You are on page 1of 134

Einfluss Der Masse und Verformung auf das Getriebe von

großen Windenergieanlagen
-

Impact of Mass and Deformation on the Gearbox of Large


Wind Turbines

Von der Fakultät für Maschinenwesen der Rheinisch-Westfälischen Technischen


Hochschule Aachen zur Erlangung des akademischen Grades eines Doktors der
Ingenieurwissenschaften genehmigte Dissertation

vorgelegt von

Tarek Magdy

Berichter: Univ.-Prof. Dr.-Ing. Georg Jacobs


Außerplanmäßiger Professor Dr.-Ing. Ralf Schelenz

Beisitz: Univ.-Prof. Dr. rer. nat. Joachim Mayer

Vorsitz: Univ.-Prof. Dr.-Ing. Christoph Broeckmann

Tag der mündlichen Prüfung: 11.11.2019


Acknowledgment
This dissertation was completed at the Center for Wind Power Drives (CWD), RWTH
Aachen University with the financial support from the German Egyptian Research
Long-Term Scholarship (GERLS).

I would like to express my appreciation to my supervisor, Professor Georg Jacobs,


whose remarks and feedback were invaluable in defining and finalizing the research
objectives. I would like to thank him for his invitation letter to come to Germany and do
my research, which made a drastic impact to my career. I would like as well to thank
Professor Ralf Schelenz for the many invaluable discussions we had through the years,
as well as for his support and encouragement during the rough phases. Also as the second
reviewer, his remarks and feedback are deeply appreciated. In addition, my thanks go to
the rest of the examination committee, professor Broeckmann and professor Mayer, for
their contribution.

Ever since I met him at Cairo University, Professor Hani Arafa has been a constant
source of knowledge, advice and support. He inspired me to pursue a Ph.D. in Germany,
and helped with conceptualizing the idea for the research topic. I owe him to a great
extent the success I have today.

My deepest gratitude goes to Christian Liewen, Stefan Kock and Björn Roscher, who
stepped in to help with my research in its most critical phases. Special thanks go to
Stefan Franzen, my first German officemate, as well as Dirk Brammertz, Vadim Krisch
and Sebastian Flock, who made me feel at home and supported me in the beginning to
adapt to the new life in Germany.

During my 4-year stay at the institute I have been blessed with a large group of
colleagues, who never spared time to help and were always willing to share knowledge
and give advice, including, but not limited to, Freia Harzendorf, Sebastian Reisch, Abdul
Baseer and Azadeh kasiri.

To all the friends who were beside me during the hard times, celebrated with me the
good times, and patiently listened to all the whining.

Finally, all of this would not have been made possible without the love and support of
my family, who had to endure a lot without my presence and let me follow my own path.

Bremen, November 2019


Abstract
The main goal is introducing technical improvements to the state-of-the-art gearboxes
to realize wind turbine gearboxes for the power class 10+ MW. Two main research
objectives are defined: gearbox optimization for nacelle mass reduction and gear mesh
optimization for gearbox reliability improvement. The first objective considers how to
reduce the sum of the gearbox and generator masses simultaneously. The second
considers two reliability-compromising challenges, namely achieving even load
distribution across the gear face, and equal load sharing among the gears splitting the
load in a given gear stage.

This work considers the state-of-the-art gearboxes as a solution space for the first
objective, namely planetary and non-planetary split-torque gearboxes. Two
mathematical models are derived that represent the gearbox and generator masses as a
function of the gearbox geometrical attributes. By applying these models, the sum of the
gearbox and generator masses for each of the state-of-the-art gearboxes can be
investigated and compared. The results show that non-planetary configurations have a
significant mass reduction in the high-speed range >1200 rpm, in addition to using
power-split modules >3. For the second objective, a self-aligning mechanism is
proposed using the flexpin bearing concept. This mechanism is envisaged by disposing
an idler gear mounted on a flexpin between each pinion and gear in the first stage, such
that the two planes of action at the two mesh points fall perpendicular to each other.

By means of finite element analysis, the hypothesis behind the self-aligning mechanism
is verified. A critical angular misalignment is assigned to the shafts, at different
orientations of the planes of action on the idler, at different combinations of contact
pattern shift across both flanks, and using both internal and external gears. The results
confirmed the hypothesis, where a reduction in the load distribution factor 𝐾𝐻𝛽 was as
high as 40%. The self-aligning mechanism yields the best results when the contact
pattern shifts across both flanks of the idler in opposite directions and vice versa. The
results show no significant differences between external and internal gears.

After merging the bearing concept with the non-planetary configuration, an optimization
procedure is provided for the final proposed gearbox construction based on a brute force
calculation and an embedded elimination-principle, and the results are compared with
the state-of-the-art planetary gearboxes. The results show that despite the mass penalty
of the added idlers, the proposed configuration still offers a mass reduction potential
against planetary gearboxes in the high-speed range, however at a higher number of
power-split modules >5.
Kurzfassung
Das Hauptziel der Arbeit ist die Einführung der technischen Modifikationen im Stand
der Technik der Zahnradgetriebe zur Realisierung von Windkraftgetriebe in der
Leistungsklasse 10+ MW. Im Rahmen der Arbeit wird die Getriebeoptimierung zur
Reduzierung der Gondelmasse sowie die Zahnkontaktoptimierung zur Verbesserung der
Zuverlässigkeit betrachtet. Das erste Forschungsziel betrachtet die Gesamtsumme der
Getriebe- und Generatormasse. Das zweite Forschungsziel beschäftigt sich mit den zwei
Hauptherausforderungen wie die Optimierung der Lastaufteilung auf die Zahneingriffe
und die Breitenlastverteilung an den Flanken.

Beim ersten Forschungsziel ist der Stand der Technik der Zahnradgetriebe mit
Planetengetriebe und leistungsverzweigte Getriebe in Stirnbauweise als Lösungsraum
gegeben. Zwei Berechnungsmodelle für die Getriebe- und Generatormasse werden in
Abhängigkeit von den geometrischen Attributen einer vorgegebenen Getriebestruktur
abgeleitet. Ausgehend von den Berechnungsmodellen wird der Einfluss der
Getriebestrukturen auf die Gondelmasse untersucht. Die Ergebnisse zeigen, dass das
größte Gewichtsreduzierungspotential durch das leistungsverzweigte Getriebe in
Stirnbauweise mit schnelldrehenden Generatoren (>1200 U/min) in Kombination von
einer Anzahl von Leistungsverzweigungs-baumodulen höher als drei möglich ist. Beim
zweiten Forschungsziel wird ein selbstausrichtendes Lagerkonzept vorgestellt. Ein
Zwischenrad auf ein Flexpin ist zwischen jedem Ritzel und Hauptzahnrad vorgesehen,
sodass die zwei Eingriffsebene am Zwischenrad senkrecht zueinander stehen.

Mit Hilfe der Finite-Elemente-Analyse ist die Hypothese hinter dem vorgeschlagenen
Lagerkonzept verifiziert. Eine kritische Winkelfehlausrichtung ist als Input für die
Zahnräder in verschiedenen Kombinationen der Richtungen der Tragbildverlagerung an
den zwei Flanken des Zwischenrades und mit dem Einsatz vom außen- und
innenverzahnten Hauptzahnrad angegeben. Die Ergebnisse konnten die Hypothese
nachweisen, wobei die Reduzierung vom Breitenlastfaktor 40% erreicht wurde. Das
Lagerkonzept ergibt die besten Ergebnisse, wenn die Tragbildverlagerung an den zwei
Flanken des Zwischenrades in unterschiedlichen Richtungen verlaufen. Der Einsatz
vom außen- und innenverzahnten Hauptzahnrad macht keinen Unterschied auf das
Lagerkonzept.

Nach der Fusion vom nach Stand der Technik leistungsverzweigten Getriebe in
Stirnbauweise mit dem vorgestellten Lagerkonzept wurde ein Optimierungsverfahren
für die finale Getriebestruktur dargestellt. Das Verfahren ist die Brute-Force-Methodik
in Kombination mit einem eingebetteten Entfernungsprinzip. Die Ergebnisse zeigen,
dass das Getriebe trotz des Massennachteils der eingeführten Zwischenräder dennoch
bei einer höheren Anzahl von Leistungsverzweigungs-baumodulen (mehr als fünf)
Massenreduzierungspotential bietet.
I

Table of Contents

List of Figures .............................................................................................................. IV


List of Tables ................................................................................................................ X
Nomenclature .............................................................................................................. XI
1 Introduction ........................................................................................................... 1
1.1 Motivation: The Potential of Offshore Wind .................................................... 1
1.2 Upscaling Turbine Capacity .............................................................................. 1
1.3 Problem Statement: Limitations to Turbine Upscaling ..................................... 3
1.4 Solution Approach: Supplementing Upscaling with Technical Innovations .... 4
1.5 Research Objectives .......................................................................................... 5
1.5.1 Gearbox optimization for nacelle mass reduction ...................................... 6
1.5.2 Gear mesh optimization for gearbox reliability improvement ................... 7
2 State-of-the-Art Wind Turbine Gearboxes ........................................................ 8
2.1 Planetary Gearbox Configurations .................................................................... 8
2.1.1 Kinematics of a planetary stage .................................................................. 8
2.1.2 Planet carrier designs ................................................................................ 10
2.1.3 Series planetary gearboxes........................................................................ 12
2.1.4 Differential planetary gearboxes ............................................................... 13
2.1.5 Stepped planetary gearboxes .................................................................... 15
2.2 Non-planetary Split-Torque Gearbox Configurations .................................... 16
2.3 Concepts for Equal Load Sharing ................................................................... 18
2.3.1 Load sharing in planetary gearboxes ........................................................ 19
2.3.2 Load sharing in non-planetary split-torque gearboxes ............................. 21
2.4 Concepts for Even Load Distribution across Gear Face ................................. 22
2.4.1 Lead modification ..................................................................................... 22
2.4.2 Flexible planet shaft with asymmetrical stiffness ..................................... 24
2.5 Summary ......................................................................................................... 25
3 Concepts Evaluation for Mass Reduction and Gear Mesh Optimization ..... 27
3.1 Proposed Idea for Gearbox Configuration Selection Method ......................... 27
3.1.1 Mathematical representation of system mass ........................................... 27
II

3.1.2 Mathematical representation of gearbox mass ......................................... 28


3.1.3 Mathematical representation of generator(s) mass ................................... 30
3.2 Applying Proposed Method to State-of-the-Art Gearboxes ............................ 31
3.2.1 Mathematical model derivation for a simple planetary stage ................... 31
3.2.2 Mathematical model derivation for a stepped planetary stage ................. 33
3.2.3 Mathematical model derivation for a dual power-split module................ 35
3.2.4 Model application to state-of-the-art gearboxes ....................................... 39
3.2.5 Results and gearbox configuration selection ............................................ 40
3.3 Self-Aligning Load-Sharing Concepts ............................................................ 47
3.3.1 Modes of gear mesh misalignment ........................................................... 48
3.3.2 Self-aligning mechanism of the proposed bearing concept ...................... 51
3.4 The Outline of the Proposed Gearbox Concept .............................................. 54
4 Self-Aligning Mechanism Verification .............................................................. 57
4.1 Development of the Modelling Concept ......................................................... 57
4.2 FE-Model Description ..................................................................................... 60
4.2.1 Gear model geometry ................................................................................ 60
4.2.2 Flexpin assembly geometry ...................................................................... 61
4.2.3 Definition of angular misalignment as an input ....................................... 62
4.2.4 Definition of load and boundary conditions ............................................. 63
4.2.5 Contact definition...................................................................................... 63
4.2.6 Output request ........................................................................................... 64
4.3 Results and Conclusions .................................................................................. 66
4.3.1 Load distribution results ........................................................................... 66
4.3.2 Structural deformations............................................................................. 70
4.3.3 Difference between external and internal gears ........................................ 72
4.3.4 Final remarks and conclusions .................................................................. 73
5 Gearbox Geometry Optimization ...................................................................... 74
5.1 Gearbox Configuration Variants ..................................................................... 75
5.1.1 Central gear ............................................................................................... 75
5.1.2 First stage design....................................................................................... 76
5.1.3 Third stage design ..................................................................................... 76
III

5.2 Spatial Constraints ........................................................................................... 77


5.3 Optimization Control Parameters .................................................................... 79
5.3.1 Specific sliding and profile shifting .......................................................... 79
5.3.2 Hunting ratio ............................................................................................. 82
5.3.3 Tooth profile ............................................................................................. 82
5.4 Mass Index Equations ..................................................................................... 84
5.5 Design Constraints .......................................................................................... 85
5.6 Assumptions and Input Loads ......................................................................... 87
5.7 Independent Parameters .................................................................................. 88
5.7.1 Pinion teeth number .................................................................................. 88
5.7.2 Stage ratios ................................................................................................ 89
5.8 Outline of Optimization Procedure ................................................................. 90
5.9 Optimization Results and Conclusions ........................................................... 93
6 Summary and Outlook ....................................................................................... 98
6.1 Summary ......................................................................................................... 98
6.2 Outlook .......................................................................................................... 101
6.2.1 Helical gears in a semi-pivoting gear system ......................................... 101
6.2.2 Application of lead correction in a semi-pivoting gear system .............. 101
6.2.3 Applying the semi-pivoting concept to a planetary stage ....................... 102
6.2.4 Reduction of multi-generators mass penalty through innovation ........... 103
7 References .......................................................................................................... 104
Appendix A Gearbox Optimization Results ........................................................ 112
IV

List of Figures

Figure 1-1: Offshore LCOE components of a total 159 £/MWh and AEP
3991 MWh/MW/year for a 500 MW farm using 4 MW turbines. “Others”
include wind farm development and decommissioning costs [12]. ......... 2
Figure 1-2: Rotor diameter, rotational speed and input torque as functions of the rated
power according to Betz law, assuming wind speed 11m/sec, tip speed
ratio 7.5 and total drivetrain efficiency 90%. ........................................... 3
Figure 1-3: Relation between the turbine rated power in MW and the LCoE
components of a 500 MW offshore wind farm. Dashed lines represent the
extrapolated values [12]. ........................................................................... 4
Figure 1-4: Schematic representation of the impact of innovation-based upscaling on
the scaling rate of (a) LCoE components and (b) total LCoE. Black
represents classical upscaling and colors represent expected impact of
innovation. ................................................................................................ 5
Figure 2-1: General kinematics of a planetary stage. .................................................. 9
Figure 2-2: Difference in ring and sun proportions between (a) 3 planets and (b) 8
planets. .................................................................................................... 10
Figure 2-3: (a) Single-flange planet carrier [37]. (b) Double-flange (or cage-type)
planet carrier [38]. .................................................................................. 10
Figure 2-4: Representation of (a) overhang pins and (b) straddle-mounted pins
according to the beam theory. ................................................................. 11
Figure 2-5: Two arrays of planet gears, each array is cantilevered on both sides of a
single-flange carrier [40]. ....................................................................... 12
Figure 2-6: Schematic of a PPS planetary gearbox. .................................................. 13
Figure 2-7: Speed relations in a differential planetary stage with ring and carrier in
opposite rotation. .................................................................................... 13
Figure 2-8: (a) Schematic of the MAAG concept. (b) Schematic of the Bosch concept.
................................................................................................................ 14
Figure 2-9: Schematic of a stepped planetary gearbox. ............................................ 15
Figure 2-10: (a) Schematic of a staggered stepped planetary configuration. (b)
Proposed stepped planetary with staggered stepped-planets for 8 MW
[48]. ......................................................................................................... 16
Figure 2-11: Dual power-split module [51]. ................................................................ 17
V

Figure 2-12: (a) 2-stage 4-output 2.5 MW gearbox [59]. (b) Schematic of half a 4-stage
2-output 6.5 MW gearbox....................................................................... 18
Figure 2-13: (a) Construction of the flexpin bearing concept comprises a planet 1,
bearing 2, sleeve 3, planet shaft 4, carrier flange 5. (b) Governing
parameters of the flexpin concept: sleeve free length L, resultant bearing
force F, planet shaft end deflection ∆, restoring moment M [66]. ......... 20
Figure 2-14: Detailed illustration of the quill shaft design [73]. ................................. 21
Figure 2-15: Design of an asymmetrical 3-helical gear intermediate shaft. ............... 22
Figure 2-16: Main 3 types of lead modification: (a) lead correction, (b) end relief and
(c) crowning [75]. ................................................................................... 23
Figure 2-17: Progression of the load distribution factor across a given load spectrum
for different optimum tooth modifications at 50% to 200% of the nominal
load. Translated from original in German [76]....................................... 23
Figure 2-18: (a) Construction of an asymmetrical straddle-mounted planet shaft [77].
(b) Simulating the deformations in a double-flange carrier under load [78].
................................................................................................................ 24
Figure 2-19: Planet axis deflection in the direction of carrier rotation of (a) a symmetric
state-of-the-art straddle-mounted planet shaft and (b) the proposed
asymmetric planet shaft. E Youngs modulus, I second moment of area, Vc
carrier angular twist, Tin driving torque at primary flange, Δb deflection
between two shaft ends, F1b & F2b bearing forces [77]. ......................... 25
Figure 3-1: Parameters of a simple gear set. ............................................................. 28
Figure 3-2: Breakdown of the torque transfer in a stepped planetary stage. ............. 33
Figure 3-3: (a) Schematic of dual power-split module. (b) Breakdown of the torque
transfer in a power-split module. ............................................................ 35
Figure 3-4: First stage of a planetary (left) and non-planetary (right) assuming the
same envelope size. ................................................................................ 37
Figure 3-5: Maximum first stage ratio of a non-planetary stage equivalent to a
planetary stage. ....................................................................................... 38
Figure 3-6: Impact of number of planets 𝑁𝑝 in a simple planetary stage on the mass
index 𝑀𝑔𝑏𝑥 ∗ and the stage ratio. .......................................................... 41
Figure 3-7: (a) Impact of 1st stage number of planets 𝑁𝑝′ of a series planetary PPS
configuration on the individual masses of each stage, with the example of
total gearbox ratio 𝑖𝑡𝑜𝑡𝑎𝑙 = 100. (b) Development of gearbox mass index
and optimum number of planets in the first stage 𝑁𝑝′ across a total
VI

gearbox ratio 𝑖𝑡𝑜𝑡𝑎𝑙 = 30: 350. ............................................................. 42


Figure 3-8: Gearbox mass index 𝑀𝑔𝑏𝑥 ∗ of state-of-the-art multiple-stage planetary
gearboxes. ............................................................................................... 43
Figure 3-9: (a) Impact of the 1st stage ratio 𝑢′ of a non-planetary configuration on the
individual masses of each stage with the example of total gearbox ratio
𝑖𝑡𝑜𝑡𝑎𝑙 = 100. (b) Development of gearbox and generator mass indices of
a non-planetary configuration across a total gearbox ratio 𝑖𝑡𝑜𝑡𝑎𝑙 =
30: 300.................................................................................................... 44
Figure 3-10: (a) Schematic of a non-planetary first stage with pinions arranged inwards
using a ring gear. (b) Schematic of a dual power path module comprising
two bull gears. ......................................................................................... 45
Figure 3-11: Impact of the first stage design on the (a) limit on the 1st stage ratio and
(b) on the mass index of the gearbox, with example of 𝑖𝑡𝑜𝑡𝑎𝑙 = 100... 45
Figure 3-12: (a) Gearbox mass and (b) generator mass of a 10 MW wind turbine for
different planetary and non-planetary configurations. ........................... 46
Figure 3-13: Sum of gearbox and generator masses 𝑀𝑠𝑦𝑠 for planetary and non-
planetary gearboxes, where 𝑁𝑚 is the number of power-split modules. 47
Figure 3-14: (a) Definition of coordinate system to reference gear angular
misalignment. (b) e2 is angular misalignment normal to the plane of action
and e3 is angular misalignment in the plane of action [94]. ................... 49
Figure 3-15: (a) Gear angular misalignment normal to the plane of action [91]. (b) Gear
angular misalignment in the plane of action [96]. .................................. 49
Figure 3-16: (a) Contact pattern under angular misalignment normal to the plane of
action e2, (b) root stress under angular misalignment normal to the plane
of action e2, (c) contact pattern under angular misalignment in the plane
of action e3, (d) root stress under angular misalignment in the plane of
action e3 [94]. ......................................................................................... 50
Figure 3-17: Schematic of the tilting moment generated by critical angular
misalignment in the plane of action between two mating flanks. Edited
from original in [96]. .............................................................................. 51
Figure 3-18: (a) Schematic of a 3-gear arrangement with two perpendicular flank
tangents 𝐹𝑇, and perpendicular base tangents 𝐵𝑇. (b) Proposed bearing
for the idler gear in [97]. ......................................................................... 52
Figure 3-19: (a) Spherical bearing [101], (b) toroidal (or barrel-shaped) bearing [101]
and (c) contact geometry of a spherical bearing showing contact angle 𝛼,
diametral Play 𝑆𝑑, raceway curvature 𝑟0 and endplay 𝑃𝑒 [102]. .......... 53
VII

Figure 3-20: General features of the power-split module in the proposed gearbox. ... 54
Figure 4-1: Schematic of the concave-convex contact in an internal gear mesh. ..... 58
Figure 4-2: Sign convention of the angular misalignment 𝜃𝐵𝑇 with respect to the
flexpin assembly. .................................................................................... 60
Figure 4-3: Abaqus model of the gears of the first stage of the proposed gearbox... 60
Figure 4-4: Geometry of the flexpin components in the Abaqus model. Dimensions
are in mm. ............................................................................................... 62
Figure 4-5: Local coordinates for gear members as a frame of reference for the
definition of angular misalignment. ........................................................ 62
Figure 4-6: Torque application to the gears via rigid kinematic coupling. ............... 63
Figure 4-7: Nodal forces at multiple nodes at the same axial position across the face
width. ...................................................................................................... 64
Figure 4-8: Nodes path through the center line of the pin. Deformation scale factor is
500:1. ...................................................................................................... 65
Figure 4-9: Deformation of the idler body and the spindle, showing the contact
pressure CPRESS in MPa. The zero-pressure zone is marked in red boxes.
Deformation scale of spindle and idler is 1000:1 and 100:1 respectively.
................................................................................................................ 65
Figure 4-10: Deflections on the spindle cylindrical surface in mm, showing location of
nodes line for reading deflections. Deformation scale is 1000:1. .......... 66
Figure 4-11: Load distribution factors for all the misalignment scenarios of the bull gear
model. ..................................................................................................... 66
Figure 4-12: Load distribution factors for all the misalignment scenarios of the ring
gear model............................................................................................... 67
Figure 4-13: Change in the load distribution factor KHβ following equation (4-3) in both
the bull and ring models with flexpin in dependence of the gear contact
shift direction. ......................................................................................... 67
Figure 4-14: Polar plot representing the flexpin deflection on the radius in mm and the
angle of the plane of flexpin deflection in degrees for a bull gear model
with (a) parallel BT and (b) perpendicular BT arrangements. Numeral tags
from 1 to 8 represent the misalignment scenarios Mis1-Mis8. F is the gear
force in the plane of action. .................................................................... 69
Figure 4-15: Deviation in the pin bending plane relative to the no misalignment
condition Mis1. ....................................................................................... 69
Figure 4-16: Representation of the ovalisation of the idler gear for both orientations of
VIII

the base tangents under the possible misalignment scenarios for the bull
gear model............................................................................................... 70
Figure 4-17: Difference of the idler ovalisation between perpendicular and parallel base
tangents for the bull gear model. ............................................................ 71
Figure 4-18: Spindle surface deflection under all misalignment scenarios at parallel
base tangents orientation for a bull gear model. ..................................... 72
Figure 5-1: (a) Representation of the possible teeth number 𝑧 combinations for all
three stages. (b) Microgeometry sub-matrix for each teeth combination 𝑧1
and 𝑧2 of a given stage 𝑘 including the teeth module 𝑚𝑛 and profile shift
distribution 𝑥1 and 𝑥2. ........................................................................... 74
Figure 5-2: Geometrical parameters of a power-split module, (a) with a bull gear
design and (b) a ring gear design. ........................................................... 75
Figure 5-3: Possible designs of the central gears; either (a) single gear or (b) double
gear.......................................................................................................... 76
Figure 5-4: Schematic of third stage variants with example of 𝑁𝑚 = 2. (a) Concept
A: each power-split module remains uncoupled, and (b) concept B: power
is summed on one output pinion 1′′′....................................................... 77
Figure 5-5: Possible collision situations in the (a) first, (b) second and (c) third stage.
................................................................................................................ 78
Figure 5-6: (a) Common area of micropitting. (b) Relative direction of slipping S and
rolling R between meshing teeth [110]. .................................................. 80
Figure 5-7: Comparing the impact of using specific sliding and DIN recommendation
to set the boundaries on the profile shift, with the example of z1=21, z2=27
and f,max = -1.5. Red area represents the solution space after applying
equations (5-9) to (5-14). ........................................................................ 82
Figure 5-8: Geometrical dimensions of tooth root critical section [47]. ................... 83
Figure 5-9: Schematic of the applicable pinion teeth number [118]. ........................ 89
Figure 5-10: Relation between the mass index of first stage 𝑀𝑔𝑏𝑥 ∗ ′ and the idler-to-
pinion ratio 𝑢12. Using 𝑁𝑚 = 6 and 𝑁3′ = 1 as an example............... 90
Figure 5-11: Gearbox mass results of the proposed gearbox against a state-of-the-art
series planetary PPS configuration. The numbers represent the power-split
modules 𝑁𝑚, A and B represent third stage design variants according to
section 5.1.3. ........................................................................................... 94
Figure 5-12: Generator mass results of the proposed gearbox against a state-of-the-art
series planetary PPS configuration. The numbers represent the generators.
IX

A and B represent third stage design variants according to section 5.1.3.


................................................................................................................ 95
Figure 5-13: System mass results of the proposed gearbox against a state-of-the-art
series planetary PPS configuration. The numbers represent the generators,
A and B represent third stage design variants according to section 5.1.3.
................................................................................................................ 95
Figure 5-14: Comparison between the central gear variants, with example of a power-
summing third stage-stage (concept B). ................................................. 96
Figure 5-15: Increase of first stage ratio 𝑢′ with the number of power-split modules
𝑁𝑚. A and B represent third stage design variants according to section
5.1.3. ....................................................................................................... 97
Figure 6-1: Possible influence of a self-aligning concept on the load distribution
deviation under a load spectrum. .......................................................... 102
Figure 6-2: (a) Relative orientation of base tangents BT in a planetary stage. (b)
Asymmetric teeth [115]. ....................................................................... 102
Figure 6-3: Schematic of a mechanical-hydraulic drivetrain [120]......................... 103
X

List of Tables

Table 3-1: Mass indices 𝑀𝑔𝑏𝑥 ∗ for state-of-the-art wind turbine gearbox
configurations ......................................................................................... 39
Table 4-1: Schematic of the simulation models showing the relative orientation of the
base tangents BT ..................................................................................... 59
Table 4-2: Misalignment scenarios showing different combinations of misalignment
in the base tangent direction ................................................................... 59
Table 4-3: Summary of the gears geometry for both models with a bull and a ring
gear.......................................................................................................... 61
Table 5-1: Summary of the control parameters applied to the proposed optimization
procedure ................................................................................................ 79
Table 5-2: Example of hunting and non-hunting gear ratios ................................... 82
Table 5-3: Mass index equations for the three stages of the proposed gearbox ...... 84
Table 5-4: Factors assumed for the preliminary and detailed strength calculation of
gears ........................................................................................................ 87
Table 5-5: Operational and geometrical assumptions required for the strength
calculations ............................................................................................. 88
Table A-1: Optimization results showing the minimum system mass 𝑀𝑠𝑦𝑠 for each
number of power-split modules 𝑁𝑚 and the corresponding geometry per
stage for a bull gear model and a simple third stage (concept A) ........ 112
Table A-2: Optimization results showing the minimum system mass 𝑀𝑠𝑦𝑠 for each
number of power-split modules 𝑁𝑚 and the corresponding geometry per
stage for a bull gear model and a power-summing third stage (concept B)
.............................................................................................................. 112
Table A-3: Optimization results showing the minimum system mass 𝑀𝑠𝑦𝑠 for each
number of power-split modules 𝑁𝑚 and the corresponding geometry per
stage for a ring gear model and a simple third stage (concept A) ........ 113
Table A-4: Optimization results showing the minimum system mass 𝑀𝑠𝑦𝑠 for each
number of power-split modules 𝑁𝑚 and the corresponding geometry per
stage for a ring gear model and a power-summing third stage (concept B)
.............................................................................................................. 113
XI

Nomenclature

Symbol Description Units

Acronyms
𝐴𝐸𝑃 Annual energy production MWh/MW/yr.
𝐵𝑜𝑃 Balance of plant €/MW
𝐵𝑇 Base circle tangent -
𝐶𝐴𝑃𝐸𝑋 Capital expenditure €/MW
𝐹𝑇 Flank tangent -
𝐿𝐶𝑜𝐸 Levelized cost of energy €/MWh
𝑂𝑃𝐸𝑋 Operational expenditure €/MW/yr.
𝑃𝑃𝑆 Three-stage gearbox (planetary-planetary-simple) -

Variables
𝑎 Center distance 𝑚𝑚
𝑏 Gear face width 𝑚𝑚
𝑐∗ Tooth tip clearance coefficient -
𝑑 Gear reference diameter 𝑚𝑚
𝑑𝑎 Gear addendum diameter 𝑚𝑚
𝑑𝑓 Gear root diameter 𝑚𝑚
𝑑𝑠𝑝,𝑖 Spindle inner diameter 𝑚𝑚
𝑑𝑠𝑝,𝑜 Spindle outer diameter 𝑚𝑚
𝐷𝐻 Housing diameter 𝑚𝑚
𝑒 Power split factor -
(in multiple-stage planetary gearboxes)
𝐸 Young’s modulus 𝑁/𝑚𝑚2
𝐹 Force 𝑁
𝐺𝑆 Geometrical function of a simple gear stage -
𝐺𝑃 Geometrical function of a planetary gear stage -
𝐺𝐶 Geometrical function of a compound gear stage -
𝐺𝑠𝑝𝑙𝑖𝑡 Geometrical function of a splitting gear stage -
𝐺𝑠𝑢𝑚 Geometrical function of a summing gear stage -

ℎ𝑎𝑝 Addendum coefficient -

ℎ𝑓𝑝 Dedendum coefficient -
ℎ𝐹𝑒 Bending moment relevant to load application at the 𝑚𝑚
outer point of single pair tooth contact
𝑖𝑡𝑜𝑡𝑎𝑙 Gearbox total ratio -
XII

𝐼 Second moment of area 𝑚𝑚4


𝐾𝐴 Application factor -
𝐾𝛾 Load sharing factor -
𝐾𝐻𝛽 Load distribution factor (or face load factor) -
𝐾𝐻𝛼 Transverse load factor -
𝐾𝑣 Dynamic factor -
𝐿 Length 𝑚𝑚
𝑚𝑛 Normal module 𝑚𝑚
𝑀 Bending moment 𝑁𝑚𝑚

𝑀𝑔𝑏𝑥 Gearbox mass index -
𝑀gbx Gearbox mass 𝑘𝑔

𝑀𝑔𝑒𝑛 Generator mass index -
𝑀𝑔𝑒𝑛 Generator mass 𝑘𝑔
𝑀𝑠𝑦𝑠 System mass (gearbox + generator) 𝑘𝑔
𝑛 Rotational speed 𝑟𝑝𝑚
𝑛𝑔𝑒𝑛 Generator rated speed 𝑟𝑝𝑚
𝑛𝑅 Rotor rated speed 𝑟𝑝𝑚
𝑁 Number (of gear members) -
𝑁𝑔𝑒𝑛 Number of generators -
𝑁𝑚 Number of power-split modules -
𝑁𝑝 Number of planets -
𝑃𝑖𝑛 Input mechanical power 𝑘𝑊
𝑃𝑔𝑒𝑛 Rated electric power of generator 𝑘𝑊
𝑃𝑒𝑙𝑒𝑐 Wind turbine nameplate power capacity 𝑘𝑊
𝑅 Pitch circle radius of shafts location 𝑚𝑚
𝑠𝑎∗ Tooth top width coefficient -
𝑠𝐹𝑛 Tooth root chord at the critical section 𝑚𝑚
𝑆 Solution matrix -
𝑆𝐻 Safety factor for pitting (ISO 6336) -
𝑆𝐹 Safety factor for tooth breakage (ISO 6336) -
𝑇 Torque 𝑁𝑚𝑚
𝑢𝑖𝑗 Gear ratio (𝑧𝑗 /𝑧𝑖 >1) -
𝑣 Gear pitch velocity 𝑚𝑚/𝑠𝑒𝑐
𝑉 Volume 𝑚𝑚3
𝑤 Force distribution (force per distance) 𝑁/𝑚𝑚
𝑥 Gear profile shift -
𝑌𝐷𝑇 Deep tooth factor -
𝑌𝐹 Tooth form factor -
XIII

𝑌𝑀 Alternating bending factor (mean stress influence -


factor)
𝑌𝑁𝑇 Life factor for tooth root stress -
𝑌𝑅 𝑟𝑒𝑙 𝑇 Relative surface factor -
𝑌𝑆 Stress correction factor -
𝑌𝑋 Size factor (tooth root) -
𝑌𝛽 Helix angle factor -
𝑧 Number of teeth -
(positive for external and negative for internal gears)
𝑍𝐵 Single pair tooth contact factor -
𝑍𝐸 Elasticity factor √𝑁/𝑚𝑚2
𝑍𝐻 Zone factor -
𝑍𝐿 Lubricant factor -
𝑍𝑁𝑇 Life factor for contact stress -
𝑍𝑅 Roughness factor -
𝑍𝑉 Velocity factor -
𝑍𝛽 Helix angle factor -
𝑍𝜀 Contact ratio factor -

𝛼 Pressure angle 
𝛼𝐹𝑒𝑛 Load direction angle at the outer point of single pair 
tooth contact
𝛼𝑔𝑏𝑥 Gearbox mass coefficient -
𝛼gen,1 Generator mass coefficient -
𝛼gen,2 Generator mass coefficient -
𝛼𝑛 Normal pressure angle 
𝛼𝑤 Working pressure angle 
𝛽 Helix angle 
𝛾 Gears orientation angle 
∆ Deflection (beams, shafts, pins, etc.) Mm
𝜖𝛼 Profile contact ratio -
𝜁 Specific sliding -
𝜃 Angle 
𝜃𝐵𝑇 Angular misalignment in the plane of action 
𝜚 Radius of relative curvature of flank at mesh point Mm

𝜌𝑓𝑝 Tooth root radius coefficient -
σF Tooth root stress 𝑁/𝑚𝑚2
σF0 Nominal tooth root stress 𝑁/𝑚𝑚2
XIV

σF lim Nominal stress number (bending) 𝑁/𝑚𝑚2


σFP Permissible tooth root stress 𝑁/𝑚𝑚2
σH Contact stress 𝑁/𝑚𝑚2
σH0 Nominal contact stress 𝑁/𝑚𝑚2
σH lim Allowable stress number (contact) 𝑁/𝑚𝑚2
σHP Permissible contact stress 𝑁/𝑚𝑚2

Subscripts
1 Intermediate pinion or sun
2 Intermediate gear, idler or ring
3 Central gear
𝑝 Planet
𝑝1 Pinion of a stepped planet
𝑝2 Gear of a stepped planet
𝑠 Planet carrier

Superscripts
(k) Stage index
′, ′′ , ′′′ First, second, third stage
Introduction 1

1 Introduction

1.1 Motivation: The Potential of Offshore Wind

By the end of 2017, the offshore wind energy stands at a share of 8 % of the worldwide
installed wind capacity [1]. However, it is still believed that offshore wind has not yet
shown its full potential. This is mainly attributed to their favorable power production
characteristics in terms of higher and more consistent wind characteristics. Offshore
energy has the highest load hours amongst other renewables at 4660 full load hours per
year, and operating hours more than 8700, which corresponds to about 363 days of
production [2], in contrast to onshore energy with full load hours around 2440 (German
national average) [3]. Consequently, this leads to higher energy yield with lower energy
production fluctuations. Offshore sites are also characterized by a relatively lower
inflow turbulence intensity, leading to relatively lower fatigue loads on the turbine [4].
Besides, the logistics of offshore are also advantageous, whereby the transport of large
cargos of towers, blades and nacelles is more practical in the sea than on land. The
European offshore share from the total wind investments has been at 50 % in 2015,
followed by 66 % in 2016 and 34 % in 2017 [5]. In 2017 the European offshore industry
hit an all-time growth record by adding 3 GW, which was double the added capacity in
2016 [1]. The Governments have been also seen providing legal frameworks and
agreements to stimulate the expansion of the offshore market to reach growth milestones
already set till 2050 [6]. These include, for example, the Renewable Energy Act in
Germany [7], and the Renewable Energy Roadmap in the UK [8].

1.2 Upscaling Turbine Capacity

Despite its huge promising potential, the offshore energy in 2017 supplied Europe only
with 1.4 % of its 2906 TWh energy consumption [5]. The problem is basically that the
offshore wind energy is still to this day not yet economically competitive with onshore
[9]. However, recent reports [10] have shown that this gap is continuously closing. In
their effort to trace down the reasons behind this cost reduction, multiple studies, such
as [11–13], have found that turbine upscaling had the strongest impact on achieving this
cost reduction. The upscaling concept aims to utilize fewer larger turbines in a given
wind farm instead of multiple smaller turbines. The reason this concept works is its
understanding of the offshore energy cost composition, and where the cost reduction
potential lies. The cost of energy, denoted by the levelized cost of energy LCoE,
comprises three main aspects: CAPEX, the capital expenditure; OPEX, the operations
and maintenance expenditure; and AEP, the net annual energy production. The CAPEX
is further divided into two subgroups: the turbine CAPEX, which includes the rotor-
2 Introduction

nacelle assembly; and the balance-of-plant BoP costs, which include the costs of the
support (tower and foundation) and electrical infrastructure as well as the installations.
Figure 1-1 shows that the offshore turbine CAPEX has a small contribution to the LCoE
in comparison to the BoP costs and the OPEX, both of which are strongly affected by
the number of turbines.

Figure 1-1: Offshore LCOE components of a total 159 £/MWh and AEP 3991 MWh/MW/year
for a 500 MW farm using 4 MW turbines. “Others” include wind farm development
and decommissioning costs [12].

As concluded by studies [11–13], the OPEX would drop by decreasing the number of
turbines that need maintenance. Installing fewer turbines, foundations and electricity
transmission infrastructure would reduce the BoP costs. These on the other hand come
on the expense of the turbine CAPEX slightly increasing. However, this increase is
overcompensated by the reduction in the other LCoE components, leading to a total
LCoE reduction. This scenario is also supported by the results of comparative studies in
[11; 12; 14].

These findings actually match the current trend of turbine size growth and explain the
ongoing efforts to continue this trend for future projects. Offshore turbines have
experienced a steady growth in size reaching an average of 6.8 MW by the end of 2018;
15 % larger than in 2017 [15]. After almost 10 years, the 8.8 MW turbine took over the
title of largest offshore turbine from the 6 MW turbine in 2018 [15]. Finally, by the end
of 2019, the prototype installation of the first 12 MW turbine is in its final stages [16].
Also, the research sector supports the industry in this venue with R&D projects for
commercially realizing 10-20 MW turbines in the future. These include the European
Wind Initiative [17], the agenda published by the European Wind Energy Technology
Platform TPWind [18] as well as the largest EU-funded R&D projects UpWind [19] and
its follow-up Innwind.EU [20].
Introduction 3

1.3 Problem Statement: Limitations to Turbine Upscaling

Simply building larger versions of the smaller ones, i.e. using same technology, is given
the term classical upscaling. The projects [19; 21] found that this type of upscaling can
face technological and economical barriers. This means it can reach a power barrier
beyond which classical upscaling will not continue to result in cost reduction, due to
negative implications on the loads and masses of the turbine components. This can be
attributed to the non-linear relation, governed by Betz law, between the generated
power, rotor dimensions and the input torque. As seen in Figure 1-2, upscaling the
turbine power demands a larger rotor (blue curve), whose speed under the same
aerodynamic limitations, i.e. tip speed and site conditions, drops proportionally to the
square root of power (green curve). This results in the rotor torque growing
exponentially with the rated power, i.e. 𝑇~𝑃1.5 , (red curve).

Figure 1-2: Rotor diameter, rotational speed and input torque as functions of the rated power
according to Betz law, assuming wind speed 11m/sec, tip speed ratio 7.5 and total
drivetrain efficiency 90%.

Based on theoretical upscaling and gathered data from actual turbines, the project [21]
found that the offshore LCoE components are actually affected by this non-linear
behavior. The reduction in the BoP costs was found to be proportional to the power of
the turbine, however the increase in the turbine CAPEX was found to be proportional to
the input torque [21]. In other words, classical upscaling forces the turbine CAPEX to
increase at a rate higher than the cost reduction in the other LCoE components. As a
numerical example, the report [12] estimated the cost components for classical upscaling
from 4 to 8 MW for a 500 MW offshore farm using wind turbines with conventional
drivetrains comprising three-stage planetary gearboxes and high-speed generators at
1500 rpm. These results were compiled here and extrapolated up to 12 MW and
4 Introduction

presented in Figure 1-3. The extrapolation was done using the curve fitting tool in
MATLAB by applying the exponential scaling relations from project [21]. The figure
shows that at a certain point, the increase in the turbine CAPEX overcomes the reduction
in the BoP and OPEX costs, producing an inflexion point on the LCoE curve, beyond
which further classical upscaling yields an increase in the total LCoE.

Figure 1-3: Relation between the turbine rated power in MW and the LCoE components of a
500 MW offshore wind farm. Dashed lines represent the extrapolated values [12].

1.4 Solution Approach: Supplementing Upscaling with Technical Innovations

To overcome this barrier, studies [11; 12; 21; 22] came to the conclusion that upscaling
needs to be supplemented with technical innovations. The reports [11; 12] were more
specific and listed the possible turbine development areas, and put the innovation in the
nacelle on the top for having the most expected contribution to LCoE reduction. The
same reports went on to state that the two main drivers behind any proposed innovation
in the nacelle must be mass reduction and reliability improvement. First, the importance
of the nacelle mass reduction is that it flows in the mass calculations of the downstream
components such as the tower and the foundation as well. Thus, a reduction in the
nacelle mass could impact the scaling rate, as represented in red in Figure 1-4 (a), of
Introduction 5

both the turbine CAPEX and the BoP costs. Second, an improved reliability would lead
to a reduction in failure rates, which in turn could reflect on the scaling rate of the OPEX,
as represented in blue in Figure 1-4 (a). Innovation-based upscaling could thus lead to
two simultaneous impacts, as represented in Figure 1-4 (b). First, the LCoE at a given
rated power could be decreased, which can be seen by shifting the LCoE curve
downwards in the direction of the Y-axis. Second, the economic feasibility of turbine
upscaling could be extended to a higher power rating, which can be seen by pushing the
inflexion point to the right along the X-axis. Figure 1-4 serves only as a schematic of
the expectations and is not to be taken as a calculated forecast. These expectations are
based on the predictions presented in the Innwind.EU project [21].

(a) (b)
Figure 1-4: Schematic representation of the impact of innovation-based upscaling on the scaling
rate of (a) LCoE components and (b) total LCoE. Black represents classical
upscaling and colors represent expected impact of innovation.

It is thus the purpose of this work to follow the recommendations of the studies [11; 12;
21; 22], and find the right drivetrain innovation that can unlock the hidden potential of
turbine upscaling through nacelle mass reduction and improving drivetrain reliability.

1.5 Research Objectives

This research follows the recommendations of reports [12; 13] by working on a solution
to the problem of reducing the mass without compromising the reliability of the
drivetrain of upscaled offshore wind turbines. The course of this research is dedicated
to upscaled power in the 10+ MW class, as it is still an uncharted territory where there
6 Introduction

is not yet a consensus in the industry on the optimum drivetrain concept for this power
class. The only attempt to enter this power class is the recently announced 12 MW direct
drive turbine model [23]. All calculations are done with the example of a generic 10 MW
turbine. The input load data of this generic turbine is extracted from an internal 10 MW
low wind rotor project at the Center for Wind Power Drives CWD of the RWTH Aachen
University [24].

The focus of this research is to address this challenge from the point of view of the
gearbox, which then gives rise to the following two research objectives.

1.5.1 Gearbox optimization for nacelle mass reduction

The gearbox mass is usually considered alone as the main optimization parameter, e.g.
[25–27]. However, here this is expanded to include the mass of the generator, which is
strongly influenced by the gearbox design as well. Both the gearbox and the generator
take up to nearly 40 % of the Turbine CAPEX with respect to a 4 MW drivetrain [12].
As a natural consequence to the pressing need to reduce the mass (and LCOE) of multi-
MW nacelles, the mass of the gearbox-generator combination becomes the main interest.
That is why the attention has started to shift towards incorporating the impact of the
gearbox design on the generator side in the decision making as well. The NDTC project
by ZF [28] is among the few that paid attention to the dependency of the LCOE on the
choice of the gearbox-generator combination, and considered the influence of the
gearbox ratio on the generator torque as a parameter in the gearbox optimization process.

Another aspect of the generators that needs to be considered is their number. The growth
of the turbine capacity beyond 10 MW exceeds the offered power range by most
manufacturers. The well-established power class delivered by most generator
manufacturers is around the 2 - 5 MW range. Exceeding this range would correspond to
a significant jump in the price of the units and their delivering time, and simultaneously
a drop in the reliability as a consequent to being a new technology. The lack of on-shelf
10+ MW electric generators increases the need to consider the multiple-generator
solution alongside the conventional single-generator concept. In other words, the
number of output shafts in a gearbox configuration becomes an additional parameter as
well.

Thus, the decision to stick with (or depart from) planetary gearboxes for the 10+ MW
power class is rather oriented towards mass reduction on a system level, i.e. the gearbox-
generator combination. Thus, this defines the first research objective as follows:

“Which gearbox configuration with how many generators can achieve the minimum
gearbox-generator mass?”
Introduction 7

The solution space for answering this research objective is limited to the available
gearbox design options.

1.5.2 Gear mesh optimization for gearbox reliability improvement

The gearbox failure has a significant contribution to the maintenance costs of the
turbine. This research focuses on the gearbox reliability problems that would only
intensify because of the exponentially increasing loading conditions of a 10+ MW
turbine. Two reliability-compromising challenges take place within the gearbox in this
power class, namely achieving even load distribution across the gear face, and equal
load sharing among the gears splitting the load in a given stage.

First, the challenge of achieving load distribution evenness stems from the significant
growth of the components size, while the range of the acceptable alignment tolerance
does not grow simultaneously. What adds to the challenge is the increased uncertainty
associated with the estimation of input loads, i.e. torque and non-torque loads, and the
resultant loads at the gear teeth and bearings. This leads to significant unsymmetrical
elastic deformations that would cause misalignment of gears and bearings, leading to
unknown and unpredicted (possibly critical) contact of the teeth and bearings. This is
commonly referred to as edge loading, which in turn would shorten the service life of
the gears and bearings. Second, in order to handle an input torque in the multi-MNm
level, the gearboxes are usually of the power-splitting nature to decrease the loads at the
teeth and bearings. The problem inherent to this type of gearing is that the power-
splitting feature is mainly dependent on machining accuracy, which is difficult to
maintain alongside the growing dimensions of the gearbox components [29]. In addition
to overloading the gear teeth and bearings, the inequality of the load sharing was found
to be a possible source of edge loading as well [30].

Since these two challenges are linked and believed to be dominant in this rated power
level, the second research objective can be defined as follows:

“What are the possible bearing concepts for the selected gearbox configuration from the
first research question to achieve even load distribution across the gear face, and equal
load sharing among the power-split paths?”

Finally, the system mass optimization of the final construction that results from the
merger of the chosen gearbox structure from the first research question, and the chosen
bearing concept from the second research question will be provided and compared
against the dominant state-of-the-art gearbox configurations.
8 State-of-the-Art Wind Turbine Gearboxes

2 State-of-the-Art Wind Turbine Gearboxes

The first 5 MW turbine was introduced by Areva with the Aerogear gearbox in 2005
[31], then the first 6 MW by REpower with a high speed gearbox in 2006 [32] and lately
the 8 MW by Vestas with a medium speed gearbox in 2014 [33], which has also been
successfully upgraded to 9.5 MW in 2017 [34]. Until the direct drive generator becomes
competitive with gearbox in terms of mass and costs, it is assumed that the gearbox
would retain its share of success in larger turbines.

This chapter covers the state-of-the-art gearbox configurations, as well as the state-of-
the-art bearing concepts that have been proposed to achieve load distribution evenness
across the gear face and load sharing equity among the power-split branches. Two main
types of gearboxes were proposed for wind turbines: planetary and non-planetary split-
torque gearboxes.

2.1 Planetary Gearbox Configurations

The planetary gearbox has been dominating the wind turbine market up to 6 MW, and
currently operates successfully in an 8 MW [35]. The building unit of these gearboxes
is the planetary stage, and the following sections explain the features and characteristics
pertaining to planetary stages, as well as the different ways the planetary stages can be
coupled in a multiple-stage gearbox.

2.1.1 Kinematics of a planetary stage

A single planetary stage splits power among multiple planet gears that are anchored in
the walls of a carrier and are in mesh with a ring and a sun gear simultaneously. With
reference to Figure 2-1, the general relation between the velocities of the planetary stage
members is given by equation (2-1).

𝑣𝑠 = 0.5 (𝑣1 + 𝑣2 ) (2-1)

Therefore, the general equation that governs the rotational speeds of all stage members
can be derived from equation (2-1), and is given by equation (2-2). The equation takes
the positive sign with the same sense of rotation of sun and ring and vice versa.
𝑛1 𝑛2
𝑛𝑠 = ± 𝑧1
|𝑧 | (2-2)
(1 + 2 ) (1 + |𝑧2 |)
𝑧1

In a wind turbine gearbox, which serves as a step-up gearbox, the stages are usually
State-of-the-Art Wind Turbine Gearboxes 9

realized either as planetary or star sets. A planetary stage operates with a fixed ring, and
the input on the carrier, i.e. 𝑛2 = 0. A star stage operates with a fixed carrier, and the
input on the ring, i.e. 𝑛𝑠 = 0. In both cases the output is delivered at the sun.

Figure 2-1: General kinematics of a planetary stage.

Generally, the planetary set has more advantages over the star set. First, the planetary
set for the same ring-to-sun ratio |𝑧2 |⁄𝑧1 would exhibit a higher gear ratio. The stage
ratio by both arrangements is derived from equation (2-2), and yields equations (2-3)
and (2-4).

𝑃𝑙𝑎𝑛𝑒𝑡𝑎𝑟𝑦: 𝑛𝑜𝑢𝑡⁄𝑛 𝑛1 |𝑧 |
𝑖𝑛
= ⁄𝑛𝑠 = 2 ⁄𝑧1 + 1 (2-3)

𝑆𝑡𝑎𝑟: 𝑛𝑜𝑢𝑡⁄𝑛 𝑛1 |𝑧 |
𝑖𝑛
= ⁄𝑛2 = 2 ⁄𝑧1 (2-4)

Besides, for the same component sizes and number of planets, a planetary set exhibits
lower tooth and bearing loads than a star set by factor of the planetary-to-star ratio given
in equation (2-5), as once highlighted before in [36]. With reference to equation (2-5),
increasing the number of planets, consequently decreases the ratio |𝑧2 |⁄𝑧1 , which then
increases this bearing load ratio. This gives the planetary set more advantage over the
star set for stages comprising a larger number of planets.

|𝑧2 |⁄
𝐹𝑃𝑙𝑎𝑛𝑒𝑡 𝐵𝑒𝑎𝑟𝑖𝑛𝑔 (𝑆𝑡𝑎𝑟) 𝑧1 + 1
= (2-5)
𝐹𝑃𝑙𝑎𝑛𝑒𝑡 𝐵𝑒𝑎𝑟𝑖𝑛𝑔 (𝑃𝑙𝑎𝑛𝑒𝑡𝑎𝑟𝑦) |𝑧2 |⁄
𝑧1

Whether it is a planetary or a star set, a common feature is the trade-off between the
stage ratio and the number of planets. As seen in Figure 2-2, increasing the number of
10 State-of-the-Art Wind Turbine Gearboxes

planets demands changing the proportions of the ring and sun diameters to fit more
planets in the space in between. Therefore, increasing the number of planets, decreases
the ring-to-sun ratio, hence the stage ratio. A decisive factor for the number of planets
is the construction of the planet carrier.

(a) (b)

Figure 2-2: Difference in ring and sun proportions between (a) 3 planets and (b) 8 planets.

2.1.2 Planet carrier designs

One significant component that determines the design of the planetary stage is the planet
carrier. It strongly impacts the number of planets, the planet bearing design and the width
of the planet gears, hence the load capacity of the whole stage. There are two main
carrier designs: the single-flange and the double-flange carrier, seen in Figure 2-3 (a)
and (b) respectively. The double-flange, also known as the cage-type carrier, is the
dominant design in wind turbine gearboxes.

(a) (b)

Figure 2-3: (a) Single-flange planet carrier [37]. (b) Double-flange (or cage-type) planet carrier
[38].

The choice of the carrier design is strongly impacted by the number of planets. In a cage-
type carrier, the bridges that couple the two flanges are located in the space between
each two successive planets, which diminishes with increasing the number of planets.
State-of-the-Art Wind Turbine Gearboxes 11

The carrier design also impacts in return the planet bearing concept. In a cage-type
carrier, the planet pins are supported at both ends, i.e. straddle-mounted planets, as
opposed to a single-flange carrier where the planet pins are supported from one end only,
i.e. overhang planets. This means that the planets in both designs would experience
completely different deflections. As seen in Figure 2-4, according to the beam theory,
the maximum pin deflections for the same pin length and load are calculated by
equations (2-6) and (2-7).
4
Figure 2-4 (a): ∆𝑚𝑎𝑥 = 𝑤. 𝐿 ⁄8𝐸𝐼 (2-6)
4
Figure 2-4 (b): ∆𝑚𝑎𝑥 = 𝑤. 𝐿 ⁄384𝐸𝐼 (2-7)

This means that the overhang planet would exhibit 48 times higher deflections – for the
same support length and loads – than the straddle-mounted planet. For this reason, the
face width of an overhang planet is usually reduced by a factor of 4√384/8 ≅ 2.6 to
maintain the same magnitude of deflections. The reduction in the face width would
however come on the expense of the load carrying capacity of the gears.

𝑤 𝑤

∆𝑚𝑎𝑥 ∆𝑚𝑎𝑥
(a) (b)
Figure 2-4: Representation of (a) overhang pins and (b) straddle-mounted pins according to the
beam theory.

For these reasons, a cage-type design is favored for high-ratio stages comprising a lower
number (3 to 5) of wide-face planets. A single-flange carrier is usually seen in low-ratio
stages comprising multiple narrow-face planets (above 5 planets). In other words, the
carrier decides if the torque capacity is increased by increasing the width of gears or the
number of planets, but both cannot be increased simultaneously. In an attempt to
overcome the compromise between the number and width of planets, a double array
design, seen in Figure 2-5, has been proposed few times for wind turbine gearboxes, e.g.
[39; 40]. The design proposes anchoring the planets on both sides of a single-flange and
compensating for the narrow-face planets by doubling the number of planets in the axial
direction.

Regardless of the construction of the single stage, the stages can be coupled in different
forms in a multiple-stage gearbox. Multiple stages are always required since the gearbox
12 State-of-the-Art Wind Turbine Gearboxes

total ratio required to drive mid- or high-speed generators cannot be delivered by a single
stage. The following sections give an insight into the various forms of multiple-stage
planetary gearboxes.

Ring (stationary)
Planets 2nd array
(generator side)

Planets 1st array


(rotor side) Sun (output)

Carrier driving
Planet carrier flange
(input)

Figure 2-5: Two arrays of planet gears, each array is cantilevered on both sides of a single-flange
carrier [40].

2.1.3 Series planetary gearboxes

This is the simplest and most dominant form of multiple-stage planetary gearboxes. The
schematic of this design can be seen in Figure 2-6. The output of the first stage, i.e. the
sun gear, would drive the planet carrier of the second stage, then finally the sun gear of
the second stage drives a simple parallel-shaft third stage. The construction with only
the two planetary stages dominates medium-speed gearboxes, which are integrated with
a medium-speed generator in the same housing. This integrated design has already been
proposed for 3 MW and 7 MW by one manufacturer [41], and for 4 MW [42] and 8 MW
[43] by another manufacturer. The two stages are complemented by the third stage to
add to the total step-up ratio and serve as a high-speed gearbox. These are currently
operating in multi-MW turbines up to 6 MW [32]. The total ratio of this design is simply
the multiplication of the stage ratios, as given by equation (2-8).

𝑇𝑖𝑛 |𝑧2′ | |𝑧2′′ | 𝑧2′′′


𝑖𝑡𝑜𝑡𝑎𝑙 = = (1 + ′ ) . (1 + ′′ ) . ′′′ (2-8)
𝑇𝑜𝑢𝑡 𝑧1 𝑧1 𝑧1

A common term for two planetary stages coupled in series is the PP configuration, and
when it is complemented with the simple third stage is called the PPS configuration.
State-of-the-Art Wind Turbine Gearboxes 13

Figure 2-6: Schematic of a PPS planetary gearbox.

2.1.4 Differential planetary gearboxes

The term differential here refers to a planetary stage in a 3-shaft operation; meaning no
shaft is stationary. The stages are coupled such that the ring and carrier of that
differential stage rotate in opposite directions, as shown in Figure 2-7. Thus equation
(2-2) applies with the negative sign, and consequently the sun gear significantly speeds
up.

Figure 2-7: Speed relations in a differential planetary stage with ring and carrier in opposite
rotation.

This has already been introduced in two different designs for multi-MW wind turbines.
The first design in Figure 2-8 (a), commonly known as the MAAG concept, is a two-
stage gearbox, where the carrier and the ring of the first stage are coupled to the ring
and the sun gear of the second stage respectively. The differential stage in this case is
14 State-of-the-Art Wind Turbine Gearboxes

the first stage. The total ratio of the two-stage design is governed by equation (2-9). A
main consequence of this arrangement is that only a fraction of the total power flows
through the star stage. This fraction is denoted by the parameter 𝑒 and can be calculated
by equation (2-10). A lower value of 𝑒 reflects lower loads acting on the respective
stage.

𝑇𝑖𝑛 |𝑧2′ | |𝑧2′ | |𝑧2′′ |


𝑖𝑡𝑜𝑡𝑎𝑙 (𝑀𝐴𝐴𝐺) = = (1 + ′ ) + ′ . ′′ (2-9)
𝑇𝑜𝑢𝑡 𝑧1 𝑧1 𝑧1
|𝑧2′ |
𝑃2 (1 + ⁄𝑧 ′ )
𝑒𝑀𝐴𝐴𝐺 = =1− 1 (2-10)
𝑃𝑖𝑛 𝑖𝑡𝑜𝑡𝑎𝑙

The second design in Figure 2-8 (b), commonly known as the Bosch concept, is a three-
stage planetary gearbox. It is a modification to the first design, where the carrier of the
first stage and the ring of the second are directly coupled, but the sun gears of the first
and the second stages are coupled indirectly through a third stage. This third stage in
this case serves as the differential stage.

(a) (b)
Figure 2-8: (a) Schematic of the MAAG concept. (b) Schematic of the Bosch concept.

The total ratio of the three-stage design is given by equation (2-11). Similar to the
MAAG concept, the first and second stages transmit only a fraction of the power, 1 − 𝑒
and 𝑒 respectively. The parameter 𝑒 for this design can be calculated by equation (2-12).

𝑇𝑖𝑛 |𝑧2′ | |𝑧2′′′ | |𝑧2′′ | |𝑧2′′′ |


𝑖𝑡𝑜𝑡𝑎𝑙 (𝐵𝑜𝑠𝑐ℎ) = = (1 + ′ ) . (1 + ′′′ ) + ′′ . ′′′ (2-11)
𝑇𝑜𝑢𝑡 𝑧1 𝑧1 𝑧1 𝑧1
|𝑧2′′ | |𝑧2′′′ |
𝑃2 ⁄ ′′
𝑧1 . ⁄𝑧 ′′′
𝑒𝐵𝑜𝑠𝑐ℎ = = 1 (2-12)
𝑃𝑖𝑛 𝑖𝑡𝑜𝑡𝑎𝑙
State-of-the-Art Wind Turbine Gearboxes 15

The two-stage design has been introduced before for a 1.5 MW wind turbine [44]. The
three-stage design has been successfully deployed in multi-MW turbines from 1.5 up to
5 MW as well as the Vestas 8 MW turbine. There is a major difference however between
differential and series planetary gearboxes in term of the obtained total gearbox ratio.
As seen in equations (2-9) and (2-11), the total ratio in the differential design is no longer
the multiplication of the individual stage ratios, which results in a significantly lower
total ratio. In other words, splitting the power on two stages comes at the expense of
reducing the total gearbox ratio.

2.1.5 Stepped planetary gearboxes

Stepped planetary gearboxes, also known as compound planetary gearboxes, have also
been deployed in multi-MW wind turbines, e.g. in a 1.8 MW turbine [45] as well as a
5 MW turbine [46]. The schematic of this design can be seen in Figure 2-9.

Figure 2-9: Schematic of a stepped planetary gearbox.

The term stepped planet, alternatively known as compound planet, refers to a gear
cluster comprising two gears with different diameters, namely a pinion 𝑧𝑝1 and a gear
𝑧𝑝2 . Similar to a simple planetary stage, the stage can operate either as a star or a
planetary stage. The gearbox step-up ratio is given by equations (2-13) and (2-14).

| | 𝑧
𝑃𝑙𝑎𝑛𝑒𝑡𝑎𝑟𝑦: 𝑛𝑜𝑢𝑡⁄𝑛𝑖𝑛 = 𝑛1⁄𝑛𝑠 = 𝑧2 ⁄𝑧𝑝 . 𝑝2⁄𝑧1 + 1 (2-13)
1

| | 𝑧
𝑆𝑡𝑎𝑟: 𝑛𝑜𝑢𝑡⁄𝑛𝑖𝑛 = 𝑛1⁄𝑛2 = 𝑧2 ⁄𝑧𝑝 . 𝑝2⁄𝑧1 (2-14)
1

The main advantage the stepped planetary has over the other planetary configurations is
the elimination of reverse bending. Reverse bending happens when the two flanks of the
tooth of a simple planet are alternately loaded as the planet engages the ring and sun
gears, i.e. a complete reversal of stress is experienced by each planet tooth root. In a
16 State-of-the-Art Wind Turbine Gearboxes

stepped planetary arrangement, none of the flanks are subjected to this. According to
ISO 6336 [47], this bi-directional loading demands reducing the permissible bending
stress by a factor 𝑌𝑀 = 0.7. Eliminating this reverse bending would thus increase the
allowable bending strength with a factor of 10/7.

The disadvantage however of this arrangement is that the number of stepped planets is
only limited to four. The sun can only accommodate up to four larger gears at its
circumference, and increasing this number is only possible when the gears are smaller
than the sun, which does not make sense in a step-up ratio gearbox; 𝑧𝑝2 ⁄𝑧1 must be
larger than one. In an attempt to eliminate this disadvantage, another variant of this
configuration features two groups of staggered stepped planets, as shown in Figure 2-10.
The assembly features two coaxial sun gears rather than one as in the conventional
design, where the planet gears of each group engage one of the sun gears.

(a) (b)
Figure 2-10: (a) Schematic of a staggered stepped planetary configuration. (b) Proposed stepped
planetary with staggered stepped-planets for 8 MW [48].

The idea behind this contraption is to double the power splitting in the gearbox, while
maintaining a high planet-to-sun gear ratio 𝑧𝑝2 ⁄𝑧1 . This has been proposed multiple
times for wind turbine gearboxes in patents [49; 50] and conference papers [48], but has
never been manufactured yet.

2.2 Non-planetary Split-Torque Gearbox Configurations

Non-planetary split-torque gearboxes – less popularly referred to as double-stage


locked-train gearing – can be envisaged in a myriad of configurations, all being common
in that they comprise the same building unit called the dual power-split module. A dual
power-split module, seen in Figure 2-11, is a two-stage locked-train gear set that features
power-splitting among two parallel fixed-axis intermediate shafts between two non-
coaxial gears, namely a high-speed low-torque pinion and a low-speed high-torque bull
gear. For a wind turbine gearbox, the arrangement is realized in a step-up fashion, where
State-of-the-Art Wind Turbine Gearboxes 17

the rotor input mechanical power is dumped at the bull gear, and the generator would be
located downstream. The two gear members of the intermediate shaft are termed the
intermediate pinion and the intermediate gear.

Figure 2-11: Dual power-split module [51].

Non-planetary split-torque gearboxes have been proposed for wind turbines in patents
[52–55], in development projects [56; 57], as a 6.5 MW prototype [58] and in
commercial use in the Liberty 2.5 MW turbine [59]. Two of these designs can be seen
in Figure 2-12. The design in Figure 2-12 (a) is the gearbox of the Liberty turbine. This
design features two dual power-split modules locked in parallel, forming a closed loop
arrangement. The four output pinions drive four high-speed generators. The second
design is built as a 6.5 MW gearbox by Winergy called the Multiduored. For visual
simplicity, a schematic of half the gearbox is shown in Figure 2-12 (b). This design is a
four-stage gearbox comprising four dual power-split modules in the first stage, where
the intermediate gears of each module are arranged on opposite sides of the bull gear.
Each two of the four output pinions form a consecutive power-split module. This
gearbox drives two high-speed generators.
18 State-of-the-Art Wind Turbine Gearboxes

Power-split
Module

(a) (b) Output

Bull Gear
Figure 2-12: (a) 2-stage 4-output 2.5 MW gearbox [59]. (b) Schematic of half a 4-stage 2-output
6.5 MW gearbox.

As seen in both designs, it is relatively difficult to sum the power to a single output. It
would require multiple stages, i.e. >3, to cover the large center distances pertaining to
this type of gearboxes. For this reason, this design has always been proposed for
multiple-generator drivetrains. The non-planetary design however did not share the
same success as planetary gearboxes. The 6.5 MW gearbox in Figure 2-12 (b) remains
as a prototype to this date. The 2.5 MW gearbox in Figure 2-12 (a) was much more
unfortunate and suffered from multiple failures in the field [60]. Despite their lack of
success in the wind turbine industry, this gearbox configuration is dominant in helicopter
main transmissions. Even the gearbox of the most powerful heavy-lift helicopter built
to date is a non-planetary split-torque gearbox [61; 62]. That is why it is believed that
this configuration has not yet shown its full potential and this configuration is further
investigated in the course of this thesis.

2.3 Concepts for Equal Load Sharing

Relying heavily on the manufacturing accuracy to deliver acceptable, not perfect, load
sharing has been deemed an acceptable practice for both planetary and non-planetary
gearboxes. However, resorting only to manufacturing accuracy does not suffice with the
increasing size of components. Achieving equal load sharing becomes difficult with the
increasing number of power-split paths, whether these are planets [63] or intermediate
shafts [51]. This leads to an increase in the inequality of power-splitting among the
power branches. In a power branch, the deviation of load from its ideal share is
considered during strength calculations via the load sharing factor 𝐾𝛾 [64]. An increase
State-of-the-Art Wind Turbine Gearboxes 19

in this factor would manifest itself in an increase in mass and volume of gears and
bearings. Extra measures have been proposed across the years for industrial gearboxes
– not just for wind turbines – to ensure equal load sharing without having to impose
stringent manufacturing tolerances.

2.3.1 Load sharing in planetary gearboxes

In planetary stages, increasing the number of planets above 3 renders the gear stage
statically indeterminate [29; 63]. This problem manifests itself as an unequal load
sharing among the planets, which increases proportionally with the number of planets.
Reducing the stiffness of the planet shafts is known as a classic method to improve load
sharing in a simple planetary stage [65]. The deflections from the reduced stiffness
allows compensating for the position errors of the planet shafts. However,
overexploiting this concept would force the planets out of alignment, and leads to
uneven load distribution across the gear face and its respective bearings. Therefore, this
solution is limited by the compromise between the equity of load sharing among the
planets and the evenness of the tooth load distribution across the gear face.

The flexpin is a bearing concept proposed by Hicks in the 60’s as a solution to overcome
this compromise [66]. It allows significantly high values of planet shaft deflections
without causing misalignment of the planet gear. The flexpin, as shown in Figure 2-13,
is a bearing assembly realized as a double-cantilever, where a relatively flexible pin
projects as a cantilever from the flange of a single-flange type carrier. The planet is
mounted on a relatively rigid sleeve, which is cantilevered from the remote end of the
pin. As seen in Figure 2-13 (b), the planet is located axially in the midplane of the free
length of the spindle 𝐿. This would exert equal bending moments on both ends of the
planet pin. This generates that S-shape deflection of the planet pin seen in the figure,
and therefore the planet under load would translate unskewed. The amount of deflection
∆ is designed to compensate for the planet position errors and achieve improved load
sharing equity among the planets.
20 State-of-the-Art Wind Turbine Gearboxes

(a) (b)
Figure 2-13: (a) Construction of the flexpin bearing concept comprises a planet 1, bearing 2,
sleeve 3, planet shaft 4, carrier flange 5. (b) Governing parameters of the flexpin
concept: sleeve free length L, resultant bearing force F, planet shaft end deflection
∆, restoring moment M [66].

The idea behind the flexpin is in line with the basics of the beam theory. Applying the
theory to the force analysis in Figure 2-13 (b) gives equation (2-15).

𝑑2 𝑦 𝑀𝑥 𝑀 − 𝐹. 𝑥
= = (2-15)
𝑑𝑥 2 𝐸𝐼 𝐸𝐼
By integration, equation (2-16) represents the beam deflection angle at a distance 𝑥:

𝐹
𝑑𝑦 𝑀. 𝑥 + . 𝑥 2
𝜃= = 2 (2-16)
𝑑𝑥 𝐸𝐼
Thus, the deflection angle at the free end of the pin is given by:

𝐿2
𝑀. 𝐿 − 𝐹. (2-17)
𝜃𝑓𝑟𝑒𝑒 𝑒𝑛𝑑 = 𝜃@𝑥=𝐿 = 2
𝐸𝐼
To achieve parallel translation of the planet under load, it is desired to achieve a zero-
deflection angle at the end of the pin. By equating equation (2-17) to zero, the required
restoring moment at the end of the pin is given by equation (2-18). This is achieved
when the planet is positioned axially on the spindle at the center of the free length of the
spindle.

𝐿
𝑀 = 𝐹. (2-18)
2
This bearing concept has already been investigated multiple times in the literature for
high number of planet above 5, and has shown improved load sharing [67–70]. The
State-of-the-Art Wind Turbine Gearboxes 21

flexpin has been also used in industrial planetary gearboxes with high number of planets
up to 8 [44; 66; 71]. This construction is however only applicable to spur gears. When
helical planets are desired, a zero-deflection angle at the end of the pin is not achieved
because of the overturning moment generated by the opposite axial forces at the sun and
ring gear meshes.

2.3.2 Load sharing in non-planetary split-torque gearboxes

Similar to planets, reducing the stiffness has been proposed as a solution to achieve
equal load sharing among multiple intermediate shafts. The only difference is that the
solution here refers to torsional stiffness instead of bending stiffness [72]. The most
common method to achieve this is the quill shaft design. The design, seen in Figure 2-14,
features two separate quill shafts, each supports an intermediate pinion and gear, and
each is supported by a respective set of bearings. Then the two quill shafts are coupled
via a third torsionally compliant shaft. This design has found its way into marine
gearboxes and has been also proposed for wind turbine gearboxes [52] but has never
been yet manufactured.

Figure 2-14: Detailed illustration of the quill shaft design [73].

The only proposed solution for equal load sharing in non-planetary wind turbine
gearboxes is the design concept seen in Figure 2-15. The concept of axially floating
intermediate shafts has been introduced twice for achieving equal power splitting among
intermediate shafts in non-planetary wind turbine gearboxes. This concept can be found
in both gearboxes from Figure 2-12. With reference to Figure 2-15, each intermediate
shaft comprises two intermediate pinions and an intermediate gear, where all gear
members are helical, and the helix angles are chosen asymmetrical for the intermediate
pinions to cancel out the axial forces by the intermediate helical gear.
22 State-of-the-Art Wind Turbine Gearboxes

Figure 2-15: Design of an asymmetrical 3-helical gear intermediate shaft.

The concept is that the shafts are left axially floating, under the assertion that the axial
float would balance the torque division among the two intermediate pinions. The design
also claims that the bearings would benefit from the cancelled axial loads. However,
what has been overlooked is that when three gear flanks; regardless helical or not, are
integrated to the same shaft, the design is in this case overconstrained. This means, for
example, that the two pinion halves can lock the shaft at an axial position, where the
teeth on the gear are not necessarily aligned properly with the teeth of its mating pinion.
With the example of the gearbox in Figure 2-12 (a), the high-speed pinions would have
to pivot to properly align its teeth with the two driving intermediate gears. This pivoting
action would force the gear axes out of parallelism, and thus lead to uneven load
distribution across the pinion face. Another disadvantage expected from this design is
the problem of axial shuttling. Frequent acceleration and deceleration scenarios are
inevitable in the variably loaded wind turbine gearboxes. In such operation conditions
where axial shuttling is permitted, severe hammering would be expected in the case of
torque reversal [74].

2.4 Concepts for Even Load Distribution across Gear Face

The following solutions have been proposed to try to maintain central tooth loads
throughout the service life of the gearbox.

2.4.1 Lead modification

Lead modification is a very well-established tooth correction method that can


compensate for possible flank deviations. The common three types of lead modification
are lead correction, end relief and crowning, which can be seen in Figure 2-16 (a), (b)
and (c) respectively [75].
State-of-the-Art Wind Turbine Gearboxes 23

(a) (b) (c)


Figure 2-16: Main 3 types of lead modification: (a) lead correction, (b) end relief and (c) crowning
[75].

The main shortcoming of applying this method to wind turbine gearboxes is that the
modifications are calculated only for the load level that contributes the most to the flank
line deviations, usually the nominal load. The first problem that faces lead modification
is that a wind turbine gearbox does not operate at a single load point. The wind subjects
the gearbox to a variable load spectrum. Sfar worked in his Dissertation [76] on
optimizing the tooth modifications for a given load spectrum. He has shown that the
modifications are a compromise between improving the running characteristics and load
capacity. As shown in Figure 2-17, modifications optimized at 50% of the nominal
torque, i.e. K50%, would exhibit unfavorable high values of the load distribution factor
𝐾𝐻𝛽 at higher load levels and vice versa. Uneven load distribution at lower load levels
would lead to higher noise and vibrations, and uneven load distribution at higher load
levels would lead to a significant increase in contact and root stress and shorten the
lifetime of the gearbox.

Figure 2-17: Progression of the load distribution factor across a given load spectrum for different
optimum tooth modifications at 50% to 200% of the nominal load. Translated from
original in German [76].
24 State-of-the-Art Wind Turbine Gearboxes

The second problem is that the gearbox manufacturer rarely designs the gearbox for a
single turbine. A gearbox is usually designed to cater for multiple platforms operating
at different sites that can have different nominal loads. This deviation in nominal input
loads is however kept a minimum and is usually in the order of magnitude of ±3-7%.

2.4.2 Flexible planet shaft with asymmetrical stiffness

This concept, seen in Figure 2-18 (a), is proposed to wide-face planets that are straddle-
mounted in cage-type planet carriers. This has been proposed recently for wind turbine
gearboxes [77; 41]. It assumes that the torsional deformation of the carrier, as seen in
Figure 2-18 (b), is the main source of gears misalignment. The torsional deformation is
a result of the stiffness asymmetry of the load paths driving the two ends of the planet
shaft. One path is directly through the primary flange, while the second path is through
the carrier bridges to the secondary flange. The proposed concept, as seen in Figure
2-18 (a), balances the two stiffnesses by assigning asymmetrical bending stiffness to the
two ends of the planet shaft.

Primary flange

Planet
Shaft

Bridges

Secondary
flange
(a) (b)
Figure 2-18: (a) Construction of an asymmetrical straddle-mounted planet shaft [77].
(b) Simulating the deformations in a double-flange carrier under load [78].

In the case of a classic symmetrical planet shaft, it will deflect in the manner seen in
Figure 2-19 (a), where the primary end advances ahead of the secondary end, forcing
the planet out of alignment, leading to a significant deflection ∆𝑏 between both ends of
the planet shaft. But in the proposed concept, the shaft would bend in the manner seen
in Figure 2-19 (b), because the asymmetrical shaft stiffness would balance the
asymmetrical load paths, leading to a reduced deflection ∆𝑏 . Thus, in the proposed
concept the planet axis approaches a straight line in the central section. This concept
offers a complete understanding of the influence of the carrier design on the planet
alignment. In addition to improving the load distribution across the gear flanks, this
State-of-the-Art Wind Turbine Gearboxes 25

concept offers other advantages as well. First, it would allow reducing the stiffness of
the carrier and achieve a significant mass reduction in the carrier mass. Second, the
reduced stiffness of the planet shaft would also help reduce the deviations in load sharing
and reduce the load sharing factor 𝐾𝛾 . However, this is only limited to the cage-type
carrier, which means this is practically applicable only to a number of planets up to 5 at
maximum.

(a)

Tin

(b)

Tin
Figure 2-19: Planet axis deflection in the direction of carrier rotation of (a) a symmetric state-of-
the-art straddle-mounted planet shaft and (b) the proposed asymmetric planet
shaft. E Youngs modulus, I second moment of area, Vc carrier angular twist, Tin
driving torque at primary flange, Δb deflection between two shaft ends, F1b & F2b
bearing forces [77].

2.5 Summary

A 10+ MW wind turbine gearbox would benefit from the collective advantages of the
presented gearbox configurations and bearing concepts. It would require the
combination of power-splitting for gearbox mass reduction, a high gearbox ratio for
generator mass reduction, and a bearing concept to achieve equal load sharing between
the gears, as well as optimizing the gear contact patterns despite the influence of
structural deformations and manufacturing tolerances.

As presented, each concept suffers from a certain limitation. There are various gearbox
configurations, each having different power-splitting features, total ratios and different
influence on the number of generators. Planetary stages have shown a compromise
between the number of planets and their stage ratio. This compromise extends to affect
26 State-of-the-Art Wind Turbine Gearboxes

all forms of multiple-stage planetary gearboxes. Non-planetary split-torque gearboxes


were found to be free from this compromise, however no successful bearing concept has
been proposed to achieve equal load sharing and even load distribution at the gears.

The chapter also presented multiple bearing concepts and has shown that they are
strongly dependent on the gearbox configuration, i.e. planetary or non-planetary, and on
the construction of the components, i.e. single- or double-flange carrier. The flexpin
design by Hicks has proven to be effective to achieve equal load sharing among high
number planets, but no reports on its ability to accommodate unexpected deformations
or non-central tooth loads. The asymmetric straddle-mounted flexpin design would
protect the planet alignment from the torsional deflection of the carrier, but it is only
applicable to low number of planets in a cage-type carrier design. On the other side for
non-planetary configurations, there has been nothing proposed except for the floating
intermediate shafts arrangement, which is expected to achieve neither equal power split
nor self-alignment. Therefore, it is the task of the next chapter to find the best gearbox
configuration and the best bearing design out of the presented concepts and integrate
their advantages into a single design.
Concepts Evaluation for Mass Reduction and Gear Mesh Optimization 27

3 Concepts Evaluation for Mass Reduction and Gear


Mesh Optimization

In this chapter, a gearbox design is proposed that attempts to deliver all the desired
research objectives, and the concept behind the choice of this proposed construction is
described. The gearbox design is developed on two fronts: the mass optimization of the
gearbox configuration and flank load optimization of the bearing concept. The first step
is to decide on the gearbox configuration; whether it is a planetary or a non-planetary
split-torque design. Depending on the outcome of the first step, the decision is made on
the bearing concept afterwards.

3.1 Proposed Idea for Gearbox Configuration Selection Method

3.1.1 Mathematical representation of system mass

It is the purpose of this section to derive a mathematical model that describes the system
mass 𝑀𝑠𝑦𝑠 in terms of the gearbox main attributes. Naturally, this mathematical model
is the sum of the gearbox and generator masses as given by equation (3-1)

𝑀𝑠𝑦𝑠 = 𝑀𝑔𝑏𝑥 + 𝑀𝑔𝑒𝑛 (3-1)

Consequently, this demands deriving a mathematical model of the gearbox and


generator masses. The mathematical model of the gearbox mass shall describe the
gearbox mass in terms of the gearbox total ratio, its distribution on the individual stages
and the number of power-split branches in each stage. The number of power-split
branches can either be the number of planets in a planetary gearbox or the number of
power-split modules in a non-planetary split-torque gearbox. The mathematical model
of the generator mass shall describe the generator mass in terms of the gearbox total
ratio and the number of output shafts, i.e. the number of generators.

This would help understand how the optimum split of the gearbox total ratio on the
individual stages is calculated. Besides, it can also help understand how the system mass
develops with the gearbox total ratio for each gearbox configuration. In other words,
this mathematical model of the system mass can generate characteristic curves for each
gearbox configuration, and then easily compare them to identify which one delivers the
minimum system mass.

The following sections describe how the mathematical models of the gearbox and
generator masses are derived.
28 Concepts Evaluation for Mass Reduction and Gear Mesh Optimization

3.1.2 Mathematical representation of gearbox mass

A gear model developed by Willis [79] in the 60’s can be used, which has been recently
revisited by FVA in 2013 [80]. This model considers the surface fatigue limit of the gear
material as the determinate criterion to represent the solid rotor volume of the gears of
a simple gear set, as the one shown in Figure 3-1, in terms of the pinion torque 𝑇1 and
gear ratio 𝑢.

Figure 3-1: Parameters of a simple gear set.

According to the model, the relationship between the dimensions of a simple gear set,
the gear ratio 𝑢 and the pinion torque 𝑇1 is governed by equation (3-2).

2
𝑢
𝑃 = 𝑇1 . 𝑛1 = 𝜎𝐻𝑙𝑖𝑚 . 𝑏. 𝑎2 . .𝑛 (3-2)
(𝑢 + 1)3 1

The allowable stress number 𝜎𝐻𝑙𝑖𝑚 accounts in this case for the material, its surface
hardness and heat treatment. The center distance 𝑎 in equation (3-2) can be expressed
in terms of the gear ratio 𝑢 using equation (3-3).

𝑎 = 0.5(𝑑1 + 𝑑2 ) = 0.5𝑑1 . (𝑢 + 1) (3-3)

Thus by substituting equation (3-3) into (3-2), the respective face width 𝑏 and pitch
diameter 𝑑1 of the pinion can be expressed as:

4𝑇1 1
𝑏𝑑12 = 2 . (1 + ) (3-4)
𝜎𝐻𝑙𝑖𝑚 𝑢

Then the respective dimensions of the mating gear are calculated from the dimensions
of the pinion to satisfy the gear ratio 𝑢 of the gear set:

4𝑇1 1 2 4𝑇1
𝑏𝑑22 = 𝑏(𝑑1 . 𝑢)2 = 2 . (1 + ). 𝑢 = 2 . (𝑢2 + 𝑢) (3-5)
𝜎𝐻𝑙𝑖𝑚 𝑢 𝜎𝐻𝑙𝑖𝑚
Concepts Evaluation for Mass Reduction and Gear Mesh Optimization 29

Thus the mass of the gearset 𝑀𝑔𝑏𝑥 can be represented in terms of the sum of gear
volumes 𝑉𝑔𝑏𝑥 in the form given in equation (3-6).

𝜋 𝜋𝑇1 1
𝑀𝑔𝑏𝑥 ~ 𝑉𝑔𝑏𝑥 = Σ𝑏𝑑 2 = 2 . (𝑢2 + 𝑢 + 1 + ) (3-6)
4 𝜎𝐻𝑙𝑖𝑚 𝑢

The work in this thesis will take the mathematical model in equation (3-6), and derive
generic models that can describe the geometrical attributes of all state-of-the-art gearbox
designs for comparative purposes. The pinion torque 𝑇1 is substituted in terms of the
input torque 𝑇𝑖𝑛 , and then both the allowable stress number 𝜎𝐻𝑙𝑖𝑚 and the input torque
𝑇𝑖𝑛 are taken as a common factor. This can be done since all state-of-the-art gearboxes
under comparison share the same material and are subjected to the same input torque.

This defines a new parameter, the gearbox mass index 𝑀𝑔𝑏𝑥 , given by equation (3-7).


Σ𝑏𝑑 2 𝑇
𝑀𝑔𝑏𝑥 ~ 𝑀𝑔𝑏𝑥 = = 𝐺𝑆 . 1⁄𝑇
4𝑇𝑖𝑛 𝑖𝑛 (3-7)
⁄𝜎 2
𝐻𝑙𝑖𝑚

Where 𝑇1 ⁄𝑇𝑖𝑛 is the share of the pinion torque from the input torque, and the factor 𝐺𝑆
includes all the geometrical attributes of that respective simple stage, and is defined as:

1
𝐺𝑆 = 𝑢2 + 𝑢 + 1 + (3-8)
𝑢

From equation (3-7) it can be seen that this mass index 𝑀𝑔𝑏𝑥 represents the gearbox
mass only in terms of the gearset attributes. Similarly, this index can be derived to all
gearbox configurations. Each gear stage, depending on its construction, would possess
its own 𝐺 factor that defines how the mass changes with the ratio and the power-splitting
in that stage. The order of this stage in a multiple-stage arrangement would determine

its torque share from the input torque 𝑇1 ⁄𝑇𝑖𝑛 . The gearbox mass index 𝑀𝑔𝑏𝑥 is then
simply the sum of mass indices of each stage as expressed in equation (3-9).

∗ ∗ (𝑘) 𝑇1 (𝑘)⁄
𝑀𝑔𝑏𝑥 = ∑ 𝑀𝑔𝑏𝑥 = ∑ 𝐺 (𝑘) . 𝑇𝑖𝑛
(3-9)

When the exact gearbox mass 𝑀𝑔𝑏𝑥 is desired at a given input torque value, it can be

calculated from the gearbox mass index 𝑀𝑔𝑏𝑥 using equation (3-10).


𝑀𝑔𝑏𝑥 (𝑘𝑔) = 𝑇𝑖𝑛 (𝑘𝑁𝑚) . 𝛼𝑔𝑏𝑥 𝑀𝑔𝑏𝑥 (3-10)

Since the mass index 𝑀𝑔𝑏𝑥 was derived from a volumetric expression of the gears, it can
be considered as the torque-to-volume ratio of the gearbox, and the constant 𝛼𝑔𝑏𝑥
30 Concepts Evaluation for Mass Reduction and Gear Mesh Optimization

converts the torque-to-volume ratio into torque-to-weight ratio. To estimate 𝛼𝑔𝑏𝑥 , first

the gearbox mass index equations 𝑀𝑔𝑏𝑥 are derived for the state-of-the-art gearboxes in
section 3.2.4. The available data of a few multi-MW gearboxes are used to calculate
their mass indices. Knowing their nominal input torque 𝑇𝑖𝑛 and mass 𝑀𝑔𝑏𝑥 , the typical
range for the constant 𝛼𝑔𝑏𝑥 can be estimated and will be further used in the comparative
study.

3.1.3 Mathematical representation of generator(s) mass

A mathematical representation of the generator mass as a function of its rated speed and
power is required. Thus, by comparing gearbox configurations having different total
ratios and number of output shafts, the impact on the mass of the generator side can also
be considered. It is already a common knowledge that the mass of the generator is
strongly affected by its rated speed. For the same rated power, the mass of the generator
drops when it is driven at a higher speed. This has already been the driver of a recent
research project [81], which aimed for utilizing high-speed generators in the 5000 rpm
range for the sake of reducing mass, costs and volume. Nonetheless, the knowledge of
how the mass develops with both the rated speed and power is still not well established.
The only available source of information is published data sheets by generator
manufacturers. One manufacturer in particular published enough sum of data for
medium- and high-speed permanent magnet generators [82; 83], which serves as a good
population size for curve fitting. Similar to the gearbox, it is expected that the mass of
the generator is directly proportional to its rated torque. Therefore, the proposed
mathematical representation governing the generator mass 𝑀𝑔𝑒𝑛 in kg, generator rated
power 𝑃𝑔𝑒𝑛 in MW and speed 𝑛𝑔𝑒𝑛 in rpm is the exponential assumption given in
equation (3-11), where 𝛼gen,1 and 𝛼gen,2 are the equation constants:
𝛼gen,2
𝑃𝑔𝑒𝑛 (𝑀𝑊)
𝑀𝑔𝑒𝑛 (𝑡𝑜𝑛) = 𝛼gen,1 . ( ⁄𝑛 ) (3-11)
𝑔𝑒𝑛 (𝑟𝑝𝑚)

Using the curve fitting tool in MATLAB, the best fit gave the values 123900 and 0.4639
for the equation constants 𝛼𝑔𝑒𝑛,1 and 𝛼𝑔𝑒𝑛,2 respectively, with adjusted R-square statistic
of 0.9814 (values closer to 1 indicate a better fit). By substituting the generator power
𝑃𝑔𝑒𝑛 in terms of the total turbine power 𝑃𝑒𝑙𝑒𝑐 and the number of generators 𝑁𝑔𝑒𝑛 , and by
substituting the generator rated speed 𝑛𝑔𝑒𝑛 in terms of the turbine rotor speed 𝑛𝑅 and
the gearbox ratio 𝑖𝑡𝑜𝑡𝑎𝑙 , the mass of the generator(s) 𝑀𝑔𝑒𝑛 can be expressed as:
Concepts Evaluation for Mass Reduction and Gear Mesh Optimization 31

0.4639
𝑃𝑒𝑙𝑒𝑐
⁄𝑁
𝑔𝑒𝑛
𝑀𝑔𝑒𝑛 = 123900 𝑁𝑔𝑒𝑛 ( ) (3-12)
𝑛𝑅 . 𝑖𝑡𝑜𝑡𝑎𝑙


By rearranging equations (3-12) we can define a generator mass index 𝑀𝑔𝑒𝑛 that can be
used to examine the dependency of the mass on the total gearbox ratio 𝑖𝑡𝑜𝑡𝑎𝑙 and the
number of the generators 𝑁𝑔𝑒𝑛 for a given gearbox configuration.


𝑁𝑔𝑒𝑛 0.5361
𝑀𝑔𝑒𝑛 = (3-13)
𝑖𝑡𝑜𝑡𝑎𝑙 0.4639

When it is desired to estimate the exact mass of the generator side 𝑀𝑔𝑒𝑛 , the parameters
of the turbine are then substituted in equation (3-14).

123900 𝑃𝑒𝑙𝑒𝑐 (𝑀𝑊) ∗


𝑀𝑔𝑒𝑛 (𝑘𝑔) = . 𝑀𝑔𝑒𝑛 (3-14)
𝑛𝑅 (𝑟𝑝𝑚)

3.2 Applying Proposed Method to State-of-the-Art Gearboxes

The work in this section is dedicated to converting the Willis equations into the
mathematical form proposed in equation (3-9). The geometrical factor 𝐺, and the

gearbox mass index 𝑀𝑔𝑏𝑥 are first derived for the building units of the state-of-the-art
gearboxes, namely a simple planetary stage, a stepped planetary stage and a dual power-
split module. The limitations on the gear ratio and the power splitting are also presented
for each building unit. Then the outcome is generalized and applied to the multiple-stage
configurations. In the final section 3.2.5, the attributes of each gearbox configuration
are explained, and the final results of the comparative study are presented.

3.2.1 Mathematical model derivation for a simple planetary stage

According to Willis [79], equation (3-4) is first applied to the planet-sun gear mesh. This
yields equation (3-15) for the sun gear dimensions, where 𝑢1𝑝 is the planet-to-sun ratio
𝑢1𝑝 = 𝑑𝑝 ⁄𝑑1 , 𝑇1 is the sun shaft torque and 𝑁𝑝 is the number of planets.

𝑇
4 1⁄𝑁 1 4𝑇1 1 1
𝑝 (3-15)
𝑏𝑑12 = 2 . (1 + ) = 2 .( + )
𝜎𝐻𝑙𝑖𝑚 𝑢1𝑝 𝜎𝐻𝑙𝑖𝑚 𝑁𝑝 𝑁𝑝 𝑢1𝑝

The total volume of 𝑁𝑝 planets is derived in relation to the sun gear dimensions as:
32 Concepts Evaluation for Mass Reduction and Gear Mesh Optimization

2 4𝑇1 1 1 2
𝑁𝑝 . 𝑏𝑑𝑝2 = 𝑁𝑝 . 𝑏(𝑑1 𝑢1𝑝 ) = 𝑁𝑝 . 2 .( + ) . 𝑢1𝑝
𝜎𝐻𝑙𝑖𝑚 𝑁𝑝 𝑁𝑝 𝑢1𝑝
4𝑇1 2
= 2 . (𝑢1𝑝 + 𝑢1𝑝 )
𝜎𝐻𝑙𝑖𝑚 (3-16)

Similarly, the dimensions of the ring are derived from the ring-to-sun ratio 𝑢12 = 𝑑2 ⁄𝑑1 .

4𝑇1 1 1
𝑏𝑑22 = 𝑏(𝑑1 𝑢12 )2 = 2 .( + 2
) . 𝑢12
𝜎𝐻𝑙𝑖𝑚 𝑁𝑝 𝑁𝑝 𝑢1𝑝
2 2
4𝑇1 𝑢12 𝑢12
= 2 .( + ) (3-17)
𝜎𝐻𝑙𝑖𝑚 𝑁𝑝 𝑁𝑝 𝑢1𝑝

By substituting equations (3-15), (3-16) and (3-17) into equation (3-7), a generalized

model of the mass index 𝑀𝑔𝑏𝑥 of a planetary stage is given by equation (3-18). As seen
in the equation, Willis estimates the actual volume of the ring gear and the structural
components at only a fraction of 0.4 of the ring hypothetical total solid volume.

∗ (𝑘)
𝑏𝑑12 + 𝑁𝑝 . 𝑏𝑑𝑝2 + 0.4𝑏𝑑22 (𝑘) 𝑇 (𝑘)
𝑀𝑔𝑏𝑥 = = 𝐺𝑃 . 1 ⁄𝑇
4𝑇𝑖𝑛 𝑖𝑛 (3-18)
⁄𝜎 2
𝐻𝑙𝑖𝑚

Where for a given stage (𝑘), the factor 𝑇1(𝑘) ⁄𝑇𝑖𝑛 is the ratio of sun torque to the input
(𝑘)
torque, and the factor 𝐺𝑃 includes all the geometrical attributes of that respective
planetary stage as:

(𝑘) 2 (𝑘) 2
(𝑘) 1 1 (𝑘) 2 (𝑘) 0.4𝑢12 0.4𝑢12
𝐺𝑃 = ( (𝑘)
+ (𝑘) (𝑘)
+ 𝑢1𝑝 + 𝑢1𝑝 + (𝑘)
+ (𝑘) (𝑘)
) (3-19)
𝑁𝑝 𝑁𝑝 . 𝑢1𝑝 𝑁𝑝 𝑁𝑝 . 𝑢1𝑝

The theoretical number of planets 𝑁𝑝 per stage (𝑘) can be approximately estimated by
expressing it as the maximum number of adjacent pitch circles that can fit in the space
between the ring and the sun. The maximum number of planets can thus be calculated
for each ring-to-sun ratio 𝑢12 as:
(𝑘)
(𝑘) −1
𝑢12 − 1
𝑁𝑝 = 𝜋 / sin ( (𝑘)
) (3-20)
𝑢12 + 1

The practical values of the maximum number of planets would be lower, since a
minimum clearance would be required between the adjacent addendum circles. The
planet-to-sun ratio 𝑢1𝑝 can also be expressed in terms of the ring-to-sun ratio 𝑢12 as:
Concepts Evaluation for Mass Reduction and Gear Mesh Optimization 33

(𝑘) 1 (𝑘) 1
𝑢1𝑝 = .𝑢 − (3-21)
2 12 2

Both equations (3-20) and (3-21) render the mass index 𝑀𝑔𝑏𝑥 in equation (3-18) as a
function of a single parameter, i.e. the ring-to-sun ratio 𝑢12 . In the practice of wind
turbine gearboxes, the ring-to-sun ratio would usually range from 2.4 to 6.4 [84]. The
state-of-the-art cage-type carrier would limit the number of planets up to 5. Therefore,
the 2.4 to 6.4 ring-to-sun ratio range would correspond to a range of 3 to 5 planets.

3.2.2 Mathematical model derivation for a stepped planetary stage

The number of teeth and the torque values used in the equations can be understood with
reference to Figure 3-2.

Figure 3-2: Breakdown of the torque transfer in a stepped planetary stage.

Similar to a simple planetary stage the derivation starts with the planet-sun gear mesh.
The dimensions of the sun gear are given by equation (3-22), where 𝑢1𝑝2 is the planet
gear-to-sun ratio 𝑑𝑝2 ⁄𝑑1 , 𝑇1 is the sun shaft torque and 𝑁𝑝 is the number of stepped
planets.

𝑇
4 1⁄𝑁 1 4𝑇1 1 1
𝑝 (3-22)
𝑏𝑑12 = 2 . (1 + ) = 2 .( + )
𝜎𝐻𝑙𝑖𝑚 𝑢1𝑝2 𝜎𝐻𝑙𝑖𝑚 𝑁𝑝 𝑁𝑝 𝑢1𝑝2

The total volume of 𝑁𝑝 planet gears is derived in relation to the sun gear volume as:

2 4𝑇1 1 1 2
𝑁𝑝 . 𝑏𝑑𝑝22 = 𝑁𝑝 . 𝑏(𝑑1 𝑢1𝑝2 ) = 𝑁𝑝 . 2 . ( + ) . 𝑢1𝑝
𝜎𝐻𝑙𝑖𝑚 𝑁𝑝 𝑁𝑝 𝑢1𝑝2 2

4𝑇1 2 (3-23)
= 2 . (𝑢1𝑝2
+ 𝑢1𝑝2 )
𝜎𝐻𝑙𝑖𝑚
34 Concepts Evaluation for Mass Reduction and Gear Mesh Optimization

The total volume of 𝑁𝑝 planet pinions is derived in equation (3-24), where 𝑢𝑝12 is the
ring-to-planet pinion ratio 𝑑2 ⁄𝑑𝑝1 and 𝑇𝑝1 is the stepped planet torque.

4𝑇𝑝1 1
𝑁𝑝 . 𝑏𝑑𝑝21 = 𝑁𝑝 . 2 . (1 + )
𝜎𝐻𝑙𝑖𝑚 𝑢𝑝12 (3-24)

Similarly, the dimensions of the ring are derived from the planet pinion as:

2 4𝑇𝑝1 1 4𝑇𝑝1 (3-25)


𝑏𝑑22 = 𝑏(𝑑𝑝1 𝑢𝑝12 ) = 2 . (1 + ) . 𝑢 2
𝑝 2 = 2 . (𝑢𝑝212 + 𝑢𝑝12 )
𝜎𝐻𝑙𝑖𝑚 𝑢𝑝12 1
𝜎𝐻𝑙𝑖𝑚

The share of the stepped planet torque 𝑇𝑝1 of the sun shaft torque 𝑇1 given by:

𝑇
𝑇𝑝1 = 1⁄𝑁 . 𝑢1𝑝2 (3-26)
𝑝

By substituting equations (3-22) to (3-26) into equation (3-7), a generalized model of



the mass index 𝑀𝑔𝑏𝑥 of a stepped planetary stage is thus given by equation (3-27). Here
again the actual mass of the ring gear and the structural components is estimated at a
fraction of 0.4 of the ring hypothetical total solid volume.

∗ (𝑘)
𝑏𝑑12 + 𝑁𝑝 . 𝑏𝑑𝑝21 + 𝑁𝑝 . 𝑏𝑑𝑝22 + 0.4𝑏𝑑22 (𝑘) 𝑇 (𝑘)
𝑀𝑔𝑏𝑥 = = ∑ 𝐺𝐶 . 1 ⁄𝑇
4𝑇𝑖𝑛 𝑖𝑛 (3-27)
⁄𝜎 2
𝐻𝑙𝑖𝑚

Where for a given stage (𝑘), the factor 𝑇1(𝑘) ⁄𝑇𝑖𝑛 is the ratio of sun gear torque to the
(𝑘)
input torque, and the factor 𝐺𝐶 includes all the geometrical attributes of that respective
stepped planetary stage as:

(𝑘) (𝑘) (𝑘) 2 (𝑘)


(𝑘) 1 1 (𝑘) 2 (𝑘) 𝑢1𝑝2 0.4𝑢1𝑝2 (𝑢𝑝12 + 𝑢𝑝1 2 )
𝐺𝐶 = (𝑘)
+ (𝑘) (𝑘)
+ 𝑢1𝑝2 + 2𝑢1𝑝2 + (𝑘)
+ (𝑘)
(3-28)
𝑁𝑝 𝑁𝑝 𝑢1𝑝2 𝑢𝑝12 𝑁𝑝

The number of stepped planets in this case is determined by the maximum number of
planet gears in mesh with the sun without collision, and can be expressed by:

(𝑘) 2
(𝑘) 𝑢1𝑝2
𝑁𝑝 = 2𝜋 / cos −1 [1 − 2 ( (𝑘)
) ] (3-29)
1 + 𝑢1𝑝2


Substituting equation (3-29) into (3-28) shows that the mass index 𝑀𝑔𝑏𝑥 of a stepped
planetary stage is dependent on the distribution of the total ratio on both the ring-to-
planet pinion 𝑢𝑝12 and sun-to-planet gear 𝑢1𝑝2 ratios. The possible combinations of these
Concepts Evaluation for Mass Reduction and Gear Mesh Optimization 35

two ratios are however limited as once examined in [85]. This is because, as shown in
Figure 3-2, the center distance 𝑎 is constant across the stage. Increasing the ring-to-
planet pinion 𝑢𝑝12 would increase the total ratio and decrease the sun torque 𝑇1 , and as
per equations (3-22) and (3-23) the stepped gear and sun should decrease in size.
However, this is not the case because on the contrary, increasing the ring-to-planet
pinion 𝑢𝑝12 would increase the center distance 𝑎, forcing the sizes of the stepped gears
and sun to increase. This limits the increase of ring-to-planet pinion 𝑢𝑝1 2 up to around
5 and consequently the total ratio is limited between 10 to 15. The number of stepped
planets is limited to 4, as previously explained in section 2.1.5.

3.2.3 Mathematical model derivation for a dual power-split module

The schematic of a dual power-split module can be seen in Figure 3-3. In this section a
generic mathematical model is derived for 𝑁𝑚 number of power-split modules. The total
volume of 2𝑁𝑚 intermediate pinions in the first stage is derived in equation (3-30)
similar to equation (3-4).

2 4𝑇1′ 1
2𝑁𝑚 . 𝑏𝑑1′ = 2𝑁𝑚 . 2 . (1 + ′ ) (3-30)
𝜎𝐻𝑙𝑖𝑚 𝑢

The volume of the central bull gear is derived in equation (3-31) relative to the pinion
volume to satisfy the stage ratio 𝑢′ .

2 4𝑇1′ 2
𝑏𝑑2′ = 𝑏(𝑑1′ . 𝑢′ )2 = 2 . (𝑢′ + 𝑢′ ) (3-31)
𝜎𝐻𝑙𝑖𝑚

(a) (b)

Figure 3-3: (a) Schematic of dual power-split module. (b) Breakdown of the torque transfer in
a power-split module.
36 Concepts Evaluation for Mass Reduction and Gear Mesh Optimization

By substituting equations (3-30) and (3-31) into (3-7), a generalized model of the index

𝑀𝑔𝑏𝑥 of the first power-splitting stage is thus given by equation (3-32). There is no
available data to tell how the actual volume of the bull gear and housing relates to its
solid volume, however assuming the same factor 0.4 like in a planetary stage is
considered a plausible assumption.
2 2
∗ ′
2𝑁𝑚 . 𝑏𝑑1′ + 0.4𝑏𝑑2′ 𝑇′
𝑀𝑔𝑏𝑥 = = 𝐺𝑆𝑝𝑙𝑖𝑡 . 1⁄𝑇 (3-32)
4𝑇𝑖𝑛 𝑖𝑛
⁄𝜎 2
𝐻𝑙𝑖𝑚

Where for the first stage, the factor 𝑇1′ ⁄𝑇𝑖𝑛 is the ratio of the intermediate shaft torque
to the input torque, and the factor 𝐺𝑆𝑝𝑙𝑖𝑡 includes all the geometrical attributes of that
respective stage as:

1 2
𝐺𝑆𝑝𝑙𝑖𝑡 = 2𝑁𝑚 . (1 + ) + (0.4𝑢′ + 0.4𝑢′ ) (3-33)
𝑢′
Similarly, the dimensions of the 𝑁𝑚 power-summing pinions of the second stage are
also derived by applying equation (3-4), which gives:

𝑇1′′⁄
4 1
= 𝑁𝑚 . 2 2 . (1 + ′′ )
2 (3-34)
𝑁𝑚 . 𝑏𝑑1′′
𝜎𝐻𝑙𝑖𝑚 𝑢

Using the second stage ratio 𝑢′′ the total dimensions of the 2𝑁𝑚 intermediate gears is
given by:

𝑇1′′⁄
4
= 2𝑁𝑚 . 𝑏(𝑑1′′ . 𝑢′′ )2 = 2𝑁𝑚 . 2 2 . ( 𝑢′′ + 𝑢′′ )
2 2 (3-35)
2𝑁𝑚 . 𝑏𝑑2′′
𝜎𝐻𝑙𝑖𝑚

By substituting equation (3-34) and (3-35) into (3-7), a generalized model of the mass
∗ ′′
index 𝑀𝑔𝑏𝑥 of the second power summing stage is thus given by:

2 2
∗ ′′
𝑁𝑚 . 𝑏𝑑1′′ + 2𝑁𝑚 . 𝑏𝑑2′′ 𝑇 ′′
𝑀𝑔𝑏𝑥 = = 𝐺𝑆𝑢𝑚 . 1 ⁄𝑇 (3-36)
4𝑇𝑖𝑛 𝑖𝑛
⁄𝜎 2
𝐻𝑙𝑖𝑚

Where for the second stage, the factor 𝑇1′′ ⁄𝑇𝑖𝑛 is the ratio of the high-speed pinion torque
to the input torque, and the factor 𝐺𝑆𝑢𝑚 includes all the geometrical attributes of that
respective stage as:

𝑁𝑚 1 2
𝐺𝑆𝑢𝑚 = . (1 + ′′ ) + 𝑁𝑚 . ( 𝑢′′ + 𝑢′′ ) (3-37)
2 𝑢
Concepts Evaluation for Mass Reduction and Gear Mesh Optimization 37

Unlike planetary gearboxes – whether simple or stepped – there is no relation between


the first stage gear ratio 𝑢′ and the number of power-split modules 𝑁𝑚 ; both parameters
can be set independently from each other. The only thing that would limit increasing
both parameters is the limit on the gearbox housing size. For a fair comparison with
planetary configurations, the limit on the first stage ratio should be the spatial equivalent
of the ratio limits on a planetary stage, as seen in Figure 3-4.

Figure 3-4: First stage of a planetary (left) and non-planetary (right) assuming the same
envelope size.

With reference to Figure 3-4, the first stage ratio in the non-planetary configuration
should fulfill the condition given by equation (3-38).

[𝑑1 + 2𝑑𝑝 ] = [2𝑑1 + 𝑑2 ]𝑁𝑜𝑛−𝑃𝑙𝑎𝑛𝑒𝑡𝑎𝑟𝑦


𝑃𝑙𝑎𝑛𝑒𝑡𝑎𝑟𝑦

[𝑑1 ]𝑠𝑢𝑛 . (1 + 2𝑢1𝑝 ) = [𝑑1 ]𝑝𝑖𝑛𝑖𝑜𝑛 . (2 + 𝑢′ ) (3-38)

To solve equation (3-38), a similar study [86] proposed the simple idea of rewriting the
volumetric expression of the pinions in terms of its face width-to-diameter ratio 𝑏⁄𝑑1 .
The minimum volume gears would favor the highest values of this ratio, however its
increase results in a higher torsional deflection of the teeth. For this reason an upper
limit is set for this ratio around 1.2 [87]. This would transform the sun volume equation
(3-15) into equation (3-39), and the intermediate pinion volume equation (3-30) into
equation (3-40).

4𝑇1 1 1
[𝑑13 ]𝑠𝑢𝑛 = .( + )
2
𝜎𝐻𝑙𝑖𝑚 . (𝑏⁄𝑑 ) 𝑁𝑝 𝑁𝑝 𝑢1𝑝
1

4 1 1 𝑇𝑖𝑛
= .( + ). (3-39)
2
𝜎𝐻𝑙𝑖𝑚 . (𝑏⁄𝑑 ) 𝑁𝑝 𝑁𝑝 𝑢1𝑝 (1 + 𝑢12 )
1
38 Concepts Evaluation for Mass Reduction and Gear Mesh Optimization

4𝑇1′ 1
[𝑑13 ]𝑝𝑖𝑛𝑖𝑜𝑛 = . (1 + ′ )
2
𝜎𝐻𝑙𝑖𝑚 . (𝑏⁄𝑑 ) 𝑢
1

4 1 𝑇𝑖𝑛
= . (1 + ). (3-40)
. (𝑏⁄𝑑 ) 𝑢 2𝑁𝑚 𝑢′
2 ′
𝜎𝐻𝑙𝑖𝑚
1

2
The input torque 𝑇𝑖𝑛 and material 𝜎𝐻𝑙𝑖𝑚 are taken as common, and it is assumed that
both gears are designed to their maximum limit of the ratio 𝑏⁄𝑑1 . Therefore, substituting
equations (3-39) and (3-40) into equation (3-38) yields the expression:

1 1 1
3( + ) 3
(1 + ′ )
√ 𝑁𝑝 𝑁𝑝 𝑢1𝑝 𝑢 . (2 + 𝑢 ′ ) (3-41)
. (1 + 2𝑢1𝑝 ) = √
(1 + 𝑢12 ) 2𝑁𝑚 𝑢′

As discussed previously, the maximum ring-to-sun ratio 𝑢12 is set at 6.4, and the
corresponding number of planets 𝑁𝑝 and planet-to-sun ratio 𝑢1𝑝 can be calculated from
equations (3-20) and (3-21). This renders equation (3-41) as a function of the first stage
ratio 𝑢′ and the number of power-split modules 𝑁𝑚 . By applying this condition, the
maximum limit on the first stage ratio for a different number of power-split modules 𝑁𝑚
can be seen in Figure 3-5. This is due to the fact that increasing the power-splitting, i.e.
higher 𝑁𝑚 , in the gearbox reduces the loads, thus decreases the gear dimensions, which
in turn gives more space to increase the ratio of the first stage 𝑢′ within the same gearbox
housing dimensions.

Figure 3-5: Maximum first stage ratio of a non-planetary stage equivalent to a planetary stage.
Concepts Evaluation for Mass Reduction and Gear Mesh Optimization 39

3.2.4 Model application to state-of-the-art gearboxes

The mass index equations of the state-of-the-art gearboxes are compiled in Table 3-1.

The table shows the mathematical model of the mass index 𝑀𝑔𝑏𝑥 of each stage in the
state-of-the-art wind turbine gearboxes from chapter 2. The torque share of each stage
from the input torque is also given in the table. For the sake of comparison, a fixed
number of three stages is held constant for the gearbox configurations under comparison.
In case the third stage is not planetary, the third stage is usually a simple gear stage that

practically has a ratio up to 7. For calculating the generator mass index 𝑀𝑔𝑒𝑛 for each
gearbox configuration, the number of the driven generators are also listed in the table.

∗ ∗ (𝑘) 𝑇1(𝑘) 𝑁𝑔𝑒𝑛


Configuration 𝑀𝑔𝑏𝑥 = ∑ 𝑀𝑔𝑏𝑥 ⁄𝑇
𝑖𝑛

|𝑧2 | ′
∗ ′ 𝑇1′ 𝑇1′
𝑀𝑔𝑏𝑥 = 𝐺𝑃′ . ⁄𝑇 ⁄𝑇 = 1⁄(1 + ′ )
𝑖𝑛 𝑖𝑛 𝑧 1
PPS
|𝑧2 | |𝑧2 | ′ ′′
1 Series ∗ ′′ 𝑇1′′ 𝑇1′′
𝑀𝑔𝑏𝑥 = 𝐺𝑃′′ . ⁄𝑇 ⁄𝑇 = 1⁄[(1 + ′ ) . (1 + ′′ )]
𝑖𝑛 𝑖𝑛 𝑧 𝑧 1 1
planetary
∗ ′′′ 𝑇1′′′ 𝑇1′′′
𝑀𝑔𝑏𝑥 = 𝐺𝑆′′′ . ⁄𝑇 ⁄𝑇 = 1⁄𝑖𝑡𝑜𝑡𝑎𝑙
𝑖𝑛 𝑖𝑛

|𝑧2 | |𝑧2 | ′
|𝑧2 | ′ ′′
∗ ′ 𝑇1′ 𝑇1′
𝑀𝑔𝑏𝑥 = 𝐺𝑃′ . ⁄𝑇 ⁄𝑇 = 1⁄[(1 + ′ ) + ( ′ ) . ( ′′ )]
𝑖𝑛 𝑖𝑛 𝑧 𝑧 𝑧1 1 1
2-stage
2 differential ∗ ′′ 𝑇1′′ 𝑇1′′ |𝑧2′ | |𝑧2′ | |𝑧2′ | |𝑧2′′ |
𝑀𝑔𝑏𝑥 = 𝐺𝑃′′ . ⁄𝑇 ⁄𝑇 = ⁄[(1 + ) + ( ′ ) . ( ′′ )]
𝑖𝑛 𝑖𝑛 𝑧1′ 𝑧1′ 𝑧1 𝑧1
+1 simple
∗ ′′′ 𝑇1′′′ 𝑇1′′′ 1
𝑀𝑔𝑏𝑥 = 𝐺𝑆′′′ . ⁄𝑇 ⁄𝑇 = 1⁄𝑖𝑡𝑜𝑡𝑎𝑙
𝑖𝑛 𝑖𝑛

|𝑧2 | ′′′
∗ ′ 𝑇1′ 𝑇1′
𝑀𝑔𝑏𝑥 = 𝐺𝑃′ . ⁄𝑇 ⁄𝑇 = (1 + ′′′ )⁄𝑖𝑡𝑜𝑡𝑎𝑙
𝑖𝑛 𝑖𝑛 𝑧 1
3-stage |𝑧2 | ′′′
3 ∗ ′′ 𝑇1′′ 𝑇1′′
𝑀𝑔𝑏𝑥 = 𝐺𝑃′′ . ⁄𝑇 ⁄𝑇 = ( ′′′ )⁄𝑖𝑡𝑜𝑡𝑎𝑙
differential 𝑖𝑛 𝑖𝑛 𝑧 1

∗ ′′′ 𝑇 ′′′ 𝑇1′′′


𝑀𝑔𝑏𝑥 = 𝐺𝑃′′′ . 1 ⁄𝑇 ⁄ 𝑇𝑖𝑛 = 1⁄𝑖𝑡𝑜𝑡𝑎𝑙
𝑖𝑛

Stepped ∗ ′ 𝑇1′ 𝑇1′ |𝑧2 | 𝑧𝑝2


𝑀𝑔𝑏𝑥 = 𝐺𝐶′ . ⁄𝑇 ⁄𝑇 = 1⁄(1 + . )
𝑖𝑛 𝑖𝑛 𝑧𝑝 𝑧1 1
4 planetary
∗ ′′ 𝑇 ′′ 𝑇1′′
+1 simple 𝑀𝑔𝑏𝑥 = 𝐺𝑆′′ . 1 ⁄𝑇 𝑇𝑖𝑛 = 1⁄𝑖𝑡𝑜𝑡𝑎𝑙

𝑖𝑛

|𝑧2 | ′
∗ ′ ′ 𝑇1′ 𝑇1′
𝑀𝑔𝑏𝑥 = 𝐺𝑆𝑝𝑙𝑖𝑡 . ⁄𝑇 ⁄𝑇 = 1⁄(2𝑁𝑚 . ′ )
𝑖𝑛 𝑖𝑛 𝑧 1
3-stage non- |𝑧2′ | |𝑧2′′ |
5 ∗ ′′ ′′ 𝑇1′′ 𝑇1′′ 𝑁𝑚
𝑀𝑔𝑏𝑥 = 𝐺𝑆𝑢𝑚 . ⁄𝑇 ⁄𝑇 = 1⁄(𝑁𝑚 . ′ . ′′ )
planetary 𝑖𝑛 𝑖𝑛 𝑧1 𝑧1
∗ ′′′ 𝑇1′′ 𝑇1′′′
𝑀𝑔𝑏𝑥 = 𝐺𝑆′′′ . ⁄𝑇 𝑁𝑚 ⁄𝑇 = 1⁄(𝑁𝑚 . 𝑖𝑡𝑜𝑡𝑎𝑙 )
𝑖𝑛 𝑖𝑛

Table 3-1: Mass indices 𝑴∗𝒈𝒃𝒙 for state-of-the-art wind turbine gearbox configurations.
40 Concepts Evaluation for Mass Reduction and Gear Mesh Optimization

To estimate the constant 𝛼𝑔𝑏𝑥 in equation (3-10), the available data of a few number of
multi-MW wind turbine gearboxes alongside a research 6 MW gearbox [88] have been
used. The constant is found to fall in the range of 4 to 6.5. The deviation in the value of
the constant is attributed to a great extent to the safety factors. Some manufacturers
design their gearboxes exactly at the minimum safety factor limit demanded by the
guidelines and some target a reserve safety factor, which has its mass penalty. The work
in this dissertation would proceed with the conservative limit, i.e. upper limit, of this
factor, i.e. 𝛼𝑔𝑏𝑥 = 6.5.

3.2.5 Results and gearbox configuration selection

3.2.5.1 Characteristics of a simple planetary stage



Applying the proposed mathematical model of the gearbox mass index 𝑀𝑔𝑏𝑥 to a simple
planetary stage has helped to a great extent understand how the number of planets
influence its mass and ratio. First, with reference to Figure 3-6, a planetary stage suffers
from a compromise between its torque capacity and its gear ratio. Adding more planets
increases the power splitting, which in turn decreases the gear and bearing loads, and
leads to mass reduction. But on the other side, increasing the planets reduces the gear
ratio, thus increases the output torque. Consequently, this increase in output torque
would cause a mass penalty on the subsequent stages.

Second, the shaded region in Figure 3-6 represents the unexploited mass reduction
potential from a planetary stage, because of the fear of unevenness of load distribution
across the gear width. For the reasons explained in section 2.1.2, more than 5 planets
would demand a change in the state-of-the-art cage-type carrier into a single-flange
design. This would lead to narrower gears in order to reduce the excessive planet pin
deflection attributed to this carrier design. Reducing the gear face width, as per equation
(3-39), would lead to a volume penalty.

Finally, the mass index 𝑀𝑔𝑏𝑥 model can be used to explain the impact of load sharing in
planetary gearboxes on their mass reduction potential. The inequality of load sharing,
according to ISO 6336 [64], is denoted by the factor 𝐾𝛾 . According to publications, e.g.
[63], and wind turbine gearbox certification guidelines [89; 90], this factor increases
with the increase in planets. This factor can be incorporated in the mass index model

𝑀𝑔𝑏𝑥 in equation (3-9) by substituting the torque ratio 𝑇1 ⁄𝑇𝑖𝑛 with 𝐾𝛾 𝑇1 ⁄𝑇𝑖𝑛 . Equation
(3-42), for example, shows the relation between the number of planets 𝑁𝑝 and the load
sharing factor 𝐾𝛾 [72]. As seen in Figure 3-6, there is a mass penalty from incorporating
the load sharing factor 𝐾𝛾 , and how it offsets the mass reduction by increasing the
number of planets 𝑁𝑝 .
Concepts Evaluation for Mass Reduction and Gear Mesh Optimization 41

𝐾𝛾 = 1 + 0.25√𝑁𝑝 − 3 (3-42)

Figure 3-6: Impact of number of planets 𝑵 in a simple planetary stage on the mass index 𝑴∗𝒈𝒃𝒙
and the stage ratio.

These observations can actually explain the current gearbox design trends on the market.
All multi-MW gearboxes, with very few exceptions, have a maximum of 4 planets in
the first stage. Wind turbine gearbox manufacturers would shy away from going for a
first stage with 5+ planets, and only adopt a sturdy cage-type carrier design that
accommodates only 3 to 4 wide-face planets. In other words, the manufacturers would
rather sacrifice the mass reduction potential for saving the gearbox reliability.

3.2.5.2 Characteristics of a multi-stage planetary gearbox

The trade-off between the mass reduction and gear ratio in a single planetary stage
repeats itself in all types of multiple-stage planetary configurations. The results of

applying the mass index 𝑀𝑔𝑏𝑥 mathematical models have shown that the minimum mass
of a multiple-stage planetary gearbox – regardless which configuration – favors
increasing the number of planets in the high-torque stage over increasing its ratio. In
other words, increasing the number of planets in the high-torque stage has priority over
increasing the gear ratio of this stage. The high-torque stage is the first stage in the series
planetary PPS and stepped planetary configurations and is the star stage in the
configurations with a differential stage, i.e. MAAG and Bosch concepts. Figure 3-7 (a),
with the example of a series planetary PPS configuration, shows that increasing the
number of planets in the first stage decreases the total mass of the gearbox. The same
figure also shows that this increase leads to a slight mass penalty in the subsequent
stages, because their input torque increases with decreasing the first stage ratio.
42 Concepts Evaluation for Mass Reduction and Gear Mesh Optimization

However as evident in the figure, increasing the number of planets in the first stage has
the upper hand. Figure 3-7 (b) shows how this trade-off impacts the total gearbox mass
across a given range of gearbox total ratio. Increasing the total ratio is achieved by
increasing the individual stage ratios, which in turn comes at the expense of losing more
planets and reducing the torque capacity. This explains the mass increase associated
with the increase of the total ratio in Figure 3-7 (b).

𝑴∗𝒈𝒃𝒙,𝒕𝒐𝒕𝒂𝒍

𝑴∗𝒈𝒃𝒙
𝑵′
′′′ ′′
𝑴∗𝒈𝒃𝒙 𝑴∗𝒈𝒃𝒙

𝑴∗𝒈𝒃𝒙

(a) (b)
Figure 3-7: (a) Impact of 1st stage number of planets 𝑵′ of a series planetary PPS configuration
on the individual masses of each stage, with the example of total gearbox ratio
𝒊𝒕𝒐𝒕𝒂𝒍 = 𝟎𝟎. (b) Development of gearbox mass index and optimum number of
planets in the first stage 𝑵′ across a total gearbox ratio 𝒊𝒕𝒐𝒕𝒂𝒍 = 𝟎: 𝟓𝟎.

Finally, the equations from Table 3-1 are applied to all the state-of-the-art multiple-stage
planetary gearboxes. As seen in Figure 3-8, the equations are applied across a range of
gearbox total ratio from 30 to 350. The figure shows that the three-stage differential
gearbox, i.e. the Bosch concept, has the advantage over the other concepts up to a total
ratio around 50, above this value increasing the gearbox ratio demands losing planets
from both the first two stages causing the strong spike in the gearbox mass. This could
be the explanation why this gearbox in the high-speed range is complemented in practice
by a fourth simple stage. The figure also shows how the limitation on the number of
stepped planets reflects itself on the stepped planetary gearbox mass. Finally, the figure
shows the series planetary PPS configuration as the dominant design in the high-speed
range, which matches the current trend on the market.
Concepts Evaluation for Mass Reduction and Gear Mesh Optimization 43

3-stage differential
Stepped 2-stage differential+
1 simple stage

Series planetary
PPS

Figure 3-8: Gearbox mass index 𝑴∗𝒈𝒃𝒙 of state-of-the-art multiple-stage planetary gearboxes.

3.2.5.3 Characteristics of non-planetary split-torque gearboxes



Applying the proposed mathematical model of the gearbox mass index 𝑀𝑔𝑏𝑥 to the non-
planetary split-torque design has shown different characteristics. Unlike planetary
gearboxes, this design is completely free from the compromise between power splitting
and stage ratio. As seen in Figure 3-9 (a), the stage ratio and the number of power-split
modules 𝑁𝑚 are two completely independent parameters. For a given number of power-
split modules 𝑁𝑚 , the first stage ratio 𝑢′ can be increased freely, and is only constrained
by the limit on the center distance imposed by the limits on the gearbox envelope size
as previously explained in section 3.2.3, Figure 3-5. This independent increase of the
first stage ratio impacts all three stages positively. The first stage itself would suffer
from a negligible increase in mass because of the increase in the bull gear size, which is
mostly compensated by the reduction in the masses of its mating pinions. In addition,
the increase in the first stage ratio would reduce the torque transmitted to the following
stages, and consequently reducing their masses as well. With reference to the dotted line
in Figure 3-9 (a), increasing the number of the power-split modules 𝑁𝑚 would allow
further increasing the first stage ratio 𝑢′ and allow further exploitation of its benefits.

As shown in Figure 3-9 (b), there is an increase in the gearbox mass index 𝑀𝑔𝑏𝑥 with
the increase in the gearbox total ratio 𝑖𝑡𝑜𝑡𝑎𝑙 . This stems mainly from the increase in the
third stage ratio to deliver the increasing total ratio.

This gearbox configuration does however suffer from another type of compromise;
increasing the number of power-split modules 𝑁𝑚 produces a mass penalty but on the
side of the generator. Figure 3-9 (b) shows the reduction in the gearbox mass resulting
from increasing the number of power-split modules 𝑁𝑚 and the mass penalty it reflects
on the generators mass. These results show that the main mass reduction potential lies
44 Concepts Evaluation for Mass Reduction and Gear Mesh Optimization

within increasing the first stage ratio, while keeping the number of power-split modules
𝑁𝑚 , hence the number of generators, low.


𝑴𝒈𝒃𝒙 (𝑵 = )

𝑴∗𝒈𝒃𝒙,𝒕𝒐𝒕𝒂𝒍
𝑴∗𝒈𝒃𝒙,𝒕𝒐𝒕𝒂𝒍 𝑴𝒈∗ (𝑵 = )
′′
𝑴∗𝒈𝒃𝒙 𝑴𝒈∗ (𝑵 = )
∗ ′
𝑴𝒈𝒃𝒙 ∗ ′
𝑴𝒈𝒃𝒙 ∗
𝑴𝒈𝒃𝒙 (𝑵 = )
′′′
𝑴∗𝒈𝒃𝒙

(a) (b)
Figure 3-9: (a) Impact of the 1st stage ratio 𝒖′ of a non-planetary configuration on the individual
masses of each stage with the example of total gearbox ratio 𝒊𝒕𝒐𝒕𝒂𝒍 = 𝟎𝟎. (b)
Development of gearbox and generator mass indices of a non-planetary
configuration across a total gearbox ratio 𝒊𝒕𝒐𝒕𝒂𝒍 = 𝟎: 𝟎𝟎.

There are two possible design modifications that can further exploit the mass reduction
potential of the non-planetary configurations. The first idea is to substitute the main bull
gear in the first stage with a ring gear, similar to a stepped planetary configuration.
Arranging the pinions inwards gives more room in the first stage to increase its ratio.
With reference to Figure 3-10 (a), equation (3-38) is thus changed into equation (3-43).

[𝑑1 + 2𝑑𝑝 ] = [𝑑2 ]𝑁𝑜𝑛−𝑃𝑙𝑎𝑛𝑒𝑡𝑎𝑟𝑦


𝑃𝑙𝑎𝑛𝑒𝑡𝑎𝑟𝑦

[𝑑1 ]𝑠𝑢𝑛 . (1 + 2𝑢1𝑝 ) = [𝑑1 ]𝑝𝑖𝑛𝑖𝑜𝑛 . 𝑢′ (3-43)

The second idea is taken from the gearbox of the largest heavy lift helicopter ever built,
i.e. the Mi-26 [61; 62]. The idea in this case is to use two bull gears in the first stage, as
in Figure 3-10 (b). This idea hits two birds with one stone: first, the doubled power
splitting in the first stage reduces the loads, thus allows increasing the stage ratio in the
first stage; second, this increase in power splitting does not reflect on the number of
output shafts. It does not produce a mass penalty on the generator side because the
number of power-split modules 𝑁𝑚 , hence generators, remains unchanged. The
additional power splitting by adding a second bull gear is worked into equations (3-30)
and (3-31), and then equation (3-38) is applied again.
Concepts Evaluation for Mass Reduction and Gear Mesh Optimization 45

(a) (b)

Figure 3-10: (a) Schematic of a non-planetary first stage with pinions arranged inwards using a
ring gear. (b) Schematic of a dual power path module comprising two bull gears.

The impact that these two possible modifications have on the maximum first stage ratio
is seen in Figure 3-11 (a). Arranging the pinions inwards using a ring gear achieved the
highest ratio possible in the first stage, followed by the double bull gear solution.
However as evident in Figure 3-11 (b), increasing the ratio does not give a steady
decrease in mass. Increasing the ratio comes to a point where the decrease in the mass
of the pinions does not cover the increase in the central gear member, whether it is a bull
or a ring gear.

𝑵 =

𝑵 =
𝑵 =

(a) (b)
Figure 3-11: Impact of the first stage design on the (a) limit on the 1st stage ratio and (b) on the
mass index of the gearbox, with example of 𝒊𝒕𝒐𝒕𝒂𝒍 = 𝟎𝟎.

3.2.5.4 Comparison between planetary and non-planetary configurations

The previous sections showed the main characteristics attributed to each group and how
these influence the masses of both the gearbox and the generator. In this section it is of
interest to see how these characteristics impact the system mass 𝑀𝑠𝑦𝑠 of each
46 Concepts Evaluation for Mass Reduction and Gear Mesh Optimization

configuration. The system masses 𝑀𝑠𝑦𝑠 are calculated by substituting the mass indices
∗ ∗
𝑀𝑔𝑏𝑥 and 𝑀𝑔𝑒𝑛 into equations (3-10) and (3-14). The operational data for an exemplary
10 MW turbine can be taken from a 10 MW low wind rotor project [24]. For a 10 MW
wind turbine, the rated power 𝑃𝑒𝑙𝑒𝑐 is 10 MW, and the corresponding rotor torque 𝑇𝑖𝑛 is
12656 kNm at a rated rotor speed 𝑛𝑅 of 8.3 rpm. The resulting gearbox and generator
masses are all compiled in Figure 3-13 (a) and (b) for both planetary and non-planetary
gearboxes. Figure 3-13 (a) shows that non-planetary gearboxes have a significant mass
reduction advantage over planetary gearboxes. This can be attributed to the ability to
combine both high power splitting and high ratio in the first stage, unlike planetary
gearboxes where there is a conflict between these two attributes. However, as seen in
Figure 3-13 (b), this mass advantage is offset by the mass penalty imposed by sharing
the power among multiple generators.

3-stage
differential 𝑵 =
Series planetary
PPS
𝑵 = 𝑵 =
𝑵 = 𝑵 =
𝑵 =
𝑵 =
𝑵 =

𝑵 = 𝑵 =

(a) (b)
Figure 3-12: (a) Gearbox mass and (b) generator mass of a 10 MW wind turbine for different
planetary and non-planetary configurations.

By summing the results of Figure 3-12 (a) and (b), the system mass 𝑀𝑠𝑦𝑠 of each gearbox
configuration can be seen in Figure 3-13. For non-planetary gearboxes, the figure shows
that the mass penalty from increasing the number of power split modules 𝑁𝑚 , i.e. adding
more generators, is more than overcompensated by the gearbox mass reduction from the
additional power-splitting and the increased first stage ratio.
Concepts Evaluation for Mass Reduction and Gear Mesh Optimization 47

3-stage
differential
Series planetary
PPS
𝑵 =

𝑵 =

𝑵 =
𝑵 = 𝑵 =
𝑵 =

Figure 3-13: Sum of gearbox and generator masses 𝑴𝒔𝒚𝒔 for planetary and non-planetary
gearboxes, where 𝑵 is the number of power-split modules.

In addition, Figure 3-13 also shows how the whole system benefits from the increased
first stage ratio gained from switching to the ring gear design. The curves of the double
bull gear design almost coincide with the curves of the ring design and have been spared
from Figure 3-13 for clarity. But the main conclusion from this figure, is the mass
advantage of non-planetary split-torque gearboxes over planetary gearboxes. In
Figure 3-13, the series planetary PPS line defines the upper boundary of the solution
space of non-planetary split-torque configurations. This solution space begins in the
high-speed range with the total ratio > 50, in combination with using multiple generators
> 3. The mass reduction potential of the non-planetary configuration further increases
with increasing the number of power-split modules 𝑁𝑚 . Thus, it can be concluded that
non-planetary split-torque configurations offer a significant mass reduction potential
against the state-of-the-art PPS configuration, and is chosen for the gearbox design
proposal in this thesis.

3.3 Self-Aligning Load-Sharing Concepts

The bearing concept must address the two reliability compromising challenges that
would dominate the gearbox in the 10+ MW power class, namely achieving even load
distribution across the gear face, and equal load sharing among the power-split paths.
Developing the bearing concept begins with understanding the basic requirements to
achieve equal load sharing and load distribution evenness.
48 Concepts Evaluation for Mass Reduction and Gear Mesh Optimization

For equal load sharing, what the proposed concepts have in common is introducing
additional flexibility as commonly known in the available literature, e.g. [65; 72].
Unequal load sharing originates from manufacturing and assembly deviations and
deflection of all surrounding structures that ends up as position errors of the gears, and
the additional flexibility would allow the gears to reposition themselves und compensate
for these errors. However, the problem of achieving even load distribution across the
gear faces is not as straight forward and requires a better understanding of the problem.
It is first required to define the critical misalignment modes and understand the
kinematics of edge loading. This would in turn help determine the restoring action, i.e.
the self-aligning mechanism, which the bearing concept should provide whenever gear
misalignment takes place.

3.3.1 Modes of gear mesh misalignment

The causes leading to gear mesh misalignment are numerous and are listed in [64; 91].
The ISO 6336 standard [64] groups these causes under three main categories, namely
load-induced elastic deformations, manufacturing deviations and thermal distortions.
Whenever any of these causes takes place, the gear flanks are forced out of their ideal
positions and one or a combination of the following three modes of misalignment might
take place:

Axial misalignment:

This represents the axial shift of one gear member relative to the other in a manner that
would only affect the effective face width of the gear. This is usually accounted for by
making the smaller member, i.e. sun or pinion, around10 to 20 mm wider than the other.

Parallel misalignment:

This refers to the deviation in the center distance with the axes remaining parallel. This
does not have any significant effect since involute teeth are insensitive to the change in
center distance. This mode of misalignment would only result in an increase in the
profile contact ratio with the decrease in the center distance and vice versa.

Angular misalignment:

The angular misalignment can only be defined with reference to a certain coordinate
system. Annex E of the ISO 6336 standard [64], among other publications [91–94] as
well, came to the conclusion that the gear misalignment should be measured with
reference to the coordinate system defined by the common tangent to the base circles
and the common tangent to the flanks, as shown in Figure 3-14. This is based on the
basic understanding of the geometry of the gear teeth. By virtue of being involute, the
Concepts Evaluation for Mass Reduction and Gear Mesh Optimization 49

flanks at any point of contact can be represented as two cylinders in contact, and thus
the main coordinate system that can define their relative positions is their common
normal and common tangent.

(a) (b)
Figure 3-14: (a) Definition of coordinate system to reference gear angular misalignment. (b) e2 is
angular misalignment normal to the plane of action and e3 is angular misalignment
in the plane of action [94].

Based on the theory of Hertz [95], two cylinders with their axes inclined in the plane of
their common tangent gives a contact area in the form of an ellipse, as seen in
Figure 3-15 (a). The major axis of the ellipse defines the span of the loaded face width
of the gear. According to [91], the major axis would still span the gear face width and
this type of misalignment has a limited influence on the contact stress. Thus, the angular
misalignment normal to the plane of action (e2) is deemed not critical. However the
angular misalignment in the plane of action (e3) is considered by the ISO 6336 standard
[64] to be the detrimental factor that directly influences the load distribution across the
gear face. As seen in Figure 3-15 (b), this misalignment would create a gap between the
flanks at one end, forcing the gear loads to shift towards the other end of the face width.

(a) (b)
Figure 3-15: (a) Gear angular misalignment normal to the plane of action [91]. (b) Gear angular
misalignment in the plane of action [96].
50 Concepts Evaluation for Mass Reduction and Gear Mesh Optimization

The study [94] investigated in detail the influence of the direction, i.e. the plane of
misalignment, on the development of the flank and root stresses in the gears. As seen in
Figure 3-16 (a), under an angular misalignment in the non-critical direction (e2), there
is no change in the size of the contact pattern, and the contact stress covers the entire
width of the flank. The load distribution remained central, which can be seen in the
Figure 3-16 (b). The study reports that the effect of this misalignment in the strength
calculations can even be ignored. On the contrary, the same study [94] shows in
Figure 3-16 (c), the angular misalignment in the critical direction (e3) leads to a drastic
change in the contact pattern. Under this form of misalignment, the root stress in
Figure 3-16 (d) can be seen rising at one side of the gear face. Despite the angle of
critical misalignment (e3) is 0.1 of the angle of non-critical misalignment (e2), the root
stress in the critical misalignment plane is around double the root stress value in the non-
critical plane. The main conclusion that can be made here is that only one mode is critical
to the gears and that is the angular misalignment in the plane of action.

(a) (c)

(b) (d)
Figure 3-16: (a) Contact pattern under angular misalignment normal to the plane of action e2,
(b) root stress under angular misalignment normal to the plane of action e2, (c)
contact pattern under angular misalignment in the plane of action e3, (d) root stress
under angular misalignment in the plane of action e3 [94].

Besides, with reference to Figure 3-17, when this critical misalignment takes place the
contact pattern, hence gear loads, shift to one side of flank. This load shift exerts an
Concepts Evaluation for Mass Reduction and Gear Mesh Optimization 51

additional tilting moment that acts in the plane of action. As shown in Figure 3-17, the
critical misalignment would exert the shown tilting moment on the driven flank in the
shown direction. Therefore, the gears do not have to be exact parallel throughout the
operation, and the self-aligning bearing must only account for this additional tilting
moment.

Figure 3-17: Schematic of the tilting moment generated by critical angular misalignment in the
plane of action between two mating flanks. Edited from original in [96].

3.3.2 Self-aligning mechanism of the proposed bearing concept

Based on the results of the previous section, the outline of the self-aligning bearing
concept can be derived. Since a perfect alignment of meshing gears cannot be achieved
due to the compliance of the gearbox components, the solution thus would be to realize
a semi-pivoting gear system, where the gears are allowed to go out of parallelism,
however only in a non-critical direction, i.e. in the plane of the common tangent to the
contacting flanks.

While searching the available literature for concepts on how to conceive this idea of a
semi-pivoting gears system, one concept was found most promising. Quenneville [97],
with reference to Figure 3-18, proposed disposing an idler 𝑑2 between a driving and a
driven gear, i.e. 𝑑1 and 𝑑3 , in a location such that the two flank tangents at the two mesh
points are perpendicular. This location is defined at an angle 𝛾 between the two center
lines that satisfies the condition in equation (3-44), where 𝛼 is the pressure angle.

𝛾 = 90 + 𝛼12 + 𝛼23 (3-44)

He additionally proposed supporting that idler on a spherical seat assembly, as seen in


Figure 3-18 (b). In this arrangement, whenever the critical angular misalignment takes
52 Concepts Evaluation for Mass Reduction and Gear Mesh Optimization

place at one of the two gear meshes, the resulting shift in loads on the flank would exert
a tilting moment on the idler. The tilting moment would act in the direction of the plane
of action 𝐵𝑇, which is parallel to the flank tangent 𝐹𝑇 at the other mesh point. Due to
this proposed arrangement, the tilting direction of the idler would lead to a non-critical
angular misalignment with respect to the other mesh point because it is normal to the
plane of action (e2) and therefore not critical. Since the orientation of the plane of action
is dependent on the direction of rotation of the gears, this concept applies only in the
shown direction of rotation 𝑛3 of the driving gear 𝑑3 ; the concept is unidirectional.

𝒃
𝒃
Idler
𝑻
γ
𝑻 𝑩𝑻
Spherical
𝑩𝑻 seat
𝒃

(a) (b)
Figure 3-18: (a) Schematic of a 3-gear arrangement with two perpendicular flank tangents 𝑻,
and perpendicular base tangents 𝑩𝑻. (b) Proposed bearing for the idler gear in [97].

However, applying this concept exactly in the form proposed in [97] for a power-
splitting wind turbine gearbox would suffer from a number of drawbacks. Any self-
aligning rolling element bearing, whether it is the spherical seat in [97] or any of the
state-of-the-art bearings as in Figure 3-19 (a-b), would face the following challenges:

▪ With reference to Figure 3-19 (c), a main parameter of this type of bearings is the
osculation, which is the ratio of the roller contour (𝑟 = 𝑟0 − 𝑆𝑑 ⁄2) to the raceway
radius of curvature 𝑟0 , i.e. 𝑟/𝑟0 [98]. This parameter is chosen to optimize the contact
stress in the bearings; a low osculation ratio leads to a stress concentration in the
roller center, and a higher ratio would lead to high edge stress on the rollers [99].
Optimizing this parameter is complicated for a wind turbine gearbox subjected to a
load spectrum from extreme wind loads down to no-load idling.

▪ As once highlighted in [100], this type of roller bearings can also impose the risk of
Concepts Evaluation for Mass Reduction and Gear Mesh Optimization 53

bearing failure by skidding, where slipping results in surface damage. The bearing
endplay in combination with the change of load and speed would change the traction
in the rollers and force the outer race to slip against the rollers in their axial direction,
resulting in increased shear stress at the surface.

(a) (b) (c)


Figure 3-19: (a) Spherical bearing [101], (b) toroidal (or barrel-shaped) bearing [101] and
(c) contact geometry of a spherical bearing showing contact angle , diametral Play
𝑺 , raceway curvature 𝒓𝟎 and endplay 𝑷 [102].

▪ Finally, a self-aligning rolling-element bearing does not offer special measures for
ensuring equal load sharing in a gearbox with power splitting. Quenneville [97]
proposed his concept for a single idler and not for power-splitting among multiple
idlers.

This dissertation offers a solution via a minor modification to the proposed design in
[97], by mounting the idler on a flexpin. This modification would offer the following:

▪ Instead of the rolling elements, the flexibility of the flexpin would allow the pivoting
action of the idler under the influence of the tilting moment exerted by the critical
angular misalignment.

▪ A high bearing load capacity with less risk of skidding damage is achievable by
combining the flexpin with well-established bearings, e.g. tapered roller bearings
[69] or journal bearings [66].

▪ Besides, the flexpin will deliver its state-of-the-art function to achieve equal load
sharing by allowing large deflections, which will allow the idler to reposition itself
when it is overloaded.
54 Concepts Evaluation for Mass Reduction and Gear Mesh Optimization

3.4 The Outline of the Proposed Gearbox Concept

In the attempt to answer all the research objectives introduced in chapter 1, this chapter
has investigated three main features of the gearbox design:

▪ Gearbox configuration: the state-of-the-art planetary and non-planetary split-torque


gearboxes were compared, and the maximum reduction in the system mass 𝑀𝑠𝑦𝑠
(sum of gearbox and generator mass) was achievable via the non-planetary split-
torque gearboxes.

▪ Load distribution evenness across gear flanks: in the attempt to find a bearing
concept that can deliver a self-aligning function, the characteristics of gear
misalignment were investigated. The outcome was the idea of a semi-pivoting gear
system, where a gear is forced to pivot only normal to the line of action and cancel
out any critical angular misalignment in the line of action.

▪ Equal load sharing: the state-of-the-art flexpin bearing concept was found to have
the experience with number of planets above 5 [44; 66; 71]

This dissertation proposes a gearbox concept that attempts to merge all these three
features in a single design. Thus, with reference to Figure 3-20, the proposed gearbox
will have the following main features:

Intermediate Gear

Intermediate Pinion
High-speed pinion
Idler Gear + Flexpin

Bull Gear

Figure 3-20: General features of the power-split module in the proposed gearbox.

▪ The gearbox configuration is a non-planetary split-torque design, and its building


unit is the power-split module.

▪ Achieving equal sharing is assigned to the flexpin bearing concept. The proposed
power-split module is a modified construction where the intermediate pinions are
not directly in mesh with the bull gear. However, an idler gear mounted on a flexpin
Concepts Evaluation for Mass Reduction and Gear Mesh Optimization 55

is disposed between each intermediate pinion and bull gear.

▪ Achieving load distribution evenness is assigned to the idler gear. The pinion-idler
and idler-bull center lines are arranged at an angle 𝛾 that satisfies equation (3-44).
This angle would force the idler gear to pivot – by virtue of the flexpin flexibility –
only in the direction of the flank tangents at its gear meshes with the pinion and bull
gear.

The choice of bearings for the flexpin – whether rolling element or journal bearings – is
irrelevant to the proposed self-aligning mechanism, and both can be used. However, in
the context of the proposed gearbox, the work proceeds with journal bearings for the
following reasons:

▪ Journal bearings offer a spatial advantage: the flexpin assembly takes space, and
simultaneously the idler diameters must be minimized to reduce the mass penalty.
This would come at the expense of the available space to dispose a roller bearing
between the flexpin and the idler. Journal bearings only require a clearance in the
order of magnitude of fraction of a millimeter for the development of adequate oil
film thickness.

▪ Journal bearings introduce higher damping: the load sharing capability of the flexpin
would be augmented by the damping offered by journal bearings.

Consequently, it is the purpose of the following chapters to confirm the hypothesis


behind the idea of the proposed construction. In chapter 4, the proposed self-aligning
mechanism is going to be verified by means of a finite-element model. The following is
going to be investigated:

▪ The behavior of the flexpin assembly and its characteristics at the proposed angle 𝛾.

▪ Exploring the difference (or similarity) between using an external bull gear or an
internal ring gear on the behavior of the proposed self-aligning mechanism.

▪ The bearing concept verification is limited to spur teeth. The axial loads on a helical
idler would lead to an additional tipping moment in a different plane. It is not yet
known how the axial loads will affect the proposed self-aligning concept.

The outcome of this chapter would show if indeed the flexpin assembly performs a self-
aligning action when the gears are arranged at the angle 𝛾 described in equation (3-44).

Afterwards, chapter 5 presents an optimization procedure to calculate the gearbox


geometry that delivers the minimum system mass 𝑀𝑠𝑦𝑠 . This optimization procedure
shall cover the following design variants:

▪ The design of the central gear, whether it is an external bull gear or an internal ring
56 Concepts Evaluation for Mass Reduction and Gear Mesh Optimization

gear

▪ The number of central gears, i.e. single or double

▪ The possible variants of the third stage, and its influence on the number of generators

The outcome of this chapter would show how the system mass 𝑀𝑠𝑦𝑠 develops with the
number of power-split modules, and how the proposed design compares to the state-of-
the-art planetary gearboxes.
Self-Aligning Mechanism Verification 57

4 Self-Aligning Mechanism Verification

It is the purpose of this chapter to provide a proof of plausibility of the self-aligning


mechanism proposed in the previous chapter. The verification of the mechanism has
been pushed before the gearbox geometry optimization because of its direct effect on
the gears arrangement. If this chapter concludes that the self-aligning mechanism is only
true at a certain angle between the center lines, i.e. certain orientation of the gears, it
would decide how the spatial parameters, e.g. center distances and profile shifts, would
be incorporated in the optimization procedure.

4.1 Development of the Modelling Concept

The classic approach for performing contact analysis of a gearbox is to model the
gearbox components in an FE-program and define the input loads, and then observe how
the components deform under load and observe its influence on the gear contact patterns.
The master thesis [103] is a good example of this approach. However, when it comes to
verifying a self-aligning mechanism, the simulation approach is different. In this case, a
worst-case scenario of misalignment is deliberately assigned to the gears as a given
input. Then the behavior of the gears, and the distribution of contact stress across the
engaging flanks is observed. The analysis is going to cover the following aspects:

The difference between the behavior of external and internal gears in a semi-pivoting
gear system. The misalignment in an internal gear mesh is one of most under-
researched topics. It is not expected that the concave-convex flank contact in an
internal gear mesh would behave similarly to the convex-convex flank contact in an
external gear mesh. Figure 4-1 offers a good visual reference to a concave-convex
contact between an internal gear 2 and an external gear 1. In an internal gear mesh
at any point on the line of action 𝐵𝑇, the external gear flank is defined by a cylinder
with radius 𝜚1 sitting on the inner surface of a larger cylinder with radius 𝜚2 , which
represents the internal gear flank. There is no mention in the literature if the angular
misalignment in the direction of the flank tangent 𝐹𝑇 is also non-critical as in the
case of an external gear mesh.
The direction of misalignment, i.e. the direction of contact pattern shift across the
gear flank. Therefore, the models would also include different combinations of
misalignment directions at the two mesh points on the idler gear.
58 Self-Aligning Mechanism Verification

Figure 4-1: Schematic of the concave-convex contact in an internal gear mesh.

For these reasons, there are going to be two general modelling cases, namely with a bull
and a ring gear. Each case is modelled twice: first, as the proposed semi-pivoting system
with the base tangents perpendicular to each other (𝐵𝑇12 ⊥ 𝐵𝑇23 ); and second, with the
two base tangents parallel to each other (𝐵𝑇12 ∥ 𝐵𝑇23 ). The second case is the exact
opposite of the semi-pivoting concept to show if the flexpin behavior under load changes
with the gears orientation, i.e. 𝛾, or not. If the results from both base tangent orientations,
i.e. 𝐵𝑇12 ⊥ 𝐵𝑇23 and 𝐵𝑇12 ∥ 𝐵𝑇23 , are the same, then the flexpin in itself has a self-
aligning function, and the whole concept of the semi-pivoting system would be
worthless. The four models are summarized in Table 4-1.
Self-Aligning Mechanism Verification 59

Model A Model B
𝐵𝑇12 ⊥ 𝐵𝑇23 𝐵𝑇12 ∥ 𝐵𝑇23
CASE I (bull gear)
CASE II (ring gear)

Table 4-1: Schematic of the simulation models showing the relative orientation of the base
tangents BT.

Finally, for all the four models, 8 different misalignment scenarios are defined as an
input, then the contact pressure and the load distribution across the flanks are registered
for comparison. The misalignment scenarios can be seen in Table 4-2, and the sign
convention of the misalignment angles is defined with reference to Figure 4-2.

Mis1 Mis2 Mis3 Mis4 Mis5 Mis6 Mis7 Mis8


𝜃𝐵𝑇,12 0° 0° 0° +0.1° −0.1° +0.1° −0.1° +0.1°
𝜃𝐵𝑇,23 0° +0.1° −0.1° 0° 0° +0.1° −0.1° −0.1°
Table 4-2: Misalignment scenarios showing different combinations of misalignment in the base
tangent direction

With reference to Figure 4-2, a positive misalignment angle would shift the contact
towards the free end of the flexpin and vice versa. Here it is necessary to note that the
chosen misalignment angle is significantly high at 0.1
60 Self-Aligning Mechanism Verification

+𝜽𝑩𝑻 −𝜽𝑩𝑻

Figure 4-2: Sign convention of the angular misalignment 𝜽𝑩𝑻 with respect to the flexpin
assembly.

The exact geometry of the components of the model, as well as the construction of the
model are described in the following sections.

4.2 FE-Model Description

4.2.1 Gear model geometry

The four models of Table 4-1 share the exact same gear geometry, which can be found
summarized in Table 4-3. These are not optimum designs, however these are a good
representative of a power branch in the first stage in terms of loads and ratio. It is
important to keep the geometry constant across all four models, such that any spotted
differences in the results can be attributed to the difference in behavior between external
and internal gears under misalignment. The geometry is then fed into an Abaqus plugin
developed at the institute for machine elements and system engineering (iMSE) in
Aachen called “Zahnplugin”, which in turn generated the 3D models of the gear
members, as seen in Figure 4-3.

Pinion

Idler

Bull
Figure 4-3: Abaqus model of the gears of the first stage of the proposed gearbox.
Self-Aligning Mechanism Verification 61

The scope of investigation here is only the behavior of the flexpin assembly and the idler
gear between the misaligned gear members. Therefore, only the idler and its flexpin
assembly are modelled completely, and the other gear members are simplified down to
a single tooth. The critical angular misalignment is already given as an input to these
gear members and the deformation of their respective gear bodies is not of interest.

First Stage
General:
Input power [kW] 550
Stage ratio [-] 9.96

Bull(ring) idler pinion


Torque [kNm] 633 63.55
Speed [rpm] 8.3 50.41 82.7

Geometry:
No. of teeth 𝑧 249 41 25
Helix angle 𝛽 [] 0
Normal module 𝑚𝑛 [mm] 12
Gear width 𝑏 [mm] 340
Table 4-3: Summary of the gears geometry for both models with a bull and a ring gear.

4.2.2 Flexpin assembly geometry

As already pointed out in [104] there is no analytical or numerical method for scaling
the components of the flexpin assembly, and FEM is the only common tool. The
geometry of the components is governed by the trade-off between a number of crossing
parameters. With reference to Figure 4-4, the space defined by the root diameter of the
idler gear 𝑑𝑓 is shared by the pin diameter 𝑑𝑝𝑖𝑛 , the idler gear rim thickness (𝑑𝑓 − 𝑑𝑖 )/2,
clearance between the spindle and the pin 𝑑𝑠𝑝,𝑖 − 𝑑𝑝𝑖𝑛 , the spindle rim thickness
(𝑑𝑠𝑝,0 − 𝑑𝑝𝑖𝑛 )/2, and the clearance of the journal bearing 𝑑𝑖 − 𝑑𝑠𝑝,𝑜 . The pin diameter
𝑑𝑝𝑖𝑛 needs to be large enough to withstand the loads, and small enough to produce the
desired large deflections. Besides, the pin deflection should not be too large to cause
gear mesh interference. Increasing the idler inner diameter 𝑑𝑖 is limited by the
recommended minimum limit for the idler gear rim thickness which is limited to 2.5~3
times the normal module [47]. The spindle inner diameter 𝑑𝑠𝑝,𝑖 needs to be large enough
to accommodate the bending deflection of the pin. This restricts increasing the spindle
rim thickness, which needs to be high enough such that the surface of the spindle takes
the same slope of the free end of the pin. The best combination of these dimensions has
been found by running the model iteratively in Abaqus, with the gear axes in parallel
position, i.e. Mis1.
62 Self-Aligning Mechanism Verification

Idler 𝑑𝑓 = 470.72
𝑑𝑖 = 398
Spindle 𝑑𝑠𝑝,𝑜 = 396.02
Pin 𝑑𝑠𝑝,𝑖 = 248
𝑑𝑝𝑖𝑛 = 238
𝑙 = 100
Figure 4-4: Geometry of the flexpin components in the Abaqus model. Dimensions are in mm.

4.2.3 Definition of angular misalignment as an input

The definition of the angular misalignment of the gears as well as the definition of the
input torque requires a frame of reference. Besides, defining the angular misalignment
as an input requires rotating the gears around a given axis. Defining the frame of
reference and the misalignment axis is explained with reference to Figure 4-5. In the
assembly module the rotation axis of the gear is defined as a child feature to its
respective gear member. Afterwards two datum planes parallel to the base tangent and
flank tangent planes are added through the rotation axis. The two planes are used to
define a local coordinate system for that respective gear at the midpoint of the gear
rotation axis. Finally, an axis perpendicular to the rotation axis in the plane parallel to
the flank tangent is added to serve as the misalignment axis. By virtue of being a child
feature of the gear, the coordinate system would rotate with the gear member such that
the Z-axis would remain aligned with the rotation axis of the gear. This would later
facilitate the definition of the torque and reading the reaction moments at the end of the
simulation. This procedure is applied to the gears that would be assigned an angular
misalignment as a given input, namely the pinion and the bull (or the ring) gear.

Misalignment axis
Flank
Tangent Plane FT Local
Coordinates

Rotation axis

Base Tangent
Plane BT

Figure 4-5: Local coordinates for gear members as a frame of reference for the definition of
angular misalignment.
Self-Aligning Mechanism Verification 63

4.2.4 Definition of load and boundary conditions

A quasi-static state representation of the gears can be done by defining both input and
output, i.e. active and reaction torques as a given in the model. The idler gear would
have to balance itself between these two acting torques. The torque is applied at the
reference point of the local coordinate, shown in Figure 4-5, of the respective gear. The
torque is transmitted from the reference point to the gear body via a rigid kinematic
coupling, as shown in Figure 4-6. The torque can be seen symmetrically applied at mid-
length of the gear width. The torsion experienced by the gear body is not of interest here,
since the angular misalignment is already an input.

Figure 4-6: Torque application to the gears via rigid kinematic coupling.

4.2.5 Contact definition

Good results would only come as a result of correct definition of the gear contact. In the
interaction module in Abaqus, the definition of the surface contact characteristics and
the settings of the gear surfaces interaction properties follow the steps of the master
thesis [103]. However, the main problem with the gears is the contact initialization. One
solution offered by Abaqus is establishing additional “imaginary” damping elements
between the flanks in case a contact is not established at the start of the simulation. This
damping element is called the contact control, and the key here is to define a value of
the “automatic stabilization factor” that is high enough to achieve convergence, and
simultaneously not too high in order to achieve realistic values. By trial and error, a
value of 0.01 was found satisfactory. In combination with a minimum step increment of
0.01, the model converged best and the fraction of the damping energy (ALLSD in
Abaqus) of the total system energy (ALLSE in Abaqus) was 0.6 % which is much lower
than the maximum limit of 10% mentioned in the documentation of the program.
64 Self-Aligning Mechanism Verification

Therefore, additional automatic stabilization in the step module was not required. The
journal bearing contact definition is exactly similar to the gear contact except the contact
control option is not required, because the contact initialization of a cylindrical contact
can be done in the assembly module before starting the simulation.

4.2.6 Output request

To understand the behavior of the proposed self-aligning mechanism, two parameters


are required to be investigated, namely the contact pressure and the deflection. First, the
contact pressure at the gear flanks is required for calculating the load distribution factor
𝐾𝐻𝛽 . This factor is calculated here according to the definition by the ISO 6336 standard
[64], represented by equation (4-1).

𝑚𝑎𝑥. 𝑢𝑛𝑖𝑡 𝑙𝑜𝑎𝑑 𝑝𝑒𝑟 𝑢𝑛𝑖𝑡 𝑓𝑎𝑐𝑒 𝑤𝑖𝑑𝑡ℎ (𝐹⁄𝑏)


𝑚𝑎𝑥
𝐾𝐻𝛽 = = (4-1)
𝑎𝑣𝑒𝑟𝑎𝑔𝑒 𝑙𝑜𝑎𝑑 𝑝𝑒𝑟 𝑢𝑛𝑖𝑡 𝑓𝑎𝑐𝑒 𝑤𝑖𝑑𝑡ℎ (𝐹⁄𝑏)
𝑚𝑒𝑎𝑛

In case of a calculation using an FE model as seen in Figure 4-7, the factor (𝐹⁄𝑏), at a
given axial position 𝑍 across the gear face, is the sum of the nodes at that given location
as expressed by equation (4-2), where 𝑁𝑛𝑜𝑑𝑒𝑠 is the number of nodes at location Z.
𝑁𝑛𝑜𝑑𝑒𝑠

(𝐹⁄𝑏) = ∑ (𝐹⁄𝑏) (4-2)


@𝑍 𝑖
𝑖=1

In Abaqus there are two possibilities to request the unit contact force 𝐹 ⁄𝑏 in equations
(4-1) and (4-2), either the nodes contact stress value CPRESS or the nodal contact force
CNORMF. The program documentation recommended using the nodal force for more
reliable results.

Figure 4-7: Nodal forces at multiple nodes at the same axial position across the face width.
Self-Aligning Mechanism Verification 65

Second, the deflections are requested for the two components that define the flexpin
assembly, namely the pin and the spindle, as well as the idler gear body. The deflections
can be found under parameter 𝑈 in Abaqus. For the pin, requesting the deflections will
show the magnitude as well as the direction of the pin bending deflection. As seen in
Figure 4-8, the deflection is requested at the center line of the pin.

Figure 4-8: Nodes path through the center line of the pin. Deformation scale factor is 500:1.

The local deflections of the spindle are also as equally as important as the pin. They will
show how the deflection of the pin is transformed into the deflection of the idler gear
body. The spindle is not infinitely rigid and its local surface deflections affect the
alignment of the idler. The deflection of the cylindrical surface of the spindle is irregular
in a manner similar to the idler gear body as seen in Figure 4-9. Therefore, the deflection
readings depend on the selection of the nodes line at which the deflection values are
requested. The results have shown that the deformation of the spindle cylindrical surface
is related to the pressure distribution in general, and the zero-pressure zone in particular.
As seen in the same figure, this zero-pressure zone is located circumferentially at the
point of maximum ovalisation of the idler gear body.

Figure 4-9: Deformation of the idler body and the spindle, showing the contact pressure
CPRESS in MPa. The zero-pressure zone is marked in red boxes. Deformation scale
of spindle and idler is 1000:1 and 100:1 respectively.
66 Self-Aligning Mechanism Verification

With reference to Figure 4-10, the maximum deflection of the spindle cylindrical is
located at this zero-pressure zone, and the deflections are read on the nodes line in this
zone. The degree of ovalisation in the idler is simply the deviation in the gear inner
diameter.

Figure 4-10: Deflections on the spindle cylindrical surface in mm, showing location of nodes line
for reading deflections. Deformation scale is 1000:1.

On a side note, this pressure distribution at the idler-spindle interface is similar to the
results of Harris’s simulations [105; 106] for roller bearings of a planet gear with a thin
rim thickness. Bearings with a flexible outer ring experience a drop in the bearing loads
in the mid distance between the points of load application on the outer ring, i.e. the
locations of the two gear meshes.

4.3 Results and Conclusions

4.3.1 Load distribution results

The results of the load distribution factors 𝐾𝐻𝛽 are compiled in Figure 4-11 for the bull
gear model and in Figure 4-12 for the ring gear model.

Figure 4-11: Load distribution factors for all the misalignment scenarios of the bull gear model.
Self-Aligning Mechanism Verification 67

Figure 4-12: Load distribution factors for all the misalignment scenarios of the ring gear model.

To quantify the influence of changing the orientation of the base tangents 𝐵𝑇, i.e. the
angle 𝛾, on the load distribution factor 𝐾𝐻𝛽 , equation (4-3) is applied to the results. This
equation compares the worst, i.e. highest, load distribution factor out of the two mesh
points. The equation is applied for both the bull and ring gear models. The results of
applying equation (4-3) can be seen in Figure 4-13.

max {𝐾𝐻𝛽12 , 𝐾𝐻𝛽23 }𝐵𝑇12 ⊥𝐵𝑇23


∆𝐾𝐻𝛽 𝑚𝑎𝑥 % = 100 × ( − 1) (4-3)
max {𝐾𝐻𝛽12 , 𝐾𝐻𝛽23 }𝐵𝑇12 ∥𝐵𝑇23

Figure 4-13: Change in the load distribution factor KHβ following equation (4-3) in both the bull
and ring models with flexpin in dependence of the gear contact shift direction.
68 Self-Aligning Mechanism Verification

The preliminary outtake from Figure 4-13 confirms the influence of the angle 𝛾 on the
load distribution. The load distribution factor 𝐾𝐻𝛽 is significantly reduced in the semi-
pivoting system up to 40% – except for the misalignment case Mis7. Second, the figure
also shows that the direction of contact shift has a strong influence on the semi-pivoting
system. The figure shows that the misalignment scenarios Mis1-Mis8 exhibit
significantly deviating results. Finally, the figure however does not show a significant
difference between the behavior of bull and ring gears.

For a further interpretation of Figure 4-13, Figure 4-14 provides an insight into the
deflection of the flexpin in both cases. The requested deflection of the pin can be used
to determine the maximum deflection at the free end, and the direction of the pin bending
plane. Since the results of the ring model almost coincide with the bull gear model
results, the figure shows only the results of the bull gear model for clarity. The first
remark that can be made clearly in Figure 4-14 is the plane in which the flexpin deflects.
Figure 4-14 (a) shows that in the case of parallel base tangents, the flexpin is restricted
to deflect in a certain direction parallel to the base tangents. The reason is that the acting
tilting moments from misalignment at both mesh points act in the same direction. In
other words, in this orientation a tilting moment at one flank would lead to another tilting
at the other flank. On the other hand, Figure 4-14 (b) shows a relatively higher deviation
in the pin bending plane direction with reference to the no misalignment scenario Mis1.
A misalignment tilting moment on one flank tilts the idler in a direction that is not
resisted at the other flank. Thus, the idler body can be seen assuming slightly different
positions with the change in the misalignment scenarios. For this reason, for the same
misalignment scenario, the deflections are much higher in (a) than in (b). The flexpin
exhibits higher deflections in (a) in response to the higher developed misalignment
tilting moments on the idler.
Self-Aligning Mechanism Verification 69

6 6
3 83 48 4
7 7 1 21 2
5 5
7 75 5
1 21 2
3 3 6 6
8 48 4

(a) (b)
Figure 4-14: Polar plot representing the flexpin deflection on the radius in mm and the angle of
the plane of flexpin deflection in degrees for a bull gear model with (a) parallel BT
and (b) perpendicular BT arrangements. Numeral tags from 1 to 8 represent the
misalignment scenarios Mis1-Mis8. F is the gear force in the plane of action.

Figure 4-15 further examines the direction of the pin bending plane. The figure shows
how far the pin deflection plane for each misalignment deviates from the no
misalignment condition Mis1. Cases Mis6 and Mis7 – regardless with parallel or
perpendicular base tangents – exhibit the smallest deviation. Nevertheless, the load
distribution factor 𝐾𝐻𝛽 at these two misalignment scenarios is not similar to Mis1 and
is much higher, as seen in Figure 4-11 and Figure 4-12. On the other side, in case of
Mis8 with perpendicular base tangents the maximum deviation from Mis1 was obtained.
This implies that the direction of the angular misalignment also has an impact on the
behavior of the idler gear.

Figure 4-15: Deviation in the pin bending plane relative to the no misalignment condition Mis1.
70 Self-Aligning Mechanism Verification

The scenarios Mis6, Mis7 and Mis8 actually behave as the resultant of their individual
components. For example, the components of Mis6 are Mis2 and Mis4, which can be
seen acting in different directions. Similar case with Mis7 and its components Mis3 and
Mis5. Therefore, when both misalignment titling moments act on the same side of the
gear width, i.e. Mis6 and Mis7, they resist each other. Thus, the resultant tilting moment
on the idler is not high enough to adjust its orientation between the misaligned gears.
However, for Mis8, its components Mis3 and Mis4, are acting in the same direction.
Thus, they add up together leading to a higher resultant tilting moment on the idler.
However as seen in Figure 4-15, the tilting action of the idler is only allowed when the
base tangents are perpendicular.

4.3.2 Structural deformations

Consequently, it is suspected that when the misalignment tilting moments are not
translated into a tilting action of the idler, they lead to structural deformations in the gear
assembly. Two main forms of structural deformations were observed: the ovalisation of
the idler gear, and the deformation of the cylindrical surface of the spindle. First, the
ovalisation of the idler impacts the relative angular positions of the idler teeth in mesh
with the driver and driven gear members. This impacts the relative orientation of the
base tangents (angle 𝛾), hence the integrity of the proposed concept. Figure 4-16 shows
the results of the detailed analysis of the idler gear ovalisation. It can be seen that the
degree of ovalisation changes significantly with the orientation of the base tangents, and
to a lesser degree with each misalignment scenario. The ovalisation is significantly
higher when the base tangents are perpendicular to each other.

Figure 4-16: Representation of the ovalisation of the idler gear for both orientations of the base
tangents under the possible misalignment scenarios for the bull gear model.
Self-Aligning Mechanism Verification 71

Figure 4-17 shows the difference of the ovalisation between parallel and perpendicular
base tangent orientations. The difference here offers a chance to rank the degrees of
ovalisation under each misalignment scenario, and it can help understand the influence
of ovalisation on the integrity of the proposed concept. The results of Figure 4-17
actually correlate with the results of Figure 4-13. It can be seen that the improvement in
the load distribution factors is best where the ovalisation difference is minimum, and the
advantage of the perpendicular orientation of the base tangents is maximized, as seen
under scenario Mis8. However, the improvement in the load distribution factor keeps
decreasing with the increase in the ovalisation difference until the perpendicular
orientation of the base tangents completely loses its advantage as evidently seen under
scenario Mis7. The ovalisation findings in Figure 4-16 and Figure 4-17 also apply to the
ring gear model and are excluded from the figures for clarity.

Figure 4-17: Difference of the idler ovalisation between perpendicular and parallel base tangents
for the bull gear model.

Second, as seen in Figure 4-18, the deformation of the spindle also plays a role in the
effectiveness of the proposed concept. Regardless of the orientation of the base tangents,
it can be seen that this deformation is always higher when the edge loading shifts
towards the free end of the spindle, i.e. Mis3 > Mis2, Mis5 > Mis4 and Mis7 > Mis6.
These findings actually reflect the same pattern on the load distribution in Figure 4-13,
i.e. ∆𝐾Hβ,Mis2 > ∆𝐾Hβ,Mis3 , ∆𝐾Hβ,Mis4 > ∆𝐾Hβ,Mis5 and ∆𝐾Hβ,Mis6 > ∆𝐾Hβ,Mis7 . This
could be attributed to the fact that the stiffness of the spindle at the connection with the
pin is significantly higher. Therefore, the spindle deformation would affect how the idler
tilting moment is delivered to the pin. At positive misalignments, i.e. Mis2 and Mis4,
the tilting moment is directly translated to the pin, while at negative misalignments, i.e.
Mis3 and Mis5, a fraction of the tilting moment is gone in the deformation of the spindle.
Figure 4-18 represents the case of parallel base tangent of a bull gear model, however
same results apply to both base tangent orientations and to the ring gear model as well.
72 Self-Aligning Mechanism Verification

Figure 4-18: Spindle surface deflection under all misalignment scenarios at parallel base tangents
orientation for a bull gear model.

From the analysis of the structural deformations in the gear assembly it can be
understood how the direction of the gear contact shift on both flanks influences the
integrity of the proposed self-aligning mechanism. One might be tempted to increase
both the idler and spindle rim thicknesses to reduce their deformation. However, the
results showed that the thin-rim idler and spindle performed well, and it is only at certain
combinations of gear contact shift (Mis6 and Mis7), where the deformation of the idler
and spindle violated the concept of the self-aligning mechanism. This information can
be used as a guideline when applying the lead correction to the flanks. The lead
correction should be optimized to avoid a condition where the gear contact shifts on both
flanks of the idler in the same direction (Mis6 and Mis7).

4.3.3 Difference between external and internal gears

No major differences were found between the behavior of external and internal gears.
As concluded throughout the previous sections, the load distribution change with the
misalignment scenarios (Mis1-Mis8), as well as the structural deformation in the system
were almost identical for both gear designs. The suspicions made in the beginning with
reference to Figure 4-1 were proven wrong. In a semi-pivoting system, i.e. 𝐵𝑇12 ⊥ 𝐵𝑇23 ,
the misalignment tilting moment at the pinion led to a non-critical misalignment and
acceptable 𝐾𝐻𝛽 – Figure 4-11 and Figure 4-12 – at both the bull and the ring gear. The
load distribution factor 𝐾𝐻𝛽 at both the bull and ring gears in scenarios Mis4 and Mis5
were almost identical. This could be explained by the relative size of the involute
curvature radii 𝜚1 and 𝜚2 of the idler and the ring respectively. Due to the high ring-to-
idler ratio, the involute curvature of the ring was significantly higher than that of the
Self-Aligning Mechanism Verification 73

idler gear, i.e. 𝜚2 ≫ 𝜚1 , it almost behaved as a flat surface. It is thus expected that an
angular misalignment in the plane of flank tangent 𝐹𝑇 would only start to be critical at
lower gear ratios. This will however never be the case in the proposed gearbox since the
idler gear diameter must be minimized in order to reduce the mass penalty of adding the
idlers.

4.3.4 Final remarks and conclusions

At this point, a number of main conclusions and remarks can be made:

The orientation of the gears relative to each other (angle 𝛾) is a decisive factor to the
load distribution across the flanks. Arranging the gears in a semi-pivoting system
has been shown favorable for the load distribution. In this arrangement the load
distribution factors have been decreased by up to 40 % as seen in the case of the
misalignment scenario Mis8 in Figure 4-13.
The direction of the angular misalignment is a decisive factor as well. The best
improvement in load distribution was achieved with opposite shift in the gear contact
at the two flanks (Mis8). While the worst case is where the gear contact shifted at
both flanks away from the pin free end (Mis7). This serves as a guideline when
applying the lead correction to the flanks.
The function of the flexpin concept can also be better understood from these results.
The clear advantage of the flexpin can be seen in the case, where the tooth loads are
central, i.e. Mis1. As seen in Figure 4-14, the flexpin allows high parallel deflections
of the pin, ~ 300 𝜇𝑚, while at the same time achieve low load distribution factor 𝐾𝐻𝛽
~1.4 as seen in Figure 4-11 and Figure 4-12. This is a very low factor for this
deflection magnitude, while also keeping in mind that the lead and profile corrections
were excluded from the models. This shows that the flexpin provides the idler with
high deflections without violating the proposed semi-pivoting concept.
74 Gearbox Geometry Optimization

5 Gearbox Geometry Optimization

The purpose of this chapter is providing a method to calculate the gearbox geometry of
the proposed gearbox configuration in chapter 3. The outcome shall be the geometry
that delivers the minimum total mass of the gearbox and generator(s). This means
finding the optimum total gearbox ratio, its optimal ratio distribution on each gear stage
and the gear geometry of each stage. However, the classic optimization methods
available in the literature do not directly apply to the proposed gearbox in this work.

The unique problem to optimizing non-planetary split-torque gearboxes is that not every
solution for a stage can be assembled with all the solutions of the other stages without
collision of gears and shafts. In other words, the geometry of a given stage affects and
limits the feasible solutions of the preceding as well as the succeeding stage. This
problem is absent in multiple-stage planetary gearboxes since planetary stages are
coaxial by nature. This coaxiality does not force any special spatial constraints on each
two successive planetary stages.

The goal of the proposed optimization procedure in this chapter is to reach a feasible
solution early enough in the calculation process, in order to save time and disk space.
As seen in Figure 5-1 (a), the geometry of a given stage would branch out into multiple
possible solutions of the succeeding stage, each having different combinations of teeth
numbers 𝑧1 and 𝑧2 . Besides, as shown in Figure 5-1 (b), each stage ratio in itself
discloses a sub-matrix of multiple possible combinations of teeth module 𝑚𝑛 and profile
shift distributions 𝑥1 and 𝑥2 . Filtering through this enormous solution space to find an
optimum would be so time-consuming it is almost impossible.

𝑧1′′′ 𝑧2′′′

𝑧1′′ 𝑧2′′
𝑚𝑛 𝑥1 𝑥2
(𝑘) (𝑘)
𝑧1 𝑧2
𝑧1′ 𝑧2′ 𝑧3′

(a) (b)
Figure 5-1: (a) Representation of the possible teeth number 𝒛 combinations for all three stages.
(b) Microgeometry sub-matrix for each teeth combination 𝒛 and 𝒛 of a given stage
𝒌 including the teeth module and profile shift distribution 𝒙 and 𝒙 .

The optimization procedure presented in this chapter is a brute force calculation


(generate and test) based on an elimination-principle embedded throughout the
procedure. The choice of using brute force is mainly attributed to the mass index
Gearbox Geometry Optimization 75

equations derived in Chapter 3. After generating the solution space, the mass index
equations can be applied and arrange the solution space ascendingly, such that the
solutions with the lowest system mass 𝑀𝑠𝑦𝑠 are tested first. This would spare the need
to test every solution in the generated solution space. The main effort would be to
generate a search space with the least failing solutions as possible, and this would mainly
depend on setting the elimination rules and a good definition of the problem boundaries.
The details of this procedure are explained in the following sections.

5.1 Gearbox Configuration Variants

The gearbox general layout described in chapter 3 can be realized in a combination of


multiple variations. These are mainly dependent on the design of the central gear, the
arrangement of gears in the first stage, as well as the possible variants of the third stage,
i.e. the way of delivering the output power.

5.1.1 Central gear

The central gear can be a bull or a ring gear, i.e. external or internal. The bull gear and
the ring gear variants can be seen in Figure 5-2 (a) and (b) respectively. The radii 𝑅′ and
𝑅′′ , in the same figure, are two important geometrical parameters that will be referred
to frequently in this chapter. They define the pitch circles that define the centers of the
output, i.e. the high-speed pinions, in each stage. Pinions 1′ are located on 𝑅′ , and
pinions 1′′ are located on 𝑅′′ . The semi-pivoting concept will influence the center
distances through the angle 𝛾, and influence the radius 𝑅′ .

(a) (b)
Figure 5-2: Geometrical parameters of a power-split module, (a) with a bull gear design and (b)
a ring gear design.
76 Gearbox Geometry Optimization

5.1.2 First stage design

As previously discussed, the power-splitting in the first stage can be increased without
increasing the number of power-split modules 𝑁𝑚 by doubling the number of central
gears. For the proposed gearbox in this dissertation Figure 5-3 shows the possible
arrangements of a single and a double bull gear. A similar arrangement with ring gears
is also possible.

(a) (b)

Figure 5-3: Possible designs of the central gears; either (a) single gear or (b) double gear.

5.1.3 Third stage design

For the third stage there are many variants of how the power of the second stage pinions
can be coupled to form the third stage. Each variant would exhibit different properties
in terms of the achievable total ratio, power splitting, as well as the possible number of
generators. This chapter investigates two possibilities, which will be referred to onwards
as concept A and B, and both can be seen in Figure 5-4 (a) and (b) respectively. In
concept A, the second stage pinion 1′′ of each power-split module drives a simple stage.
In other words, each power-split module drives a single generator. This was the variant
covered previously in chapter 3. Since section 3.2.5.4 observed a mass penalty to
increasing the number of generators, concept B is investigated as a possible solution.
Concept B cuts the number of generators in concept A in half. This is achieved by
coupling the power from each two successive power-split modules on one output high-
speed pinion 1′′′ . Comparing concept A against B will allow investigating how the
design of the third stage can influence the whole system mass 𝑀𝑠𝑦𝑠 for the same number
of power-split modules 𝑁𝑚 .
Gearbox Geometry Optimization 77

′ 1′′′
′′′
2 1′′ 2 ′′′ 1′′ 1′′′ 2′′′ 1′′

(a) (b)

𝑅′′ 𝑅′′ 𝑅′′


Figure 5-4: Schematic of third stage variants with example of 𝑵 = . (a) Concept A: each
power-split module remains uncoupled, and (b) concept B: power is summed on one
output pinion ′′′ .

5.2 Spatial Constraints

The radii 𝑅′ and 𝑅′′ define the interface between the stages. 𝑅′ is the interface between
the first stage pinion 1′ and second stage gear 2′′ , and similarly 𝑅′′ is the interface
between the second stage pinion 1′′ and the third stage gear 2′′′ . Numerous sorts of
collisions can occur between the gears, and any combination of the three stages that does
not meet the spatial constraints needs to be excluded from the solution space. There are
three main forms of spatial constraints that must be fulfilled.

Circumferential collision constraints:

The power-split modules are arranged circumferentially about the central gear.
Therefore, the gears of each power-split module must fit into their respective angular
space, which can be represented by equation (5-1). This equation should apply to each
of the three stages, i.e. Figure 5-5 (a-c).

𝜃 (𝑘) < 2𝜋⁄𝑁 (5-1)


𝑚

Gear collision:

This can take place in the first stage between the idler gear and the neighboring pinion,
as shown in Figure 5-5 (a). The mathematical representation that needs to be fulfilled in
this case is given by equation (5-2).

𝐿′ − 1⁄2 (𝑑𝑎′ 1 + 𝑑𝑎′ 2 ) ≥ 𝑚𝑖𝑛 𝑐𝑙𝑒𝑎𝑟𝑎𝑛𝑐𝑒 (5-2)

There are no concrete recommendations in the available literature on the allowable


78 Gearbox Geometry Optimization

distance between neighboring gears. In this dissertation a value of 1.5 times the normal
tooth module is assumed as an acceptable limit.

𝑅′ 𝐿′ 𝑅′ 𝑅′

𝑅′′ 𝑅′′ 𝑅′′

𝜃′ 𝜃 ′′ 𝜃 ′′′

(a) (b) (c)


Figure 5-5: Possible collision situations in the (a) first, (b) second and (c) third stage.

Housing dimension:

The housing dimension 𝐷𝐻 denotes the envelope that encloses the whole gearbox
configuration. The mathematical representation of the housing constraint is given by:
′′
2 × 𝑅′ + 𝑑𝑎,2 ≤ 𝐷𝐻 (5-3)
′′′
2 × 𝑅′′ + 𝑑𝑎,2 ≤ 𝐷𝐻 (5-4)

An additional constraint is applied in the case of a ring gear, where the diameter of the
ring gear also needs to fit within the housing envelope 𝐷𝐻 . Usually the optimization of
a gearbox considers minimizing the housing volume to achieve a desired total ratio as
the main target. In the proposed optimization procedure in this dissertation, the total
ratio is unknown, and the total gearbox ratio leading to a minimum system mass 𝑀𝑠𝑦𝑠 is
supposed to be the output of the calculation. Second, the calculation scheme does not
consider the mass of the gearbox housing. This dissertation takes another approach; the
gearbox housing is considered as a fixed parameter, and the gear dimensions that can fit
within that fixed housing is considered as the solution space. The assumptions of the
gearbox housing dimension are extrapolated from the dimensions of gearboxes of
similar construction, which would differ between a bull and a ring gear. As already
concluded in section 3.2.3 from equations (3-39) and (3-40), the diameter of the gears
relates to the input torque as in equation (5-5).

𝐷 3 ~ 𝑇𝑖𝑛 (5-5)

This is used to extrapolate the housing constraint from the dimensions of gearboxes of
Gearbox Geometry Optimization 79

similar construction. The variant with an external bull gear is extrapolated from a
6.5 MW gearbox with a similar bull gear [58].
3 3
𝐷𝐻,1 𝑇𝑖𝑛,1 𝐷𝐻,1 12.667
3 = → = → 𝐷𝐻,𝑏𝑢𝑙𝑙 = ∅4750 𝑚𝑚 (5-6)
𝐷𝐻,2 𝑇𝑖𝑛,1 36003 5.5

The variant with an internal ring gear is then extrapolated from a generic 6 MW
planetary gearbox with a series planetary PPS configuration [88].
3 3
𝐷𝐻,1 𝑇𝑖𝑛,1 𝐷𝐻,1 12.667
3 = → 3
= → 𝐷𝐻,𝑟𝑖𝑛𝑔 = ∅4000 𝑚𝑚 (5-7)
𝐷𝐻,2 𝑇𝑖𝑛,1 2500 5.15

The conclusion from this section is that all spatial constraints are interdependent, and
each constraint is always dependent on the geometrical aspects of more than just one
stage.

5.3 Optimization Control Parameters

Control parameters are those values that would determine the boundaries of the solution
space. They would trim down the solution space from all non-feasible solutions and
prevent the optimization procedure from wandering off into pointless iterations. The
control parameters are summarized in Table 5-1 and are explained in detail in the
following sections.

Parameter Symbol Limits (or Range)


Negative specific sliding 𝜁𝑓,𝑚𝑎𝑥 -1.5
Profile ratio 𝜖𝛼 1.3 – 1.9

Addendum height coefficient ℎ𝑎𝑝 0.8 – 1.2

Dedendum depth coefficient ℎ𝑓𝑝 1.4

Root radius coefficient 𝜌𝑓𝑝 0.38

Tooth top land coefficient 𝑠𝑎,𝑚𝑖𝑛 0.4

Tooth clearance coefficient 𝑐𝑚𝑖𝑛 0.25
Table 5-1: Summary of the control parameters applied to the proposed optimization
procedure.

5.3.1 Specific sliding and profile shifting

The center distances are the main geometrical parameters that strongly impact the spatial
constraints in the gearbox. For every teeth combination their respective center distance
is dependent on the profile shift sum. As a novel approach in this work, the specific
sliding 𝜁 can be used a control parameter to set the limits on the profile shift for every
teeth combination. The reason behind choosing this parameter is its possession of two
advantageous features; one is mathematical, and the other is operational.
80 Gearbox Geometry Optimization

The operational advantage of specific sliding is its direct impact on the micropitting load
capacity of the gears. It is already known in the literature that specific sliding is an
important influencing parameter [107; 108]. Specific sliding describes the ratio of
slipping to rolling occurring between mating gear flanks along the contact line. Positive
values denote same direction of rolling and slipping and vice versa, while a zero value
indicates pure rolling, which naturally takes place at the pitch point. However, the
regions of negative specific sliding are where micropitting is most likely to occur, as
shown in Figure 5-6 (a). This is the critical region since the direction of slipping tends
to open the micropitting cracks on the driven flank as seen in Figure 5-6 (b). From the
equations for micropitting calculation [109], and knowing the range of contact pressure
and surface roughness values common in wind turbine gearboxes, it would be
advantageous to propose a conservative value of 𝜁𝑓,𝑚𝑎𝑥 = −1.5 as a limit on negative
specific sliding. A further reduction is expected after introducing tip and root relief as
profile modifications, which is not a part of this assessment.

(a) (b)

Figure 5-6: (a) Common area of micropitting. (b) Relative direction of slipping S and rolling R
between meshing teeth [110].

The mathematical advantage of specific sliding is its dependence only on the number of
teeth. This means for each teeth combination the limits on the working pressure angle
𝛼𝑤 , hence profile shift sum Σ𝑥, can be calculated. The formulae of the negative specific
sliding at the roots, according to [75], are given by:

𝑇2 𝐴 𝑢. 𝑇2 𝐸
𝜁𝑓1 = 1 − ; 𝜁𝑓2 = 1 − (5-8)
𝑢. 𝑇1 𝐴 𝑇1 𝐸

By substituting the radii of involute curvature in equation (5-8) in terms of the gears
basic geometry yields an expression of the maximum limit on the profile shift of both
gears x1 and x2:

1 1 ∗
𝑥1,𝑚𝑎𝑥 = 𝑧1 cos 𝛼𝑛 . √1 + 𝐶𝑥1 . tan2 𝛼𝑤 − 𝑧1 − ℎ𝑎𝑝,1 (5-9)
2 2
Gearbox Geometry Optimization 81

1 1 ∗
𝑥2,𝑚𝑎𝑥 = 𝑧2 cos 𝛼𝑛 . √1 + 𝐶𝑥2 . tan2 𝛼𝑤 − 𝑧2 − ℎ𝑎𝑝,2 (5-10)
2 2

Where both constants 𝐶𝑥1 and 𝐶𝑥2 are calculated as:


2 2
( 𝜁𝑓2 − 1) . (𝑢 + 1)2 (𝜁𝑓1 − 1) . (𝑢 + 1)2
𝐶𝑥1 = 2 ; 𝐶𝑥2 = 2 (5-11)
(𝑢 + 1 − 𝜁𝑓2 ) (𝜁𝑓1 . 𝑢 − 𝑢 − 1)

The minimum limit on the profile shift of both gears can simply be found from the
profile shift sum as follows:

𝑥1,𝑚𝑖𝑛 =  𝑥 − 𝑥2,𝑚𝑎𝑥 (5-12)


𝑥2,𝑚𝑖𝑛 =  𝑥 − 𝑥1,𝑚𝑎𝑥 (5-13)

Finally, the solution space of the acceptable profile shift sum would be defined by the
region where the condition in equation (5-14) is fulfilled.

𝑥𝑚𝑖𝑛 ≤ 𝑥𝑚𝑎𝑥 (5-14)

In contrast with the published papers, e.g. [111; 112], usually a general limit is set to the
profile shift 𝑥 regardless of the number of teeth as seen in equation (5-15). These
publications follow the recommendations from [113].

−0.5 ≤ 𝑥 ≤ 1 (5-15)

The impact of the new approach against the state-of-the-art recommendations is


explained with reference to the example given in Figure 5-7. For the given combination
of teeth, applying equations (5-9) to (5-14) would corner the applicable solution into the
small shaded area shown in the figure, while the DIN recommendation would assume
the whole area bound between the two lines of equation (5-15) as an acceptable solution
space. The recommendations from DIN can be advantageous if only the running
characteristics, such as vibrations reduction, are the only optimization target. However,
when stringent spatial constraints need to be applied, using specific sliding as a control
parameter proves to be superior to the optimization procedure.
82 Gearbox Geometry Optimization

𝑥𝑚𝑎𝑥,𝐷 𝑁
𝑥1,𝑚𝑖𝑛

𝑥2,𝑚𝑖𝑛
𝑥1,𝑚𝑎𝑥 𝑥2,𝑚𝑎𝑥

𝑥𝑚𝑖𝑛,𝐷 𝑁

Figure 5-7: Comparing the impact of using specific sliding and DIN recommendation to set the
boundaries on the profile shift, with the example of z1=21, z2=27 and f,max = -1.5.
Red area represents the solution space after applying equations (5-9) to (5-14).

The specific sliding is complemented by other classical control parameters that can
further trim the profile shift solution space and also set the limit on the maximum profile
shift sum, i.e. the right side of the shaded area. These additional parameters are also
mathematically convenient like specific sliding in the sense that they only depend on the
teeth combination. These include: the profile contact ratio 𝜀𝛼 , tooth top land width
coefficient 𝑠𝑎∗ and tip clearance coefficient 𝑐 ∗ . The recommended values for these
parameters are listed in Table 5-1.

5.3.2 Hunting ratio

A hunting ratio is where one tooth would mesh with all the teeth on the mating gear
before it meshes with the same tooth again. This would avoid uneven wear of the gears,
since any given geometrical deviation in one tooth is then distributed evenly across all
the mating teeth and not one tooth in particular. A hunting ratio is achieved when the
maximum common divisor of the teeth numbers is one. An example of a hunting and a
non-hunting ratio can be explained in Table 5-2.

𝑧1 𝑧2 Common divisors Hunting or non-hunting ratio


24 120 2, 3, 4 Non-hunting
23 124 1 hunting
Table 5-2: Example of hunting and non-hunting gear ratios.

5.3.3 Tooth profile

In the attempt to keep the idler diameter at a minimum, non-standard tooth profiles can
help achieve acceptable safety factors at smaller dimensions. The term tooth profile
refers to the proportions of the tooth height above and below the pitch circle, i.e. the
addendum ℎ𝑎𝑝 and the dedendum ℎ𝑓𝑝 heights respectively. Standard cutting tools can
manufacture the tooth profiles listed in DIN 867 [114]. In the meanwhile gear teeth with
Gearbox Geometry Optimization 83

custom proportions have been gaining wide attention, especially in the field of aviation
gear transmissions [115]. This would allow adapting the gear geometry to the specific
needs of the application at hand.

One proposed design of non-standard gear profiles is involute teeth with unequal
addenda [116]. This is a design where the two mating gears possess different addenda,
where the shorter addendum is assigned to the gear member with the lower bending
safety factor 𝑆𝐹 . This is for example the case with planet gears, since their bending
capacity is always reduced by a factor of 𝑌𝑀 = 0.7 due to alternating bending [47]. An
unequal addenda design in this case would assign a lower addendum height coefficient

0.8 ≤ ℎ𝑎𝑝 ≤ 1 to the planet, and a higher coefficient to the other mating gear

1 ≤ ℎ𝑎𝑝 ≤ 1.2. This is in contrast to the standard involute profiles [114] where all gears

share the same addendum coefficient ℎ𝑎𝑝 = 1. The driver behind this concept is its
direct impact on the reduction of the tooth form factor 𝑌𝐹 , equation (5-16), which flows
in the root stress calculation [47]. The reduction is achieved from reducing the bending
moment arm ℎ𝐹𝑒 , seen in Figure 5-8.

6ℎ𝐹𝑒⁄
𝑚𝑛 . cos 𝛼𝐹𝑒𝑛
𝑌𝐹 = 2 (5-16)
𝑠
( 𝐹𝑛⁄𝑚𝑛 ) . cos 𝛼𝑛

The concept of unequal addenda indirectly improves the flank load capacity as well by
allowing higher positive profile shifting. It would allow increasing the tooth root cord
𝑠𝐹𝑛 , while maintaining a short bending moment arm ℎ𝐹𝑒 . For this reason, a non-standard
involute profile is adopted for the proposed gearbox, and the addendum coefficient of
the gears is defined within a given range and used also as a control parameter.

Figure 5-8: Geometrical dimensions of tooth root critical section [47].


84 Gearbox Geometry Optimization

5.4 Mass Index Equations



The derived mass index equations in chapter 3 of the gearbox 𝑀𝑔𝑏𝑥 and the generator

𝑀𝑔𝑒𝑛 are applied to the proposed gearbox, then the solution space is arranged
ascendingly such that the combinations yielding the smallest system mass 𝑀𝑠𝑦𝑠 are
checked first for detailed calculation. This way there is no need to check the strength
and spatial constraints of all the combinations seen in Figure 5-1 (a). The mass equations
of the three stages are listed in Table 5-3.

Description Equations Remarks


First Stage
Share of pinion torque of input 𝑇𝑖𝑛
𝑇1′ =
torque 2𝑁𝑚 𝑢′ 𝑁3′
2 4𝑇𝑖𝑛 1 1 𝑁1′ = 2𝑁𝑚 𝑁3′
Volume of pinions 𝑁1′ . 𝑏𝑑1′ = 2 ( ′ + ′ ′ )
𝜎𝐻𝑙𝑖𝑚 𝑢 𝑢 𝑢12
2 4𝑇𝑖𝑛 𝑢12 ′ 2 ′
𝑢12 𝑁2′ = 2𝑁𝑚 𝑁3′
Volume of idlers 𝑁2′ . 𝑏𝑑2′ = 2 ( ′ + ′ )
𝜎𝐻𝑙𝑖𝑚 𝑢 𝑢

2 4𝑇𝑖𝑛 0.4𝑢 0.4𝑢′ 𝑁3′ = 1 𝑜𝑟 2
Volume of central gear(s) 𝑁3′ . 𝑏𝑑3′ = 2 ( + ′ )
𝜎𝐻𝑙𝑖𝑚 2𝑁𝑚 2𝑁𝑚 𝑢12
′ 2 ′
∗ ′
1 1 𝑢12 𝑢12 0.4𝑢′ 0.4𝑢′
Mass index of first stage 𝑀𝑔𝑏𝑥 = ′+ ′ ′ + ′ + ′ + + ′
𝑢 𝑢 𝑢12 𝑢 𝑢 2𝑁𝑚 2𝑁𝑚 𝑢12
Second Stage
Share of pinion torque of input 𝑇 𝑖𝑛
𝑇1′′ =
torque 2𝑁𝑚 𝑢′ 𝑢′′
2 4𝑇𝑖𝑛 1 1 𝑁1′′ = 𝑁𝑚
Volume of pinions 𝑁1′′ . 𝑏𝑑1′′ = 2 ( ′ ′′ + ′ ′′ 2 )
𝜎𝐻𝑙𝑖𝑚 2𝑢 𝑢 2𝑢 𝑢
′′
2 4𝑇 𝑖𝑛 𝑢 1 𝑁2′′ = 2𝑁𝑚
Volume of gears 𝑁2′′ . 𝑏𝑑2′′ = 2 ( ′ + ′ )
𝜎𝐻𝑙𝑖𝑚 𝑢 𝑢
∗ ′′
1 1 𝑢′′ 1
Mass index of second stage 𝑀𝑔𝑏𝑥 = + + +
2𝑢′ 𝑢′′ 2𝑢′ 𝑢 ′′ 2 𝑢′ 𝑢′
Third stage – Concept A
Share of pinion torque of input 𝑇𝑖𝑛
𝑇1′′′ =
torque 𝑁𝑚 . 𝑖𝑡𝑜𝑡𝑎𝑙
2 4𝑇𝑖𝑛 1 1 𝑁1′′′ = 𝑁𝑚
Volume of pinions 𝑁1′′′ . 𝑏𝑑1′′′ = 2 (1 + ′′′ ) .
𝜎𝐻𝑙𝑖𝑚 𝑢 𝑖𝑡𝑜𝑡𝑎𝑙
2 4𝑇 𝑖𝑛 2 1 𝑁2′′′ = 𝑁𝑚
Volume of gears 𝑁2′′′ . 𝑏𝑑2′′′ = 2 (𝑢′′′ + 𝑢′′′ ).
𝜎𝐻𝑙𝑖𝑚 𝑖𝑡𝑜𝑡𝑎𝑙
∗ ′′′ 2 1 1
Mass index of third stage 𝑀𝑔𝑏𝑥 = (𝑢′′′ + 𝑢′′′ + 1 + ′′′ ) .
𝑢 𝑖𝑡𝑜𝑡𝑎𝑙
Third stage – Concept B
Share of pinion torque of input 𝑇𝑖𝑛
𝑇1′′′ =
torque (per mesh) 𝑁𝑚 . 𝑖𝑡𝑜𝑡𝑎𝑙
2 4𝑇𝑖𝑛 1 1 1 𝑁1′′′ = 𝑁𝑚 /2
Volume of pinions 𝑁1′′′ . 𝑏𝑑1′′′ = 2 ( + ′′′ ) .
𝜎𝐻𝑙𝑖𝑚 2 2𝑢 𝑖𝑡𝑜𝑡𝑎𝑙
2 4𝑇 𝑖𝑛 2 1 𝑁2′′′ = 𝑁𝑚
Volume of gears 𝑁2′′′ . 𝑏𝑑2′′′ = 2 (𝑢′′′ + 𝑢′′′ ).
𝜎𝐻𝑙𝑖𝑚 𝑖𝑡𝑜𝑡𝑎𝑙
∗ ′′′ 2 1 1 1
Mass index of third stage 𝑀𝑔𝑏𝑥 = (𝑢′′′ + 𝑢′′′ + + ′′′ ) .
2 2𝑢 𝑖𝑡𝑜𝑡𝑎𝑙
Table 5-3: Mass index equations for the three stages of the proposed gearbox.
Gearbox Geometry Optimization 85

5.5 Design Constraints

Pitting and bending fatigue are both used as the design constraints that would help define
the main dimensions of the gear components and generate the solution space. According
to the certification guidelines for wind turbine gearboxes [89; 90], only the ISO 6336
standard is accepted for the flank and root strength calculations of gears. The equations
governing the surface strength according to ISO 6336-part 2 [117] are:

σH lim . ZNT
σHP = ZL . Zv . ZR . ZW . ZX (5-17)
SH

Ft u + 1
σH0 = ZH . ZE . Zϵ . Zβ . √ . (5-18)
d1 . b u

σH = ZB,D . σH0 √K A . K v . K Hβ . K Hα (5-19)

Where σH is the contact stress describing the load condition and 𝜎𝐻𝑃 is the permissible
contact stress describing the material strength capacity. Similarly, the equations
governing the root strength according to ISO 6336-part 3 [47] are:

𝜎𝐹𝐸 . 𝑌𝑁𝑇
𝜎𝐹𝑃 = . 𝑌𝛿𝑟𝑒𝑙𝑇 . 𝑌𝑅𝑟𝑒𝑙𝑇 . 𝑌𝑋 (5-20)
𝑆𝐹
𝐹𝑡
𝜎𝐹0 = .𝑌 .𝑌 .𝑌 .𝑌 .𝑌 .𝑌 (5-21)
𝑚𝑛 . 𝑏 𝐹 𝑆 𝐵 𝛽 𝐷𝑇 𝑀
𝜎𝐹 = 𝜎𝐹0 . 𝐾𝐴 . 𝐾𝑣 . 𝐾𝐻𝛽 . 𝐾𝐻𝛼 (5-22)

A design would fulfil the strength criteria when the safety factors meet (or exceed) a
given acceptable minimum value 𝑆𝐻 𝑙𝑖𝑚 , 𝑆𝐹 𝑙𝑖𝑚 . In the literature, e.g. [80], the gear face
width 𝑏 would be determined iteratively to meet a required safety factors. Here a
different approach is adopted where the minimum safety factors are given as an input
and two strength control parameters, namely the face width-to-diameter ratio 𝑏⁄𝑑1 , the
face width-to-module ratio 𝑏⁄𝑚𝑛 are used as a safety test. The equations for both
parameters are derived from the strength equations. Deriving these parameters from the
flank strength equations (5-17) to (5-19) yields:

𝑏 𝐹𝑡 . (𝑍1 + 𝑍2 ). (𝑆𝐻 𝑙𝑖𝑚 . 𝑍𝐵,𝐷 . 𝑍𝐻 . 𝑍𝐸 . 𝑍𝜖 . 𝑍𝛽 )2 . 𝐾𝐴 . 𝐾𝑣 . 𝐾𝐻𝛽 . 𝐾𝐻𝛼


( ) = (5-23)
𝑚𝑛 𝐹𝑙𝑎𝑛𝑘 (𝜎𝐻 𝑙𝑖𝑚 . 𝑍𝑁𝑇 . 𝑍𝐿 . 𝑍𝑣 . 𝑍𝑅 . 𝑍𝑊 . 𝑍𝑋 )2 . 𝑍2 . 𝑍12 . 𝑚𝑛2
86 Gearbox Geometry Optimization

And when driven from the root strength equations (5-20) to (5-22) yields:

𝑏 𝐹𝑡 . (𝑆𝐹 𝑙𝑖𝑚 𝑌𝐹 . 𝑌𝐹 ). 𝐾𝐴 . 𝐾𝑣 . 𝐾𝐻𝛽 . 𝐾𝐻𝛼


( ) = (5-24)
𝑚𝑛 𝑅𝑜𝑜𝑡 (𝜎𝐹 𝑙𝑖𝑚 . 𝑌𝑀 . 𝑌𝑁𝑇 . 𝑌𝑆𝑇 . 𝑌𝛿𝑟𝑒𝑙𝑇 . 𝑌𝑅𝑟𝑒𝑙𝑇 . 𝑌𝑋 ). 𝑚𝑛2

And the control parameters used for calculation in this case would be the larger, i.e. the
less safe, of the two values:

𝑏 𝑏 𝑏 𝑏
( ) = max {( ) ,( ) } ≤( ) (5-25)
𝑚𝑛 𝐶𝑎𝑙𝑐𝑢𝑙𝑎𝑡𝑖𝑜𝑛 𝑚𝑛 𝐹𝑙𝑎𝑛𝑘 𝑚𝑛 𝑅𝑜𝑜𝑡 𝑚𝑛 𝐴𝑙𝑙𝑜𝑤𝑎𝑏𝑙𝑒
𝑏 𝑏 1 𝑏
( ) = ( ) × ≤ ( ) (5-26)
𝑑1 𝐶𝑎𝑙𝑐𝑢𝑙𝑎𝑡𝑖𝑜𝑛 𝑚𝑛 𝐶𝑎𝑙𝑐𝑢𝑙𝑎𝑡𝑖𝑜𝑛 𝑍1 𝑑1 𝐴𝑙𝑙𝑜𝑤𝑎𝑏𝑙𝑒

The factors 𝑌, 𝑍 and 𝐾 serve as bridge between the ideal and the practical loading
conditions of the gears. The problem is that these factors depend as well on the
geometrical modifications from profile shifting. So in order to check a solution space as
large as the one depicted in Figure 5-1, a round of preliminary calculation can eliminate
failing solutions without having to perform profile shifting on any of the teeth
combinations. Preliminary design is based on predefined factors that are based on
experience and would eliminate the solutions that would fail even after optimizing the
profile shifting. Report [80] proposed initial values for the factors for a preliminary
design, which can be seen summarized in Table 5-4. The values for the factors 𝑏⁄𝑚𝑛
and 𝑏⁄𝑑1 are based on experience with gearboxes that have shown good load
distribution factors 𝐾𝐻𝛽 on the test bench at nominal load. Higher values in special cases
are also possible, but it is not the purpose of this dissertation to investigate the maximum
limits on these factors.
Gearbox Geometry Optimization 87

Factor Description Value Remarks / Source


K-Factors
𝐾𝐴 Application factor 1
𝐾𝑣 Dynamic factor 1.05
[80]
𝐾𝛽 Load distribution factor 1.5
𝐾𝛼 Transverse load factor 1
Y-Factors
𝑌𝐹 Form factor 3
𝑌𝑆 Stress correction factor 1
𝑌𝑀 Alternative bending factor 1 (0.7 for
Preliminary

idler) [80]
𝑌𝛿𝑟𝑒𝑙𝑇 Relative notch sensitivity factor 1
𝑌𝑅𝑟𝑒𝑙𝑇 Relative surface factor 1
𝑌𝑋 Size factor 1
Z-Factors
𝑍𝐻 Zone factor 2.3
𝑍𝐵 Single contact factor 1
𝑍𝐸 Elasticity factor 191.6
𝑍𝜀 Contact ratio factor 0.93
[80]
𝑍𝑁𝑇 life factor 1
𝑍𝐿 Lubricant factor 1
𝑍𝑣 Velocity factor 1
𝑍𝑅 Roughness factor 1
𝑏⁄𝑚𝑛 Width-to-module ratio 25
Experience
factors
Safety

𝑏⁄𝑑1 Width-to-diameter ratio 1.2


𝑆𝐻 𝑙𝑖𝑚 Flank safety factor 1.2
[89]
𝑆𝐹 𝑙𝑖𝑚 Root safety factor 1.5
Table 5-4: Factors assumed for the preliminary and detailed strength calculation of gears.

5.6 Assumptions and Input Loads

As previously mentioned, this dissertation serves as a blueprint for designing gearboxes


in the next range of power class 10-20 MW. However, to get a numerical estimation of
output, the calculations are performed with the example of a 10 MW turbine. Also, a
number of assumptions were required to cover all the parameters needed to perform a
complete calculation. All the assumptions and parameters can be found summarized in
Table 5-5. As previously mentioned, the input loads are taken from an internal 10 MW
low-wind rotor project [24]. The operational assumptions here refer to the parameters
that are not directly relevant to the investigation of the impact of the new gearbox design.
These parameters are going to be predefined and assumed fixed throughout the
optimization procedure. These are practical values used in the wind turbine gearbox
practice and are going to be accepted in this dissertation without further investigation.
88 Gearbox Geometry Optimization

Parameter Value
Mechanical input power 11 MW

Loads
Input
Rated rotor speed 8.3 RPM
Rated input torque 12.656 MNm
Electrical Power 10 MW
Gear material 18 CrNiMo 7-6
assumptions
Operational

Gearbox oil Mobilgear SHC XMP 320


Gear quality 6
Surface Roughness (root / flank) 9.6 μm / 4.0 μm
Normal pressure angle 20
Maximum normal module 25
Table 5-5: Operational and geometrical assumptions required for the strength calculations.

5.7 Independent Parameters

The starting matrix of each stage is all the possible combinations of teeth numbers and
would take the form in equation (5-27). Generating such matrix would first require the
definition of the acceptable range of a number of parameters, namely the pinion teeth
number 𝑧1 and the stage ratio 𝑢.
𝑧1 𝑧2
𝑍 (𝑘) = [ ] (5-27)

5.7.1 Pinion teeth number

The minimum limit is defined by the gear mesh interference. This limit is found by
equation (5-28), which links the minimum pinion teeth number with the normal pressure
angle. For a standard cutter with a normal pressure angle of 20, equation (5-28) yields
18 as a minimum pinion teeth number.

2
𝑧1,𝑚𝑖𝑛 ≥ (5-28)
sin2 𝛼𝑛

Figure 5-9 shows how the applicable range of pinion teeth number is usually
determined. For a minimum volume gearset, an increase in the pinion teeth number
would reduce the normal module. Since the flank contact stress is mainly defined by the
relative diameters of the mating gears, the limit set by the pitting is seen almost constant
in the figure. However, the drop in module would cause the bending limit to fall with
the increase in the number of pinion teeth.
Gearbox Geometry Optimization 89

Figure 5-9: Schematic of the applicable pinion teeth number [118].

Increasing the number of teeth – with a simultaneous decrease in gear module – has the
advantage of increasing the profile contact ratio, hence lower noise levels. For this
reason, the maximum is set at 35, increased to 45 for the high-speed third stage.
Therefore, the vector of the pinion teeth number can be represented as:

𝑍1 (1) = 𝑍1 (2) = 18: 35; 𝑍1 (3) = 18: 45 (5-29)

5.7.2 Stage ratios

The optimum gearbox total ratio as well as its optimum distribution on the three stages
is the expected output of the optimization procedure. The possible range of each stage
is however the input of the procedure. It has already been proven in section 3.2.5.3 that
the mass reduction potential is achieved by maximizing the ratio of the first stage. The
maximum value would always be limited by the housing envelope assumed. For the
housing dimensions assumed in section 3, the limit on the first stage ratio can be defined
in the range:

𝑢(1) = 4: 25 (5-30)

The increase in the first stage ratio would increase the pitch circle 𝑅′ . This would allow
a small space for the second stage, since their gears 2′′ would have to fit between 𝑅′ and
the housing envelope 𝐷𝐻 . This forces the ratio of the second stage to vary in a very tight
range, as in equation (5-31).

𝑢(2) = 1: 3 (5-31)

In the third stage it would be advantageous to use larger ratios to increase the total ratio
of gearbox. The reduced torque in this stage would lead to smaller gear dimensions and
allow significantly larger ratios. Therefore, the limit set on the third stage ratio is
relatively high as seen in equation (5-32).
90 Gearbox Geometry Optimization

𝑢(3) = 1: 7 (5-32)

The ratio that would need additional investigation to estimate is the idler-to-pinion ratio
𝑢12 = 𝑧2′ ⁄𝑧1′ . The idler has a mass penalty and it has only been added to serve the bearing
concept previously explained in chapter 4. But reducing this mass penalty does not mean
necessarily minimizing this ratio. The flank contact stress, according to equation (5-18),
decreases by increasing the ratio, which in turn leads to smaller gear dimensions. To
understand the effect of the idler-to-pinion ratio on the mass of the stage, the mass index
equations from Table 5-3 are applied to the first stage. Afterwards the idler-to-pinion
ratio is plotted against the mass of the first stage at different first stage ratios, as seen in
Figure 5-10. The figure shows that the minimum mass is not always at the minimum
idler-to-pinion ratio, and would fall in the range:

𝑢12 = 1: 2.5 (5-33)

The figure shows the trend for a given assumption of 𝑁𝑚 and 𝑁3′ . Different values yield
however the same trend.

𝑢′ = 5

𝑢′ = 10

𝑢′ = 20


Figure 5-10: Relation between the mass index of first stage 𝑴∗𝒈𝒃𝒙 and the idler-to-pinion ratio
𝒖 . Using 𝑵 = and 𝑵′ = as an example.

5.8 Outline of Optimization Procedure

The elements of the optimization procedure explained in the previous sections are
integrated together in the following steps:

Similar to equation (5-27), the teeth matrix of each stage 𝑍 (𝑘) is generated using the
independent parameters in section 5.7. Here an elimination process takes place,
where the matrices 𝑍 (𝑘) are scanned for hunting ratios, see section 5.3.2, and any
Gearbox Geometry Optimization 91

non-hunting ratio combinations are eliminated.


The control parameters from section 5.3 are applied to each matrix 𝑍 (𝑘) , and the
profile shift sum limits, i.e. Σ𝑥𝑚𝑖𝑛 and Σ𝑥𝑚𝑎𝑥 , are calculated for each tooth
combination in each matrix, as in equations (5-34) to (5-36).
′ ′ ′ ′
′′′
𝑧1′ 𝑧2′ 𝑧3′ Σ𝑥12,𝑚𝑖𝑛 Σ𝑥12,𝑚𝑎𝑥 Σ𝑥23,𝑚𝑖𝑛 Σ𝑥23,𝑚𝑎𝑥
𝑍 =[ ] (5-34)

′′
′′′
𝑧1′′ 𝑧2′′ Σ𝑥𝑚𝑖𝑛 ′′
Σ𝑥𝑚𝑎𝑥
𝑍 =[ ] (5-35)

′′′
′′′
𝑧1′′′ 𝑧2′′′ Σ𝑥𝑚𝑖𝑛 ′′′
Σ𝑥𝑚𝑎𝑥
𝑍 =[ ] (5-36)

The preliminary design phase starts by applying the equations from section 5.5 to
the first stage matrix 𝑍 ′ in equation (5-34). As a result, the minimum normal module
𝑚𝑛′ for each teeth combination is calculated. The maximum limit on the normal

module 𝑚𝑛,𝑚𝑎𝑥 should yield a first pitch circle radius 𝑅′ within the housing envelope
limits. The solution matrix 𝑆 ′ of the first stage is given by equation (5-37).
′ ′ ′

𝑧1′ 𝑧2′ 𝑧3′ 𝑚𝑛,𝑚𝑖𝑛 𝑚𝑛,𝑚𝑎𝑥 𝑅𝑚𝑖𝑛 ′
𝑅𝑚𝑎𝑥
𝑆 =[ ] (5-37)


Teeth combinations with 𝑅𝑚𝑖𝑛 beyond the housing envelope limits are eliminated.

The information in equation (5-37) allows proceeding with the preliminary


dimensioning of the second stage. The ratio 𝑢′ = 𝑧3′ ⁄𝑧1′ determines the torque in the
second stage, which then sets the lower limit on the normal module 𝑚𝑛′′ in the second
stage. The range of the radius 𝑅′ defines the available space to the second stage, thus
sets the upper limit on the gear diameters, hence the normal module. Thus, the range
on the second pitch circle radius 𝑅′′ can be estimated. Now the solution matrix 𝑆 ′′
combines each solution from 𝑆 ′ with a matrix of the possible solutions of the second
stage.
′′ ′′ ′′
′ ′ 𝑧1′′ 𝑧2′′ 𝑚𝑛,𝑚𝑖𝑛 𝑚𝑛,𝑚𝑎𝑥 𝑅𝑚𝑖𝑛 ′′
𝑅𝑚𝑎𝑥
𝑢′ 𝑢12 𝑅𝑚𝑖𝑛 ′
𝑅𝑚𝑎𝑥 [ ]
𝑆 ′′ = [ ] (5-38)

Checking the spatial constraints on each combination at this stage is not practical.
Here the elimination is done by applying the spatial constraints to the convenient
limits. This means for example, the housing constraint – equation (5-3) – is applied
′′
to the smallest second stage gears, i.e. 𝑚𝑛,𝑚𝑖𝑛 , at the smallest first pitch circle, i.e.

𝑅𝑚𝑖𝑛 . The circumferential constraint – equation (5-1), Figure 5-5 (a) – is applied to
92 Gearbox Geometry Optimization

′′ ′
the smallest second stage gears, i.e. 𝑚𝑛,𝑚𝑖𝑛 , at the largest first pitch circle, i.e. 𝑅𝑚𝑎𝑥 .
This relaxed application of the spatial constraints is more than enough to filter
“most” and not “all” failing solutions from equation (5-38).

Similarly, the preliminary dimensioning of the third stage requires the ratios of the
preceding stages to determine the load, and the second pitch circle radius 𝑅′′ to
determine the available space. The third stage solution matrix is not calculated for
each solution in the second stage solution matrix in equation (5-38). To avoid a third
level branching of the solution space, as in Figure 5-1 (a), the third stage solution is
calculated for the best convenient conditions, lowest load, i.e. highest ratio 𝑢′ × 𝑢′′ ,
and the largest possible range of the second radius 𝑅′′ from matrix 𝑆 ′′ in equation
(5-38).
′′′ ′′′
′′ 𝑧1′′′ 𝑧2′′′ 𝑚𝑛,𝑚𝑖𝑛 𝑚𝑛,𝑚𝑎𝑥
𝑢′ ′′
𝑢𝑚𝑎𝑥 min (𝑅min ′′
) max (𝑅𝑚𝑎𝑥 )[ ]
𝑆 ′′′ =[ ] (5-39)

Similar to step 4, the relaxed application of spatial constraints is done here again to
matrix 𝑆 ′′′ in equation (5-39).

Finally, the first stage solution from 𝑆 ′ , and its respective second stage matrix from
𝑆 ′′ , and their respective third stage matrix from 𝑆 ′′′ are combined together in matrix
𝑆, equation (5-40). This matrix represents all the possible total ratio distributions
across the three stages.

𝑢′′ (1,1)
′′′ ′′′
𝑆= 𝑢′ ′
𝑢12 [𝑢12 (1,1) … 𝑢12 (1,𝑗) ] (5-40)
′′
[𝑢 (𝑖,1) ]

[ ]
The mass index equations in Table 5-3, followed by the system mass equations
(3-10), (3-14) and (3-1), are applied to the ratio combinations in matrix 𝑆. This yields
equation (5-41), which gives the system mass 𝑀𝑠𝑦𝑠 for all the possible ratio
combinations.
Gearbox Geometry Optimization 93

𝑀𝑠𝑦𝑠 (1,1) … 𝑀𝑠𝑦𝑠 (1,𝑗)

𝑀𝑠𝑦𝑠 = ⋱ (5-41)
[ 𝑀𝑠𝑦𝑠 (𝑖,1) … 𝑀𝑠𝑦𝑠 (𝑖,𝑗) ]
(𝑘,1)

[ ]
The minimum 𝑀𝑠𝑦𝑠 values has the indices (𝑖, 𝑗, 𝑘), where 𝑘 corresponds to the order
of the first stage solution in 𝑆 ′ , 𝑖 is the row index in the corresponding second stage
solution matrix in 𝑆 ′′ , and 𝑗 is the column index in the corresponding third stage
solution matrix in 𝑆 ′′′ . By using these indices, all the ratios, hence teeth numbers the
limits on profile shift are extracted from equations (5-34) to (5-36). The teeth
modules are subsequently extracted from equations (5-37) to (5-39).
The entry of the final step is then generated by the permutations of the profile shifts
and normal modules, as seen in equation (5-42). The last step is applying the spatial
constraints, from section 5.2, and the exact design constraints, from section 5.5. If
equation (5-42) after elimination gives an empty matrix, the next lowest system mass
from equation (5-41) is selected, and step 8 is repeated until equation (5-42) yields a
final solution.
𝑚𝑛′ 𝑧1′ 𝑧2′ 𝑧3′ Σ𝑥12
′ ′
Σ𝑥23 𝑚𝑛′′ 𝑧1′′ 𝑧2′′ Σ𝑥 ′′ 𝑚𝑛′′′ 𝑧1′′′ 𝑧2′′′ Σ𝑥 ′′′
[ ] (5-42)

The steps from 1 to 9 are carried out for a given combination of number of split modules
𝑁𝑚 and number of central gears 𝑁3′ . The steps are repeated as much as the number of
proposed combinations.

5.9 Optimization Results and Conclusions

The optimization procedure described in steps 1 to 9 were applied to all the proposed
gearbox variants in section 5.1. The number of the power-split modules
𝑁𝑚 have been varied from 1 up to 10, where 10 is simply chosen as an arbitrary value.
The results of the gearbox, generators and system masses are plotted against the gearbox
total ratio in Figure 5-11, Figure 5-12 and Figure 5-13 respectively. The data for these
plots are compiled and presented in Appendix A. For a comparison against state-of-the-
art wind turbine gearboxes, the classic series planetary PPS configuration is chosen as a
benchmark.

To facilitate reading the figures and for future references: black markers are bull, white
94 Gearbox Geometry Optimization

markers are ring; solid lines are single central gear and dashed lines are double; circles
are third stage concept A and diamonds are third stage concept B. Finally, a border line
can be seen splitting Figure 5-13 into two halves at gearbox ratio ~144, which
corresponds to a generator speed of 1200 rpm. Thus, the left half of Figure 5-13
represents the mid-speed drivetrain, and the right half represents the high-speed.

The results match to a great extent the predictions made in section 3.2.5.4. Here are a
number of remarks on the similarities and differences with the previous predictions:

Non-planetary gearboxes are superior in the high-speed range:

In the gearbox ratio region above 150, non-planetary gearboxes exhibit a lower mass
𝑀𝑔𝑏𝑥 than the conventional PPS. As seen in Figure 5-11, non-planetary gearboxes start
showing their mass advantage over the PPS configuration first at 𝑁𝑚 > 5. This is
however a notable difference in the results between this section and section 3.2.5.4,
where the mass advantage started earlier at 𝑁𝑚 > 3. The reason is attributed to the mass
penalty by adding the idler gears between each pinion and the central gear. This reflects
similarly on the system mass 𝑀𝑠𝑦𝑠 .

2 2

3
3
3
5
4 4 4
4
4
4 5 5
6 6
5 6
7 6
6 6
6 6 7 7
8 9 7 8 8 8
8 9
8 10 8 9
8 9
10 10 10
10 10 10

Figure 5-11: Gearbox mass results of the proposed gearbox against a state-of-the-art series
planetary PPS configuration. The numbers represent the power-split modules 𝑵 ,
A and B represent third stage design variants according to section 5.1.3.
Gearbox Geometry Optimization 95

5
6
7
8 9
10
3
10
1 3
4
5 5 6 7 8 9
4 8
3 7 10 10
4 5 6 9
2 4 5 5 6 9
3 7 8
4 4
2 3 3 5
5
1 2 3 4

Figure 5-12: Generator mass results of the proposed gearbox against a state-of-the-art series
planetary PPS configuration. The numbers represent the generators. A and B
represent third stage design variants according to section 5.1.3.

1
1
3
5
3
3

6 4
4
4
2 2
3 5
7 5 5
2 6
8 6
7 6
9 7
3 10 7
4 3 8 8 8
3 9 9
10 9
5 4 4 10 10
4
5
5 5

Figure 5-13: System mass results of the proposed gearbox against a state-of-the-art series
planetary PPS configuration. The numbers represent the generators, A and B
represent third stage design variants according to section 5.1.3.

The lowest system mass 𝑀𝑠𝑦𝑠 by a bull gear configuration is at 130.7 ton, drives five
generators, with two central gears and has a total ratio of 295.5. The lowest system mass
𝑀𝑠𝑦𝑠 by a ring gear configuration is also at 130.6 ton, drives five generators, with two
central gears and has a total ratio of 344.4.
96 Gearbox Geometry Optimization

The influence of the central gear design in the first stage is confirmed:

The proposed central gear designs are compared against the classic single bull gear in
Figure 5-14. The most important remark here is that the ring gear solutions are for a 4 m
housing diameter, against 4.75 m for the bull gear. The ring gear has a compactness
advantage against the bull gear design, that even makes it spare with planetary
gearboxes. This compactness advantage also manifests itself in the gearbox variants
with low power-splitting, e.g. one generator, where the bull gear solution did not yield
any results that fit within the defined housing diameter in contrast to the ring gear
solution. The results in Figure 5-14 are for the gearbox variant with a power-summing
third stage (concept B), and concept A shares the same trend.

Figure 5-14: Comparison between the central gear variants, with example of a power-summing
third stage-stage (concept B).

The advantage of a high ratio in the first stage:

These presented results can all be linked to the ratio of the first stage. The advantage of
non-planetary against planetary can be attributed to the possibility of increasing the first
stage ratio and power-split modules simultaneously. The advantages of double central
gears against a single gear, and the ring against the bull design is also attributed to the
difference in the achievable first stage ratio. To show this feature, the results are
summarized in Figure 5-15.

The same figure can also explain how strongly the spatial constraints influence the
design of all stages. The slight difference in the first stage ratio is attributed to the
difference in the construction and spatial requirements between concept A and B at the
third stage. This is more evident in the ring gear design, since the gears are more
crowded, being arranged inwards, and the chances of collision are higher.
Gearbox Geometry Optimization 97

Figure 5-15: Increase of first stage ratio 𝒖′ with the number of power-split modules 𝑵 . A and
B represent third stage design variants according to section 5.1.3.

Impact of generators on system mass is confirmed:

As seen in Figure 5-12, the total generator mass is dependent on the number of
generators, and the gearbox total ratio. In the figure, all the diamonds, i.e. concept B,
can be seen significantly lighter than their circle counterparts, i.e. concept A. This shows
that the choice of the third stage – despite being the lightest – has a significant impact
on the generator side. With the example of using 10 power-split modules and a single
ring gear, the change in the third stage from concept A to B does not have any influence
on the gearbox mass, as seen in Figure 5-11. However, the change in the third stage led
to a 10.5 ton reduction on the generator side from total mass of 40.6 ton down to 30.1
ton, as seen in Figure 5-12.
98 Summary and Outlook

6 Summary and Outlook

6.1 Summary

Offshore wind energy has the technical potential to supply Europe with its required share
of electricity by 2030 [119]. The main problem standing in the way of fully exploiting
this potential is the offshore energy cost. In the pursuit to bring down the offshore cost
of energy, studies and technical reports concluded that replacing multiple smaller
turbines with fewer larger turbines can have the strongest contribution to the cost
reduction [11–14]. A number of studies [17–20] consequently recommended upscaling
the turbine capacity into the 10-20 MW range in the future. Studies [11; 12; 21; 22]
added that upscaling must be supplemented with technical innovation. Technical
improvements in the nacelle are expected to have the strongest contribution to cost
reduction, provided that the improvements are oriented towards two goals: mass
reduction and reliability improvement. This dissertation is dedicated to achieving these
two goals from the point of view of the main gearbox, with the example of a 10 MW
turbine.

The first research objective is how the gearbox configuration can contribute to the mass
reduction of the nacelle. This objective considers how the generator mass is impacted
by the gearbox design, and how to reduce both masses simultaneously. The end result
of this objective is the selection of a gearbox configuration, its total ratio, the distribution
of this ratio among the gearbox stages and the number of driven generators. Since a
relatively endless spectrum of gearbox configurations is possible, the dissertation limits
the selection to the state-of-the-art gearboxes, i.e. planetary and non-planetary split-
torque gearboxes.

The second research objective is how to develop a bearing concept that renders the gear
teeth insensitive to the critical and asymmetrical deformations in the gearbox. The
concept must also provide a method for the gears to achieve equal load sharing among
the power-split paths within the gearbox. The end result of this objective is a new bearing
construction that provides the gears with a self-aligning mechanism, and also provides
enough flexibility to compensate for the gear position errors due to manufacturing and
assembly tolerances.

Chapter 3 is dedicated to determining the topology of the gearbox. The process starts
with deciding on the gearbox configuration between planetary and non-planetary
gearboxes. Two mathematical models are derived. The first model is a mathematical
representation of the gearbox mass derived from the Willis equations, commonly known
as the solid rotor volume method [79]. The model represents the gearbox mass as a
Summary and Outlook 99

function of the input torque, the total ratio, the stage ratios and the geometrical attributes
of each gear stage. Applying this model to the state-of-the-art gearboxes renders the
mass of each gearbox configuration as a function of its general attributes. Thus, the mass
of multiple gearbox configurations can be compared across a desired range of total ratio.
The second model is a mathematical representation of the generator mass as a function
of the generator torque and speed. The generator torque and speed in this model can be
substituted in terms of the influencing gearbox parameters: total ratio and number of
output shafts. By applying these two models to the state-of-the-art gearbox
configurations, the system mass through each gearbox configuration can be investigated
and compared across a desired range of gearbox total ratio. The results of the
comparative study show that non-planetary split-torque gearboxes have a significant
advantage over planetary gearboxes in the high-speed range, i.e. > 1200 rpm, in addition
to using multiple generators > 3. The ability to simultaneously increase both the number
of pinions, i.e. power splitting, in the first stage, and the ratio of the first stage gives a
significant advantage to non-planetary gearboxes.

Deciding on the bearing concept starts by taking advantage of the success of the state-
of-the-art flexpin bearing concept in achieving equal load sharing in planetary
gearboxes. The concept is adapted to a non-planetary gearbox by virtue of the kinematic
analogy between planets and idlers. Therefore, for the chosen gearbox configuration,
the first stage is equipped with idlers mounted on flexpins between each pinion and the
central gear. In order to develop a self-aligning mechanism using flexpins, all types of
gear misalignments were first identified. One type of gear misalignment was found to
be critical, where the misalignment angle falls in the plane of action of the meshing
teeth. This misalignment leads to edge loading of the flanks in contact, which generates
an additional parasitic tilting moment on the gears. However, in this work, this moment
is used in combination with the flexibility from the flexpin as a restoring action for the
idler. An idler is thus positioned between the central gear and each pinion such that the
restoring action from a mesh point causes an uncritical misalignment at the second mesh
point. The angle that satisfies this condition is where the two planes of action at the two
mesh points fall perpendicular to each other.

Chapter 4 proceeds with verifying the hypothesis behind the proposed self-aligning
bearing concept. The self-aligning mechanism is verified by changing the orientation of
the two flank tangents at the two gear meshes on the idler. For each relative orientation
of the flank tangents, a critical angular misalignment of the shafts is given as an input,
and the face load factor is determined at both mesh points for each misalignment. Two
FE-models are built for both the external and internal gear variants to understand the
difference between their behavior under angular misalignment. To understand the
impact of the direction of contact pattern shift across both flanks, each model is
100 Summary and Outlook

investigated at a total of 8 different combinations of angular misalignment directions.


The results confirmed the hypothesis, and the position that satisfies the condition is
where the two planes of action fall perpendicular to each other. The improvement in the
load distribution factor 𝐾𝐻𝛽 in one of the misalignment scenarios reached 40 %. The
proposed concept was also found sensitive to the direction of contact shift at each flank.
The proposed concept worked best, when the contact shift at both idler gear meshes
were in opposite directions. However, the concept yielded the worst results when both
misalignments shifted the contact in the same direction. The results of the simulations
showed no differences between the behavior of the bull and ring gear under angular
misalignment.

Chapter 5 provides an optimization procedure for the geometry of the proposed gearbox.
The challenge in optimizing this design is fulfilling the spatial constraint conditions
without having to examine each possible combination of the three stages of the gearbox.
An optimization procedure is proposed in this chapter based on a combination of brute
force calculations and an embedded elimination-principle. The optimization considers
different variants of the proposed gearbox: an external bull and an internal ring as central
gears; with a single and two central gears; with coupled and uncoupled power-split
modules in the third stage. The size of the housing is assumed constant and extrapolated
from similar state-of-the-art gearboxes; 4.75 m for the bull gear design and 4 m for the
ring gear design. Specific sliding is proposed here as a novel control parameter that can
set the limits on the profile shift sum for each teeth combination and can eliminate the
teeth combinations that do not satisfy the spatial constraints early in the calculation
procedure. The procedure starts by generating a solution space with all the possible teeth
combinations within the limits of preset ranges for each stage ratio. The non-feasible
teeth combinations are filtered out by means of spatial constraint conditions, control
parameters and strength requirements. The mass index equations derived in Chapter 3
are finally applied to the remaining teeth combinations and the solutions yielding the
lowest system mass 𝑀𝑠𝑦𝑠 are examined first. The input loads for this procedure are taken
from an internal 10 MW rotor project at the RWTH Aachen [24], and the number of
power-split modules is varied from 1 to 10, the total gearbox ratio is varied from 50 to
350, and the number of stages is fixed to 3.

The outcome confirms the hypothesis made in chapter 3; non-planetary gearboxes have
the advantage over planetary gearboxes in the high-speed range, i.e. generator speed
>1200 rpm. Because of the mass penalty of adding idlers in the first stage, the proposed
design has the advantage over planetary gearboxes only with power-split modules above
5. The optimum bull gear configuration drives five generators, with two central gears
and has a total ratio of 295.5. The optimum ring gear configuration drives also five
generators, with two central gears and has a total ratio of 344.4. The results have shown
Summary and Outlook 101

the importance of including the generators in the optimization procedure. One example
showed two gearboxes having the same mass, but a difference of around 10.5 ton on the
generator side, simply by changing the number of generators through changing how the
gears are coupled in the third stage.

6.2 Outlook

Both planetary and non-planetary configurations can benefit from the results of this
work. The results have recognized different improvement potentials, which can open the
door for new research questions. These are presented in the following sections.

6.2.1 Helical gears in a semi-pivoting gear system

The presented semi-pivoting gear system has been investigated only through using spur
gears. Helical gears would behave differently in this system, however it is not known to
which extent. The helical idler would be subjected to an additional tipping moment, in
a similar fashion to a helical planet, because of the opposite axial forces acting on both
mesh points. In this context, the helical angle should not be investigated as a standalone
parameter, and should be considered together with the idler gear dimensions in the
manner shown in equation (6-1). The width 𝑏 influences the distribution of axial forces,
and the diameter 𝑑 is the arm of the tipping moment exerted by the axial forces.

𝛽
𝑏⁄ (6-1)
𝑑𝑖𝑑𝑙𝑒𝑟

It is still not known if introducing the helical gears would also influence the gears
orientation angle 𝛾.

6.2.2 Application of lead correction in a semi-pivoting gear system

The lead correction has been excluded from the investigation of the semi-pivoting gear
system in chapter 4. Similar to the study [76], lead correction optimization for a load
spectrum should be also investigated in the proposed semi-pivoting gear system. It is
expected that the combination of lead correction and the proposed bearing concept can
yield interesting results. As Figure 6-1 shows, lead correction is usually optimized for
the nominal load, and the gears will experience an unfavorable stress distribution at both
higher and lower loads. But with a self-aligning bearing concept, the lead correction can
be optimized for the lower loads, and the self-aligning bearing concept takes over
balancing the load distribution at higher loads.
102 Summary and Outlook

Figure 6-1: Possible influence of a self-aligning concept on the load distribution deviation under
a load spectrum.

6.2.3 Applying the semi-pivoting concept to a planetary stage

Planetary stages can still benefit from the outcome of investigating the semi-pivoting
gear system. In a planetary stage as seen in Figure 6-2 (a), the angle of gears orientation
𝛾 is fixed to 180 and cannot be changed. However, the relative orientation of the base
tangent can still be a variable and is influenced by the profile shift sum at both gear
meshes. Reaching a perpendicular relative position between the base tangents is
impossible, as it means that both pressure angle must be around 45 to achieve 𝛾 = 180.
The angle 𝛾 ∗ between the base tangents can be influenced by the choice of profile
shifting at the ring and sun mesh points. It can be influenced even to a greater extent by
the use of asymmetric teeth as in Figure 6-2 (b), where two different pressure angles can
be assigned to the two flanks of the planet.

𝑩𝑻 𝒍𝒂 𝒕−𝒔𝒖

γ ∗

𝑩𝑻 𝒍𝒂 𝒕−𝒓𝒊 𝒈

(a) (b)
Figure 6-2: (a) Relative orientation of base tangents BT in a planetary stage. (b) Asymmetric
teeth [115].

This can open the door for a novel application of asymmetric teeth. This would add a
new objective function to the optimization of pressure angles on the sun and ring flanks
of the planet.
Summary and Outlook 103

6.2.4 Reduction of multi-generators mass penalty through innovation

Chapter 5 has shown that a slight modification in the design of the third stage can cut
down the number of generators into half and reduce the mass penalty on the generator
side (obviously with respect to an even number of power-split modules 𝑁𝑚 ). However,
a further reduction in the driven generators can be difficult because summing the power
to a single output in a non-planetary design would require multiple gear stages to cover
the large center distance.

However, there are other innovative solutions for the third stage that can sum the power
on a single output and in a single stage. There are two plausible solutions that might
have a promising potential. One solution, in Figure 6-3 (a), has been proposed before in
[120]. It comprises a hydrostatic transmission, where multiple output shafts can drive
multiple pumps, and their output can drive a single hydraulic motor, hence a single
generator.

Figure 6-3: Schematic of a mechanical-hydraulic drivetrain [120].

A belt-drive can also span significantly large center distances in a single stage. The
reduced torque in the third stage should not impose a challenge to the load capacity of
the belt. However, it is not yet known how the belt drive would behave dynamically in
the case of summing the power from multiple drivers to a single output.
104 References

7 References

[1] Global Wind Energy Council GWEC: Global Wind Report; 2018.
[2] Knorr, K.; Horst, D., et al.: Energiewirtschaftliche Bedeutung der Offshore-
Windenergie für die Energiewende, Update 2017; Fraunhofer IWES, 2017.
[3] Lütkehus, I.; Salecker, H.: Onshore Wind Energy Potential in
Germany; Current Study by the Federal Environment Agency on the
Nationwide Area and Output Potential; Umweltbundesamt, Germany, 2013.
[4] Saranyasoontorn, K.; Manuel, L.: A Comparison of Wind Turbine Design
Loads in Different Envirnonments Using Inverse Reliability Techniques;
ASME Transactions, Journal of Solar Energy Engineering, Vol. 126, pp. 1060-
1068, 2004.
[5] Wind Europe: Wind in Power 2017; 2018.
[6] Lacal Arantegui, R.; Serrano Gonzalez, J.: 2014 JRC Wind Status Report;
European Commission, Joint Research Centre, Institute for Energy and
Transport, 2014.
[7] Bundesministeriums der Justiz und für Verbraucherschutz: Gesetz für den
Ausbau erneuerbarer Energien; 2014.
[8] Department of Energy and Climate Change: UK Renewable Energy Roadmap;
2011.
[9] Milborrow, D.: At the Tipping Point - 2017 Wind Cost Analysis; Windpower
Monthly, 2018.
[10] Watanabe, J.: Giant Fall in Generating Costs from Offshore Wind; Bloomberg
New Energy Finance, 2016.
[11] Valpy, B.; Hundleby, G., et al.: Future Renewable Energy Costs - Offshore
Wind; BVG Associates, 2017.
[12] BVG Associates: Offshore Wind Cost Reduction Pathways - Technology Work
Stream; 2012.
[13] Hobohm, J.; Krampe, L., et al.: Cost Reduction Potentials of Offshore Wind
Power in Germany; Prognos AG and The Fichtner Group, 2013.
[14] Arwas, P.; Charlesworth, D., et al.: Offshore Wind Cost Reduction Pathways
Study; The Crown Estate, 2012.
[15] Wind Europe: Wind Energy in Europe in 2018: Trends and Statistics; 2019.
[16] Skopljak, N.: First Blade Installed on Haliade-X 12MW; 2019.
https://www.offshorewind.biz/2019/10/14/first-blade-installed-on-haliade-x-
12mw/. Access on: 24 Nov 2019.
[17] The European Wind Energy Association EWEA: The European Wind
Initiative; 2010.
References 105

[18] European Wind Energy Technology Platform: Strategic Research Agenda and
Market Deployment Strategy; 2014.
[19] The European Wind Energy Association EWEA: UpWind - Design Limits and
Solutions for Very Large Wind Turbines; 2011.
[20] Chaviaropoulos, P. K.; Natarajan, A.; Jensen, P. H.: Key Performance
Indicators and Target Values for Multi-Megawatt Offshore Turbines;
Proceedings of EWEA 2014, Barcelona, 2014.
[21] Chaviaropoulos, P. K.; Natarajan, A.: Definition of Performance Indicators
(PIs) and Target Values; Deliverable D1.2.2, INNWIND.EU, 2014.
[22] Klinger, F.: UPSCAL - Beschreibung von Gesetzmäßigkeiten des
Größenwachstums von Windkraftanlagen und Lösungsansätze für die sich
daraus ergebenden Problemstellungen; Hochschule für Technik und Wirtschaft
des Saarlandes, 2002.
[23] Wind Europe: Wind Energy in Europe - Outlook to 2022; 2018.
[24] Berroth, J.; Busch, C., et al.: 10 MW Low Wind Turbine - Load Calculations
for Design Purposes; 14th World Wind Energy Conference & Excibition,
Jerusalem, Israel, 2015.
[25] Schulze, T.: Design and Optimization of Planetary Gears Considering All
Relevant Influences; Gear Technology Magazine, pp. 96-102, 2013.
[26] Giger, U.; Anaudov, K.: Redesign of a gearbox for 5 MW Wind Turbines;
Proceedings of the ASME 2011 International Design Engineering Technical
Conferences & Computers and Information in Engineering Conference,
Washington DC, USA, 2011.
[27] Li, Y. F.; Xu, Y. X.; Li, G. X.: Optimization Design of the Wind Turbine
Gearbox Based on Genetic Algorithm Method; Materials Science Forum, Vol.
697-698, pp. 697-700, 2012.
[28] Leimann, O.; Götz, M., et al.: New Drive Train Concepts for Multi-Megawatt
Wind Turbines; Conference for Wind Power Drives CWD 2015, Aachen,
Germany, 2015.
[29] Gold, P.: Stand der Technik Windenergieanlagen mit Getriebe;
Antriebstechnik, Vol. 6, pp. 42–49, 2005.
[30] Guo, Y.; Keller, J.; LaCava, W.: Planetary Gear Load Sharing of Wind Turbine
Drivetrains Subjected to Non-Torque Loads; Wind Energy, Vol. 18, Issue 4, pp.
757–768, 2015.
[31] Boiger, P.: Gear Units for Wind Power Plants; Renk AG.
[32] REpower Systems AG: 6M - The Further Development of the Successful
Offshore Turbine REpower 5M.
[33] Quilter, J.: Vestas find its way through the rough times; Wind Power Monthly,
2014.
106 References

[34] Eize de Vries: MHI Vestas Launches 9.5 MW V164 Turbine in London;
Windpower Monthly, 2017.
[35] MHI Vestas Offshore Wind: World’s most powerful wind turbine once again
smashes 24 hour power generation record as 9 MW wind turbine is launched;
2017. http://www.mhivestasoffshore.com/new-24-hour-record/. Access on: 4
Nov 2018.
[36] Arafa, H.; Bedewy, M.: Quasi-Exact-Constraint Design of Wind Turbine
Gearing; Proceedings of ASME 2010 Power Conference POWER2010,
Chicago, Illinois, USA, 2010.
[37] Abousleiman, V.; Velex, P.; Becquerelle, S.: Modelling of Spur and Helical
Gears with Flexible Rings and Carrier; ASME Transactions, Journal of
Mechanical Design, Vol. 129, pp. 95–106, 2007.
[38] Klein-Hitpass, A.: Planetenträger für ein Planetengetriebe; EP1186804, 2002.
[39] Litzba, J.: ZF Wind Power “Bogie concept” - Full integration of a Planetary
Gearbox in the Wind Turbine Structure; Conference for Wind Power Drives
CWD 2013, Aachen, Germany, 2013.
[40] Smook, W.; Bogaert, R.: Gear Transmission Unit with Planetary Gear;
US7806799, 2010.
[41] Uusitalo, K.: Fusiondrive from 3 MW to Offshore Class; 4th International
Conference Drivetrain Concepts for Wind Turbines, Bremen, Germany, 2013.
[42] Deicke, M.: Field Experience with the Winergy HybridDrive; 5th International
Conference Drivetrain Concepts for Wind Turbines, Bremen, Germany, 2014.
[43] Eize de Vries: Turbines of the Year - Drivetrains; Windpower Monthly, 2016.
[44] Raeber, R.; Weller, U.; Amato, R.: A New Gearbox Generation for Vertical
Roller Mills; MAAG Gear AG, 2006.
[45] GE Transportation Drivetrain Technologies: CP 1.8 Wind Turbine Gearbox -
Reliability and Efficiency by Design; 2007.
[46] AREVA Wind GmbH: M5000 Technical Data; 2010.
[47] ISO 6336: Calculation of Load Capacity of Spur and Helical Gears - Part 3:
Calculation of Tooth Bending Strength; International Organization for
standardization, Second Edition 2006.
[48] Pinnekamp, B.; Ludas, B.; Schulz, M.: Multi-Megawatt Offshore Wind Turbine
Drive System without Converter; Conference for Wind Power Drives CWD
2013, Aachen, Germany, 2013.
[49] Benito, P.: Planetary Gearbox Having Multiple Sun Pinions; EP2187091, 2010.
[50] Pischel, K.; Meimann, V.: Planetary Gear Wind Generator Comprising a
Planetary Gear and Use of a Planetary Gear; EP2884100, 2015.
[51] Krantz, T. L.; Delgado, I. R.: Experimental Study of Split-Path Transmission
References 107

Load Sharing; Technical Report ARL–TR–1067, NASA Technical


Memorandum 107202, 1996.
[52] Schäfer, W.: Windkraftgetriebe für eine Windkraftanlage höherer Leistung;
EP1619386, 2006.
[53] Mikhail, A.; Hahlbeck, E.: Distributed Powertrain that Increases Electric Power
Generator Density; US6731017, 2004.
[54] Mikhail, A.; Hahlbeck, E.: Distributed Power Train (DGD) with Multiple
Power Paths; US7069802, 2006.
[55] Hulshof, F.: Leistungsverweigtes Getriebe für eine Windkraftanlage;
EP2431632, 2012.
[56] Mikhail, A.: Distributed Generation Drivetrain for High Torque Wind Turbine
Applications; CEC-500-2011-002, 2011.
[57] Mikhail, A.: Low Wind Speed Turbine Development Project Report;
NREL/SR-500-43743, NREL National Renewable Energy Laboratory, 2009.
[58] Winergy AG: Multi Duored Gearbox - The future for Offshore Applications;
Company Catalogue,
[59] Mikhail, A.: University of Minnesota Clipper Liberty Research Turbine;
Presentation, 2010.
[60] Ragheb, A.; Ragheb, M.: Wind Turbine Technologies; Proceedings of the First
International Nuclear and Renewable Energy Conference INREC10, Amman,
Jordan, 2010.
[61] Smirnov, G.: Multiple-Power-Path Nonplanetary Main Gearbox of the Mi-26
Heavy Lift Transport Helicopter; Vertiflite, Vol. 36, No. 3, pp. 20–23, 1990.
[62] Tishchenko, M.: Mil Design Bureau Heavy-Lift Helicopters; Journal of
American Helicopter Society, 1997.
[63] Ligata, H.; Kahraman, A.; Singh A.: An Experimental Study of the Influence of
Manufacturing Errors on the Planetary Gear Stresses and Planet Load Sharing;
ASME Transactions, Journal of Mechanical Design, Vol. 130, pp. 1–9, 2008.
[64] ISO 6336: Calculation of Load Capacity of Spur and Helical Gears - Part 1:
Basic Principles, Introduction and General Influence Factors; International
Organization for standardization, Second Edition 2006.
[65] Montestruc, A.: A Numerical Approach to Calculation of Load Sharing in
Planetary Gear Drives; ASME Transactions, Journal of Mechanical Design,
Vol. 132, pp. 1–4, 2010.
[66] Hicks, R.: Experience with Compact Orbital Gears in Service; Proceedings of
the Institution of Mechanical Engineers, Vol. 184, Issue 15, pp. 85–94, 1969.
[67] Neubauer, B.: Lastverteilung und Anregungsverhalten in
Planetengetriebesystemen; Technische Universität München, Dissertation,
2016.
108 References

[68] Bärtsch, R.; Dinner, H.; Röthlin, C.: Flexpin Design for Maximised Power
Density;
[69] Fox, G. P.; Jallat, E.: Use of the Integrated Flexpin Bearring for Improving the
Performance of Epicyclic Gear Systems; Proceedings of DETC'03 ASME 2003
Design Engineering Technical Conferences and Computers and Information in
Engineering Conference, Chicago, Illinois, USA, 2003.
[70] Giger, U.: Planetenkoppelgetriebe in Windenergieanlagen mit flexibler
Planetenlagerung; Dresdner Maschinenelemente Kolloquium, Dresden, 2003.
[71] Wikov Wind: W2000 spg Specification.
[72] DNVGL-CG-0036: Calculation of Gear Rating for Marine Transmissions;
DNV GL, 2015.
[73] Dudley, W.: Gear Handbook - The design, Manufacture, and Application of
Gears; McGraw-Hill, 1962.
[74] Arafa, H.: Mechanical Design Pitfalls; Proceedings of the Institution of
Mechanical Engineers, Part C: Journal of Mechanical Engineering Science,
Vol. 220, pp. 887–899, 2006.
[75] DIN ISO 21771: Zahnräder - Zylinderräder und Zylinderradpaare mit
Evolventenverzahnung - Begriffe und Geometrie (ISO 21771:2007); DIN
Deutsches Institut für Normung e. V., 2014.
[76] Sfar, M.: Bestimmung von Verzahnungskorrekturen und Lagerkräften in
Planetengetrieben für Lastkollektive; Ruhr-Universität Bochum, Dissertation,
2011.
[77] Lahtinen, P.; Ryymin, A.: Arrangement in a Planetary Gearing and a Planetary
Gear; US8241172, 2012.
[78] Oyague, F.: Gearbox Reliability Collaborative Experimental Data Overview
and Analysis; NREL/PR-500-48176, NREL National Renewable Energy
Laboratory, 2010.
[79] Willis, R.: New Equations and Charts Pick Off - Lightest-Weight Gears;
Product Engineering, pp. 64–75, 1963.
[80] Forschungsvereinigung Antriebstechnik: Erweiterung Getriebeauslegungs-
programm III - Abschlussbericht; Forschungsvorhaben Nr. 421 III, Heft 1070,
2013.
[81] Serowy, S.; Andrei, C., et al.: 6 MW Wind Turbine with Multiple High-Speed
Generators; Conference for Wind Power Drives CWD 2015, Aachen, Germany,
2015.
[82] A Yaskawa Company (The Switch): High-Speed Permanent Magnet
Generators; Company Catalogue, 2018.
[83] A Yaskawa Company (The Switch): Medium-Speed Permanent Magnet
Generators; Company Catalogue, 2018.
References 109

[84] Blank, H.: Konzeptvergleich von Windkrafthauptgetrieben der


Multimegawattklasse unter dem Aspekt der Bauraum- und
Gewichtsoptimierung; RWTH Aachen, Diplomarbeit, 2010.
[85] Manegold, M.: Konzeptvergleich zwischen Stirnradpaaren und Stirnrad-
Beveloidradpaaren zum Einsatz in Stufenplaneten Windkraftgetrieben; RWTH
Aachen, Masterarbeit, 2015.
[86] Olver, A. V.: The Effect of Configuration in the Design of Geared Transmission
Systems; ASME Transactions, Journal of Mechanical Design, Vol. 131, Issue
7, pp. 1–5, 2009.
[87] Schultz, C. D.: The Effect of Gearbox Architecture on Wind Turbine Enclosure
Size; AGMA Technical Paper, 2009.
[88] Berroth, J.: Design and Modelling of the structural components of an offshore
wind energy plant for multibody simulation; RWTH Aachen, Diplomarbeit,
2011.
[89] GL 2010: Guideline for the Certification of Wind Turbines; Germanischer
Lloyd Industrial Services GmbH, 2010.
[90] IEC 61400: Wind Turbines - Part 4: Design Requirements for Wind Turbine
Gearboxes; International Organization for standardization, Ed. 1 2012.
[91] Houser, D.; Harianto, J.; Talbot, D.: Gear Mesh Misalignment; Gear Solutions
Magazine, pp. 34–43, 2006.
[92] Palermo, A.; Mundo, D., et al.: Multibody Element for Spur and Helical Gear
Meshing Based on Detailed Three-Dimensional Contact Calculations;
Mechanism and Machine Theory, Vol. 62, pp. 13–30, 2013.
[93] Jablonski, P.; Czajka, P., et al.: Case Study of Assembly Errors Influence on
Stress Distribution in Spur Gear Train; Proceedings of the 7th International
Conference on Mechanics and Materials in Design, Portugal, 2017.
[94] Li, S.: Effects of Machining Errors, Assembly Errors and Tooth Modifications
on Loading Capacity, Load-Sharing Ratio and Transmission Error of a pair of
Spur Gears; Mechanism and Machine Theory, Vol. 42, pp. 698–726, 2007.
[95] Johnson, K.: Contact Mechanics; Third EditionCambridge University Press,
2003.
[96] Li, S.: Effects of Misalignment Error, Tooth Modifications and Transmitted
Torque on Tooth Engagements of a Pair of Spur Gears; Mechanism and
Machine Theory, Vol. 83, pp. 125–136, 2015.
[97] Quenneville, R. N.: Gear Alignment Means; US3434365, 1969.
[98] Kleckner, R, J.; Dyba, G.: High Speed Spherical Roller Bearing Analysis and
Comparison with Experimental Performance; Proceedings of a symposium
sponsored by Propulsion Laboratory, Cleveland, Ohio, USA, 1987.
[99] Cao, M.: A Refined Double-row Spherical Roller Bearing Model and its
110 References

Application in Performance Assessment of Moving Race Shaft Misalignments;


Journal of Vibration and Control, Vol. 13, pp. 1145-1168, 2007.
[100] Ionescu, L.; Pontius, T.: Main SHaft Support for Wind Turbine with a Fixed
and Floating Bearing Configuration; 2009.
[101] SKF Catalogue: Rolling bearings; 2018.
[102] Harris, T.; Kotzalas, M.: Rolling Bearing Analysis - Essential Concepts of
Bearing Technology; Fifth EditionCRC Press, 2006.
[103] Zhao, J.: Entwicklung eines vereinfachten FE-Modells zur Strukturanalyse von
Planetengetrieben in Windenergieanlagen; RWTH Aachen, Diplomarbeit,
2014.
[104] Yeh, H.; Tsai, S.; Yu, Y.: Numerical and Experimental Analysis of Flexible Pin
Mechanisms for Load Balancing in Planetary Gear Drives; Proceedings of the
ASME 2013 International Design Engineering Technical Conferences and
Computers and Information in Engineering Conference IDETC/CIE 2013,
Portland, Oregon, USA, 2013.
[105] Mignot, L.; Bonnard, L.; Abousleiman, V.: Analysis of Load Distribution in
Planet-Gear Bearings; AGMA Technical Paper, 2010.
[106] Harris, T.; Kotzalas, M.: Rolling Bearing Analysis - Advanced Concepts of
Bearing Technology; Fifth Edition, 2006.
[107] Forschungsvereinigung Antriebstechnik: Einfluß von Verzahnungsgeometrie
und Betriebsbedingungen auf die Graufleckentragfähigkeit von Zahnrad-
getrieben; Forschungsvorhaben Nr. 259/I, Heft 583, 1999.
[108] Forschungsvereinigung Antriebstechnik: Einfluß der Graufleckigkeit auf die
Grübchentragfähigkeit einsatzgehärteter Zahnräder; Forschungsvorhaben Nr.
459, Heft 844, 2008.
[109] ISO/TR 15144: Calculation of Micropitting Load Capacity of Cylindrical Spur
and Helical Gears; International Organization for standardization, 2014.
[110] Errichello, R. L.: Morphology of Micropitting; Gear Technology Magazine, pp.
74–81, 2012.
[111] Marjanovic, N.; Isailovic, B.; et. al.: A Practical Approach to the Optimization
of Gear Trains with Spur Gears; Mechanism and Machine Theory, Vol. 53, pp.
1–16, 2012.
[112] Forschungsvereinigung Antriebstechnik: Wissensbasiertes System zur
Auslegung von anregungsarmen beanspruchungsgerechten Verzahnungen;
Forschungsvorhaben Nr. 201/IIb, Heft 512, 1997.
[113] DIN 3992: Profilverschiebung bei Stirnrädern mit Außenverzahnung; DIN
Deutsches Institut für Normung e. V., 1964.
[114] DIN 867: Bezugsprofile für Evolventenverzahnungen an Stirnrädern
(Zylinderrädern) für den allgemeinen Maschinenbau und den
References 111

Schwermaschinenbau; DIN Deutsches Institut für Normung e. V., 1986.


[115] Kapelevich, A.; Kleiss, R.: Direct Gear Design for Spur and Helical Involute
Gears; Gear Technology Magazine, pp. 29–35, 2002.
[116] Savage, M.; Coy, J.; Townsend, D.: The optimal Design of Involute Gear Teeth
with Unequal Addenda; TR 82-C-7, NASA Technical Memorandum, 1982.
[117] ISO 6336: Calculation of Load Capacity of Spur and Helical Gears - Part 2:
Calculation of Surface Durability (Pitting); International Organization for
standardization, Second Edition 2006.
[118] American Gear Manufacturers Association AGMA: A Rational Procedure for
the Preliminary Design of Minimum Volume Gears; AGMA 901-A92, 1992.
[119] European Environment Agency: Europe's onshore and offshore wind energy
potential; EEA Technical Report No. 6/2009, 2009.
[120] Brughuis, F.: Design of the mechanical-hydraulic drivetrain of the Hydrautrans
HT10 wind turbine; 4th International Conference Drivetrain Concepts for Wind
Turbines, Bremen, Germany, 2014.
112
First stage Second Stage Third Stage Ratio Masses (ton)
′ ′ ′ ′ ′′ ′′ ′′ ′′ ′′′ ′′′ ′′′ ′′′
𝑁3′ 𝑁𝑚 𝑚𝑛′ 𝑧1′ 𝑧2 𝑧3 𝑢 𝑢12 𝑚𝑛 𝑧1 𝑧2 𝑢 𝑚𝑛 𝑧1 𝑧2 𝑢 𝑖𝑡𝑜𝑡𝑎𝑙 𝑀𝑔𝑏𝑥 𝑀𝑔𝑒𝑛 𝑀𝑠𝑦𝑠
1 5 22 21 22 125 5.95 1.05 19 20 41 2.05 8.5 43 183 4.3 51.9 179.5 56.6 236.1
1 6 20 20 21 151 7.55 1.05 12 27 58 2.15 10 20 99 5.0 80.3 154.6 51.0 205.6
1 7 18 20 21 178 8.90 1.05 10.5 27 59 2.19 6.5 35 194 5.5 107.8 139.9 48.3 188.2
Appendix A

1 8 16.5 20 21 202 10.10 1.05 9.5 28 61 2.18 7.5 20 123 6.2 135.3 130.0 46.7 176.7
1 9 16 19 21 212 11.16 1.11 9 28 61 2.18 5.5 32 209 6.5 158.8 122.0 46.2 168.2
1 10 15 19 21 233 12.26 1.11 8 29 64 2.21 6.5 19 132 6.9 188.0 115.2 45.2 160.4
2 3 20 20 23 144 7.20 1.15 18 21 41 1.95 11 32 121 3.8 53.2 191.5 42.6 234.1
2 4 15.5 21 26 205 9.76 1.24 14.5 21 43 2.05 8 36 167 4.6 92.7 164.2 38.4 202.6
2 5 14 20 27 236 11.80 1.35 12 23 49 2.13 8 23 123 5.3 134.4 147.6 36.4 184.0
2 6 12.5 20 27 275 13.75 1.35 10 25 54 2.16 7.5 20 123 6.2 182.7 136.3 34.8 171.1
2 7 12 19 26 293 15.42 1.37 9.5 26 57 2.19 6.5 20 137 6.9 231.6 127.6 33.9 161.5
2 8 10.5 20 29 339 16.95 1.45 8 30 67 2.23 5 31 203 6.5 247.9 118.4 35.3 153.7
2 9 10.5 19 26 343 18.05 1.37 7 35 79 2.26 4.5 33 226 6.8 279.1 112.3 35.6 147.8
2 10 9.5 19 31 386 20.32 1.63 7 30 67 2.23 4 33 226 6.8 310.7 106.8 35.8 142.6
Table A-1: Optimization results showing the minimum system mass 𝑴𝒔𝒚𝒔 for each number of power-split modules 𝑵
and the corresponding geometry per stage for a bull gear model and a simple third stage (concept A)

First stage Second Stage Third Stage Ratio Masses (ton)



𝑁3′ 𝑁𝑚 𝑚𝑛′ 𝑧1′ 𝑧2′ 𝑧3′ 𝑢′ 𝑢12 𝑚𝑛′′ 𝑧1′′ 𝑧2′′ 𝑢′′ 𝑚𝑛′′′ 𝑧1′′′ 𝑧2′′′ 𝑢′′′ 𝑖𝑡𝑜𝑡𝑎𝑙 𝑀𝑔𝑏𝑥 𝑀𝑔𝑒𝑛 𝑀𝑠𝑦𝑠
1 6 20 20 21 151 7.55 1.05 13.5 23 51 2.22 8 37 143 3.9 64.7 148.5 38.9 187.3
1 8 16.5 20 21 202 10.10 1.05 11 23 51 2.22 6 39 184 4.7 105.7 124.3 36.1 160.5
1 10 15 19 21 233 12.26 1.11 9 25 56 2.24 6 23 126 5.5 150.5 110.4 34.6 145.0
2 4 16.5 20 23 193 9.65 1.15 14 22 47 2.14 11.5 21 76 3.6 74.6 160.0 29.3 189.3
Gearbox Optimization Results

2 6 12.5 20 27 277 13.85 1.35 10.5 23 51 2.22 8 20 97 4.9 148.9 132.3 26.4 158.8
2 8 10.5 20 29 340 17.00 1.45 8 29 66 2.28 5 37 209 5.6 218.5 116.2 25.8 142.0
2 10 9 21 29 401 19.10 1.38 6 41 97 2.37 4 37 242 6.5 295.5 105.4 25.3 130.7
Table A-2: Optimization results showing the minimum system mass 𝑴𝒔𝒚𝒔 for each number of power-split modules 𝑵
and the corresponding geometry per stage for a bull gear model and a power-summing third stage
(concept B)
Appendix A
Appendix A

First stage Second Stage Third Stage Ratio Masses (ton)



𝑁3′ 𝑁𝑚 𝑚𝑛′ 𝑧1′ 𝑧2′ 𝑧3′ 𝑢′ 𝑢12 𝑚𝑛′′ 𝑧1′′ 𝑧2′′ 𝑢′′ 𝑚𝑛′′′ 𝑧1′′′ 𝑧2′′′ 𝑢′′′ 𝑖𝑡𝑜𝑡𝑎𝑙 𝑀𝑔𝑏𝑥 𝑀𝑔𝑒𝑛 𝑀𝑠𝑦𝑠
1 3 24.5 19 26 163 8.58 1.37 12.5 43 100 2.33 10 27 112 4.1 82.8 186.4 34.7 221.1
1 4 21 19 27 190 10.00 1.42 10.5 43 99 2.30 10 20 99 5.0 114.0 164.2 34.9 199.1
1 5 18.5 19 29 216 11.37 1.53 10 37 84 2.27 9 20 109 5.5 140.7 149.0 35.7 184.7
1 6 17 19 28 235 12.37 1.47 9.5 34 77 2.26 6.5 25 149 6.0 166.9 137.2 36.3 173.5
1 7 15.5 19 31 258 13.58 1.63 8.5 33 74 2.24 7 20 131 6.6 199.4 129.7 36.3 166.0
1 8 13.5 21 29 296 14.10 1.38 8.5 30 67 2.23 6.5 20 139 7.0 218.8 121.2 37.4 158.6
1 9 14 19 27 284 14.95 1.42 8 29 64 2.21 6.5 19 132 6.9 229.2 114.6 39.0 153.5
1 10 13.5 19 25 296 15.58 1.32 8 27 59 2.19 6 19 132 6.9 236.5 108.7 40.6 149.3
2 3 15 23 35 254 11.04 1.52 11.5 43 100 2.33 8.5 31 142 4.6 117.6 187.1 29.5 216.5
2 4 15.5 19 29 247 13.00 1.53 10 43 100 2.33 6.5 40 213 5.3 161.0 164.2 29.7 193.9
2 5 13.5 19 30 289 15.21 1.58 9 43 100 2.33 7.5 20 123 6.2 217.5 149.3 29.1 178.4
2 6 13 18 29 305 16.94 1.61 8 41 95 2.32 5.5 27 181 6.7 263.2 137.7 29.4 167.1
2 7 12 18 29 333 18.50 1.61 7 43 99 2.30 4.5 37 246 6.6 283.2 127.6 30.9 158.4
2 8 11 18 29 363 20.17 1.61 7 37 85 2.30 4.5 33 226 6.8 317.3 119.9 31.5 151.4
2 9 9.5 19 33 421 22.16 1.74 6.5 37 84 2.27 4 33 226 6.8 344.5 113.8 32.3 146.0
2 10 9 20 31 444 22.20 1.55 6.5 34 77 2.26 4 33 226 6.8 344.3 107.2 34.1 141.4
Table A-3: Optimization results showing the minimum system mass 𝑴𝒔𝒚𝒔 for each number of power-split modules 𝑵
and the corresponding geometry per stage for a ring gear model and a simple third stage (concept A)
113
114

First stage Second Stage Third Stage Ratio Masses (ton)



𝑁3′ 𝑁𝑚 𝑚𝑛′ 𝑧1′ 𝑧2′ 𝑧3′ 𝑢′ 𝑢12 𝑚𝑛′′ 𝑧1′′ 𝑧2′′ 𝑢′′ 𝑚𝑛′′′ 𝑧1′′′ 𝑧2′′′ 𝑢′′′ 𝑖𝑡𝑜𝑡𝑎𝑙 𝑀𝑔𝑏𝑥 𝑀𝑔𝑒𝑛 𝑀𝑠𝑦𝑠
1 2 24.5 26 35 163 6.27 1.35 18.5 30 67 2.23 18 31 48 1.5 21.7 219.0 35.8 254.8
1 4 21 19 27 190 10.00 1.42 10.5 43 99 2.30 11 20 77 3.9 88.6 160.0 27.0 187.1
1 6 17 19 28 235 12.37 1.47 9.5 34 77 2.26 10 18 83 4.6 129.2 133.0 28.2 161.2
1 8 13.5 21 29 296 14.10 1.38 8.5 30 67 2.23 7.5 20 107 5.4 168.4 116.9 29.1 146.0
1 10 13.5 19 25 296 15.58 1.32 8 27 59 2.19 6 20 119 6.0 202.6 106.1 30.1 136.2
2 2 20 22 31 181 8.23 1.41 15 41 95 2.32 17 20 57 2.9 54.3 220.6 23.4 244.0
2 4 12.5 25 37 312 12.48 1.48 10 43 100 2.33 7.5 33 134 4.1 117.9 160.0 23.7 183.7
2 6 10 25 38 395 15.80 1.52 8 42 97 2.31 6 35 177 5.1 184.5 133.2 23.9 157.1
2 8 9 23 36 443 19.26 1.57 7 38 87 2.29 7 20 121 6.1 266.8 117.8 23.5 141.3
2 10 9 20 31 444 22.20 1.55 6.5 34 77 2.26 5.5 20 137 6.9 344.4 107.1 23.5 130.6
Table A-4: Optimization results showing the minimum system mass 𝑴𝒔𝒚𝒔 for each number of power-split modules 𝑵
and the corresponding geometry per stage for a ring gear model and a power-summing third stage
(concept B)
Appendix A

You might also like