m/s)
X-dir. Displacement : present study (5sections aerodynamic force, 7m/s) X-dir. Displacement : present study (5sections aerodynamic force, 20m/s)
70.0
X-dir. Displacement : NREL (3D expriment, aerodynamic force, 7m/s) NREL(5sections) X-dir. Displacement : NREL (3D expriment, aerodynamic force, 20m/s)
25.0 ---------------------------------------------------------------------------------------------
Y-dir. Displacement : present study (FSI, 7m/s) ---------------------------------------------------------------------------------------------
Y-dir. Displacement : present study (16sections aerodynamic force, 7m/s) 60.0 Y-dir. Displacement : present study (FSI, 20m/s)
Y-dir. Displacement : present study (5sections aerodynamic force, 7m/s) FSI Y-dir. Displacement : present study (16sections aerodynamic force, 20m/s)
Y-dir. Displacement : NREL (3D expriment, aerodynamic force, 7m/s) Y-dir. Displacement : present study (5sections aerodynamic force, 20m/s)
20.0 16sections
Y-dir. Displacement : NREL (3D expriment, aerodynamic force, 20m/s) NREL(5sections)
50.0
5sections
FSI
15.0 40.0
16sections
5sections
30.0
10.0

20.0

5.0 X-dir Disp. X-dir Disp.


10.0

Y-dir Disp. Span (m) Y-dir Disp. Span (m)


0.0 0.0
0.0 1.0 2.0 3.0 4.0 5.0 6.0 0.0 1.0 2.0 3.0 4.0 5.0 6.0

Displacement (10m/s, mm) Displacement (25m/s, mm)


50.0 80.0
X-dir. Displacement : present study (FSI, 10m/s) X-dir. Displacement : present study (FSI, 25m/s)
X-dir. Displacement : present study (16sections aerodynamic force, 10m/s) X-dir. Displacement : present study (16sections aerodynamic force, 25m/s)
X-dir. Displacement : present study (5sections aerodynamic force, 10m/s) X-dir. Displacement : NREL (3D expriment, aerodynamic force, 25m/s)
X-dir. Displacement : NREL (3D expriment, aerodynamic force, 10m/s) X-dir. Displacement : present study (5sections aerodynamic force, 25m/s)
40.0 --------------------------------------------------------------------------------------------- --------------------------------------------------------------------------------------------- NREL(5sections)
Y-dir. Displacement : present study (FSI, 10m/s) 60.0 Y-dir. Displacement : present study (FSI, 25m/s)
Y-dir. Displacement : present study (16sections aerodynamic force, 10m/s) Y-dir. Displacement : present study (16sections aerodynamic force, 25m/s)
Y-dir. Displacement : present study (5sections aerodynamic force, 10m/s) NREL(5sections) Y-dir. Displacement : present study (5sections aerodynamic force, 25m/s) FSI
Y-dir. Displacement : NREL (3D expriment, aerodynamic force, 10m/s) FSI Y-dir. Displacement : NREL (3D expriment, aerodynamic force, 25m/s) 16sections
30.0 16sections 5sections

40.0
5sections
20.0

20.0
10.0 X-dir Disp.
X-dir Disp.

Y-dir Disp. Span (m) Y-dir Disp. Span (m)


0.0 0.0
0.0 1.0 2.0 3.0 4.0 5.0 6.0 0.0 1.0 2.0 3.0 4.0 5.0 6.0

Displacement (13m/s, mm)


50.0
X-dir. Displacement : present study (FSI, 13m/s)
X-dir. Displacement : present study (16sections aerodynamic force, 13m/s)
X-dir. Displacement : present study (5sections aerodynamic force, 13m/s)
X-dir. Displacement : NREL (3D expriment, aerodynamic force, 13m/s) NREL(5sections)
40.0 ---------------------------------------------------------------------------------------------
Y-dir. Displacement : present study (FSI, 13m/s)
Y-dir. Displacement : present study (16sections aerodynamic force, 13m/s)
Y-dir. Displacement : present study (5sections aerodynamic force, 13m/s)
Y-dir. Displacement : NREL (3D expriment, aerodynamic force, 13m/s) FSI
30.0 16sections

5sections

20.0

10.0
X-dir Disp.

Y-dir Disp. Span (m)


0.0
0.0 1.0 2.0 3.0 4.0 5.0 6.0

Fig. 13. X, Y-dir. displacement along the blade span for FSI and BEM according to the wind speed.
528 K. Lee et al. / Renewable Energy 113 (2017) 512e531

X dir displacement along to the span (mm) Y dir displacement along to the span (mm)
60.0 30.0
X-dir. Displacement : present study (FSI, 25m/s) X-dir. Displacement : present study (FSI, 25m/s)
Y-dir. Displacement : present study (FSI, 25m/s)
X-dir. Displacement : present study (FSI, 20m/s) 25m/s Y-dir. Displacement : present study (FSI, 20m/s)
50.0 X-dir. Displacement : present study (FSI, 15m/s) 25.0 Y-dir. Displacement : present study (FSI, 15m/s)
X-dir. Displacement : present study (FSI, 13m/s) Y-dir. Displacement : present study (FSI, 13m/s)
X-dir. Displacement : present study (FSI, 10m/s) Y-dir. Displacement : present study (FSI, 10m/s)
X-dir. Displacement : present study (FSI, 7m/s)
20m/s
Y-dir. Displacement : present study (FSI, 7m/s)
40.0 20.0
X-dir. Displacement : present study (FSI, 5m/s) Y-dir. Displacement : present study (FSI, 5m/s)

10/13/15m/s X-dir. Displacement : 25m/s


30.0 15.0
25m/s
Tower displacement 7m/s Tower displacement 10m/s
20.0 10.0 13/20m/s
15m/s
5m/s 7m/s
10.0 5.0
5m/s
Span (m) Span (m)
0.0 0.0
0.0 1.0 2.0 3.0 4.0 5.0 6.0 0.0 1.0 2.0 3.0 4.0 5.0 6.0

Fig. 14. X, Y-dir. displacement of whole turbine along the blade span using FSI analysis.

X dir displacement along to the span (mm) Y dir displacement along to the span (mm)
60.0 30.0
X-dir. Displacement : Blade+Tower (FSI, 25m/s) Y-dir. Displacement : Blade+Tower (FSI, 25m/s)
X-dir. Displacement : Blade only (FSI, 25m/s) Blade+Tower Y-dir. Displacement : Blade only (FSI, 25m/s)
X-dir. Displacement : Blade only (FSI, 20m/s) 25m/s Y-dir. Displacement : Blade only (FSI, 20m/s)
50.0 25.0
X-dir. Displacement : Blade only (FSI, 15m/s) Y-dir. Displacement : Blade only (FSI, 15m/s)
X-dir. Displacement : Blade only (FSI, 13m/s) Y-dir. Displacement : Blade only (FSI, 13m/s)
X-dir. Displacement : Blade only (FSI, 10m/s) Y-dir. Displacement : Blade only (FSI, 10m/s)
40.0 X-dir. Displacement : Blade only (FSI, 7m/s) Blade only 20.0 Y-dir. Displacement : Blade only (FSI, 7m/s)
X-dir. Displacement : Blade only (FSI, 5m/s) Y-dir. Displacement : Blade only (FSI, 5m/s)
25m/s

30.0 20m/s 15.0


Blade+Tower
10/13/15m/s 25m/s
20.0 10.0
7m/s Blade only
25m/s
10/13/15/20m/s
10.0 5m/s 5.0
7m/s
Span (m) 5m/s
0.0 0.0
Span (m)
0.0 1.0 2.0 3.0 4.0 5.0 6.0 0.0 1.0 2.0 3.0 4.0 5.0 6.0

Fig. 15. X, Y-dir. displacement of blade alone model along the blade span using FSI analysis.

cases using the various analysis methods of this study. As a whole, of this paper, more accurate structural design can be performed by
the BEM results for NREL’s 5-sections are higher than the other avoiding 2D aerodynamic coefficients, and basically be used to
results obtained using the CFD results. As described in Fig. 14, the wide range of practical design applications like pitch angle control
displacement component in the Y direction, which is dependent on and fatigue analysis which the 3D aerodynamic structural response
the torque force, is far lower than the X direction component for all analysis should be considered.
the numerical cases. The BEM results for 16-sections are in good
agreement with the FSI results, compared to the 5-section results. 9. Conclusion
When comparing the whole turbine system, the displacement
contributions of the tower were not higher than those of the blade. In this study, structural static linear analyses were performed
The deformation response of the whole turbine could be governed using a 1-way FSI analysis for the NREL Phase VI wind turbine, and
by the blade. The blade was designed using a light material to be the characteristics of the structural analysis method were discussed
flexible and deformed by the aerodynamic force to make it rotate. by comparing the results with those obtained by the BEM method.
Fig. 17 shows the normalized values of Fig. 16, together with the For each structural analysis method (FSI and BEM), the entire
normalized aerodynamic characteristics (torques, thrust forces). wind turbine, including the nacelle and tower, was used as the
The 25 m/s wind speed was used as the reference value for wind turbine structural analysis model. The surface pressure in-
normalizing the values. The normalized structural responses in the formation of the fluid domain from a RANS CFD analysis was
X and Y directions are highly related to the normalized thrust and directly transferred to the structure using the interface in an FSI
torque force in the global coordinates, respectively. The normalized analysis. In contrast, the aerodynamic forces evaluated in global
responses of the blade alone model are close to the results for the coordinates were used for BEM.
whole turbine model, in which the tower and nacelle were According to the structural analysis results, the NREL Phase VI
considered. wind turbine was mainly governed by the thrust force, which is a
Traditionally, the main purpose of CFD analysis for wind turbine component of the X axis of the global coordinate system, and the
was considered to get accurate aerodynamic characteristics only to deformation characteristics based on its rotational force were
design or predict the practical wind turbine in commercial. The relatively insignificant, because of the blade’s geometrical char-
sectional aerodynamic coefficients were evaluated to be utilized in acteristics. The blade was designed to maximize the power
BEM process in design stage. By using the process and methodology generating efficiency of the wind turbine. It could be seen that the
K. Lee et al. / Renewable Energy 113 (2017) 512e531 529

X/Y dir maximum displacement (mm)


100.0
Tip X-dir. Displacement : present study (FSI)
Tip X-dir. Displacement : present study (16sections aerodynamic forces)
Tip X-dir. Displacement : present study (5sections aerodynamic forces)
Tip X-dir. Displacement : NREL (3D experiment, 5sections aerodynamic forces)
80.0 ---------------------------------------------------------------------------------
Tip Y-dir. Displacement : present study (FSI)
Tip Y-dir. Displacement : present study (16sections aerodynamic force)
Tip Y-dir. Displacement : present study (5sections aerodynamic force)
Tip Y-dir. Displacement : NREL (3D experiment, 5sections aerodynamic force)
--------------------------------------------------------------------------------- NREL(5sections)
60.0 Tip X-dir. Displacement : Blade only (FSI)
Tip Y-dir. Displacement : Blade only (FSI) FSI
16sections
5sections
40.0
FSI (blade only)

20.0

Wind speed (m/s)


0.0
0.0 5.0 10.0 15.0 20.0 25.0 30.0

Y dir maximum displacement (mm)


25.0
Tip Y-dir. Displacement : present study (FSI)
Tip Y-dir. Displacement : present study (16sections aerodynamic force)
Tip Y-dir. Displacement : present study (5sections aerodynamic force)
Tip Y-dir. Displacement : NREL (3D experiment, 5sections aerodynamic force)
20.0 ---------------------------------------------------------------------------------
Tip Y-dir. Displacement : Blade only (FSI)

15.0
NREL(5sections)
FSI
16sections
5sections
10.0

FSI (blade only)

5.0

Wind speed (m/s)


0.0
0.0 5.0 10.0 15.0 20.0 25.0 30.0

Fig. 16. Maximum displacement for whole turbine according to the wind speed.

deformation characteristics of the wind turbine were directly efficient for the evaluation of the aerodynamic force characteris-
affected by changes in the aerodynamic characteristics caused by tics as a structural analysis method for wind turbines. Thus, it
stall, etc., rather than by increasing the wind speed. In this case, could be said that FSI can derive more accurate analysis results in
the effects of the tower on the behaviors of the whole wind comparison with the general BEM method, which needs a larger
turbine structural responses were relatively simple, and its number of divisions on the span. A BEM analysis using valid CFD
characteristics were highly related to the aerodynamic force results can provide greater analytical applicability to choose the
characteristics. resolution of the blade element. However, rigorous and time
The FSI analysis process used in this study was shown to be consuming CFD studies must be performed to accomplish this. In
530 K. Lee et al. / Renewable Energy 113 (2017) 512e531

Normalized displacement
1.2

Normalized Torque
1.0
Normalized Max. Y dis.
(Blade only)
0.8

Normalized Max. X dis.


(Blade only)
0.6

Normalized Thrust
0.4
Normalized Max. X-Disp. : Blade only (FSI)
Normalized Max. Y-Disp. : Blade only (FSI)
Normalized Torque
0.2 Normalized Max. Y-Disp : Blade+Tower (FSI)
Normalized Max. X-Disp : Blade+Tower (FSI)
Normalized Thrust
Wind speed (m/s)
0.0
0 5 10 15 20 25 30

Fig. 17. Normalized displacement according to the wind speed.

the future, practical design applications such as pitch angle control wind turbine rotors at full scale. Part II: fluid-structure interaction modeling
with composite blades, Int. J. Numer. Methods Fluids 65 (2012) 236e253.
and fatigue analysis will be presented with transient dynamic
[10] Y. Bazilevs, M.C. Hsua, M.A. Scottb, Isogeometric fluidstructure interaction
response of the wind turbine using a 1-way and 2-way FSI analysis with emphasis on non-matching discretizations, and with application
analysis. to wind turbines, Comput. Methods Appl. Mech. Eng. 249e252 (2012) 28e41.
[11] E.P.N. Duque, M.D. Burkund, W. Johnson, Navier-Stokes and comprehensive
analysis performance predictions of the NREL phase VI experiment, J. Sol.
Acknowledgement Energy Eng. 125 (2003) 457e467.
[12] C. Tongchitpakdee, S. Benjanirat, L.N. Sankar, Numerical simulation of the
aerodynamics of horizontal Axis wind turbines under yawed flow conditions,
This research was supported by the National Science Foundation in: 43th AIAA Aerospace Sciences Meeting and Exhibit, Reno, NV, January,
(NSF) through the Center for Energy and Environmental Sustain- AIAA, 2005, pp. 281e304.
ability (CEES), a CREST Center (Award No. 1036593). And also was [13] N. Sezer-Uzol, N.L. Long, 3-D time-accurate CFD simulations of wind turbine
rotor flow fields, in: 44th AIAA Aerospace Sciences Meeting and Exhibit, Reno,
supported by a National Research Foundation of Korea (NRF) grant NV, January, AIAA, 2006, pp. 2006e2394.
funded by the Korea Government (MEST) (No. [14] M. Moshfeghi, Y.J. Song, Y.H. Xie, Effects of near-wall grid spacing on SST-K-u
2012R1A2A2A01043088). model using NREL Phase VI horizontal axis wind turbine, J. Wind Eng. In-
dustrial Aerodynamics 107e108 (2012) 94e105.
[15] J.O. Mo, Y.H. Lee, CFD Investigation on the aerodynamic characteristics of a
References small-sized wind turbine of NREL PHASE VI operating with a stall-regulated
method, J. Mech. Sci. Technol. 26 (1) (2012) 81e92.
[16] Y. Li, K.J. Paik, T. Xing, P.M. Carrica, Dynamic overset CFD simulations of wind
[1] R.R. Ramsay, M.J. Hoffman, G.M. Gregorek, Effects of Grit Roughness and Pitch
turbine aerodynamics, Renew. Energy 37 (1) (2012) 285e298.
Oscillation on the S809 Airfoil, NREL/TP-442e7817, National Renewable En-
[17] M.M. Yelmule, E. Anjuri VSJ, CFD predictions of NREL phase VI rotor experi-
ergy Laboratory (NREL), 1995.
ments in NASA/AMES wind tunnel, Int. J. Renew. Energy Res. 3 (2013)
[2] M. Hand, D. Simms, L.J. Fingersch, D. Jager, S. Larwood, J. Cotrell, S. Schreck,
250e260.
Unsteady Aerodynamics Experiment Phase VI: Wind Tunnel Test Configura-
[18] R. Lanzafame, S. Mauro, M. Messina, Wind turbine CFD modeling using a
tions and Available Data Campaigns, NREL/TP-500e29955, National Renew-
correlation-based transitional model, Renew. Energy 52 (2013) 31e39.
able Energy Laboratory (NREL), 2001.
[19] J.O. Mo, A. Choudhry, M. Arjomandi, R. Kelso, Y.H. Lee, Effects of wind speed
[3] D. Simms, S. Schreck, M. Hand, L.J. Fingersch, NREL Unsteady Aerodynamics
changes on wake instability of a wind turbine in a virtual wind tunnel using
Experiment in the NASA-Ames Wind Tunnel: a Comparison of Predictions to
large eddy simulation, J. Wind Eng. Industrial Aerodynamics 117 (2013)
Measurements, NREL/TP-500e29494, National Renewable Energy Laboratory
38e56.
(NREL), 2001.
[20] F.R. Menter, Review of the shear-stress transport turbulence model experi-
[4] J.L. Tangler, J.D. Kocurek, Wind turbine Post-Stall airfoil performance charac-
ence from an industrial perspective, Int. J. Comput. Fluid Dyn. 23 (4) (2009)
teristics guidelines for blade-element momentum methods, in: 43rd AIAA
305e316.
Aerospace Sciences Meeting and Exhibit, Reno, 2005. NV 2005 (NREL/CP-
[21] F.R. Menter, Two-Equation eddy-viscosity turbulence models for engineering
500e36900).
applications, AIAA-Journal 32 (8) (1994) 269e289.
[5] R. Lanzafame, M. Messina, Bem theory : how to take into account the radial
[22] R.B. Langtry, F.R. Menter, Transition modeling for general CFD applications in
flow inside of a 1-D numerical code, Renew. Energy 39 (2012) 440e446.
aeronautics, AIAA (2005) paper 2005-522.
[6] D. Cardenas, H. Elizalde, P. Marzocca, S. Gallegos, O. Probst, A coupled aero-
[23] F.R. Menter, R.B. Langtry, S. Vo € lker, Transition modeling for general purpose
elastic damage progression model for wind turbine blades, Compos. Struct. 94
codes, J. Flow Turbul. Combust. 77 (1e4) (2006) 277e303.
(2012) 3072e3081.
[24] F.R. Menter, R.B. Langtry, S. Likki, Y. Suzen, P. Huang, S. Vo€lker, A correlation-
[7] L.I. Lago, F.L. Ponta, A.D. Otero, Analysis of alternative adaptive geometrical
based transition model using local variables part I: model formulation,
configurations for the NREL-5 MW wind turbine blade, Renew. Energy 59
J. Turbomach. 128 (2006) 413e422.
(2013) 13e22.
[25] R.B. Langtry, F.R. Menter, S. Likki, Y. Suzen, P. Huang, S. Vo€lker, A correlation-
[8] A. Sharifi, M.R.H. Nobari, Prediction of optimum section pitch angle distribu-
based transition model using local variables part II: test cases and industrial
tion along wind turbine blades, Energy Convers. Manag. 67 (2013) 342e350.
applications, J. Turbomach. 128 (2006) 423e434.
[9] Y. Bazilevs, M.-C. Hsu, J. Kiendl, R. Wüchner, K.-U. Bletzinger, 3D simulation of
K. Lee et al. / Renewable Energy 113 (2017) 512e531 531

[26] R.B. Langtry, F.R. Menter, A correlation-based transition model using local (2005) 265e291.
variables for unstructured parallelized CFD codes, AIAA J. 47 (12) (2009). [29] M. Popovac, K. Hanjalic, Compound wall treatment for RANS computation of
[27] T. Esch, F.R. Menter, Heat transfer predictions based on two-equation turbu- complex turbulent flows and heat transfer, J. Flow Turbul. Combust. 78 (2)
lence models with advanced wall treatment, Turbul. Heat. Mass Transf. 4 (2007) 177e202.
(2003) 6333e6640. [30] K.S. Lee, Z. Huque, R. Kommalapati, S.E. Han, Evaluation of equivalent struc-
[28] G. Kalitzin, G. Medic, G. Iaccarino, P. Durbin, Near-wall behaviour of RANS tural properties of NREL phase VI wind turbine blade, Renew. Energy 86
turbulence models and implications for wall functions, J. Comput. Phys. 204 (2016) 796e818.

You might also like