During Transient Events.: Article

This is a repository copy of Dynamic modelling of wind turbine gearbox bearing loading
during transient events.
White Rose Research Online URL for this paper:

http://eprints.whiterose.ac.uk/86532/
Version: Accepted Version
Article:
Bruce, T., Long, H. and Dwyer-Joyce, R. (2015) Dynamic modelling of wind turbine
gearbox bearing loading during transient events. IET Renewable Power Generation. ISSN
1752-1416
https://doi.org/10.1049/iet-rpg.2014.0194
© 2015 The Institution of Engineering and Technology. This paper is a postprint of a paper
submitted to and accepted for publication in IET Renewable Power Generation and is
subject to Institution of Engineering and Technology Copyright. The copy of record is
available at IET Digital Library.
Reuse
Unless indicated otherwise, fulltext items are protected by copyright with all rights reserved. The copyright
exception in section 29 of the Copyright, Designs and Patents Act 1988 allows the making of a single copy
solely for the purpose of non-commercial research or private study within the limits of fair dealing. The
publisher or other rights-holder may allow further reproduction and re-use of this version - refer to the White
Rose Research Online record for this item. Where records identify the publisher as the copyright holder,
users can verify any specific terms of use on the publisher’s website.
Takedown
If you consider content in White Rose Research Online to be in breach of UK law, please notify us by
emailing eprints@whiterose.ac.uk including the URL of the record and the reason for the withdrawal request.
eprints@whiterose.ac.uk
https://eprints.whiterose.ac.uk/
IET Renewable Power Generation ISSN 1752-14 16 Received on 17th June 2014
IET Journals Revised on 16th February 2015
doi: 10.1049/iet-rpg.2014.0194 Accepted 31st March 2015
Dynamic modelling of wind turbine gearbox bearing loading

during transient events
T. Bruce, H. Long*, R. S. Dwyer-Joyce
Leonardo Centre for Tribology, Department of Mechanical Engineering, The University of

Sheffield, United Kingdom
*h.long@sheffield.ac.uk
Tel: +44 (0) 114 222 7759
Fax: +44 (0) 114 222 7890
Abstract
Wind turbine gearbox bearings (WTGBs) are the most reliability critical component in wind
turbine gearboxes (WTGs) due to their high failure rate and long downtime-per-failure. Current
design methods predict bearing failure by fatigue life models. However, premature WTGB failures
have been observed by many other modes. This study presents the development of a multibody
dynamic gearbox model, used to determine maximum bearing contact stresses from laboratory
measured shaft torque data during normal operation and shutdown conditions. The model was
validated by comparing its results to other models of the 750 kW National Renewable Energy
Laboratory (NREL) test drive train by the Gearbox Reliability Collaborative (GRC). During normal
operation, the maximum contact stress experienced by the planetary stage bearings exceeded
recommended levels by 1% and during shutdown, by 15%. High speed shaft bearings also exceeded
recommended levels during shutdown, by 18%.
Keywords
Wind turbine gearbox, bearing failure, multibody, dynamic model, overload, white etching crack.
1. Introduction
Approximately two-thirds of WTG failures initiate in the bearings [1] despite best practice
manufacturing being followed [2]. A typical onshore failure takes around 250 hours to repair and 20%
of the overall lifetime downtime of a wind turbine (WT) can be expected to be caused by gearbox
failures [3], with this percentage greatly increased for offshore applications. Current standards explain
bearing failure via rolling contact fatigue [4], but do not explain the shortened lifetimes. It is clear that
the bearing failures have other root causes [3, 5].
The number of failures increases with WT size, due to larger component deflections and
misalignment [5, 6]; although NREL found that size does not affect the mode of failure [7]. European
offshore wind capacity is expected to increase by approximately 30-40 times 2011 levels by 2020 [8],
with offshore WTs being more expensive to repair than onshore.
The purpose of the model developed in this study is to determine whether maximum bearing
contact pressures are exceeding recommended levels during normal operation and manual shutdown
conditions. A dynamic gearbox model of the NREL two-speed, stall-regulated 750 kW test drivetrain
was created, using Ricardo PLC’s VALDYN software. The model was then validated against models
of the same gearbox, created independently by the NREL GRC round-robin project. Simulation
results of bearing dynamic loading are presented and maximum Hertzian contact stresses calculated.
2. Review of Recent Research
2.1. Causes of premature failure in wind turbine gearbox bearings
Overloading may be a cause of premature failure [9], occurring when the contact pressure
between a bearing raceway and rolling element is higher than the material’s yield strength.
Overloading is caused by short-term impact loading, which may arise from: fluctuating wind loads,
non-synchronisation of blade pitch (for machines with pitching blades), sudden braking, sudden grid
drops, generator/grid engagements [9], extended periods of high rotor torque, misalignment due to
gearbox component deflections, oval compression of gearbox bearings due to gearbox support
structure deflections, oval compression of blade pitch bearings causing short term torque spikes due to
delayed blade pitching [10], non-torsional loading [11], preloading on account of tight fits, excessive
drive-up on a tapered seating, extreme events leading to torque reversal (for example emergency
shutdown), or impact during fitting [12, 13]. When rolling elements are in the unloaded zone, they
can be instantaneously loaded beyond the material’s yield strength in misaligned conditions, along
one or two contact points in the load profile [9]. These periods of heavy, dynamic loading may lead to
raceway stresses exceeding 3.1 GPa [10].
Possible damage modes may include, subsurface damaged due to localised adiabatic heating [14],
or if impact loads are high enough to cause low cycle fatigue, failure may occur over a relatively low
number of load cycles [15]. Overloading may lead to the development of hard, brittle, white etching
cracks (WECs) [9, 13]. Repetitive impact loading (hammering impact), may occur during torque
reversals, leading to many overload cycles in a short time period. If WECs propagate to the raceway
surface, they may initiate failure by spalling [9] or axial cracking [16]; a mode of damage that can
lead to bearing failure within 1-20% of the L10 design life [17]. These observations indicate
overloading over short time periods is a potential cause of the premature bearing failure, validating
the need to model a WTG during transient loading.
2.2. Dynamic system modelling
Peeters et al. [18] compared three types of WTG model.
- 1: Torsional multibody models, modelled with 1 degree of freedom (DOF);
- 2: Six DOF rigid multibody models with discrete flexible elements;
- 3: Fully flexible multibody models.
Type 1 models are useful for modelling drivetrains during the early design stage, and are suitable
for investigating dynamic torque levels, but not for modelling bearing reaction forces [19]. A study of
a type 1 gearbox model used to simulate normal operation and emergency stop conditions found that
most off-axis wind loads are absorbed by the main rotor shaft bearings before entering the gearbox
[1], a finding supported by NREL [20]. Failure-by-fatigue analysis [4] was used to conclude that the
first stage planetary bearings showed the greatest amount of fatigue damage, which was within
acceptable limits. Normal stops caused no additional damage but emergency stops led to higher
damage levels, especially immediately after grid loss. The model did not predict any excess loads on
the HSS (High Speed Shaft) [1].
Type 2 models can be used to investigate the influence of bearing stiffness on drivetrain dynamics
[18] and to model planetary and parallel stages as well as complete gearboxes [19]. It is useful to use
a ‘sliced’ model for gear tooth contact [21], dividing it into many ‘spring-damper’ force elements.
Axial forces introduced by helical gears can be well modelled using this method, which is important
due to the induced moment on the planet gears, leading to shaft bending and misalignment. A 35-slice
contact model was determined to be the optimum compromise between accuracy and computational
cost [21].
Type 3 models are able to calculate internal component stresses and strains, but with high
computational expense [18, 19]. The computational cost of each model type was investigated [20, 21]
by developing a gearbox model, initially as a two mass system (rotor and generator), and
incrementally increasing its complexity, validating each stage against a tested braking event. Type 2
model cost was 400 % higher than type 1, with type 3 a further 26 % higher than type 2. It was found
that type 1 models are time effective and that accurate modelling is possible using type 2, without
using finite element analysis. A new method of multibody dynamic modelling, streamlined for
efficiently calculating bearing loading, has been developed in this study, using the software package
“VALDYN”.
3. Development of a dynamic WTG model using VALDYN

The proposed new model is compared to existing models in Table 1 and is described as modified
type 2. It is able to model variable bearing stiffness, bearing damping and uses a sliced gear mesh
stiffness model.
Type 2 – 6 DOF
Modelled Type 1 – 1 DOF Type 3 – Flexible Proposed model
rigid multibody
stiffness torsional model multibody model in this study
model
Constant (sliced
Gear mesh Constant Constant Variable
model)
Variable axial,
Variable axial,
Constant axial, radial, radial, tilt
Bearing Rigid radial, tilt (stiffness
tilt (no damping) (stiffness),
and damping)
constant damping
Shaft Discrete torsional Discrete torsional, FEA representation Discrete torsional,
Equivalent bending Equivalent bending

Discrete torsional,
Splined Torsional spline Torsional spline Torsional spline
Discrete bending,
connection connection connection connection
Discrete tilting
Structural
Rigid Rigid FEA representation Rigid
components
Outputs
Torque time
history
Detailed
reaction forces
Internal stresses
(bearings only)
and strains
Table 1: Comparison of model types 1-3 to proposed gearbox model (adapted from [19])
Figure 1a shows a conventional WT drivetrain and the coordinate system used in this study. All
components were modelled, with the exception of the bedplate, which was assumed to be rigid, and
the main shaft bearings, as it was assumed that no off-axis wind loads were transmitted to the
gearbox, an accepted assumption [1, 21]. Gearbox components were modelled in 6 DOFs, while the
rotor and the generator were modelled as point inertias, free to rotate in the Z direction. A typical
WTG is shown schematically in Figure 1b including the description of the bearings used in the
developed model. VALDYN was used to dynamically model all components at each timestep using the
equation of motion (1), where [M], [C], and [K] are the inertia, damping and stiffness matrices
representing the modelled components respectively; and represent the acceleration,
velocity and displacement of each modelled component respectively and {F} represents the calculated
forces and moments bearing supports. These matrices are described in more detail in equations (2-19).
All required input parameters are listed in Table 5 in the appendix.
(a)
(b) Position Attached Type

bearings
Low Speed PLC-A Full complement
Shaft (LSS) cylindrical roller bearing
(CRB)
PLC-B Full complement CRB
Planet Gears PL-A CRB
PL-B CRB
Intermediate ISS1-A Full complement CRB
Speed Shaft ISS1-B Tapered roller bearing
1 (ISS1) (TRB)
ISS1-C TRB
Intermediate ISS2-A CRB
Speed Shaft ISS2-B TRB
2 (ISS2) ISS2-C TRB
High Speed HSS-A CRB
Shaft (HSS) HSS-B TRB
HSS-C TRB
Figure 1: NREL gearbox layout: (a) Typical WT drivetrain configuration and coordinate system (b) Gearbox
schematic [7]
3.1. Mass and inertias of gearbox components

The mass matrix below describes the mass and inertia of each component, where M is its mass
and IX , IY and IZ are its inertia in the X , Y and Z directions respectively. Masses and inertias were
found from component data sheets where possible, or alternatively, using CAD models.
M 0 0 0 0 0
0 M 0 0 0 0
0 0 M 0 0 0 (2)
0 0 0 IX 0 0
0 0 0 0 IY 0
0 0 0 0 0 IZ
3.2. Bearings
Bearing stiffness is described by the matrix shown below [22], where the value for kZ Z is zero,
because this is the DOF representing rotation around the shaft axis (axial friction is considered later).
Constant diagonal terms (shown in bold) assume linearly varying stiffness with bearing displacement
[18]. The modified model in this study calculated varying stiffness terms from the inputted bearing
geometry, taking into account the off-diagonal terms, which link the 6 DOFs. VALDYN is able to
model both cylindrical roller bearings (CRBs) and tapered roller bearings (TRBs) in this way, using
the theory presented in [23].

kXX kXY kXZ kXX kXY 0

kYY kYZ kYX kYY 0
kZZ kZX kZY 0
(3)
kX X kX Y 0
Symmetric kY Y 0
kZ Z
The bearing loads, F x, F y and F z, and tilting moments, Mx and My, are computed at each timestep
as functions of the relative displacements x, y, z, x , y of the outer and inner raceways, i and j
respectively. Initial displacements must be zero. oi and oj are the initial axial inner and outer race
offsets; the distance from the common centre of mass of the outer race and inner race respectively in
the z direction [24]:
(4)
(5)
(6)
(7)
(8)
Acceleration and velocity are calculated at each timestep, using these displacement results.
VALDYN then calculates the forces, F x and F y, in the X and Y directions respectively and RI, the
radius at which the force acts. The loads F and moments M acting on each of the raceways are
calculated in the following form [24]:
(9)
(10)
(11)
(12)
The axial moment caused by bearing friction is calculated below, considering the coefficient of
rotational friction . Assumed values of are 0.0011 and 0.0018 for CRBs and TRBs respectively
[25].
(13)
The most difficult bearing parameter to estimate accurately is its damping coefficient c. One
method is to consider bearing damping as a fraction of its critical damping; calculated using equation
14 where k is the bearing stiffness and I is the inertia of the rolling elements [11].
(14)
Introducing the ratio of the critical damping allowed the level of the damping to be set.
(15)
Bearing stiffness values at rated torque were used to calculate constant damping coefficients in
the five restrained DOFs. Three damping coefficients were required; axial ca , radial cr and tilt ct. Axial
damping was calculated using the axial stiffness; radial damping, using the mean radial stiffness in the
x and y directions; and tilt damping, using the mean tilt stiffness in the X and Y directions.
Combining these damping values found using in equations (16-18) creating the damping matrix (19).
(16)
(17)
(18)
cr 0 0 0 0 0
0 cr 0 0 0 0
0 0 ca 0 0 0
0 0 0 ct 0 0
(19)
0 0 0 0 ct 0
0 0 0 0 0 0
3.3. Gears
As previously discussed, a ‘sliced’ gear tooth contact model was used. The higher the number of
slices, the more accurate the results are, but at computational cost. A previous study found that the
optimum number of slices is 35 [22]. This study found this number to be a good compromise between
computational cost and accuracy.
Tooth contact stiffness were assumed because the effective tooth stiffness of a pair of spur gears
is relatively independent of the tooth and gear size when standard involute tooth profiles are used
[26]. For spur gears, c = 14 N/(mm. m) and for helical gears (helix angle, = 20 ): c = 13.1
N/(mm. m). These values were adjusted appropriately for different helix angles.
Damping was approximated, using an appropriate critical damping ratio in the same manner as in
section 3.2. It was assumed that gear contact friction is zero, as it is anticipated that it will have little
effect on bearing loading.
3.4. Shafts
Non-rigid shafts were modelled by specifying connected masses and inertias, as well as the shaft
dimensions, mass matrix and bending and torsional damping. As the damping values of the shaft were
unknown, they were approximated, again by using an appropriate value for the critical damping ratio.
3.5. Planetary carrier
The planetary carrier was assumed to be a rigid body as the computational cost of using an FE
model was too high. This assumption is valid, provided that modelling unequal load sharing between
upwind and downwind planetary bearings is not required. The magnitude of the sum of the upwind
and downwind bearing reaction forces for each planetary gear will be correct. Previously calculated
scaling factors for unequal load sharing are later introduced to approximate maximum load
magnitudes.
3.6. Splined sun shaft connection
Many WT gearboxes use a splined shaft connection to allow the planetary stage sun gear to
‘float’. A floating sun gear centres itself within its planets, encouraging load sharing. The splined
connection was modelled as a rigid connection in the X, Y, Z and Z DOFs, while the X and Y
degrees of freedom were unrestrained, allowing the sun gear move up and down. As the sun gear’s
position is restrained by the three planets, it displacements in these directions are small, and the
assumption is valid.
3.7. Gearbox casing and elastomer supports
The gearbox casing was modelled under the assumption that it is rigid, using a series of zero-
mass “connection points” at each bearing location that are linked to a mass positioned at the centroid.
The casing’s movement was unrestrained and supported by two rubber mounts. The stiffness and
damping values of these rubber mounts were assumed to be constant, which is an acceptable
assumption as casing displacements are small [21].
3.8. Generator resistance torque
A simple induction generator model was used to model the generator resistance torque curve
(Figure 2) that was accurate for most operation modes, although too simple to model the WT start-up
control system [7]. Four inputs were required: synchronous generator speed ( ), HSS rated speed,
HSS Rated torque and pullout torque; given by the following equations, where PR is the pullout ratio
(the ratio of pullout torque to HSS rated torque) [7]:
(20)
Figure 2: Generator resistance torque curve [7]
3.9. Disc braking
A trial and error process was utilised, by applying different contact loads (simulating the brake
pads) to a brake disc, created using a braking element in VALDYN, which applied a frictional
retarding force to the brake disc. Contact loads were altered until the time taken to bring the rotor to a
stop from 10 rpm was the same as that taken for the tested event measured by NREL.
3.10. Maximum Hertzian contact stress calculation
After bearing dynamic loading was calculated, the maximum contact stress on the inner raceway
of each bearing was calculated. First, the maximum load experienced by the most heavily loaded
roller raceway contact Qmax was approximated using equation (21) [27], where P max is the maximum
load experienced perpendicular to the roller/raceway contact and z is the number of rolling elements.
This calculation assumed that: the internal clearance of bearings was greater than zero, the elastic
deformation of a rolling element was never negative, and the bearing roller was a perfect cylinder.
These assumptions will lead to conservative bearing contact stress estimations as they do not take into
account stress concentrations due to edge loading and roller profiling. Standard Hertzian calculations
for line contact loading (22) were used to calculate maximum contact pressure P 0, where a is the
contact area [28].
(21)
(22)
4. Application to the NREL750 kW WTG
The NREL 750 kW gearbox, shown in Figure 2, was modelled in this study, which is a
conventional three-stage design, with a low speed, three gear planetary stage, followed by two parallel
stages. The gearbox ratio is 1:81.49, with a planetary stage ratio of 1:5.71, and two parallel ratios of
1:3.57 and 1:4.00 respectively. The sun gear is attached to the intermediate stage gear, via a shaft with
a splined connection. The WT is able to generate power at two rated speeds, using either four or six
generator poles.
As previously mentioned, it has been found that WTGB failures experienced by small-scale WTs
(500-1000 kW) are also found in modern larger turbines. This confirms that findings from a study
based on the 750 kW test-turbine, can be extrapolated to larger WTGs [7], despite their larger size,
updated design standards and different control systems. This is highly advantageous because working
with smaller turbine test facilities reduces cost and increases the availability of experimental research.
Field and laboratory measurements are widely available for older turbines, whereas many modern
designs are still protected by industry.
4.1. Model validation
The model was validated to evaluate the assumptions made during its development, against the
results that were available for comparison. The first step was to compare torque distribution
throughout the gearbox with that obtained by the anonymous GRC contributors involved in the round-
robin project [20]. The modelling method used by each of the partners involved different software
packages, by which the studies were independently performed.
The VALDYN model was loaded at rated torque and the torque distribution throughout the
gearbox was calculated and found to be very close to GRC results, as shown in Figure 3a. The
maximum error from the mean result was 1.8% for the ISS2 torque level. The GRC mean result is
possibly skewed by result D, which is slightly different from all other results. This comparison
validates the model’s torque distribution, proving that gearbox ratios and component dimensions have
been setup correctly.
The loading levels in the sun/planet gear contact were tested to check the gear tooth contact
model. The VALDYN model was run at rated torque until all vibrations caused by initial system
unbalance had dissipated. Then the contact load was calculated and found to be very close to GRC
results as shown in Figure 3b. The largest percentage difference from the mean GRC result was 2.4 %
for the radial contact load. These slight differences are probably due to the use of different modelling
methods.
Finally, the bending of the main shaft under rated load was checked to validate the assumptions
used for modelling shaft bending and bearing stiffness. A model was created in which the shaft was
split into four lengths, separated by five masses, with 6 DOF mass matrices representing the section of
shaft they represented, allowing shaft deflection to be calculated in three locations between its
supporting bearings. Main shaft bearing geometry, was input into the VALDYN model, which was
loaded at rated rotor torque and the relative displacements of the main shaft was compared to NREL
results. Figure 3c shows that the results from the GRC round-robin were inconsistent. Reading
maximum displacement values from the figure allows the mean GRC result to be calculated to be
approximately 7.3 m. The calculated maximum value in this study; 8.2 m, is around 11% higher
than the mean magnitude and only the GRC result D is closer to this mean. This suggests that
assumptions were reasonable and validates the method used to model shafts and their supports
(bearing stiffness) in VALDYN.

(a)
(b)
(c)
Figure 3: Validation of model in comparison to partners A-F (20) (a) Shaft torque levels (b) Planet-sun gear
contact load levels (c) Main shaft bending displacement
4.2. Torque loading
Input torque and rotor speed data were provided by NREL for the following conditions:
- Normal Operation: 90 seconds of operation at six-pole rated speed.
- Shut-down: the rotor begins at six-pole rated speed, then the generator is disconnected and
aerodynamic braking is initiated, until the rotor reaches 10 rpm, at which point the disc brake is
engaged, bringing the rotor to a halt.
Figure 4 shows the input torque for the two operating conditions that were investigated by the
developed multibody dynamic model.
Figure 4: Input torque for normal operation and shutdown events
5. Results from simulated model and analysis
Planetary gear bearings and HSS bearings have historically been the most problematic in wind
turbines, so were selected to be analysed in detail in this study.
5.1. Normal operation
Figure 5 displays the LSS and HSS velocities over the 96 second simulation period. Ignoring
oscillations caused by initial system unbalance during the first 5 seconds, LSS speed oscillation
magnitudes of approximately 0.8 rpm compare well with NREL measured data, and the average
velocity of the simulated LSS signal was 0.13% different from NREL measured data. No comparison
data was available for HSS velocity, which oscillates either side of the generator rated speed of 1809
rpm by approximately +/-5 rpm.
Figure 5: LSS and HSS velocity during normal operation
Figure 6a shows the resultant force acting on the planetary bearings, PL-A and PL-B, supporting
each planet gear. The maximum resultant force experienced was 280 kN. Figure 6b shows the
resultant forces acting on the HSS bearings, which in the case of HSS-B and HSS-C include both axial
and radial elements, as they are TRBs, with maximum values of 63 kN, 79 kN and 15 kN for HSS-A,
HSS-B and HSS-C respectively.

a)
b)
Figure 6: Resultant forces acting on bearings roller/raceway contacts during normal operation for:
(a) Planetary bearings (b) High speed shaft bearings
5.2. Shut-down
LSS and HSS velocities during shut-down are shown in Figure 7. The simulation was stopped after 36
seconds as rotor velocity was oscillating either side of zero and maximum bearing loading had been
experienced. Figure 8a shows that planetary bearing loading peaks after 10 seconds, when the
generator is switched offline, then the maximum loading of 358 kN is experienced when the disc
brake is engaged at 26 seconds. Figure 8b shows that the HSS bearings experience maximum resultant
force magnitudes of 110 kN for HSS-A, 102 kN for HSS-B and 63 kN for HSS-C under braking.
Figure 7: LSS and HSS velocity during shut-down

a)
b)
Figure 8: Resultant forces acting on planetary bearings roller/raceway contacts during shut-down for:
(a) Planetary bearings (b) High speed shaft bearings
5.3. Maximum Hertzian contact pressures
Maximum inner race contact pressures were calculated because most WTGBs fail on the inner
race. As previously mentioned, the developed dynamic model in this study has not taken into account
unequal load sharing between the upwind and downwind planetary bearings. Previous studies [29-31]
have found that upwind planetary bearings experience considerably higher loading; up to 1.3 times
higher [30]. This factor is taken into account when calculated planetary bearing contact stress.
Tables 2 and 3 show calculated maximum Hertzian stress experienced for normal operation and
shut-down respectively. Stresses are presented as a percentage of the maximum allowable contact
stress at Miner’s sum dynamic equivalent bearing load (P max) as listed in current wind turbine design
standards [32]; 1,500 MPa for low speed planetary bearings and 1,300 MPa for HSS bearings.
PLA PLB PLC HSSA HSSB HSSC

P (kN) 277 322 278 63 78 15
p0 (MPa) 1410 1520 1410 1170 688 305
% of Pmax reached 94 101 74 90 83 20
Table 2: Maximum bearing contact stresses during normal operation
PLA PLB PLC HSSA HSSB HSSC

P (kN) 412 408 345 110 102 63
p0 (MPa) 1720 1710 1570 1540 788 618
% of Pmax reached 115 114 105 118 61 48
Table 3: Maximum bearing contact stresses during shut-down
6. Discussion
A dynamic gearbox model has been created and successfully validated in this study. Bearing
loading results can be assumed to be accurate, taking into account the following limitations of the
model:
- Inability to model load sharing between planet gear bearing pairs;
- Inability to model off-axis loads (Z DOF) from the rotor and generator;
- Inability to model internal component deflections.
It can be seen that planetary loading is fairly evenly shared between the three planetary gears,
during both shutdown and normal operation, suggesting that the splined shaft connection does
promote load sharing in the planetary stage. Planetary bearings are loaded to higher magnitudes than
HSS bearings during both operating regimes, exceeding maximum the recommended contact stress by
15% during shutdown and by 1% during normal operation. Bearing HSSA on the HSS is also loaded
to a high percentage of the recommended level during normal operation (90%) and significantly
exceeds it during shutdown (by 18%). It is anticipated that extreme operating conditions, such as
emergency shutdown, would further increase contact stresses.
The TRB pair, HSS-B and HSS-C do not come close to exceeding recommended stress levels.
This could be because TRBs are designed to withstand high axial loading, which was not experienced
during these simulations. Radial loading is shared between the two tapered bearings, whereas axial
load is only supported by one of the two, as HSS-B is takes the positive axial load, while HSS-C takes
the negative. Loading on HSS-B is considerably higher than HSS-C. This is likely to be because HSS-
B takes a greater share of the load in the radial direction, to a greater extent than HSS-C, because it is
closer to the meshing gears.
Maximum bearing loading conditions were experienced under braking during manual shutdown,
inducing greater transient bearing loading than that during normal operation. The shutdown event also
produced large torque reversals in the gearbox, visible by the highly fluctuating bearing loading
during braking. This would likely cause hammering impact between the roller and raceway. It is
anticipated that during emergency stop events, this will be experienced at higher levels.
Another consideration is that sub-surface material defects in the bearing steel microstructure act
as stress raisers [9], locally increasing the stress that the material experiences and likely causing
plastic deformation in their close proximity. Further work will be carried out to confirm this and to
identify the loading levels at which it occurs.
This paper has assumed that all rolling elements are identically sized, which in reality is not the
case and any variation may have large implications on load sharing [33]. Bearing misalignment and
resulting roller edge loading has not been considered and would lead to a considerably reduced
contact area and therefore a considerably increased contact pressure. As previously mentioned, rollers
and raceways have been assumed to be perfect cylinders under ideal Hertzian line contact, which
equally distributes loading over the length of the line contact. The small increase in contact pressure
caused by roller profiling has not been considered.
This study has shown that under normal operating conditions such as shutdown, the bearing
loading exceeds recommended values both on the high speed shaft and the low speed planetary stage,
using a simulation method that likely underpredicts contact stresses. During extreme events such as
emergency shutdown or grid disconnection due to grid faults, the bearing loading is very likely to be
higher. As a result the bearings used in this gearbox are thought to be undersized and unable to
operate at rated torque and during manual shutdown for their design lifetime.
7. Conclusions
A multibody dynamic model of a 750 kW wind turbine gearbox has been successfully developed,
validated and used to simulate normal operation and manual shutdown. Bearing loading variations in
the time domain have been calculated and analysed. The model found that the gearbox bearings were
undersized and recommended contact stresses were exceeded on four of the six analysed bearings:
1. It was found that during normal operation, the maximum contact stress on one of the planetary
bearing inner raceways exceeded the maximum recommended level by 1 %.
2. During the braking event in the wind turbine shutdown procedure, contact stresses in the
planetary bearings exceeded the maximum recommended level in all bearings and by a maximum
of 15% in one of the bearings.
3. The upwind high speed shaft bearing HSS-A, experienced contact stresses 18% higher than
recommended levels during manual shutdown.
4. In the NREL 750 kW test wind turbine, gearbox bearings are operating at contact stress levels
higher than the recommended values during normal operating conditions and therefore may
accumulate fatigue damage when operating at rated torque and when undergoing manual
shutdown.
Acknowledgements
The authors would like to thank Ricardo for the provision of the dynamic modelling package
“VALDYN” and the NREL GRC for the provision of NREL 750 kW Gearbox design and testing
data.
References
1. Scott, K., S, Infield, D., Barltrop, N., Coultate and J., Shahaj, A. (2011) Effects of extreme and
transient loads on wind turbine drive trains, In: 50th AIAA Aerospace Sciences Meeting Including the
New Horizons Forum and Aerospace Exposition. American Institute of Aeronautics and Astronautics.
2. Musial, W., Butterfield, S. and McNiff, B. (2007) Improving wind turbine gearbox reliability,
National Renewable Energy Laboratory.
3. Ribrant, J. and Berting, L.M. (2007) Survey of failures in wind power systems with focus on
Swedish wind power plants during 1997-2005, Transactions on Energy Conversion, 22(1), 167-173.
4. British Standards Institution (2008) BS ISO 281:2007: Rolling Bearings - Dynamic load
ratings and rating life.
5. Tavner, P. J., Xiang, J. and Spinato, F. (2007) Reliability analysis for wind turbines, Wind
Energy, 10(1), 1-18.
6. Jain, S. and Hunt, H. (2011) A dynamic model to predict the occurance of skidding in wind
turbine gearbox bearings, Journal of Physics, 305(1).
7. Oyague, F., (2009) Gearbox modeling and load simulation of a baseline 750-kW wind turbine
using state-of-the-Art simulation codes. Technical Report, National Renewable Energy Laboratory,
NREL/TP-500-41160 .
8. Commision of the European Communities (2008) Offshore Wind Energy: Action needed to
deliver on the Energy Policy Objectives for 2020 and beyond. Technical Report, COM(2008) 768.
9. Evans, M. H. (2005) White structure flaking (WSF) in wind turbine gearbox bearings: effects
of 'butterflies' and white etching cracks (WEC), Materials Science and Technology, 28(1), 3-22.
10. Stadler, K. and Stubenrauch, A. (2013) Premature bearing failures in industrial gearboxes.
Technical Repoert, SKF, ATK 2013.
11. SKF Bearings (1994) Bearing failures and their causes. Technical Report, Product
information 401.
12. Sankar, S., Nataraj, M. and Raja, V. (2012) Failure analysis of bearing in wind turbine
generator gearbox, Journal of Information Systems and Communication, 3(1), 302-309.

13. Rosinski, J. and Smurthwaite, D. (2010), Toubleshooting Wind Gearbox problems. Article,
Gear Solutions.
14. Luyckx, J. (2011) WEC failure mode on roller bearings. Presentation at Wind Turbine
Tribology Seminar, Hansen Transmissions.
15. Cambell, F. (2008). Elements of Metallurgy and Engineering Alloys. Ohio: ASM
International. 243 – 262.
16. Errichello, R., Budny, R. and Eckert, R. (2013) Investigations of bearing failures associated
with white etching areas (WEAs) in wind turbine gearboxes, Tribology Transactions, 56(6), 1069-
1076.
17. Greco, A., Sheng, S., Keller, J. and Erdemir, A. (2013) Material Wear and Fatigue in Wind
Turbine Systmes, Wear, 302(1-2), 1583-1591.
18. Peeters, J. L. M., Vandepitte, D. and Sas, P. (2005) Analysis of Internal Drive Train
Dynamics in a Wind Turbine, Wind Energy, 9(1-2), 141-161.
19. Helsen, J., Vanhollebeke, F., Marrant, B., Vandepitte, D. and Desmet, W. (2011) Multibody
modelling of verying complexity for model behaviour analysis of wind turbine gearboxes, Renewable
Energy, 36(11), 3098-3113.
20. Oyague, F. (2008) Progressive dynamical drive train modeling as part of NREL Gearbox
Reliability Collaborative, Technical Report, National Renewable Energy Laboratory, NREL/CP-500-
43473.
21. Link, H., LaCava, W., van Dam, J., McNiff, B., Sheng, S., Wallen, R., McDade, M., Lambert,
S., Butterfield, S., and Oyague. F. (2011) Gearbox Reliability Collaborative project report: Findings
from phase 1 and phase 2 testing. Technical Report, National Renewable Energy Laboratory,
NREL/TP-5000-51885.
22. LaCava, W., Guo, Y., Xing, Y., and Moan, T. (2012) Determining wind turbine gearbox
model complexity using measurement validation and cost comparison, Conference Paper, National
Renewable Energy Laboratory, NREL/CP-5000-54545.
23. Houpert, L. (1997), A uniform analytical approach for ball and roller bearings calculations,
Journal of Tribology, 119(4), 851-858.

24. Ricardo PLC (2012), VALDYN User's Manual.
25. SKF (2003) General Catalogue.
26. International Organization for Standardization (2006) ISO 6336-3: Calculation of load
capacity of spur and helical gears.
27. NSK (2009) Bearing internal load distribution and displacement. Chapter from Catalogue:
Rolling Bearings.
28. Hertz, H. (1881), On the contact of elastic solids, Math, 92, (156–171), Translated and
reprinted in English in Hertz's Miscellaneous Papers, Reine Angew, J. (1896) Macmillan & Co.,
London, UK.
29. Guo, Y., Keller, J. and LaCava, W. (2012) Combined effects of gravity, bending moment,
bearing clearance, and input torque on wind turbine planetary gear load sharing. Conference Paper,
National Renewable Energy Laboratory, NREL/CP-5000-55968.
30. Lacava, W., Keller, J, Mcniff, B. (2012) Gearbox Reliability Collaborative: Test and model
investigation of sun orbit and planet load share in a wind turbine gearbox, Conference Paper,
National Renewable Energy Laboratory, NREL/CP-5000-54618.
31. Lacava, W., Xing, Y., Marks, C., Guo, Y. and Moan, T. (2013) Three-dimensional bearing
load share behavior in the planetary stage of a wind turbine gearbox, IET Renewable power
Generation, 7(4), 1752-1416.
32. International Organisation for Standardization (2012) IEC 61400-4:2012: Wind turbines --
Part 4: Design requirements for wind turbine gearboxes.
33. Yu, S., Wang, D., Dong, H. and Wang, B. (2013) A New Method for Determining Load
Distributions among Rollers of Bearing with Manufacturing Errors, Journal of Mechanical
Engineering Science, 227(11), 2402-2415.

Appendix
Planet Gears Shafts Bearings
Rotor Brake Casing Generator
carrier (All gears) (All shafts) (All bearings)
Peak normal Mass and Mass and
Mass Dimensions No. of teeth Dimensions Axial damping
force inertia (6 DOF) inertia of
Inner radius of generator
Young's Casing
Gear module Radial damping disc contact rotor (6
Moment Mass and modulus dimensions
area DOF)
of inertia inertia (6
Addendum
(z') DOF) Shear
modification Tilt damping
modulus
Outer radius of Elastomer
Mass and disc contact mount stiffness
Normal
inertia (6 Coefficient of area (6 DOF)
pressure
DOF) rotational friction
angle
Spline
Number of Static friction
Helix angle dimensions
rolling elements coefficient Elastomer
mount damping
Bending
Rolling element (6 DOF)
TIF diameter damping
diameter Coulomb
about X
friction
Bearing bore coefficient
Tip diameter Bending diameter
damping
Tooth face about Y Bearing outer Stribeck
width diameter velocity
Normal
Axial Rolling element Mass of brake
backlash
damping contact length disc
allowance
Contact Torsional
Contact angle
stiffness damping
Damping
coefficient
Friction
coefficient
Mass and
inertia (6
DOF)
Table 4: Required parameters to create the gearbox model

During Transient Events.: Article

Uploaded by

Copyright:

Available Formats

During Transient Events.: Article

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

During Transient Events.: Article

Uploaded by

Copyright:

Available Formats

This is a repository copy of Dynamic modelling of wind turbine gearbox bearing loading

during transient events.

White Rose Research Online URL for this paper:

Version: Accepted Version

Dynamic modelling of wind turbine gearbox bearing loading

T. Bruce, H. Long*, R. S. Dwyer-Joyce

Leonardo Centre for Tribology, Department of Mechanical Engineering, The University of

recommended levels during shutdown, by 18%.

the bearing failures have other root causes [3, 5].

with offshore WTs being more expensive to repair than onshore.

2. Review of Recent Research

2.1. Causes of premature failure in wind turbine gearbox bearings

raceway stresses exceeding 3.1 GPa [10].

the need to model a WTG during transient loading.

2.2. Dynamic system modelling

Peeters et al. [18] compared three types of WTG model.

- 1: Torsional multibody models, modelled with 1 degree of freedom (DOF);

- 2: Six DOF rigid multibody models with discrete flexible elements;

- 3: Fully flexible multibody models.

the HSS (High Speed Shaft) [1].

3. Development of a dynamic WTG model using VALDYN

Equivalent bending Equivalent bending

representing the modelled components respectively; and represent the acceleration,

All required input parameters are listed in Table 5 in the appendix.

(b) Position Attached Type

3.1. Mass and inertias of gearbox components

the theory presented in [23].

kXX kXY kXZ kXX kXY 0

the z direction [24]:

calculated in the following form [24]:

computational cost and accuracy.

effect on bearing loading.

3.5. Planetary carrier

3.6. Splined sun shaft connection

3.7. Gearbox casing and elastomer supports

assumption as casing displacements are small [21].

3.8. Generator resistance torque

(the ratio of pullout torque to HSS rated torque) [7]:

Figure 2: Generator resistance torque curve [7]

3.9. Disc braking

3.10. Maximum Hertzian contact stress calculation

contact area [28].

4. Application to the NREL750 kW WTG

designs are still protected by industry.

4.1. Model validation

packages, by which the studies were independently performed.

been setup correctly.

(bearing stiffness) in VALDYN.

4.2. Torque loading

- Normal Operation: 90 seconds of operation at six-pole rated speed.

engaged, bringing the rotor to a halt.

developed multibody dynamic model.

Figure 4: Input torque for normal operation and shutdown events

5. Results from simulated model and analysis

turbines, so were selected to be analysed in detail in this study.

5.1. Normal operation

rpm by approximately +/-5 rpm.

Figure 5: LSS and HSS velocity during normal operation

HSS-B and HSS-C respectively.