IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 50, NO.
6, NOVEMBER/DECEMBER 2014
Gearbox and Drivetrain Models to Study Dynamic
Effects of Modern Wind Turbines
Irving P. Girsang, Student Member, IEEE, Jaspreet S. Dhupia, Member, IEEE, Eduard Muljadi, Fellow, IEEE,
Mohit Singh, Member, IEEE, and Lucy Y. Pao, Fellow, IEEE
Abstract—Wind turbine drivetrains consist of components that
directly convert kinetic energy from the wind to electrical energy.
Therefore, guaranteeing robust and reliable drivetrain designs
is important to prevent turbine downtime. Current drivetrain
models often lack the ability to model both the impacts of electrical
transients as well as wind turbulence and shear in one package.
In this paper, the capability of the FAST wind turbine comput-
er-aided-engineering tool, developed by the National Renewable
Energy Laboratory, is enhanced through the integration of a
dynamic model of the drivetrain. The dynamic drivetrain model
is built using Simscape in the MATLAB/Simulink environment
and incorporates detailed electrical generator models. This model
can be used to evaluate internal drivetrain loads due to excitations
from both the wind and generator.
Index Terms—Gears, mechanical power transmission, reso-
nance, variable-speed drives, vibration, wind energy.
N OMENCLATURE
αrot Turbine rotor angular acceleration.
β Gear helical angle.
λopt Optimum tip-speed ratio.
ωrot
ceff
Turbine rotor angular speed.
Effective drivetrain torsional damping with respect to
the low-speed side of the gearbox.
fn Nonzero eigenfrequency of the two-mass model.
keff Effective drivetrain torsional stiffness with respect to
the low-speed side of the gearbox.
kgear Gear tooth stiffness.
kmesh Constant gear meshing stiffness.
kY Torsional stiffness of each drivetrain shaft, Y ∈ {low-
speed shaft (LSS), high-speed shaft (HSS), and lth
intermediate shaft, l = 1, 2}.
rb Gear base circle radius.
Manuscript received November 6, 2013; revised March 19, 2014; accepted
April 14, 2014. Date of publication April 29, 2014; date of current version
November 18, 2014. Paper 2013-IDC-868.R1, presented at the 2013 IEEE
Energy Conversion Congress and Exposition, Denver, CO, USA, September
16–20, and approved for publication in the IEEE T RANSACTIONS ON I NDUS -
TRY A PPLICATIONS by the Industrial Drives Committee of the IEEE Industry
Applications Society.
I. P. Girsang and J. S. Dhupia are with Nanyang Technological University,
Singapore 639798 (e-mail: irving1@e.ntu.edu.sg; djaspreet@ntu.edu.sg).
E. Muljadi and M. Singh are with the National Renewable Energy Lab-
oratory, Golden, CO 80401 USA (e-mail: eduard.muljadi@nrel.gov; mohit.
singh@nrel.gov).
L. Y. Pao is with the University of Colorado Boulder, Boulder, CO 80309
USA (e-mail: pao@colorado.edu).
Color versions of one or more of the figures in this paper are available online
at http://ieeexplore.ieee.org.
Digital Object Identifier 10.1109/TIA.2014.2321029
0093-9994 © 2014 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
3777
Jeff Effective inertia of the generator and the gearbox with
respect to the low-speed side of the gearbox.
JX Inertia of the drivetrain components, X ∈ {turbine
rotor (rot), planet gears and the carrier (P C), carrier
only (C), sun gear (S), generator (gen), planet gears
(Pm ), m = 1, . . . , M , and parallel gear components
(Gi ), i = 1, 2, 3, 4}.
M Number of planet gears.
N Overall drivetrain gear ratio.
Nj Gear ratio of each gear stage, j = 1, 2, 3, Nj ≥ 1.
Qaero Wind aerodynamic torque.
Qem Generator electromagnetic torque.
Qopp Rotor-opposing torque.
Rrot Turbine rotor radius.
Vw Effective wind speed.
I. I NTRODUCTION
T HE WIND energy industry has experienced substantial
growth in recent decades, and there has been similar
growth in the structural size of and output power from wind
turbines. The operating conditions of wind turbines are largely
determined by structural loadings such as wind turbulence and,
in offshore wind turbines, sea wave excitations. Aeroelastic
computer-aided engineering (CAE) tools such as FAST [1],
Bladed [2], and HAWC2 [3] have been developed to model
and simulate the dynamics of wind turbines in response to
different wind fields and controllers. These tools simulate the
most relevant environmental conditions and output time series
of turbine operational load variations.
Because the wind turbine drivetrain consists of components
that directly convert rotational kinetic energy from the wind to
electrical energy, ensuring the reliability of drivetrain designs
is critical to preventing wind turbine downtime. Because of
the steadily increasing size of wind turbines, larger forces
and torques bring up the influence of the gearbox and other
drivetrain flexibilities in the overall turbine dynamic response
[4], [5], often leading to failure in the drivetrain components,
particularly the gears [6]–[9]. Failure in drivetrain components
is currently listed among the most problematic failures dur-
ing the operational lifetime of a wind turbine. In particular,
gearbox-related failures are responsible for more than 20% of
the downtime of wind turbines. Although the expected lifetime
of gearboxes is usually advertised as 20 years, in practice,
gearboxes usually need to be replaced every 6 to 8 years
[10], [11].
3778 IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 50, NO. 6, NOVEMBER/DECEMBER 2014

TABLE I
M ODEL PARAMETERS OF GRC W IND T URBINE

Further insights into wind turbine drivetrain dynamics will

be helpful in understanding the global dynamic response of a
wind turbine as well as in designing and preserving its internal
drivetrain components. Thus far, however, in most of the afore- Fig. 1. Schematic diagram of a DFIG.
mentioned CAE tools, the drivetrain model is reduced to a few
(mostly two) degrees of freedom, resulting in restricted detail in TABLE II
PARAMETERS OF THE 750-kW DFIG
describing its complex dynamic behavior. Although researchers
have developed dynamic models for wind turbine drivetrains
with various levels of fidelity [12]–[21], these studies do not
provide direct insights on the dynamic interactions between
the drivetrain and other components of the entire wind turbine.
Recent studies in [20] and [21] take a decoupled approach in
which the global turbine response was first simulated using
an aeroelastic CAE tool (i.e., HAWC2). After the simulation,
the resulting loads and motion trajectories of the rotor as well
as the nacelle were used as inputs to a high-fidelity model of
the drivetrain to simulate its internal dynamic behavior. Thus,
this approach fails to capture the influence of the drivetrain
dynamics onto the overall turbine response.
This paper aims to investigate the dynamic response of a
wind turbine drivetrain by enhancing the capability of FAST
through the integration of a dynamic model of the drivetrain.
The drivetrain model is built using Simscape [22]. This tool
extends the FAST capability in the MATLAB/Simulink envi-
ronment, which can help in the drivetrain design as well as the
verification of active control strategies to mitigate the drivetrain
loads [23], [24].
This paper is organized as follows. Section II highlights
the wind turbine properties modeled in this paper. Section III
Fig. 2. Modular drivetrain configuration of wind turbine.
describes the different fidelity levels of the drivetrain model
investigated in this paper. Possible resonant frequencies of the
drivetrain are computed based on the eigenfrequency analysis the power electronics, an average model of the ac/dc/ac con-
in Section IV. The results are verified against eigenfrequencies verter is used in this paper, in which the power electronic de-
published in the literature and obtained from experiments. vices are replaced by controlled current sources. The induction
Section V discusses the integration of the developed model into generator model is based on a commercial wind turbine of the
the FAST CAE tool. Finally, concluding remarks are given in same rating available in the market. Table II summarizes the
Section VI. key parameters of the generator and the converter, which are
used to build the variable-speed WTG model in the Simscape/
SimPowerSystems environment.
II. W IND T URBINE D ESCRIPTION
For wind speeds below 12.5 m/s (the rated wind speed of this
The turbine modeled in this paper is based on the Gearbox turbine), the output power of this WTG is controlled to track
Research Collaborative (GRC) turbine [25] at the National the maximum power coefficient (CP max ) while maintaining
Renewable Energy Laboratory’s National Wind Technology constant pitch angle at its optimum (−3.5◦ for this turbine). In
Center. Table I summarizes the important properties of this variable-speed operation, the rotor speed ωrot is proportional to
turbine. the wind speed Vw
The GRC wind turbine originally employed two fixed-speed
wind turbine generators (WTGs). To model variable-speed λopt Vw
ωrot = (1)
operation, the WTG is modeled using a doubly fed induction Rrot
generator (DFIG), shown in Fig. 1. Because the mechanical
components of the drivetrain have much slower dynamics than where λopt is the desired optimum tip-speed ratio.
GIRSANG et al.: GEARBOX AND DRIVETRAIN MODELS TO STUDY DYNAMIC EFFECTS OF MODERN WIND TURBINES
Fig. 3. Five-mass model in SimDriveline to study dynamics of wind turbine drivetrain with fixed-speed induction generator.
Fig. 4. Two-mass model of wind turbine drivetrain.
III. D RIVETRAIN M ODELING
A. Five-Mass Model
This paper focuses on a commonly used modular drivetrain
configuration in operating turbines [12]. Fig. 2 shows the build-
ing blocks of the configuration. In this turbine, the multistage
gearbox consists of a planetary (epicyclic) gear set at the low-
speed side followed by two parallel gear sets assembled using
two intermediate shafts.
Fig. 3 shows the five-mass model developed in the Simscape/
SimDriveline environment. A fixed-speed generator has an
electrical torsional stiffness between the air gap magnetic field
and the generator rotor. This stiffness behaves as a spring to
the inertial reference frame of the drivetrain, which provides a
restoring torque to the rest of the drivetrain. Such stiffness arises
because of tight allowable speed variation in the fixed-speed
turbine. Effects of this stiffness are prominent in the transient
response of the generator (e.g., during start-up). Parameters
for the five-mass model, including the electrical stiffness, are
available in [24]. For a variable-speed generator, this restoring
effect does not exist, and the drivetrain model shown in Fig. 3
will have a free boundary condition on the other side of the
generator.
B. Two-Mass Model
Fig. 4 illustrates the configuration of the two-mass model
commonly used to model the drivetrain dynamics in wind
turbine aeroelastic CAE tools, such as FAST [1]. Inputs into
the model are the five parameters Jrot , keff , ceff , N , and
Jeff /N 2 . The generator electrical torsional stiffness is generally
not required in CAE tools because this stiffness is inherent to
the generator model used for the analysis. Parameters of the
3779
two-mass model can be derived from the five-mass model as
follows:

Jeff = JP C +N12 JS +JG1 +N22
 
× JG2 +JG3 +N32 (JG4 +Jgen ) (2)
1 1 1 1 1
= + 2 + + .
keff kLSS N1 k 1 (N1 N2 )2 k2 (N1 N2 N3 )2 kHSS
(3)
The effective drivetrain torsional damping ceff can be deter-
mined experimentally through several braking events [12].
The two-mass model cannot represent the different possible
drivetrain configurations because the rest of the drivetrain is
represented by one spring and one mass. Furthermore, as later
presented in this paper, this model has limitations in providing
insights on possible resonant excitations of the drivetrain as
well as in analyzing the loads experienced by each of the
drivetrain components.
C. Pure Torsional Model of Gearbox
In the two previously described drivetrain models, the mesh-
ing gear is modeled as an ideal static gain for mechanical
power (i.e., torque and speed) transmission. In reality, the gear
transmission error, which is defined as the difference between
the actual and ideal angular positions of the rotating gear caused
by the gear elastic deformation, contributes to the dynamics
of the pair meshing gears. This phenomenon contributes to the
definition of gear meshing stiffness. Several studies have been
performed to investigate the dynamics of the gearbox using
flexible multibody models built in various software packages
[12]–[21]. These studies, however, focus on the internal dynam-
ics of the gearbox, and the developed model packages are not
readily compatible with available wind turbine aeroelastic CAE
tools.
In this paper, a purely torsional model of the gearbox
with constant meshing stiffness is built in the Simscape/
SimDriveline environment. The model development and anal-
ysis on both planetary and parallel gear stages will be discussed
in this section, followed by some remarks on the integration
and simulation of this model with the aeroelastic CAE tool of
interest (i.e., FAST) in Section V.
3780 IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 50, NO. 6, NOVEMBER/DECEMBER 2014
Fig. 5. (a) Parallel gear stage, (b) dynamic model representation, and
(c) model representation in Simscape/SimDriveline.
Fig. 6. Planetary gear set with three planet gears.
1) Parallel Gear Stage: Fig. 5(a) shows a parallel gear
set, which is a torque reducer, commonly employed in wind
turbine drivetrains. Fig. 5(b) represents its flexible equivalent,
in which the meshing stiffness acts along the line of action of
the meshing gears. This meshing stiffness kmesh , with respect
to the input gear, can be represented as [15]
kmesh = kgear (rb1 cos β)2
where the gear tooth stiffness kgear can be determined accord-
ing to standards [26], [27].
2) Planetary Gear Stage: Fig. 6 shows a planetary gear set
with three planet gears, which is a similar configuration to the
one installed in the GRC turbine.
The rotational input is from the carrier of the planetary
gear stage, which provides rotational motion through the planet
gears, and finally to the sun gear. The ring gear is modeled to
have flexible coupling with the fixed gear housing. Flexibility
between the meshing planet and ring gears and between the
meshing planet and sun gears can be modeled similar to that
of a parallel gear set using (4), as shown in Fig. 5(b). Fig. 7
shows a purely torsional model of a planetary gear set built
in the Simscape/SimDriveline environment. This model can be
adapted for any M equispaced-planet gear set.
IV. E IGENFREQUENCY A NALYSIS AND VALIDATIONS
Eigenfrequencies of a system can be found either analytically
or numerically. Equations of motion of a wind turbine drivetrain
Fig. 7. Torsional model of planetary gear stage with M planet gears in
Simscape/SimDriveline.
are used to compute the eigenfrequencies in [18] and [24].
However, this approach becomes inconvenient as the order of
the dynamic model increases. The developed drivetrain model
in SimDriveline allows the usage of MATLAB/Simulink basic
packages to numerically estimate the eigenfrequencies.
The resulting eigenfrequencies were validated at both com-
ponent level (i.e., the gearbox) and integrated drivetrain level.
The eigenfrequencies of planetary gear sets were validated
against those published in the literature, while those of the
overall GRC drivetrain were validated against the results from
field measurements.
A. Eigenfrequencies of Purely Torsional Gearbox Model
(4) While some works have reported experimental measure-
ments of gearbox eigenfrequencies [13], [28], it is difficult to
justifiably replicate the results as the mechanical properties of
the gearboxes are not publicly available. Thus, to validate the
developed purely torsional gearbox, eigenfrequencies of the
SimDriveline gear set models are compared with those reported
in [15].
It is important to note that the gear parameters used in
the component-level validation in this section is based on the
ones originally used in [29], which are different from the ones
implemented in the GRC drivetrain. Eigenfrequencies of this
planetary gear set have been evaluated in other published works
[15], [30]–[33]. In [15], the analysis was performed on the
planetary gear sets with different numbers of planet gears.
Estimating the eigenfrequencies using the SimDriveline
models was done by first defining the excitation input and the
output to be monitored. For analysis and verification, the input
for the gearbox model was the torque to the carrier, while the
output was the angular velocity of the sun gear as illustrated
in Fig. 8. Based on the input–output configuration, MATLAB/
Simulink Control System Toolbox was used to compute the
linear state-space representation of the model, the frequency
GIRSANG et al.: GEARBOX AND DRIVETRAIN MODELS TO STUDY DYNAMIC EFFECTS OF MODERN WIND TURBINES
Fig. 8. Input–output configuration on SimDriveline model for eigenfrequency analysis.
Fig. 9. FRF of three-planet planetary gear stage for the gearbox presented
in [15].
TABLE III
C OMPARISONS W ITH E IGENFREQUENCIES R EPORTED IN [15]
response function (FRF) of which has been used to indicate
the eigenfrequencies. The implementation of this analysis is
detailed in [34]. Fig. 9 shows the FRF of the planetary gear set
that has three planet gears, where the sharp peaks correspond
to the response amplifications at the gearbox eigenfrequencies.
The comparisons for different numbers of planet gears show
good agreement with the results presented in [15] and are
summarized in Table III.
B. Eigenfrequencies of Overall Drivetrain Model
Ultimately, all necessary component models, including the
purely torsional gearbox model discussed earlier, were inte-
grated, and the parameters were set to build the overall GRC
3781
drivetrain model. Eigenfrequency analysis for the drivetrain
was performed with the torque to the turbine rotor as the
input and the angular position of the generator as the output.
The resulting eigenfrequencies are compared herein with the
frequency components of several measured drivetrain transient
responses [20]. The GRC turbine was equipped with a torque
transducer to measure the operating drivetrain loads. Due to the
limited sampling rate of 100 Hz, the measured response can
only be used to validate the first drivetrain eigenfrequency.
The first two transient responses shown in Fig. 10 are
attributed to the generator. Hence, the analysis on the over-
all drivetrain model included the electrical torsional stiffness.
Table IV summarizes the drivetrain eigenfrequencies both with
and without the electrical torsional stiffness. Fig. 10(a) shows
the torque measured on the low-speed shaft (i.e., the rotor-
opposing torque) during a generator start-up. The dotted box
highlights the transient response oscillating at the drivetrain
eigenfrequencies, which is revealed to be around 0.78 Hz.
Fig. 10(b) shows another transient response when the drive-
train was upshifted to a higher speed generator. It reveals an
eigenfrequency of 0.68 Hz. These two frequency readings are
very close to the one calculated from the SimDriveline model,
which is 0.84 Hz. The second mode of 4.62 Hz was not apparent
in the field measurements due to two possible reasons. First,
the energy stored in the second torsional mode is much smaller
than that stored in the first one. Second, the second mode is
attributed to the high-speed shaft. As discussed in the following
section, excitations from the high-speed shaft get attenuated
at the low-speed shaft side; thus, it may not appear as a
dominant component in a measurement that contains noise. The
experimental setup did not implement any torque measurements
at the high-speed shaft.
Fig. 11 shows a transient response, which is not attributed
to the generator. In this event, a braking torque is applied
mechanically through the disk brake to stop the turbine. The
frequency decomposition of the transient rotor-opposing torque
reveals a frequency of 1.76 Hz, which is close to an eigen-
frequency of 1.83 Hz predicted from the SimDriveline model.
These results illustrate the validity of the SimDriveline model
to predict the eigenfrequencies of both an independent gearbox
and an integrated wind turbine drivetrain.
If the overall drivetrain is simplified to a two-mass model,
the only nonzero eigenfrequency can be calculated using the
parameters of (2) and (3) as
  
1 1 1
fn = keff + . (5)
2π Jrot Jeff
3782 IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 50, NO. 6, NOVEMBER/DECEMBER 2014
Fig. 10. Transient rotor torque from field measurements during generator
transient events: (a) Start-up and (b) upshift.
TABLE IV
E IGENFREQUENCIES OF GRC D RIVETRAIN W ITH
T ORSIONAL G EARBOX M ODEL
The resulting eigenfrequency using the two-mass model for
the GRC drivetrain is 2.32 Hz, which is quite different from
the first nonzero drivetrain eigenfrequency of 1.83 Hz predicted
earlier. This discrepancy can create a significant difference in
predicting the loads experienced by the drivetrain.
V. M ODEL I NTEGRATION
Fig. 12 illustrates the proposed strategy to integrate the
described drivetrain models into the two-mass model inherent
Fig. 11. Transient rotor torque from field measurements during braking event.
inside the FAST CAE tool. For simplicity, the flexible modes
of the other turbine components modeled inside FAST, such as
those of the blades and tower, are not depicted in the schematic
diagram in Fig. 12.
In FAST, the two-mass drivetrain model is reduced to a
single-mass model consisting of solely the rotor and the rigid
shaft (as shown in the bottom part of Fig. 12). This is done
by deactivating the flexibility of the drivetrain (simulating rigid
transmission) and setting the gear ratio and the generator inertia
to unity and zero, respectively. The rotor equation of motion can
be expressed as
Jrot αrot = Qaero − Qopp . (6)
FAST internally calculates the input aerodynamic torque
Qaero from the defined wind profile but does not provide this
torque as an output. However, as the rotor acceleration αrot
is an available FAST output, the aerodynamic torque Qaero
can be reconstructed using (6) to be one of the inputs to the
external drivetrain model. In this process, the rotor inertia Jrot
is assumed constant and replicated in the Simscape drivetrain
model. The rotor inertia is connected to the flexible low-speed
shaft, the purely torsional gearbox model, the high-speed shaft,
and the generator inertia. The electrical machine and grid
model will take the generator speed and provide the generator
electromagnetic torque to the drivetrain.
The rotor-opposing torque Qopp is required as an input to the
FAST drivetrain model as well as to calculate the aerodynamic
torque Qaero in (6). In SimDriveline, this rotor-opposing torque
can be retrieved by utilizing the torque sensor element behind
the built rotor body. In general, torque, velocity, and angular-
position sensor elements can be placed flexibly within the
Simscape drivetrain model to monitor the response of the
drivetrain under various load conditions.
In the remainder of this section, simulation results showing
the effectiveness of the purely torsional gearbox model un-
der different transient load cases are presented. Simulations
using the FAST wind turbine CAE tool were conducted in
the Simulink environment. In the simulations, all available
wind turbine flexible modes in FAST—including that of the
blades, tower, and drivetrain—were activated. Zero damping is
GIRSANG et al.: GEARBOX AND DRIVETRAIN MODELS TO STUDY DYNAMIC EFFECTS OF MODERN WIND TURBINES 3783

Fig. 12. Proposed schematic of integrating the Simscape drivetrain model into the FAST aeroelastic CAE tool.

Fig. 13. Rotor speed response at wind speed of 7.25 m/s.

defined within the drivetrain model to highlight the transient

response of the drivetrain. Aerodynamic damping computed
within FAST is the only source of damping that stabilizes the
overall drivetrain simulations. The results are compared with
those using an undamped two-mass model of the drivetrain
inherent in FAST. Fig. 14. Transient response comparison of the rotor torque.

A. Transient Response Caused by Wind Excitation

This simulation was performed under a constant wind speed
of 7.25 m/s (below the rated wind speed), and the turbine speed
was initialized to be 17.9 r/min. The start of simulation effec-
tively imparts a large step input to the system that can excite all
of the drivetrain modes, particularly during the transient period.
Fig. 13 shows the turbine rotor speed using the integrated
drivetrain model as well as the inherent FAST two-mass model.
The rotor speed steadily increases to reach the optimal tip-speed
ratio. Both models are in good agreement in the speed response
of the turbine. Fig. 14 is the transient response comparison of
the rotor torque of the two models. It reveals the excitation of
the high eigenfrequency components coming from the high-
fidelity drivetrain/gearbox model.
Fig. 15 highlights the distinction between the two models in Fig. 15. Steady-state response comparison of the rotor torque.
predicting the steady-state load response of the drivetrain. It
also shows that the integrated model is able to capture dynamic the frequency of 0.9 Hz and its harmonics come from the blade
coupling from other parts of the turbine structures. The fre- passing frequency (3P). The response of the drivetrain model
quency of 0.4 Hz comes from the tower fore-aft mode, whereas with the purely torsional gearbox is particularly high at two
3784 IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 50, NO. 6, NOVEMBER/DECEMBER 2014
Fig. 16. Electromagnetic torque excitations resulting from a voltage drop on
the grid.
times the blade passing frequency (i.e., 6P) of 1.8 Hz because
it is very close to the estimated first eigenfrequency of 1.83 Hz.
Thus, it predicts the amplification of load caused by resonance
at the wind speed of 7.25 m/s. On the other hand, the two-
mass drivetrain model estimates an eigenfrequency of 2.32 Hz,
which is at some distance from the harmonics of the blade pass
frequency and hence predicts no resonance.
B. Transient Response Resulting From Grid Excitation
Another transient load can arise due to excitations from the
grid events. One example of grid excitation is simulated in this
section to predict loads on the gearbox. A voltage drop for
0.15 s, from 100% to 90% and back to 100% of the nominal root
mean square voltage, is simulated after the turbine has reached
steady state.
As shown in Fig. 16, this voltage drop results in harmonic
torque excitations onto the drivetrain with frequencies of 50.78
and 56.15 Hz. These frequencies are inherent to the gener-
ator characteristic. It is important to note that the frequency
component of this torque excitation may cause resonances if
the frequency matches any of the drivetrain eigenfrequencies.
These resonances cannot be predicted using the standard two-
mass model because the two-mass model can predict only the
lowest eigenfrequency of the drivetrain.
Fig. 17 illustrates how this load gets transmitted to each
stage of the multistage gearbox. The torque through each shaft,
except the low-speed shaft, is shown in both time and fre-
quency domains. The high-speed shaft experiences the largest
proportion of high-frequency loads caused by grid excitations
compared to other shafts. Therefore, this shaft and the gear
set directly connected to it are most prone to failures from
fatigue in the event of grid disturbances. Sudden increases in
the generator electromagnetic torque excite the two lowest and
most dominant modes of the drivetrain (i.e., 1.83 and 154 Hz).
This torque also excites the system at its excitation frequencies
of 50.78 and 56.15 Hz, but with less dominant effect than at
the eigenfrequencies. The transmitted loads are reduced as the
shaft gets further from the source of excitation, but the most
Fig. 17. Transmitted loads onto the gears because of grid excitation in (a) time
and (b) frequency domains.
dominant drivetrain eigenfrequency of 1.83 Hz prevails during
the transient regime.
The developed high-fidelity drivetrain model can also be
used in designing the drivetrain components to preserve or
extend the life of the gearbox. In simulation analysis, this high-
speed shaft stiffness was varied to 0.1 and 10 times of the nom-
inal value to investigate its influence on the load transmitted
to the gearbox. In practice, the generator and gearbox high-
speed shaft are often connected through a mechanical coupler
to provide for the misalignment of the shafts. However, this
coupler reduces the effective stiffness between the generator
and gearbox. Higher stiffness can be achieved by reducing the
length of the high-speed shaft.
The transmitted loads through the high-speed shaft to the
gearbox were evaluated and shown in Fig. 18. A lower stiffness
(i.e., using a coupler) appears to transmit more severe loads to
the gearbox. This phenomenon is a result of reduction in the
second drivetrain eigenfrequency due to the lower stiffness and
vice versa, as shown in Fig. 18. The transmitted load gets ampli-
fied as the second eigenfrequency is brought down to 92.8 Hz,
which is close to the second harmonics of the generator electro-
magnetic torque frequencies (i.e., resonance).
GIRSANG et al.: GEARBOX AND DRIVETRAIN MODELS TO STUDY DYNAMIC EFFECTS OF MODERN WIND TURBINES
Fig. 18. Transmitted loads onto the second parallel gear stage under various
high-speed shaft stiffness values in (a) time and (b) frequency domains.
VI. C ONCLUSION
In this paper, the capability of the FAST CAE tool is
enhanced through the integration of a dynamic model of a
drivetrain built using Simscape in the MATLAB/Simulink en-
vironment. As described in this paper, the implementation of
the developed drivetrain model enables the CAE tool to be
used in a number of ways. First, the model can be simulated
under different wind and grid conditions to yield further insight
into the drivetrain dynamics, particularly in terms of predicting
possible resonant excitations. Second, the tool can be used to
simulate and understand transient loads and their couplings
across the drivetrain components. Third, the model can be used
to design the various flexible components of the drivetrain such
that transmitted loads onto the gearbox can be reduced.
ACKNOWLEDGMENT
The authors would like to thank Dr. K. Nguyen and
Dr. J. Jonkman for the fruitful discussions about integrating
FAST with the external drivetrain model.
3785
R EFERENCES
[1] J. M. Jonkman and M. L. Buhl, Jr., “FAST user’s guide,” Nat. Re-
new. Energy Lab., Golden, CO, USA, Tech. Rep. NREL/EL-500-38230,
Aug. 2005.
[2] E. A. Bossanyi, “GH bladed version 3.51 user manual,” Garrad Hassan
and Partners Ltd., London, U.K., Tech. Rep. 282/BR/010, Jun. 2003.
[3] T. J. Larsen and A. M. Hansen, “How 2 HAWC2, the user’s man-
ual,” Tech. Univ. Denmark, Roskilde, Denmark, Tech. Rep. Risø-R-1597,
Dec. 2007.
[4] J. Helsen, G. Heirman, D. Vandepitte, and W. Desmet, “The influence
of flexibility within multibody modelling of multimegawatt wind turbine
gearboxes,” in Proc. Int. Conf. Noise Vibration Eng., 2008, pp. 15–17.
[5] J. Helsen, F. Vanhollebeke, D. Vandepitte, and W. Desmet, “Optimized
inclusion of flexibility in wind turbine gearbox multibody model in view
of model updating on dynamic test rig,” in Proc. 1st Joint Int. Conf.
Multibody Syst. Dyn., May 2010, pp. 1–6.
[6] L. Hong, J. S. Dhupia, and S. Sheng, “An explanation of frequency fea-
tures enabling detection of faults in equally spaced planetary gearbox,”
Mech. Mach. Theory, vol. 73, pp. 169–183, Mar. 2014.
[7] L. Hong and J. S. Dhupia, “A time domain approach to diagnose gearbox
fault based on measured vibration signals,” J. Sound Vib., vol. 333, no. 7,
pp. 2164–2180, Mar. 2014.
[8] L. Hong and J. S. Dhupia, “A time-domain fault detection method based
on an electrical machine stator current measurement for planetary gear-
sets,” in Proc. IEEE/ASME Int. Conf. Adv. Intell. Mechatronics, 2013,
pp. 1631–1636.
[9] J. Zhang, L. Hong, and J. S. Dhupia, “Gear fault detection in planetary
gearbox using stator current measurement of ac motors,” in Proc. ASME
5th Annu. Dyn. Syst. Control Conf., 2012, pp. 673–680.
[10] S. Faulstich, B. Hahn, and P. J. Tavner, “Wind turbine downtime and
its importance for offshore deployment,” Wind Energy, vol. 14, no. 3,
pp. 327–337, 2011.
[11] L. F. Villaa, A. Reñonesa, J. R. Perána, and L. J. de Miguelb, “Statistical
fault diagnosis based on vibration analysis for gear test-bench under non-
stationary conditions of speed and load,” Mech. Syst. Signal Process.,
vol. 29, pp. 436–446, 2012.
[12] F. Oyague, “Gearbox modeling and load simulation of a baseline 750-kW
wind turbine using state-of-the-art simulation codes,” Nat. Renew.
Energy Lab., Golden, CO, USA, Tech. Rep. NREL/TP-500-41160,
Feb. 2009.
[13] J. Helsen, F. Vanhollebeke, F. D. Coninck, D. Vandepitte, and W. Desmet,
“Insights in wind turbine drivetrain dynamics gathered by validating ad-
vanced models on a newly developed 13.2-MW dynamically controlled
test-rig,” Mechatronics, vol. 21, no. 4, pp. 737–752, Jun. 2011.
[14] J. Helsen, F. Vanhollebeke, B. Marrant, D. Vandepitte, and W. Desmet,
“Multibody modelling of varying complexity for modal behaviour analy-
sis of wind turbine gearboxes,” Renew. Energy, vol. 36, no. 11, pp. 3098–
3113, Nov. 2011.
[15] J. Peeters, “Simulation of dynamic drivetrain loads in a wind turbine,”
Ph.D. dissertation, Dept. Mech. Eng., K.U. Leuven, Leuven, Belgium,
Jun. 2006.
[16] Y. Guo, J. Keller, and R. Parker, “Dynamic analysis of wind turbine
planetary gears using an extended harmonic balance approach,” in Proc.
Conf. Noise Vibration Eng., Sep. 2012, pp. 1–18.
[17] Y. Guo and R. G. Parker, “Dynamic modeling and analysis of a spur
planetary gear involving tooth wedging and bearing clearance non-
linearity,” Eur. J. Mech. A, Solids, vol. 29, no. 6, pp. 1022–1033,
Nov./Dec. 2010.
[18] M. Todorov, I. Dovbrev, and F. Massouh, “Analysis of torsional oscillation
of the drivetrain in horizontal axis wind turbine,” in Proc. ELECTROMO-
TION, Jul. 2009, pp. 1–7.
[19] Y. Xing and T. Moan, “Multi-body modelling and analysis of a planet
carrier in a wind turbine gearbox,” Wind Energy, vol. 16, no. 7, pp. 1067–
1089, Oct. 2013.
[20] Y. Guo, J. Keller, T. Moan, and Y. Xing, “Model fidelity study of dynamic
transient loads in a wind turbine gearbox,” in Proc. Windpower Conf.,
May 2013, pp. 1–4.
[21] Y. Xing, M. Karimirad, and T. Moan, “Modelling and analysis of floating
spar-type wind turbine drivetrain,” Wind Energy, vol. 17, no. 4, pp. 565–
587, Apr. 2014.
[22] Simscape User’s Guide, The MathWorks, Inc., Natick, MA, USA,
Mar. 2012.
[23] I. P. Girsang, J. S. Dhupia, E. Muljadi, M. Singh, and J. Jonkman, “Mod-
eling and control to mitigate resonant load in variable-speed wind turbine
drivetrain,” IEEE J. Emerging Sel. Topics Power Electron., vol. 1, no. 4,
pp. 277–286, Dec. 2013.
3786 IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 50, NO. 6, NOVEMBER/DECEMBER 2014
[24] G. Mandic, A. Nasiri, E. Muljadi, and F. Oyague, “Active torque control
for gearbox load reduction in a variable-speed wind turbine,” IEEE Trans.
Ind. Appl., vol. 48, no. 6, pp. 2424–2432, Nov./Dec. 2012.
[25] F. Oyague, “Gearbox reliability collaborative (GRC) description and load-
ing,” Nat. Renew. Energy Lab., Golden, CO, USA, Tech. Rep. NREL/
TP-5000-47773, Nov. 2011.
[26] Calculation of Load Capacity of Cylindrical Gears, DIN 3990, 1987.
[27] Calculation of Load Capacity of Spur and Helical Gears,
Std. ISO 6336-1, 2007.
[28] S. Gade, R. Schlombs, C. Hundeck, and C. Fenselau, “Operational modal
analysis on a wind turbine gearbox,” in Proc. Conf. Expo. Structural Dyn.,
2009, pp. 1–11.
[29] J. Lin and R. G. Parker, “Analytical characterization of the unique prop-
erties of planetary gear free vibration,” J. Vib. Acoust., vol. 121, no. 3,
pp. 316–321, Jul. 1999.
[30] R. G. Parker, V. Agashe, and S. M. Vijayakar, “Dynamic response of a
planetary gear system using a finite element/contact mechanics model,”
J. Mech. Des., vol. 122, no. 3, pp. 304–310, May 1999.
[31] J. Lin and R. G. Parker, “Sensitivity of planetary gear natural frequencies
and vibration modes to model parameters,” J. Sound Vib., vol. 228, no. 1,
pp. 109–128, Nov. 1999.
[32] J. Lin and R. G. Parker, “Natural frequency veering in planetary gears,”
Mech. Struct. Mach., vol. 29, no. 4, pp. 411–429, 2001.
[33] J. Lin and R. G. Parker, “Planetary gear parametric instability caused by
mesh stiffness variation,” J. Sound Vib., vol. 249, no. 1, pp. 129–145,
Jan. 2002.
[34] M. Singh et al., “Simulation for wind turbine generator—with FAST and
MATLAB/Simulink modules,” Nat. Renew. Energy Lab., Golden, CO,
USA, Tech. Rep. NREL/TP-5500-59195, 2013.
Irving P. Girsang (S’12) received the B.Eng. degree
in mechanical engineering from Nanyang Techno-
logical University (NTU), Singapore, in 2011. He
is currently working toward the Ph.D. degree in the
School of Mechanical and Aerospace Engineering
and Energy Research Institute at NTU. His study is
funded by the National Research Foundation (Clean
Energy) Scholarship, administered by the Energy
Innovation Programme Office Singapore.
He was a Visiting Student Researcher at the Na-
tional Wind Technology Center, Boulder, CO, USA,
from May 2012 to July 2012. His research interest is in modeling and control
for load mitigation of wind turbine drivetrain.
Jaspreet S. Dhupia (M’13) received the B.Tech.
degree in mechanical engineering from the Indian
Institute of Technology-Delhi, Delhi, India, in 2001
and the M.S. and the Ph.D. degrees in mechanical
engineering from the University of Michigan, Ann
Arbor, MI, USA, in 2004 and 2007, respectively.
He has been an Assistant Professor at Nanyang
Technological University, Singapore, since 2008,
where he is leading a research group engaged in
modeling, monitoring, and controls of electrome-
chanical drivetrains. He is an author of more than
30 peer-reviewed articles.
Prof. Dhupia has been a member of the American Society of Mechanical
Engineers (ASME) since 2006. He served on the organizing committees for the
IEEE/ASME International Conference on Advanced Intelligent Mechatronics
in 2013 and the American Control Conference in 2014.
Eduard Muljadi (M’82–SM’94–F’10) received the
Ph.D. degree in electrical engineering from the Uni-
versity of Wisconsin-Madison, Madison, WI, USA.
He was with the California State University at
Fresno, Fresno, CA, USA, from 1988 to 1992. In
1992, he joined the National Renewable Energy
Laboratory, Golden, CO, USA. He is the holder
of two patents on power conversion for renewable
energy. His current research interests are in the fields
of electric machines, power electronics, and power
systems in general with an emphasis on renewable
energy applications.
Dr. Muljadi is a member of Eta Kappa Nu and Sigma Xi and an Editor of
the IEEE T RANSACTIONS ON E NERGY C ONVERSION. He is involved in the
activities of the IEEE Industry Application Society (IAS), IEEE Power Elec-
tronics Society, and IEEE Power and Energy Society (PES). He is currently a
member of various committees of the IAS and a member of the Working Group
on Renewable Technologies and the Task Force on Dynamic Performance of
Wind Power Generation, PES.
Mohit Singh (M’11) received the M.S. and Ph.D. de-
grees in electrical engineering from The University
of Texas at Austin, Austin, TX, USA, in 2007 and
2011, respectively.
He is a Researcher with the National Renewable
Energy Laboratory, Golden, CO, USA, working on
the transmission and grid integration of renewable
energy. His research is focused on the dynamic mod-
eling of wind turbine generators (WTGs). His cur-
rent interests include modeling and testing various
applications of WTGs and other renewable energy
resources.
Dr. Singh is involved in the activities of the IEEE Power and Energy Society.
Lucy Y. Pao (M’91–SM’98–F’12) received the B.S.,
M.S., and the Ph.D. degrees in electrical engineering
from Stanford University, Stanford, CA, USA, in
1987, 1988, and 1992, respectively.
She is currently the Richard and Joy Dorf Pro-
fessor in the Electrical, Computer, and Energy En-
gineering Department at the University of Colorado
Boulder, Boulder, CO, USA. Her primary research
focus is in the control systems area, with applications
to flexible structures, atomic force microscopes, disk
drives, tape systems, and wind turbines.
Prof. Pao is a Fellow of the International Federation of Automatic Control
and the Renewable and Sustainable Energy Institute. She was a member of
the U.S. Defense Science Study Group from 2010 to 2011 and the General
Chair for the 2013 American Control Conference. She is an IEEE Control
Systems Society (CSS) Distinguished Lecturer and a recipient of the 2012 IEEE
Control Systems Magazine Outstanding Paper Award. She has been a member
of the IEEE CSS Board of Governors. She has recently given plenary lectures
at the 2011 American Society of Mechanical Engineers (ASME) International
Mechanical Engineering Congress and Exposition, the 2012 ASME Dynamic
Systems and Control Conference, and the 2013 IEEE Conference on Decision
and Control.