Académique Documents
Professionnel Documents
Culture Documents
DOI: 10.1002/we.2150
RESEARCH ARTICLE
KEY W ORDS
1 | I N T RO D U CT I O N
The cost of energy from wind has declined tremendously over the past 2 decades1 owing to a combination of lower capital costs, higher availability
and production, and more efficient operation2,3; however, wind power plant operation and maintenance costs are frequently higher than antici-
pated,4 and a significant portion of these costs is related to drivetrain reliability.5 In 2007, the US Department of Energy established the Gearbox
Reliability Collaborative (GRC) with the goal of understanding the root causes of premature gearbox failures and improving gearbox reliability.6 To
date, the GRC has focused on testing and modeling a 750‐kW drivetrain, including the dedicated design and testing of 2 gearboxes. The GRC has
led to major insights relating to the detrimental effect of rotor moments on planetary load sharing, the predicted fatigue life in high‐torque condi-
tions, and the risk of planetary bearing sliding in low‐torque conditions.7
Although planetary gear and bearing failures attract much attention because of their high repair costs, the most commonly damaged compo-
nents in wind turbine gearboxes are the rolling element bearings in the high‐speed parallel stages.8 The damage is dominated by axial cracks, also
commonly called white etching cracks (WECs), rather than classic rolling contact fatigue.9 The term white etching refers to the appearance of the
steel microstructure when the bearing cross sections are polished, etched with chemicals, and examined under reflected light. This mode of failure
can occur at 5% to 20% of the predicted design life and has been observed in many industries, bearing locations, bearing types, bearing compo-
nents, and steel types.10-12 Theories about the root causes of WECs include corrosion, hydrogen‐induced embrittlement from lubricant decompo-
sition or water contamination, high‐stress and slip conditions, mechanical impact loading, tensile hoop stress from bearing seat form deviation, stray
electrical current, or a combination of these influencing factors. Despite the variety of theories about WEC formation, there is general agreement in
the industry that operating conditions prevalent in wind turbines—such as unsteady wind and gusts, idling, braking, misalignment, torque reversals,
Wind Energy. 2017;1–12. wileyonlinelibrary.com/journal/we Copyright © 2017 John Wiley & Sons, Ltd. 1
2 GUO AND KELLER
and grid faults—can lead to disturbed bearing kinematics, loading, and lubrication.11-14 Many of these conditions can also cause sliding of the rolling
elements in the bearings. Recently, WECs were generated on a 3‐ring‐on‐roller benchtop test rig, and a cumulative energy criterion related to load,
sliding, and run time was postulated15,16; however, at this time, there is only suspicion, rather than strong evidence, that high‐load and sliding con-
ditions in excess of the criterion occur during typical wind turbine operations.
To understand WEC‐related failures in wind turbine gearboxes, instrumentation was added to the GRC high‐speed shaft (HSS) and bear-
ings,17,18 and a series of field‐representative dynamometer tests were conducted.19 Gearbox motion and bearing loads in steady‐state operations
were examined first,20,21 followed by transient tests.22 This paper summarizes and extends these first‐of‐a‐kind, full‐scale testing and modeling
efforts to investigate operating conditions potentially related to WEC generation in wind turbine gearboxes.
2 | T ES T P R O G R A M
The GRC 750‐kW drivetrain has a main spherical roller bearing that supports the main shaft and rotor and 2 trunnion mounts that support the gear-
box, which is commonly called a 3‐point suspension drivetrain. The gearbox includes a planetary stage with 3 equally spaced planets and 2 parallel
stages. This configuration represents a majority of larger, utility‐scale wind turbine drivetrains. The rated rotor speed is 22.1 rpm, and the gearbox,
with a ratio of 1:81.491, increases to 1800 rpm the output speed to the generator.23
For this work, the GRC gearbox HSS 32222 J2 tapered roller bearings (TRBs)24 were instrumented with strain gauges in a Poisson configura-
tion at 2 axial and 4 circumferential locations around each bearing outer race. The TRB strain gauges are installed in machined grooves in the race,
as shown in Figure 1. Through calibration in a test fixture representative of the gearbox housing,18 the load in the remaining “bridge” of the bearing
race material is measured as the rollers move underneath it at a frequency typically called the ball‐pass frequency outer (BPFO).25 For the TRBs, the
BPFO is just over 8.5 times the HSS speed.24 The 4 circumferential locations provide a coarse indication of the load zone distribution around the
bearing outer ring, whereas instrumenting both the upwind and downwind TRBs gives an indication of the load sharing between the TRBs. The
TRBs are mounted on the HSS in a matched back‐to‐back configuration, separated by an oil feed ring. As a pair, they support both radial and axial
loads on the generator (downwind) side of the helical gear mesh, whereas a cylindrical roller bearing (CRB) supports only radial loads on the rotor
(upwind) side of the mesh. To measure the input loads to the TRBs, the HSS itself was instrumented with strain gauges in full‐bridge arrangement
for bending moments at 3 axial locations (A, B, and C), torque, and pinion tooth load distribution and with an encoder for speed and azimuth. High‐
speed shaft bending and torque measurements were also calibrated in situ.17 The installed TRBs, HSS, and environmentally protected instrumen-
tation package are shown in Figure 2. The CRB was not instrumented and thus is not shown in the figure.
The GRC drivetrain was installed in the dynamometer, as shown in Figure 3. Steady‐state, constant‐speed drivetrain operations were con-
ducted throughout a range of power levels, from off‐line to full power and torque. Several field‐representative dynamic load and transient opera-
tions were also conducted, such as a braking event and a grid‐loss event. When necessary, representative rotor loads were applied with hydraulic
actuators, generator misalignment was induced by shimming the generator, braking was accomplished by disengaging the dynamometer and apply-
ing a mechanical brake to the HSS, and grid loss operations were induced by electrically disconnecting the drivetrain from the grid.19
3 | MODELING APPROACHES
Three different models are used to compare HSS bearing loads with the measurements. The models are differentiated by their complexity, level of
fidelity in predicting bearing loads, and ability to model only steady‐state or time‐varying conditions.20,21
The first is a simple semianalytic model that derives the total bearing loads from the measured shaft‐bending moments and torque through
force and moment equilibrium at each instant in time. This model serves as an efficient mechanism to understand the underlying physical behavior
of the system.
The second is a SIMPACK multibody dynamics model that represents the bearings as spring‐damper elements. The specified bearing stiff-
nesses and predicted relative displacements of the raceways are then used to calculate the total bearing loads.
FIGURE 2 Instrumented high‐speed shaft. TRB, tapered roller bearing. Photo by Jonathan Keller, NREL 27895 [Colour figure can be viewed at
wileyonlinelibrary.com]
FIGURE 3 Dynamometer testing. Photo by Mark McDade, NREL 32734 [Colour figure can be viewed at wileyonlinelibrary.com]
The third is a Transmission3D finite element/contact mechanics model that predicts the bearing load zones and stress fields in addition to the
total bearing load in only steady‐state conditions. This model requires significantly more computational time than the other 2 approaches.
T− JΩ _
Fm ¼ ;
Rb cos β
(1)
T− JΩ_ _
T− JΩ _
T− JΩ
Fm
x ¼ tan ðϕβÞ; Fm
y ¼− sin γ; Fm
z ¼ cosγ:
Rb Rb Rb
The moment equilibrium at location C yields the moments applied by the generator coupling. These moments result from operational misalign-
ment and the brake disk weight; however, their contribution to the bearing loads is relatively small compared with the contribution from the high‐
speed pinion gear mesh force.20
4 GUO AND KELLER
FIGURE 4 High‐speed shaft free‐body diagram showing instrument locations. CRB, cylindrical roller bearing; TRB, tapered roller bearing
The moment equilibrium at location B yields the CRB reactions in terms of only the measured moments:
MBz
y ¼þ
F CB ; (4)
LB −LCB
MBy
z ¼−
F CB : (5)
LB −LCB
The axial force equilibrium and moment equilibrium at location A yields the TRB pair reactions in terms of measured torque, speed, and bending
moments:
_
T− JΩ
F BP
x ¼− tan ðϕβÞ; (6)
Rb
_
T− JΩ MBz
FBP
y ¼þ sinγ− ; (7)
Rb LB −LCB
_
T− JΩ MBy
FBP
z ¼− cosγ þ þ W; (8)
Rb LB −LCB
" #
_
T− JΩ MBy
MBP ¼ MCy −MAy − cosγ− ðLBP −LA Þ−W ðL−LC −LBP Þ; (9)
y
Rb LB −LCB
" #
_
T− JΩ MBz
MBP C A
z ¼ Mz −Mz − sinγ− ðLBP −LA Þ: (10)
Rb LB −LCB
The properties of the shaft and pinion are as follows: L = 775 mm, Rb = 53.0727 mm, α = 20°, and β = 14°. The points of action of the bearings
are LCB = 33 mm and LBP = 412.5 mm. The instrumentation locations are LA = 260.5 mm, LB = 88.5 mm, and LC = 298 mm. The brake disk weight is
W = 1 kN, and J = 8 kg m2 is the inertia.17 Unique to this formulation is the angle γ, which is the direction of the line of action connecting the base
circles between the intermediate gear and the high‐speed pinion relative to the vertical (+z) axis. This angle determines the relative magnitudes of
the mesh vertical and lateral forces, and it can be determined from the housing and gear dimensions. For the GRC gearbox and left‐handed helix of
the high‐speed pinion, γ = 349.8° with positive torque (ϕ = −1) and γ = 129.8° with negative torque (ϕ = 1).
GUO AND KELLER 5
In this section, the bearing load models are first compared with the measured, steady‐state load data for validation. Then the effects of torque,
rotor moments, and generator misalignment are examined. Finally, bearing loads and predicted contact stress during a braking event and a grid loss
event are examined for their potential contribution to WEC generation.
FIGURE 6 High‐speed shaft and bearing roller loads at 100% power calculated using the Transmission3D model. The contact areas on the rollers
are highlighted [Colour figure can be viewed at wileyonlinelibrary.com]
Figure 7. As expected, the bearing loads increase linearly with power as the gear mesh forces increase linearly with power. The models compare
very favorably with the experimental TRB measurements, with the exception of off‐line conditions (ie, 0 power) in which the loads are so low that
uncertainty in the upwind TRB load measurements is the dominating factor. The sum of the CRB and TRB pair loads equates to the theoretical gear
mesh radial load at full power (78 kN). The CRB actually supports 60% (47 kN) of the radial mesh load in this condition because it is closer to the
gear mesh than the TRB pair. The remaining 40% (31 kN) of the radial load is supported by the TRB pair, which is well below the fatigue load limit for
even a single one of the TRBs (61 kN).24 Surface contact pressures for the TRBs are relatively low in these steady‐state, pure‐torque conditions—
under 1 GPa at full power.21
FIGURE 7 Variation of high‐speed shaft bearing loads with drivetrain power. CRB, cylindrical roller bearing; TRB, tapered roller bearing [Colour
figure can be viewed at wileyonlinelibrary.com]
GUO AND KELLER 7
the shaft loads because the generator coupling relieves the weight of the brake disk.20 Generator coupling and brake disk loads, although reacted
by the HSS bearings, are simply too small in comparison with torque‐induced loads to have an effect on the loads.
FIGURE 9 Braking event in the dynamometer test. HSS, high‐speed shaft [Colour figure can be viewed at wileyonlinelibrary.com]
8 GUO AND KELLER
FIGURE 10 Tapered roller bearing (TRB) pair radial load during the braking event. The minimum requisite load is highlighted in the grey box. HSS,
high‐speed shaft [Colour figure can be viewed at wileyonlinelibrary.com]
when the torque is at its highest. Because the semianalytic model correlates closely with the other models and the experimental results, the physics
of the system are believed to be sufficiently captured by the simple equilibrium equations. However, the semianalytic model does require the mea-
sured shaft bending moments to achieve this level of accuracy, which are more difficult to measure in practice than just shaft torque and speed.
Because the drivetrain speed changes during a transient event like this, the TRB strain gauge data were resampled with respect to the shaft
azimuthal position, as measured by the encoder and examined in the frequency domain—also called order analysis.22 Figure 11 compares the order
spectrum of the downwind and upwind TRB strain gauges at 0° at 100% power and during the first 2 torque excursions (t = 3.3‐4.8 s). The BPFO
content is clearly evident in both spectra, but it is much higher and more distinct at full power than it is during braking, when the load and speed
change. The widening of the BPFO during braking is an indicator that the bearing cage speed has changed and that sliding may be occurring.
Bearing race contact stress is the driver of bearing health conditions and the accumulation of rolling contact fatigue. Additionally, in benchtop
tests, high stresses that occur simultaneously with roller sliding have been found to create WECs.15,16 Figure 12 shows the maximum inner‐race
contact stress predicted by Transmission3D during the braking event for the CRB and the individual TRBs. Contact stresses on the bearing inner
race are greater than those of the outer race,21 and WECs typically occur in the inner race; therefore, only inner‐race contact stresses are consid-
ered here. The CRB reaches contact stresses of almost 2 GPa during the torque excursions, a stress level at which WECs were successfully pro-
duced on a test rig—although only in −30% slide‐to‐roll ratio conditions over millions of cycles.15 The TRBs have slightly lower contact stresses
than the CRB, reaching a maximum of only 1.3 GPa. Stresses below 0.4 GPa occur during the periods where the radial load is below the minimum
requisite load. Although the braking behavior examined herein is specific to this model turbine and the GRC drivetrain, it is similar to commercial
turbines.
FIGURE 12 Calculated inner‐race contact stress during the braking event. CRB, cylindrical roller bearing; TRB, tapered roller bearing [Colour figure
can be viewed at wileyonlinelibrary.com]
shutdown of the dynamometer for a period of 3 minutes. Tests were conducted at 25%, 50%, 75%, and 100% power levels, but only the full‐power
condition is examined here.
Time histories of the measured HSS torque and speed immediately after the grid loss event are shown in Figure 13. The drivetrain was oper-
ating at full power when the generator was disconnected at t = 2 seconds. The torque then quickly drops and fully reverses to −60% of the rated
torque in less than 1 second, followed by additional positive and negative torque oscillations. Each time the torque reverses, there is a short period
when the gear teeth move through the backlash. The fully reversing torque oscillations continue for approximately 8 seconds, during which time
the drivetrain has decelerated to only 85% speed. After this point, the torque remains near 0 as the drivetrain slowly decelerates and finally stops
after 3 minutes.
The TRB radial loads during this grid loss event are shown in Figure 14. In general, there is strong agreement among the approaches. The max-
imum bearing load of approximately 35 kN occurs during the full‐power condition prior to the generator disconnection. After the grid loss, the bear-
ing radial load is almost always below the requisite minimum radial load of 16 kN during a period where the drivetrain speed is still very high and
when the torque is fully reversing. The effect of the torque reversal on the downwind and upwind TRB load zones is shown in Figure 15. Three
instants in time are examined: (A) right before the grid loss (t = 1.8 s), (B) when the torque first reverses (t = 2.2 s), and (C) at the maximum reversing
torque (t = 2.5 s). Before the grid loss (A), both load zones are similar to those during steady‐state full power. Likewise, when the torque is near 0 at
the point of first reversal (B), both load zones are similar to those in off‐line conditions. At the maximum reversing torque (C), the downwind TRB
load zone remains similar to off‐line conditions; however, the upwind TRB load zone has changed dramatically. The direction of its load zone is now
nearly aligned with the gear mesh radial force direction for negative torque (γ = 129.8°). The load zone area has also nearly tripled, and almost all of
the rollers are in contact. The shape of the load zone indicates that it is now supporting the axial load from the gear mesh force in addition to the
radial load. These changes all occur in less than 1 second while the drivetrain is still operating at nearly full speed. The pattern of reversing torques
repeats itself for 5 cycles, during which time the torque and bearing loads decay but the drivetrain speed is still greater than 85%.
FIGURE 13 Grid loss event in the dynamometer test. HSS, high‐speed shaft [Colour figure can be viewed at wileyonlinelibrary.com]
10 GUO AND KELLER
FIGURE 14 Tapered roller bearing (TRB) pair radial load during the grid loss event. The minimum requisite load is highlighted in the grey box
[Colour figure can be viewed at wileyonlinelibrary.com]
Figure 16 shows the resulting maximum inner‐race contact stress predicted by Transmission3D for the CRB and the individual TRBs during the
grid loss event. Since the torque is never greater than in constant‐power conditions, the grid loss event only resulted in stresses below 1 GPa on
both sets of bearings. These relatively low contact stresses are experienced when the radial load is below the minimum requisite load, but the drive-
train speed is still relatively high.
FIGURE 16 Calculated inner‐race contact stress during the grid loss event. CRB, cylindrical roller bearing; TRB, tapered roller bearing [Colour
figure can be viewed at wileyonlinelibrary.com]
GUO AND KELLER 11
5 | C O N CL U S I O N S
Many wind turbine gearbox bearings experience white etching or axial cracking, the root causes of which have been debated for nearly 2 decades.
Recent benchtop tests suggest that high contact stress combined with sliding of the bearing rolling elements is a possible root cause. This paper
examines several situations in which high bearing stress and sliding conditions occur in wind turbine gearbox operations. Bearing loads were
directly measured and predicted using a variety of modeling tools in normal operations, misaligned conditions, and transient events such as braking
and grid loss particularly prone to cause bearing sliding.
During the braking event, the gearbox torque increased abruptly to 70% above rated and resulted in maximum stresses in the bearings 45% to
60% above rated. Conversely, during the grid loss event, the torque fully reversed to −60% rated and relatively low stresses. However, the bearing
load zones changed abruptly and the loads were below the minimum requisite load during the reversals, which can potentially cause roller sliding.
The effect operational misalignment due to rotor moments and induced generator misalignment on the bearing loads was also examined. Neither
changed the loads appreciably compared with pure‐torque conditions; thus, it is not likely that rotor moments and generator misalignment are
drivers of axial cracking in wind turbine gearbox bearings. The results of this paper suggest the need to measure loads and sliding in a gearbox
throughout the full range of operating conditions for the turbine and to assess the relative potential contribution of each condition to generate
WECs, as planned in an upcoming test program.28
ACKNOWLEDGEMEN TS
This work was supported by the US Department of Energy under contract no. DE‐AC36‐08GO28308 with the National Renewable Energy Labo-
ratory. Funding for the work was provided by the US Department of Energy, Office of Energy Efficiency and Renewable Energy, Wind and Water
Power Technologies Office.
The US Government retains and the publisher, by accepting the article for publication, acknowledges that the US Government retains a non-
exclusive, paid‐up, irrevocable, worldwide license to publish or reproduce the published form of this work, or allow others to do so, for US
Government purposes.
ORCID
Y. Guo http://orcid.org/0000-0002-6413-947X
J. Keller http://orcid.org/0000-0001-7724-3885
RE FE R ENC E S
1. US Department of Energy. Wind vision: a new era for wind power in the United States. Washington, DC, 2015. http://energy.gov/sites/prod/files/
2015/03/f20/wv_full_report.pdf
2. Sheng S. Report on wind turbine subsystem reliability—a survey of various databases. NREL/PR‐5000‐59111. Golden, CO: National Renewable Energy
Laboratory, 2013.
3. Lantz E. Operations expenditures: historical trends and continuing challenges. NREL/PR‐6A20‐58606. Golden, CO: National Renewable Energy Labora-
tory, 2013.
4. Wiser R, Bolinger, M, Barbose, G, et al. 2012 wind technologies market report. DOE/GO‐102013‐3948. Washington, DC: US Department of Energy,
2013.
5. Keller J, Sheng S, Cotrell J, Greco A. Wind turbine drivetrain reliability collaborative workshop—a recap. NREL/OT‐5000‐66593. Golden, CO: National
Renewable Energy Laboratory, 2015.
6. Link H, LaCava W, van Dam J, McNiff B, Sheng S, Wallen R, McDade M, Lambert S, Butterfield S, Oyague F. Gearbox reliability collaborative project
report: findings from phase 1 and phase 2 testing. NREL/TP‐5000‐51885. Golden, CO: National Renewable Energy Laboratory, 2011.
7. Guo Y, Keller J, LaCava W. Planetary gear load sharing of wind turbine drivetrains subjected to non‐torque loads. Wind Energy. 2015;18:757‐768.
https://doi.org/10.1002/we.1731
8. Sheng S. Wind turbine gearbox reliability database, condition monitoring, and operation and maintenance research update. NREL/PR‐5000‐66028.
Golden, CO: National Renewable Energy Laboratory, 2016.
9. Sheng S. Wind turbine gearbox reliability database, condition monitoring, and O&M research update. NREL/PR‐5000‐63868. Golden, CO: National
Renewable Energy Laboratory, 2015.
10. Greco A, Sheng S, Keller J, Erdemir A. Material wear and fatigue in wind turbine systems. Wear. 2013;302:1583‐1591. https://doi.org/10.1016/j.
wear.2013.01.060
11. Evans M‐H. An updated review: white etching cracks (WECs) and axial cracks in wind turbine gearbox bearings. Mater Sci Technol. 2016;32(11):1133‐
1169. https://doi.org/10.1080/02670836.2015.1133022
12. Gould B, Greco A, Stadler K, Xiao X. An analysis of premature cracking associated with microstructural alterations in an AISI 52100 failed wind turbine
bearing using X‐ray tomography. Materials and Design. 2017;117:417‐429. https://doi.org/10.1016/j.matdes.2016.12.089
13. Stadler K, Lai J, Vegter RH. A review: the dilemma with premature white etching crack (WEC) bearing failures. Bearing Steel Technologies: 10th Volume,
Advances in Steel Technologies for Rolling Bearings, STP 1580, John M. Beswick, ed.. West Conshohocken, PA: ASTM International, 2015; 487‐508.
https://doi.org/10.1520/STP158020140046
14. Lai J, Stadler K. Investigation on the mechanisms of white etching crack (WEC) formation in rolling contact fatigue and identification of a root cause for
bearing premature failure. Wear. 2016;364‐365:244‐256. https://doi.org/10.1016/j.wear.2016.08.001
12 GUO AND KELLER
15. Gould B, Greco A. The influence of sliding and contact severity on the generation of white etching cracks. Tribology Letters. 2015;60(2):1‐13. https://doi.
org/10.1007/S11249‐015‐0602‐6
16. Gould B, Greco A. Investigating the process of white etching crack initiation in bearing steel. Tribology Letters. 2016;62(26). https://doi.org/10.1007/
s11249‐016‐0673‐z
17. Keller J, McNiff B. Gearbox reliability collaborative high‐speed shaft calibration. NREL/TP‐5000‐62373. Golden, CO: National Renewable Energy Lab-
oratory, 2014.
18. Keller J, McNiff B. Gearbox reliability collaborative high‐speed shaft tapered roller bearing calibration. NREL/TP‐5000‐60319. Golden, CO: National
Renewable Energy Laboratory, 2013.
19. Keller J, Wallen R. Gearbox reliability collaborative phase 3 gearbox 2 test report. NREL/TP‐5000‐63693. Golden, CO: National Renewable Energy Lab-
oratory, 2015. DOI: https://doi.org/10.7799/1254154
20. Keller J, Guo Y, Sethuraman L. Gearbox reliability collaborative investigation of gearbox motion and high‐speed‐shaft loads. NREL/TP‐5000‐65321.
Golden, CO: National Renewable Energy Laboratory, 2016.
21. Keller J, Guo Y. Gearbox reliability collaborative investigation of high‐speed‐shaft bearing loads. NREL/TP‐5000‐66175. Golden, CO: National Renew-
able Energy Laboratory, 2016.
22. Helsen J, Guo Y, Keller J, Guillaume P. Experimental investigation of bearing slip in a wind turbine gearbox during a transient grid‐loss event. Wind Energy.
2016. https://doi.org/10.1002/WE.1979
23. Oyague F. Gearbox reliability collaborative description and loading. NREL/TP‐5000‐47773. Golden, CO: National Renewable Energy Laboratory, 2011.
24. Tapered roller bearings, single row. [Online]. http://www.skf.com/us/products/bearings‐units‐housings/roller‐bearings/tapered‐roller‐bearings/single‐
row‐tapered‐roller‐bearings/single‐row/index.html?designation=32222%20J2
25. Austin J, Talbot D, Vijayakar S, Houser D. Finite element evaluation of bearing instrumentation method. Power Transmission Engineering. 2014;78‐84.
26. Transmission3D. [Online]. http://ansol.us/Products/TX3/
27. Vijayakar S. A combined surface integral and finite element solution for a three‐dimensional contact problem. International Journal of Numerical Methods
in Engineering. 1991;31(3):525‐545.
28. Gould B, Greco A, Keller J. Investigation of bearing axial cracking: benchtop and full‐scale test results. NREL/CP‐5000‐67523. Golden, CO: National
Renewable Energy Laboratory, 2017.
How to cite this article: Guo Y, Keller J. Investigation of high‐speed shaft bearing loads in wind turbine gearboxes through dynamometer
testing. Wind Energy. 2017;1‐12. https://doi.org/10.1002/we.2150