EES KISSsoft GmbH ++41 41 755 09 54 (Phone)

Neudorfstrasse 22 ++41 41 755 09 48 (Fax)

6313 Menzingen P.O. Box 121
Switzerland www.EES-Gear.ch

Dipl. Ing. Hanspeter Dinner, EES KISSsoft GmbH, Switzerland

1 On gear modifications for wind turbine gearboxes

1.1 Objective
The purpose of this paper is to provide a reasonable, verifiable and easy to apply guideline on
gear corrections for conventional wind gearboxes.

Conventional design Not conventional design

Gear corrections for wind gearboxes

LSS PLC on main shaft / in gearbox housing Planet carrier directly in rotor bearing (e.g.
WinWinD 3MW turbine)
ISS PLC supported in gearbox housing Planets on flexpin (e.g. Wikov gearboxes)
One planet per pin, pin supported on both Compound planets (e.g. Renk gearbox for
sides Multibrid)
No power split in helical stage Power split helical stages (e.g. Clipper
Quantum Drive / Winergy Multi Duored)
Ring gears supported on both sides Ring gears supported on one side (e.g. Bosch
Redulus)

Two steps are used during the corrections design, sizing and optimisation. Different
philosophies are used when designing corrections, below, only one is described. The
calculations should be combined with experience values and verified in tests.

Figure 1: Suboptimal tooth contact pattern in a helical stage

1.2 Abbreviations
f Mesh misalignment due to shafts
fH Helix slope deviation
fpar Parallelism error, shafts
fma Total mesh misalignment
PLC Planetary carrier
1, 2, sp, pr Index for gears in a mesh, index for sun planet and planet ring gear mesh
s, p, r Index for sun gear, planet gear, ring gear
RS, GS Rotor side, generator side
TRB, DRTRB Taper roller bearing, double row TRB
CRB Cylindrical roller bearing
LSS, ISS, HSS Low, intermediate and high speed stage
KH Face load distribution factor
I, II Sides of the gear, I=RS, II=GS
RF, LF, AF, PF Righ, left, active (working), passive (non working) flank
f, f Deviation / out of plane error, inclination / in plane error
fp Indivicual gear pitch error
HPSTC Highest point of single tooth contact, point B
, t, wt Nominal, transverse and operating transverse pressure angle
T, Tnom Torque, nominal torque
TE, PPTE Transmission error, peak to peak transmission error
b, x Face width, coordinate along face width
i, sp Index
Fbt, Ft Tooth normal, tangential force (transverse section)
LDD Load spectrum
G, Ip Shear modulus, polar moment of resistance
df, di Root diameter, inner diameter
RH, LH Right hand / left hand helix

1.3 References
[1] G. Niemann, H. Winter, Maschinenelemente, Band II, Zweite Auflage, Springer
Verlag, 1985
[2] H. Dinner, U. Kissling, An Algorithm for Robust Gear Modifications Design, Gear
Solutions, July 2012
[3] H. Dinner, Micropitting in Wind Turbine Gearboxes: Calculation of the safety factor
and optimization of the gear geometry, ICPT2011, Xian, China, 2011
[4] KISSsoft Software Release 03-2013
[5] U. Kissling, Effects of Profile Corrections on Peak-to-Peak Transmission Error,
Geartechnology, July 2010

2 Corrections design
2.1 ISS and LSS corrections
The basic mechanisms may be described as follows:

We assume the ring gear to be parallel to the gearbox axis and consider:
1) Torsional wind up of planetary carrier
2) In case of helical gears, the planet tilts, leading to an inclination error
3) Tilting of planetary carrier in two axis due to main shaft bending
4) Tilting of sun gear in two axis
Effect 1) and 2) are (assuming that the torque direction does not change) always in the same
direction and may be compensated by means of a helix angle correction.

Effect 3) and 4) can be in any direction, hence, they can not be compensated for by means of a
helix angle correction and a crowning is required on the planets.

Figure 2: Misalignment modes: Top left: Carrier torsion. Top right: Planet tilting due to helix
angle. Bottom left: sun tilting. Bottom right: Carrier tilting. Calculated using [4].
Figure 3: Contact patterns on the sun and planets after tilting of the PLC. Calculated using [4]

Then, we should apply the following corrections:

On the ring gear, a long (starting at HPSTC / Point B), curved tip relief is applied, see e.g. [5].
The amount is estimated using this formula [1] where fp is along ISO1328-1.
F f Ft
Ca  bt  p , Fbt  (1)
c * b 2 cos( t )

No other corrections are applied on the ring gear.

On the sun gear, a tip relief, using the same methodology as described above, is applied.

To compensate the sun torsional deformation, a helix angle correction and crowning are
applied. The amount of torsional deflection is estimated using below formula for the torsion
angle (x), using recommendations from ISO6336 for the equivalent diameter (blue line in
figure 4):

x T ( x) 1 xT *x 1 T * x2
 ( x)   dx  * dx  * (2)
0 G * Ip G * Ip 0 b G * Ip 2 * b

Considering the whole face width, x=b


 (b) 
T *b
2 * G * Ip
where Ip 
32

(df  0.4 * mn) 4  di 4  (3)

The required helix angle correction CH at the pitch diameter d (black line in figure 4), normal
to the flank, is:

CH   (b) * d * cos(wt ) towards RH (4)

Use about 80% of this value as not the full face width will see the theoretical line load T/b.
The crowning C (distance between black and orange line in figure 4) is calculated
(neglecting helix angle correction slope):

CH b b 1 T * b2

C      * d * cos( wt ) where     * (5)
2 2  2  G * Ip 8

The above crowning compensates the sun deformation, this value is increased by about 15m
to accommodate for misalignment between sun and gearbox axis.

Crowning, sun, LSS

Torsional angle [rad]

Torsional deformation [m]
Correction [m]
Helix angle correction only [m]

0 50 100 150 200 250 300 350 400 450 500

Length along sun gear, from RS to GS, [mm]

Figure 4: Sun shaft deformation and corrections

On the planets, a tip relief, using the same methodology as described above, is applied. In
order to minimise the PPTE, a profile crowning may be applied too, however, as the speed in
this stage is low, the benefit is questionable.

For the lead correction, we first need to know the misalignment of the planet to the gearbox
axis (e.g. due to the torsional deformation of the carrier). This is expressed as deviation error
f and inclination error f. Then, the required helix angle correction (we use a non
symmetrical tapered correction) on the planet is (assuming rotor rotating clockwise and left
hand sun helix, LF=AF on sun):

CH  cos wt  pt  * f   sin  wt  pr  * f  planet  ring mesh, towards RH

(6)
CH  cos wt sp * f   sin  wt sp * f  for sun  planet mesh, towards LH

Note that any misalignment of the planet carrier itself to the gearbox axis needs to be
compensated for by means of additional crowning on the planet.

For the ISS, the same effects as in the LSS stage are considered except
1) the planet carrier remains parallel to the gearbox axis
2) sun shaft tilts less in operation, hence, less crowning is required
2.2 HSS corrections
2.2.1 General considerations
The effect of bending of the HSS shaft is very low, in particular if the pinion is arranged in
the middle or if the HSS bearing span is low as in gearboxes with only one helical stage (in
gearboxes with one planetary and two helical stages, the bearing span on the HSS tends to be
higher because a pinion and a gear needs to be placed on the driving shaft). Use a similar
bearing clearance on the RS and GS and the HSS shaft moves in a parallel manner, giving
little misalignment. In this respect, the use of two CRB and one FPBB to take the axial loads
is more easy to handle compared to a HSS shaft where one CRB (on the RS) and a DRTRB
(on the GS) is used. The DRTRB tends to have no radial clearance under the combined radial
and axial load while the CRB on the GS keeps some clearance, resulting in a tilting of the
HSS shaft.

Due to the axial forces acting from the ISS sun on the hollow shaft, the GS TRB is loaded on
all rollers, giving a high radial stiffness and no operating clearance. The RS CRB however,
still maintains clearance and has low radial stiffness, the hollow shaft tilts. This is the main
contribution to the misalignment in the HSS mesh, while the deformation due to bending or
torsion is small for the hollow shaft.

The effect of the housing deformation and manufacturing errors are small. Typically, the
tolerance in parallelism of the hollow and the HSS shaft is approximately 50m and the gear
face width is 1/3 of the bearing span, hence, the mesh misalignment fpar is only 15m. This
may be considered in the calculation of KH, and it suffices to apply crowning to take care of
this error (also of the helix angle error of the gear due to manufacturing, helix slope deviation
fH). Applying a helix angle modification does not help hear, because the direction of the error
is random.

2.2.2 Effect of each type of modification

The effect, as shown in below figure, of each correction is
- Helix angle correction: centre the contact in the middle of the face width
- Crowning: compensate variations in load and gear misalignment
- End relief: Ensure that there is no hard contact at the end of the face width

No corrections Crowning
KH=1.52 KH=1.31
Helix angle correction, crowning Helix angle correction, crowning, end relief
KH=1.12 KH=1.13
Figure 5: Influence of the various modifications in a HSS stage. Calculated using [4].

2.2.3 Starting point for modifications

On the gear, a long, curved tip relief is applied, use formula (1) to estimate the amount.

Because the face width of the gear is typically 5mm higher than the one of the pinion, a
curved end relief is applied on side I and II, starting at x=0.1*b and x=0.9*b (10% of the face
width each). Use about two or three times the amount of crowning in the mesh, so, CI=
CII=3*C or 2*C to avoid stress concentration at the sides.

On the pinion, a long, curved tip relief is applied. To reduce the PPTE, combine the tip relief
with a profile crowning of about 10m, improving the vibration characteristics of the HSS.

Instead of the end relief, a small – but sufficient to accommodate for any variation of
misalignment due to torque variations or variations in the bearing clearances – amount of
crowning is applied. Start with 25m, to be verified in the optimisation.

To compensate the misalignment of the hollow and HSS shaft under load, a helix angle
correction CH is applied. For this, calculate the gear misalignment considering the shaft and
bearing deflection, not considering manufacturing errors, for the load Ti in the LDD giving
the highest partial damage Dimax, typically Tnom < Ti < 1.05*Tnom.

Experience shows that the housing stiffness is less relevant in normal housing designs. You
may however consider the housing deformation (to be calculated in a separate FEM) as an
initial bearing positioning error (with respect to it’s theoretical position) in the LTCA.

2.2.4 Comparison to test

The obvious question is: does the results from calculation correlate well enough to test
results?

From experience with several back to back full load tests and overload tests with gearboxes
ranging from less than 1MW to more than 3MW, the author can confirm that the calculations
yield practically usefull results. Using the results from a 2MW gearbox full load test
performed in late 2012 / early 2013 by one of our customers, we found the following:
Considering the bearing operating clearance (calculated based on theoretical bearing data) and
the bearing inner geometry (for the stiffness calcualtion), the shaft deflection and the meshing
stiffness, the following contact stress was computed using KISSsoft software, release 03-2012
in the plane of action. Note that the calculation shown below is for 100% of nominal load
while the gearbox was actually optimised for a slightly different load level not disclosed here.

At this load level used for comparison between field results and calculation, it was found that
the contact was slightly on the GS (left side in below figure). At the tranition from unmodified
to modified flank (start of tip relief), a stress increase could be observed (two ridges).

The stress concentration on the GS could have been removed somewhat if a triangular end
relief was applied, due to constraints on the manufacturing, this was however not done.

Figure 6: Calculated contact stress in the plane of action of a 2MW gearbox, HSS stage, at
100% nominal torque.

The test was then conducted using an electrical back to back arrangement at the gearbox
manufacturers site. The test was witnessed by the author on behalf of the wind turbine
company. After flushing, load levels were increased in several steps, each step was run until
thermal equilibrium was achieved in the gearbox or longer. Then, at 100% of the nominal
load, the test run lasted for several hours and the usual measurements (air born sound,
vibrations, bearing temperatures, oil temperatures, oil pressure and so on) were taken. The test
was then stopped and no overlad test was conducted. The gearboxes were disassembled and
inspected two days later. Photographs of the flank were taken, see below.
Figure 7: Calculated contact stress and photograph of the flank of the driving gear. Left side
in picture=RS, right side=GS.

The calculated contact and the picture taken after test show similar behavior: the contact is
somewhat harder on GS, we see a bit of the two areas of high loads (closer to tip and closer to
root) and the start of the end relief.

Figure 8: Calculated contact stress and photograph of the flank of the pinion. Left side in
picture=RS, right side=GS.
The calculated patterns correspond quite well to the pictures taken. Both in the calculation and
in the picture, the contact is a little harder on the GS (right side in picture, side II). In both the
calculation and in the picture, the contact does not extend to the full face width. Also, the
shape of the pattern / the end relief is well visible in both cases.

Note that from the above calculation, KHβ not considering manufacturing errors was found at
KHβ=1.17. Note that KHβ for the load level for which the modifications were optimised was
somewhat lower. With manufacturing errors (fHbeta=15um and fma=10um), KHβ was found
at KHβ=1.27. Hence, the use of KHβ=1.25 in the gear calculation is confirmed. The values
are considered as slightly high due to the relatively high crowning and end relief applied.

3 Further considerations
3.1 Optimisation of corrections
Of course, the above procedure gives a starting point and the corrections need to be optimised,
e.g. using a procedure described in [2]. Finally, the modifications calculated should be
confirmed in test.

In the test, it is recommended to use strain gauges in the ring gear root and the sun gear to
confirm the findings from the visual inspection of the contact patterns. For the HSS stage, this
is usually not done.

3.2 Cross check: micropitting and scuffing

Note that the tip relief is considered in the micropitting calculation along ISO/TR 15144-1,
method B. Hence, after selecting a tip relief as per the above recommendations, the
micropitting safety factor should be calculated to ensure that a sufficient tip relief has been
applied, in particulat on the LSS stage.

More accurate is method A, where the whole pressure distribution on the flank is considered.
This then allows for a more detailed optimisation of the corrections to achieve an optimal
safety level, see e.g. [3].

The same principle applies for the scuffing safety factor calculation along ISO/TR13989-1.
Here, in particular the HSS stage needs to be checked. It may be that the tip relief has to be re-
designed based on the results of this calculation.

3.3 KHβ calculation

KH should be calculated along ISO6336-1, Annex E. Two points are mentioned here, firstly,
the influence of the manufacturing errors and secondly the calculation for a load spectrum.

If the corrections have been designed well as per the above procedure, then, it is common to
calculate KH values at design load level of KH<1.10 if manufacturing errors are not
considered. After considering manufacturing errors as per (2), the values typically are
KH<1.20.

f ma  f H2 1  f H2 2  f 2 (7)

Note: mesh misalignment due to the shafts f includes shaft parallelism error fpar.
KH strongly depends on the load applied (blue curve in figure, using a HSS of a 2.xMW
gearbox as example). As the load level Ti changes (orange curve) in the LDD, KH needs to be
calculated for each bin separately. Note that the partial damage Di (black curve) is significant
only for a few bins in the LDD, in an area where KH may be assumed as constant at about
KH=1.13 and Tnom<T<1.07*Tnom. This means that while certification guidelines require KH
to be calculated for each bin separately, this approach has no practical relevance, a constant
KH value could be used.

KHbeta, partial damage per load bin, time fraction at torque level

3
2.8
2.6 KHb
Partial damage Di (*0.1)
2.4
Time fraction (%)
KHbeta, Di/10, Time fraction

2.2
2
1.8
1.6
1.4
1.2
1
0.8
0.6
0.4
0.2
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1 1.2
Torque level normalised with nominal torque

Figure 6: KH compared to torque level compared to partial damage