Advanced Structure Modeling ....

UNIVERSITY OF STRATHCLYDE
ADVANCED STRUCTURAL
MODELLING AND DESIGN OF WIND
TURBINE ELECTRICAL
GENERATORS
Pablo Jaen Sola
A thesis presented in fulfilment of the requirements

for the Degree of Doctor of Philosophy
Centre for Doctoral Training in Wind Energy Systems
Department of Electronic and Electrical Engineering
University of Strathclyde
Glasgow, G1 1XW
Scotland, UK
May 2017
This thesis is the result of the author’s original research. It has been composed by the
author and has not been previously submitted for examination which has led to the
award of a degree.
The copyright of this thesis belongs to the author under the terms of the United
Kingdom Copyright Acts as qualified by University of Strathclyde Regulation 3.50.
Due acknowledgment must always be made of the use of any material contained in,
or derived from, this thesis.
Signed:
Date:
I
Acknowledgments
I would like to express my most sincere gratitude to my principal supervisor Dr
Alasdair S. McDonald. I will never forget the opportunity he gave me to carry out
this project. Without his moral and technical support, this thesis would not have been
possible. I would also like to thank him for the freedom he gave me, for his excellent
proof-reading and for all the skills he taught me along this way. I am sure they all
will be very useful in my career.
I am grateful to my second supervisor, Dr Erkan Oterkus for his support and

valuable advice on materials engineering and composite modelling.
This thesis is the culmination of four years study at the Wind Energy Systems
Doctoral Training Centre within the University of Strathclyde and I would like to
thank Professor William Leithead who gave me this incredible opportunity. A big
thank you is due to Mr Drew Smith who is the person that keeps everything running
smoothly in our centre and is always there to give us a hand.
A thank you is also due to my fellow CDT colleagues that have kept me company
throughout this process. We created a nice environment where it is actually fun to
discuss work.
Finally, but most importantly, I would like to express my eternal gratitude to my

parents, Mercedes and Leopoldo and to my elder brother Javier for always being
there no matter the distance or the time, and for their unconditional love and
patience. I will never be able to explain with words how important you have been for
me in this journey and in my entire life.
II
“Our hours are minutes when we hope to know, but centuries when we know what it
is possible to learn”
Antonio Machado
III
Abstract
This thesis concentrates on direct drive electrical generators for wind energy
applications. A variety of wind turbine configurations and generator topologies are
reviewed.
Direct drive renewable energy converters introduce a low speed, high torque input
into the electrical machine. Due to this, these generators have to be larger and more
robust than their high speed counterparts. With very large airgap closing forces, a
very stiff structure capable of withstanding the stress is necessary. As a result very
heavy machines, with structural (‘inactive’) material dominating the
electromagnetically ‘active’ material are designed.
In this thesis a stiffness approach is introduced which combines electromagnetic

stiffness and structural stiffness for different modes of deflection. This is used to
minimise mass of the generator by trading stiffness of rotor and stator structures.
Design tools are presented, validated and utilised to model lightweight supporting
structures (‘inactive material’) for high torque radial flux permanent magnet
synchronous generators. Different structural layouts are statically studied, compared
and optimised. Making use of low density materials, such as composites, a simplified
generator structure is designed and contrasted with its optimised steel counterpart.
As a rotating piece machinery forming part of a bigger and more complex machine,
electrical generators are subject to dynamic and external forces coming from the
wind turbine rotor. The optimised steel design is looked at from a dynamic
viewpoint. Discussions and conclusions highlight the potential design solutions that
can be adopted to minimise the mass and therefore the cost of these machines.
IV
V
Table of Contents
Acknowledgments ................................................................................................... II
Abstract ..................................................................................................................IV
List of Figures ........................................................................................................XI
List of Tables ....................................................................................................... XVI
Nomenclature .................................................................................................... XVII
Chapter 1: Introduction ............................................................................................ 1

1.1 Introduction ........................................................................................................ 1
1.2 Research question ............................................................................................... 4
1.3 Aims of the thesis ............................................................................................... 4
1.4 Thesis layout ....................................................................................................... 5
1.5 References .......................................................................................................... 7
Chapter 2: Literature Review ................................................................................... 8

2.1 Wind energy ....................................................................................................... 9
2.2 Modern wind turbines ...................................................................................... 10
2.3 Wind turbine drivetrains ................................................................................... 12
2.4 Electrical machines ........................................................................................... 16
2.5 Wind turbine generators ................................................................................... 17
2.5.1 Electrical machines for direct drive wind turbines ................................... 18
2.5.2 Excitation techniques ................................................................................. 18
2.5.2.1 Electrically excited direct drive generators............................................ 18
2.5.2.2 Permanent magnet excited direct drive generators ................................ 20
2.6 Permanent magnet topologies .......................................................................... 22
2.6.1 Radial flux PM generators ......................................................................... 23
2.6.2 Uncommon radial flux configurations ....................................................... 24
2.6.3 Further variations ...................................................................................... 25
2.6.4 Surface mounted or buried ........................................................................ 25
2.6.3 Axial flux generators.................................................................................. 26
2.6.4 Transverse flux generators ........................................................................ 26
VI
2.6.5 PM configurations comparison ................................................................. 27
2.7 Structural Analysis, Modelling and Design of Direct Drive Generators.......... 28
2.7.1 Dimensions of direct drive generators....................................................... 28
2.7.2 Forces and moments acting on electrical machines .................................. 29
2.7.3 Integrating a direct drive generator within the wind turbine .................... 32
2.7.4 Design external loads for HAWT: International standards....................... 36
2.7.5 Configurations ........................................................................................... 37
2.7.6 Integrated design of direct drive machines................................................ 37
2.7.7 Modelling methods ..................................................................................... 44
2.7.7.1 Analytical approaches for structural modelling ..................................... 44
2.7.7.2 Numerical approaches for structural modelling .................................... 46
2.7.7.3 Structural optimization ........................................................................... 46
2.7.7.4 Modes of deflection ................................................................................. 48
2.8 The Concept of Direct Drive Generator Supporting Structures ....................... 51
2.9 References ........................................................................................................ 53
Chapter 3: Magnetic Stiffness Modelling of Wind Turbine Electrical

Generators ................................................................................................................ 58
3.1 Introduction ...................................................................................................... 58
3.2 Introduction to the Stiffness Concept ............................................................... 60
3.2.1 Mechanical Stiffness .................................................................................. 60
3.2.2 Magnetic Stiffness ...................................................................................... 63
3.2.3 Overall Stiffness ......................................................................................... 65
3.3 Magnetic Airgap Stiffness ................................................................................ 65
3.3.1 Airgap closing force per unit area ............................................................. 66
3.3.2 Validation using finite element code .......................................................... 73
3.4 Case study generator ........................................................................................ 79
3.5 Discussion ........................................................................................................ 97
3.5.1 Magnetic stiffness model ............................................................................ 97
3.5.2 Permanent magnet and wound rotor machines ......................................... 98
3.5.3 Structural stiffness model........................................................................... 98
3.5.4 Modes of deflection .................................................................................... 99
3.6 Conclusion ...................................................................................................... 100
VII
3.7 References ...................................................................................................... 102
Chapter 4: Comparison of Methods for Estimating Generator Structural

Stiffness ................................................................................................................... 103
4.1 Introduction .................................................................................................... 103
4.2 Estimating minimum generator structural stiffness ........................................ 105
4.2.1 Finding Structural Stiffness through the Combination of Sub Structures
Stiffness ............................................................................................................. 105
4.2.2 A Case Study Generator .......................................................................... 107
4.2.3 Analytical characterisation of disc and arm structures........................... 108
4.2.4 Modelling structural stiffness: Finite Element Analysis .......................... 109
4.2.5 Modelling structural stiffness using a hybrid technique: Rotor disc model
.......................................................................................................................... 113
4.2.6 Structural optimisation ............................................................................ 114
4.3 Results ............................................................................................................ 117
4.3.1 Finite element approach .......................................................................... 117
4.3.2 Analytical approach ................................................................................. 121
4.3.3 Hybrid approach ...................................................................................... 123
4.3.4 2D optimisation of simplified structures.................................................. 129
4.3.5 Structural topology optimisation ............................................................. 133
4.4 Discussion ...................................................................................................... 138
4.5 Conclusions .................................................................................................... 140
4.6 References ...................................................................................................... 142
Chapter 5: Lightweight Materials in Generator Structures .............................. 144

5.1 Introduction .................................................................................................... 144
5.2 Composite Materials....................................................................................... 145
5.2.1 What are composite materials and how do we form them? ..................... 145
5.2.2 Carbon fibre and epoxy ........................................................................... 148
5.2.3 Manufacturing processes ......................................................................... 150
5.2.4 Advantages and drawbacks of using composites ..................................... 151
5.2.5 Other factors to be considered in design ................................................. 152
5.2.6 Establishing the design conditions: loading ............................................ 153
VIII
5.2.7 Mechanically analysing a composite structure ....................................... 156
5.2.8 Classical lamination theory ..................................................................... 156
5.3 Composite Structure Modelling ..................................................................... 158
5.4 Investigation ................................................................................................... 163
5.4.1 Conventional Approach ........................................................................... 163
5.4.2. Mosaic Pattern Approach ....................................................................... 166
5.4.3 Conventional Approach vs. Mosaic Pattern Approach ........................... 170
5.5 Results ............................................................................................................ 172
5.6 Discussion ...................................................................................................... 175
5.7 Conclusions .................................................................................................... 180
5.8 References ...................................................................................................... 183
Chapter 6: Dynamics of a Direct Drive Generator ............................................. 185

6.1 Introduction .................................................................................................... 185
6.2 Methodology .................................................................................................. 189
6.2.1 Evaluating structural natural frequencies ............................................... 189
6.2.2 Dynamically designing a direct drive generator supporting structure ... 194
6.2.3 Techniques for elevating structure’s natural frequencies ....................... 197
6.2.3.1 Dimensional alteration of structures made with discs.......................... 197
6.2.3.2 Use of stiffeners .................................................................................... 203
6.2.3.3 Rotor conical structure ......................................................................... 204
6.3 Results ............................................................................................................ 205
6.4 Discussion ...................................................................................................... 214
6.5 Conclusions .................................................................................................... 217
6.6 References ...................................................................................................... 219
Chapter 7: Discussions and Conclusions ............................................................. 221

7.1 Chapter summaries ......................................................................................... 221
7.2 Discussion ...................................................................................................... 225
7.2.1 Evaluating stiffness .................................................................................. 226
7.2.2 Lightweight materials for generator supporting structures .................... 227
7.2.3 Direct drive electrical generator dynamics ............................................. 228
7.3 Revisiting the thesis research question ........................................................... 230
IX
7.4 Contribution to knowledge ............................................................................. 231
7.5 Further work ................................................................................................... 232
7.6 References ...................................................................................................... 234
Appendix ............................................................................................................. 235
X
List of Figures
Figure 1.1 Global mean temperature between 1880 and 2000 [5]........................................................ 2
Figure 2.1 Offshore wind turbine arrangement and transmission to shore .......................................... 10
Figure 2.2 Layout of a conventional fixed speed wind turbine [11] .................................................... 13
Figure 2.3 Layout of a conventional DFIG wind turbine [11]............................................................. 14
Figure 2.4 Layout of fully rated converter wind turbine [11] .............................................................. 14
Figure 2.5 Illustration of a simple electrical machine [4] .................................................................... 16
Figure 2.6 Illustration of an AC synchronous generator [11] .............................................................. 17
Figure 2.7 7 MW Enercon E-126 [14]................................................................................................. 19
Figure 2.8 4 MW GE (former Scanwind) wind turbine [23] ............................................................... 21
Figure 2.9 PMDD generator topologies; (a) Radial flux; (b) Axial flux; (c) Transverse flux [32][33]
.................................................................................................................................................... 22
Figure 2.10 PMDD radial flux machine and its components [34] ....................................................... 23
Figure 2.11 Cutaway of the Zephyros/Harakosan Europe Z72 wind turbine [35] .............................. 24
Figure 2.12 ST10 10 MW offshore wind turbine developed by Sway [36] ........................................ 24
Figure 2.13 Inner and outer rotor generator variants [13] ................................................................... 25
Figure 2.14 A single-sided surface-mounted TFPM machine [47] ..................................................... 27
Figure 2.15 Cylinder model of torque produced by a generator [34] .................................................. 29
Figure 2.16 Shear loading [34] ............................................................................................................ 30
Figure 2.17 Magnetic attraction of the moving and the stationary components of the generator [34] 30
Figure 2.18 Gravitational loading [34] ............................................................................................... 31
Figure 2.19 Thermal expansion of the generator structure [34] .......................................................... 31
Figure 2.20 Generator’s structure under centrifugal forces [34] ......................................................... 32
Figure 2.21 Zephyros/Harakosan Z72 wind turbine configuration [13] .............................................. 33
Figure 2.22 PMDD Drivetrain [31] ..................................................................................................... 34
Figure 2.23 PMDD Generator [31] ..................................................................................................... 34
Figure 2.24 MTorres 1.5MW PMDD wind turbine [13] ..................................................................... 35
Figure 2.25 GE ScanWind 4.1-113 wind turbine [13]......................................................................... 36
Figure 2.26 Interactions between electrical, thermal and mechanical design aspects of the generator
[13] ............................................................................................................................................. 38
Figure 2.27 Early design stages – traditional approach [13] ............................................................... 38
Figure 2.28 Early design stages – integrated approach [13] ................................................................ 39
Figure 2.29 Electrical design perspective [13] .................................................................................... 40
Figure 2.30 Mechanical design perspective [13] ................................................................................. 41
Figure 2.31 Thermal design perspective [13] ...................................................................................... 42
Figure 2.32 Manufacturing design perspective [13] ............................................................................ 43
Figure 2.33 Typical rotor structures [51] ............................................................................................ 44
Figure 2.34 a) Zephyros/Harakosan Europe Z72, b) MTorres 1.5 MW .............................................. 44
XI
Figure 2.35 Radial flux models: a) radial, b) axial and c) tangential deflection [13] .......................... 45
Figure 2.36 Axial flux model [13]....................................................................................................... 45
Figure 2.37 Illustrations of the three tested generator topologies: a) radial flux, b) transverse flux No.
1, c) transverse flux No. 2 [59] ................................................................................................... 47
Figure 2.38 Illustration of the variables that were utilised for the optimisation of the generator
structures; (a) The variables that describe a structure with arms; (b) The variables that describe
arms sub structure; (c) The variables that describe a rotor with discs; (d) The variables that
describe the electromagnetic model [59] .................................................................................... 47
Figure 2.39 A rotor deforming into the airgap towards a stator .......................................................... 49
Figure 2.40 Airgap clearance (in mm) for the Northern Power 1.5MW prototype [61]. The values
shown here are the mean of the airgap clearance at the upwind and downwind ends of the
machine. Clearance is plotted for different angles as seen from the upwind end of the machine
.................................................................................................................................................... 50
Figure 2.41 Total deflection from designed air-gap clearance (in mm) for the Northern Power 1.5MW
prototype [61] ............................................................................................................................. 51
Figure 2.42 Siemens Direct Drive Wind Turbine Cutaway [62] ......................................................... 52
Figure 3.1 Coupling of magnetic and structural models ...................................................................... 60

Figure 3.2 (a) Generator structure (b) Shown as stiffness [4] ............................................................. 62
Figure 3.3 Airgap Closing Force vs. Theta ......................................................................................... 63
Figure 3.4 Magnetic permeance per unit area comparison. Magnetic permeance per unit area vs.
Magnetic permeance per unit area approximation assuming infinite permeability for the back
iron and ignoring slots. ............................................................................................................... 68
Figure 3.5 Magnetic stress vs. Theta for different deflection modes................................................... 70
Figure 3.6 Magnetic airgap stiffness for a pole pair of a PM generator for a direct drive wind turbine,
based on [3]. (a) Airgap closing force on one pole pair vs. Airgap clearance (b) Airgap closing
force on one pole pair vs. Change in airgap clearance (c) Airgap closing force on one pole pair
vs. Magnetic effective airgap ...................................................................................................... 75
Figure 3.7 PM generator full model validation; (a) Detailed view of the triangular mesh; (b) Density
plot showing the behaviour of the magnetic flux at a particular moment in time ....................... 78
Figure 3.8 Comparison between analytical model and FEMM model (Mode 1 deflection) ............... 79
Figure 3.9 Measured structural deflection of rotor .............................................................................. 80
Figure 3.10 Rotor structural deflection (m) vs. Theta (degrees) ......................................................... 81
Figure 3.11 Rotor structure deflection vs. Applied Force; (a) Mode 0; (b) Mode 1; (c) Mode 2;
(d) Mode 3; (e) Mode 4. ............................................................................................................. 82
Figure 3.12 Bearing model showed as stiffness .................................................................................. 83
Figure 3.13 Bearing structure split into top and bottom parts; (a) top part; (b) bottom part ............... 84
Figure 3.14 Rotor structural stiffness for deflection modes ranging from 0 to 4 vs. Theta ................. 85
Figure 3.15 Stator structural stiffness vs. Theta .................................................................................. 86
Figure 3.16 Generator structural stiffness vs. Theta ............................................................................ 87
XII
Figure 3.17 Stiffness on beta degree section ....................................................................................... 87
Figure 3.18 Rotor structural stiffness; (a) Mode 0; (b) Mode 1; (c) Mode 2; (d) Mode 3; (e) Mode 4 90
Figure 3.19 Magnetic stiffness + Structural stiffness vs. Theta for the worst case scenario
(𝛿̅ = 0.004 m; 𝛿∆ = 0.001 m) ....................................................................................................... 91
Figure 3.20 Compliant structure stiffness vs. Theta; (a) Rotor; (b) Stator .......................................... 92
Figure 3.21 Compliant structure stiffness vs. Theta; (a) Generator structural stiffness; (b) Magnetic +
Structural stiffness ...................................................................................................................... 93
Figure 3.22 Overall stiffnesses comparison; (a) Stiff structure; (b) Compliant structure .................... 94
Figure 3.23 Wound rotor machine airgap closing force vs. Theta ...................................................... 95
Figure 3.24 Wound rotor machine magnetic stiffness vs. Theta ......................................................... 96
Figure 3.25 Wound rotor overall stiffness vs. Theta ........................................................................... 96
Figure 4.1 Rotor structure split into disc and cylinder models [9] .................................................... 105
Figure 4.2 Rotor structures with thickness dimensions as altered in this analysis (a) Disc structure (b)
Arm structure [12] .................................................................................................................... 107
Figure 4.3 Rotor model; Disc structure showing loading conditions and constraints (left side); Arm
structure (right side) .................................................................................................................. 110
Figure 4.4 Stator model; Disc structure showing loading conditions and constraints (left side); Arm
structure (right side) .................................................................................................................. 111
Figure 4.5 View of the different positions for the cone sub structure ............................................... 112
Figure 4.6 Rotor conical structure as altered in the study ................................................................. 112
Figure 4.7 Equation vs. FE disc stiffness [11] ................................................................................... 114
Figure 4.8 Rotor structure shape optimisation; (a) Rotor structure highlighting the elements to be
eliminated; (b) Cutouts of the optimised rotor structure (dimensions shown in mm)............... 115
Figure 4.9 Flowchart of the structural topology optimisation process .............................................. 116
Figure 4.10 2D optimisation for 3 MW rotor and stator disc and arm structures with structural
stiffness criterion; (1) Rotor disc structure; (2) Stator disc structure; (3) Arm rotor structure; (4)
Arm stator structure .................................................................................................................. 118
Figure 4.11 Structural Stiffness vs. Deflection Modes (Disc Rotor Structure) ................................. 120
Figure 4.12 Structural Stiffness vs. Deflection Modes (Disc Stator Structure) ................................. 120
Figure 4.13 Comparison of stiffness calculated from FE and Benham model [11]........................... 121
Figure 4.14 Comparison of stiffness estimated with analytical model and FE [6] ............................ 123
Figure 4.15 Equation vs. FE cylinder stiffness [11] .......................................................................... 124
Figure 4.16 Stator cylinder Equation vs. FE [6] ................................................................................ 125
Figure 4.17 Armed rotor cylinder Equation vs. FE [6]...................................................................... 126
Figure 4.18 Armed stator cylinder Equation vs. FE [6] .................................................................... 127
Figure 4.19 Equation Stiffness vs. FE Stiffness for complete structures; (a) Disc rotor structure; (b)
Arm rotor structure; (c) Disc stator structure; (d) Arm stator structure .................................... 129
Figure 4.20 2D optimization for 3MW rotor and stator disc structures with structural stiffness
criterion [6] ............................................................................................................................... 131
XIII
Figure 4.21 Mass optimisation result for disc structures [6] ............................................................. 131
Figure 4.22 2D optimization for 3MW rotor and stator armed structures with structural stiffness
criterion [6] ............................................................................................................................... 132
Figure 4.23 Mass optimisation result for armed structures [6] .......................................................... 133
Figure 4.24 Large Scale Rotor Structure Shape Optimisation........................................................... 135
Figure 4.25 Design Explorer Optimisation (Large Rotor Structure) ................................................. 136
Figure 4.26 Large Scale Stator Structure Shape Optimisation .......................................................... 137
Figure 4.27 Design Explorer Optimisation (Large Stator Structure) ................................................. 137
Figure 5.1 Advanced composite structure design stages [4] ............................................................. 146
Figure 5.2 Laminate elastic constants for high modulus carbon/epoxy [9] ....................................... 149
Figure 5.3 Stress-strain curves for Ductile (left) and Brittle (right) materials ................................... 152
Figure 5.4 Composite structure delamination [17] ............................................................................ 153
Figure 5.5 Composite loading: (a) Tensile; (b) Compressive; (c) Shear and (d) Flexural [6] ........... 154
Figure 5.6 Shell281 Geometry [21] ................................................................................................... 160
Figure 5.7 Loads acting on the generator rotor structure ................................................................... 161
Figure 5.8 Stator structure according to the conventional method .................................................... 163
Figure 5.9 CA composite structure design process used in this chapter ............................................ 165
Figure 5.10 Example of Composite Fibre Orientations & Stacking Sequences for the Generators
Components; (a) Following a Cartesian Coordinate Frame; (b) Following a Cylindrical
Coordinate Frame; (c) For the Cylinder Sub Structure Following a Cylindrical Coordinate
System ...................................................................................................................................... 166
Figure 5.11 View of the areas forming the mosaic pattern composite disc sub structure.................. 167
Figure 5.12 Detailed view of the composite rotor disc structure based on Morozov’s model
highlighting variable fibre orientation (small green bits) and reinforcement in purple [25] ..... 169
Figure 5.13 Contour plots highlighting stress in the radial direction (a) CA composite rotor structure
(b) Mosaic pattern composite rotor structure ............................................................................ 171
Figure 5.14 Small Scale Model (Mass vs. Type of Structure) ........................................................... 173
Figure 5.15 Large Scale Model (Mass vs. Type of Structure) ........................................................... 174
Figure 5.16 Contour plot displaying maximum deflection in the hoop direction of the mosaic pattern
approach composite stator structure ......................................................................................... 175
Figure 5.17 Potential disc sub-structure designs (a) Conventional model (b) mosaic pattern approach
disc (c) 6-unit disc (d) 8-unit disc [20] ..................................................................................... 180
Figure 6.1 Campbell diagram for a 3 MW machine [3] .................................................................... 186

Figure 6.2 Wire frame view of the rotor structure with transverse plate stiffeners [16] .................... 188
Figure 6.3 Spring-mass arrangement [19] ......................................................................................... 190
XIV
Figure 6.4 (a) First order (linear tetrahedral solid element with no nodes at the midpoints); (b) Second
order for a higher quality mesh (parabolic tetrahedral solid element with nodes at the midpoints)
[20] ........................................................................................................................................... 193
Figure 6.5 Rotor speed power spectrum of a VS conventional geared wind turbine [21] ................. 195
Figure 6.6 Campbell diagram of the system ...................................................................................... 196
Figure 6.7 Rotor structure as changed in the analysis ....................................................................... 197
Figure 6.8 Simplified Rotor Structure ............................................................................................... 198
Figure 6.9 Simplified Rotor Mode Shapes ........................................................................................ 200
Figure 6.10 First mode shape in Y direction ..................................................................................... 201
Figure 6.11 Second mode shape in Y direction ................................................................................. 201
Figure 6.12 Third mode shape in Y direction .................................................................................... 202
Figure 6.13 Rotor structures with axial stiffeners ............................................................................. 204
Figure 6.14 2D Optimisation for 3 MW rotor and stator disc structures with 1 st mode natural
frequencies criterion ................................................................................................................. 206
Figure 6.15 First 5 mode shapes of a typical rotor disc structure (1 st Mode: top left; 2nd Mode: Middle
top; 3rd Mode: top right; 4th Mode: bottom left; 5th Mode: bottom right) ................................. 208
Figure 6.16 Quantification of the effect produced by the implementation of stiffeners in a rotor disc
structure .................................................................................................................................... 209
Figure 6.17 Frequency variation as dimensions are altered; (a) With cylinder thickness maintained at
0.02 m; (b) with cone thickness kept at 0.04 m ........................................................................ 213
Figure 6.18 Deflection vs. Modes of Deflection (Cone Structure) .................................................... 214
Figure A.1 Frequency variation as dimensions are altered with rotor cylinder sub structure thickness
maintained at 0.02 m; (a) 2nd mode; (b) 3rd mode; (c) 4th mode; (d) 5th mode ..................... 236
Figure A.2 Frequency variation as dimensions are altered with rotor disc sub structure thickness
Figure A.3 Frequency variation as dimensions are altered with stator cylinder sub structure thickness
Figure A.4 Frequency variation as dimensions are altered with stator disc sub structure thickness
Figure A.5 Frequency variation as dimensions are altered with cylinder thickness maintained at 0.02
m; (a) 2nd mode; (b) 3rd mode; (c) 4th mode; (d) 5th mode......................................................... 244
Figure A.6 Frequency variation as dimensions are altered with cone thickness kept at 0.04 m; (a) 2 nd
mode; (b) 3rd mode; (c) 4th mode; (d) 5th mode ......................................................................... 246
XV
List of Tables
Table 2.1 Permanent magnet synchronous generator system advantages and disadvantages .............. 15
Table 3.1 Case study generator data .................................................................................................... 64

Table 3.2 Minimum structural stiffness per case and mode ................................................................ 90
Table 3.3 Thicknesses for stiff and compliant structures .................................................................... 91
Table 4.1 Disc and arm generator structures data .............................................................................. 128
Table 4.2 Cone structure optimisation results ................................................................................... 134
Table 5.1 Material properties for a filament wound disc [20] ........................................................... 159
Table 5.2 CA Composite Structure Stacking Sequence ..................................................................... 164
Table 5.3 Effect Caused by the Reduction of Critical Areas Thickness ............................................ 172
Table 6.1 Simplified rotor model features ......................................................................................... 199

Table 6.2 Analytical vs. FE comparison ............................................................................................ 203
Table 6.3 Cone Structure Optimisation Results ................................................................................. 211
XVI
Nomenclature
Roman letters
A Area, m2
Aarm Cross-section area of the arm, m2
Aij Extensional and shear stress stiffnesses, N/m
B Magnetic flux density, Wb/m2
Bg Peak flux density in the airgap, T
Bij Extension-bending coupling stiffnesses, N/m
Br Remanent flux density in the airgap, T
b Beam width in X direction, m
bp Magnets width, m
Ci Balancing constant
Dij Bending and torsional stiffnesses, N/m
Ds Distance between holes centres (stator further optimisation), m
ds Optimisation holes diameter (stator further optimisation), m
E Young’s modulus, Pa
Ek Kinetic energy, J
F Force caused by the normal component of the Maxwell stress, N
Fc Airgap closing force, N
F Magnetomotive force, At
F̂ Peak magnetomotive force, At
̂
F 1 Peak magnetomotive force calculated with analytical model, At
̂
F 3 Peak magnetomotive force calculated with analytical model including
third harmonic, At
̂ PM
F Peak magnetomotive force for a permanent magnet machine, At
XVII
f Structure’s natural frequency, Hz
G Shear modulus, Pa
𝑔 Acceleration due to gravity, 9.81 m/s2
g magnetically effective airgap width, m
H Intensity of magnetic field, A/m
𝐻C Composite fracture toughness, Pa
h Beam’s height in Y direction, m
hm Magnet height, m
hry Rotor yoke height, m
hsy Stator yoke height, m
i Finite number of bearing radial stiffness
I Current, A
I Second moment of area, m4
𝑘A Stiffness of component A, N/m
𝑘a Arm stiffness, N/m
𝑘ac Armed rotor cylinder sub structure stiffness, N/m
𝑘ac,s Armed stator cylinder sub structure stiffness, N/m
𝑘a,eq Arms equivalent stiffness, N/m
𝑘B Stiffness of component B, N/m

𝑘b Bearing stiffness, N/m
𝑘cr Carter factor
𝑘c Cylinder stiffness, N/m
𝑘c,a Rotor cylinder stiffness for armed structures, N/m
𝑘c,s Stator cylinder stiffness for disc structures, N/m
𝑘d Disc stiffness, N/m

𝑘eq Equivalent stiffness, N/m
𝑘eq,d Equivalent rotor stiffness, N/m
𝑘eq,r Equivalent rotor stiffness, N/m
XVIII
𝑘M Magnetic stiffness, N/m
𝑘r Bearing radial stiffness, N/m
𝑘s Generator structure stiffness, N/m
𝑘s,r Rotor stiffness, N/m
𝑘s,s Stator stiffness, N/m

L Half notch length, m
L Beam length in the Z direction, m
l Axial length of machine, m
larm Arm length, m
las,c Axial length of armed stator cylinder sub structure, m
lc,a Axial length of armed rotor cylinder sub structure m
lc,s Axial length of disc stator cylinder sub structure, m
lr Optimisation hole length, m
M Moments, Nm
M Attachment mass, kg
m Beam mass per unit length, kg/m
ms,r mass of rotor made with a disc, kg
ms,s mass of stator made with discs, kg

ms,ar mass of rotor made with arms, kg
ms,as mass of stator made with arms, kg

N Loads, N
N Total number of bearing radial stiffnesses
n Number of phases
n Number of peaks
𝑛𝑎𝑟𝑚𝑠 Number of support arms
P Power rating, W
P Frequency of the wind turbine rotational speed, Hz

𝑃̅ Mean value of airgap permeance per unit area, H/m2
𝑃∆ Amplitude of variation of the airgap permeance per unit area, H/m2
p Number of pole pairs
XIX
q Normal component of Maxwell stress, Pa
R Radius of machine, m
R Resistance, Ω
Ras,c Armed stator cylinder radius, m
Rc,a Armed rotor cylinder radius, m
Rc,s Disc stator cylinder sub structure radius, m
Rl Distance between centres (rotor further optimisation), m
Rm Radius of the large circumference (rotor further optimisation), m
Rs Radius of the small circumference (rotor further optimisation), m
r Radius of the generator’s shaft, m
T Torque, Nm
T Period of oscillation, sec
Tr Temperature of the rotor structure, degrees
Ts Temperature of the stator structure, degrees
tc Cylinder casing thickness, m
tcon Cone sub structure thickness, m
tcyl Rotor conical cylinder sub structure thickness, m
td Disc sub structure thickness, m
ta,s Stator arm’s thickness, m
tarm Rotor arm’s thickness, m
tc,a Armed rotor cylinder sub structure thickness, m
tc,as Armed stator cylinder sub structure thickness, m
ts,c Stator cylinder sub structure thickness, m
ts,d Stator disc sub structure thickness, m

V Potential energy, J
𝑤 Arm width, m
𝑦̇ Mass linear velocity, m/s
ymax Spring vertical deformation, m
XX
Greek letters
𝛼PM Variable used to simplify equation 3.22a
𝛼WR Variable used to simplify equation 3.21a
γ Angle between bearing radial stiffnesses, degrees
𝛾 Deflection mode
δ Deflection, m
δ Distance between composite fibres, m
δr Rotor deflection, m
δs Stator deflection, m
δStructural Structural deflection, m
𝛿̅ Mean deflection, m
𝛿∆ Deflection change respect to the mean, m
ε Mechanical strain
𝜀arm Arm strain
𝜀r Radial strain
κ Curvatures
θ Angular displacement, degrees
θ Cross-section area of a single fibre, m2
ν Poisson’s ratio
μ0 Permeability of free space, 4̟×10-7 H/m
μr Relative permeability
ρ Density, kg/m3
ρc Composite material density, kg/m3

𝜌PM Permanent magnets’ density, kg/m3
σ Structural stress, Pa
σ Airgap shear stress, Pa
XXI
σarm Arm structural stress, Pa
𝜎PM Stress during operation in permanent magnet machine, Pa
𝜎N∞ Residual stress from Mar-Lin formula, Pa
𝜎r Radial stress, Pa
𝜎θ Angular stress, Pa
τp Pole pitch, m
φ Phase angle, degrees
Φ Magnetic flux, Wb
ψ Cone angle, degrees
Abbreviations
AC Alternating Current
APDL ANSYS Parametric Design Language
BAT Buoyant Airborne Wind Turbine
BEM Blade Element Momentum Theory
CA Conventional Approach Disc Modelling
CAD Computer Aided Design
CDT Centre for Doctoral Training
CMC Ceramic Matrix Composites
DC Direct Current
DD Directly Driven
DFIG Doubly Fed Induction Generator
DS Danish Standard
EEDD Electrically Excited Directly Driven
EMF Electromotive force
EU European Union
XXII
FE Finite Element
FEA Finite Element Analysis
FEMM Finite Element Method Magnetics
GL Germanischer Lloyd
IEC International Electrotechnical Standard
ILSS Interlaminar Shear Strength
IGBT Insulated Gate Bipolar Transistors
HAWT Horizontal Axis Wind Turbine
HVAC High Voltage Alternating Current
HVDC High Voltage Direct Current
MMC Metal Matrix Composites
MMF Magnetomotive Force
PM Permanent Magnet
PMC Polymer Matrix Composites
PMDD Permanent Magnet Directly Driven
RF Radial Flux
RTM Resin Transfer Moulding
rpm revolutions per minute
UK United Kingdom
USA United States of America
XXIII
Introduction
Chapter 1
Introduction
1.1 Introduction
Global energy consumption has experienced an important increase since the last
century especially in emerging countries, such as China and India. The same trends
are shown for the world’s population and economic growth [1]. However, the fact
that these three factors are highly correlated makes difficult to tackle the problem of
climate change. In addition, the recent financial crisis has delayed the development
and growth of low carbon and no-carbon technologies [2]. With almost 80% of the
electricity demand worldwide coming from the combustion of conventional fossil
fuels (coal, oil and gas) [3], the level of greenhouse gases, such as carbon dioxide
1
Introduction
(CO2), methane (CH4), sulphur dioxide (SO2) and nitrous oxide (N2O) [4], released
to the atmosphere is causing significant changes in the Earth’s mean temperature.
Figure 1.1 Global mean temperature between 1880 and 2000 [5]
Figure 1.1 clearly displays an increase of the global temperature with a sharp slope
between 1980 and 2000 due to the growth of greenhouse gas emissions from human
activities.
Taking into consideration this clarifying fact, governments have turned their
attention on renewable energies. Wind energy is one of the most developed and
mature clean technology and has an important role to play in the fight against global
warming. Wind resources tend to be greater and steadier offshore and therefore large
scale renewable energy projects will be developed further away from shore. For
instance, in August 2016, the UK government gave the go-ahead for the world’s
biggest offshore wind farm off the Yorkshire coast with 300 wind turbines of 7 MW
rated power capacity each that will cover an area of more than 480 km2 in the North
Sea [6].
In this context, where the offshore wind sector is becoming a key player, a huge
effort is being made worldwide with the main aim of quickly reducing its high
levelised cost and making it capable of competing with conventional electricity
production technologies.
2
Introduction
Offshore wind turbines are placed in harsh environments where the wind speeds tend
to be higher, the air has high humidity and salt content, foundations and substations
are subject to wave and tidal current loading and access can be limited by wave
height. Some of the manufacturers’ considerations for wind turbine design are low
maintenance requirements, easy access to important components, high capacity
factors and assembly with the lightest and cheapest possible crane. Lifting large
pieces of equipment offshore, costs between £10k and £100k a day [7]. Thus,
lightweight designs are highly desirable.
With large heavy rotating machinery working under tough conditions, minimising
the number of moving parts could eventually help diminish the maintenance and
mass of the machines. The problems introduced by wind turbine gearboxes could be
observed at Scroby Sands, one of the UK’s first offshore wind farms. In 2005, 27
intermediate speed and 12 high speed gearbox bearings of the Vestas V-80 wind
turbines had to be replaced [8]. In Horns Rev offshore wind farm Vestas again had to
remove and replace 80 of V90 gearboxes [9].
Bearing this in mind, the use of brushless direct drive generators, where the electrical
machine is directly connected to the wind turbine rotor and the gearbox and the
electrical slip rings or brushes are eliminated, can be a potential solution. Without
gearbox and brushes maintenance downtimes are significantly reduced. Nonetheless,
direct drive generators are quite heavy and robust machines especially designed to
withstand large torques and other typical loads present during operation. Therefore, it
is also necessary to find a method to minimise its mass. The traditional way is to
compare the torque per unit mass between the competing machines. However, these
comparisons are usually based only on the machine active mass, which consists of
the copper in the windings, permanent magnets and the rotor and stator back iron.
According to Hartkopf et al. two thirds of the mass of a radial flux electrical machine
corresponds to the inactive mass, also known as supporting structure, so this must
also be included for an accurate comparison [10]. Significant mass savings leading to
machine’s substantial cost decrease can be achieved by considering the supporting
structure mass during the design stage. In order to carry out this task, it is necessary
3
Introduction
to account with versatile tools capable of estimating the minimum generator

structural requirements in a quick and precise manner.
1.2 Research question
Can electromagnetic and structural stiffness models be used effectively to minimise

the mass of electrical generators for direct drive wind turbines? If so, are there
limitations to such an approach?
1.3 Aims of the thesis
The main aims of this thesis are to explore and develop the potential options
available for lightweight design of the supporting structure of a direct drive wind
turbine generator using stiffness as a framing device. Objectives are as follows:
 To produce design tools which can be used during the design process to
accurately estimate the necessary structural stiffness of the machine
 To compare different structural geometries in order to find the lightest
configuration capable of dealing with major loads acting on the generator
structure
 To optimise the generator steel structure (rotor and stator) with a view to
minimise its mass
 To analyse the optimised steel structure from a dynamic viewpoint and
generate design tools which can be utilised when looking at the dynamics of
the electrical machine
 To explore the distinct structural configurations available so as to identify the
one presenting the best characteristics dynamically speaking
 To find an optimum structural generator design using low density materials
and compare it with the optimised steel structure
4
Introduction
1.4 Thesis layout
The next chapter will introduce the reader to renewables and renewable energy
converters concentrating all the attention on the wind energy sector and the different
existing types of wind turbines and electrical generators. This section also contains a
detailed review of the latest progress in the technology.
Chapter 3 includes the theory behind the design of a radial flux direct drive PM
electrical machine supporting structure. A parametric model that couples the
electromagnetic and mechanical design has been derived in this chapter, taking disc
structures that are utilised to model radial deflection with the main aim of accurately
estimating the minimum structural stiffness of the components forming the radial
flux generator. The approach has been validated using finite element modelling
techniques.
Three distinct techniques for modelling the required machine structural stiffness (FE,
analytical and hybrid) are explained in Chapter 4, including a hybrid method
produced by the authors which combines the results obtained from dimensional
homogeneity studies and the data retrieved from FE analyses.
A comparison between two different types of structural configurations, one made

with discs and one made with arms, is presented with a view of finding out the
lightest layout. The most suitable structure is studied under different deflection
modes and further optimised aiming the minimisation of its mass. Moreover, a
conical structural geometry is proposed for study and optimisation taking into
consideration its inherent axial stiffness and its excellent radial characteristics.
In Chapter 5, radial flux direct drive PM generator structures are designed and
optimised at a small scale and at a large scale using low density materials, such as
composites, and compared with optimised steel structures. Using disc sub structures
with fibres following conventional and mosaic pattern orientations, significant mass
savings could be achieved.
Chapter 6 shows the results achieved from a complete set of modal analyses carried
out over the optimised disc steel structure. The available options to avoid the
5
Introduction
resonant frequencies of the generator structure are explored and discussed in this
chapter. A conical structural configuration is also dynamically tested as its distinct
layout showed potential for further dynamic enhancements.
Further discussions on the work completed in this thesis and the conclusions drawn
are presented in Chapter 7. Potential improvements and suggestions are made for
possible future work.
6
Introduction
1.5 References
[1] BP Statistical Review (2011). BP Energy Outlook 2030. London, UK:

World Energy.
[2] Bloomberg, http://www.bloomberg.com/news/articles/2016-03-03/brazil-s-

clean-energy-boom-running-out-of-steam. Last accessed [Online] 19/08/2016.
[3] O. Edenhofer et al., “IPCC, 2011: Summary for policymakers. In: IPCC
special report on renewable energy sources and climate change mitigation”,
Cambridge University Press, Cambridge, United Kingdom and New York,
NY, USA, 2011.
[4] M. G. Salameh, “Can renewable and unconventional energy sources bridge the
global energy gap in the 21st century?”, Applied Energy Journal, No. 75, pp.
33–42, 2003.
[5] J. Hansen et al., “A closer look at United States and global surface
temperature change”, Geophysical Research Journal, Volume 106, No. 20, pp.
23,947-64, 2001.
[6] The guardian,

https://www.theguardian.com/environment/2016/aug/16/hornsea-project-two-
windfarm-second-phase-grimsby. Last accessed [Online] 16/11/2016.
[7] A. S. McDonald, “Structural analysis of low speed, high torque electrical
generators for direct drive renewable energy converters”, Ph.D. dissertation,
University of Edinburgh, Edinburgh, 2008.
[8] E. de Vries, ‘Trouble spots – Gearbox failures and design solutions’,

Renewable Energy World, Mar. 2006, 9, (2), pp.37-47.
[9] J. Wallace, M. Jackson, S. Rogers, “The problem with O&M”, Renewable

Energy Focus Journal, Volume 9, No. 7, pp. 22-7, January-February 2009.
[10] T. Hartkopf, M. Hofmann, S. Jöckel, ‘Direct-drive generators for megawatt

wind turbines’. European Wind Energy Conf., Dublin, Ireland, 1997, pp. 668-
671.
7
Structural Stiffness Modelling of Wind Turbine Electrical Generators
Chapter 2
Literature Review
Climate change and global warming are issues that need to be addressed if we want
to leave a sustainable planet to future generations. The energy sector is one of the
key subjects that must be dramatically redirected. Fossil fuels are rapidly increasing
the CO2 emissions level. According to [1], the World CO2 emissions released have
doubled between the years 1973 and 2010 with 30,000 Mtonnes per year. With an
increasing energy demand, caused by the world population growth, a decrease in the
use of fossil fuels and a rise in the renewable sources availability have become vital.
In this context, wind energy has an important role to play. Harvesting wind power
has been a significant energy source for many years, especially during the 9th
8
Century when wind mills were utilised to mill grain [2]. At present, modern wind
turbines are used to generate electrical power and are usually seen as complete wind
farms. Current wind turbines are complex large rotating devices that challenge
engineers of all disciplines.
2.1 Wind energy
Wind is produced by the differences in atmospheric pressure, as air moves from high
pressure to low pressure regions. Since the Sun’s rays reach the Earth’s surface with
distinct angles, different temperatures between the equator and the poles are set up.
Warmer air rises (leading to high pressure), whereas colder air sinks (leading to low
pressure). Wind is the flow of air between areas of high and low pressure.
Wind energy is a natural resource present all over the planet that concentrates in
certain regions. A steady technological progress in the area, alongside investments
and economic support have lead the wind industry to be the quickest growing
renewable energy source with a total installed capacity of 432.9 GW at the end of
2015 [3].
Wind turbines can be located both onshore and offshore, although it is true that more
than 97 % of the worldwide installed wind capacity is placed on land [3]. This is
because the cost of energy is lower onshore due to lower capital costs.
Wind speeds are slower onshore than at the sea surface due to the terrain’s roughness
that generates friction and creates turbulence. Offshore, the sea roughness can be
considered almost zero with good weather conditions (waves influence must be
estimated) so that wind speeds are generally higher. In addition, onshore wind farms
face certain drawbacks that offshore devices do not. For example, no size and noise
restrictions exist for turbines installed offshore making them a suitable alternative for
developers globally.
Nevertheless, some significant challenges have to be considered when planning and

installing offshore wind parks. The sea is a harsh environment that does not always
allow an easy access to the turbines for installation or maintenance. In the installation
9
case, both the wind speed and wave height must be below certain levels so that
cranes and vessels can work safely. It is worth pointing out that if a scheduled
maintenance cannot be carried out because of extreme conditions a very high cost
produced by the downturn in the turbine availability will be paid by the owner. This
is why manufacturing reliable elements at a minimum cost is so important.
Lightweight and easy to install structures are of interest although not always possible
since generators and other components such as gearboxes can be very complicated
and heavy.
2.2 Modern wind turbines
Modern turbines take out the power of the wind by transforming kinetic energy into
pressure energy at the rotor plane. Then, by making use of an electrical generator the
rotational energy and rotating power of the turbine’s main shaft can be converted into
electrical energy to be sent to the end consumer through the transmission and
distribution systems. As depicted in Figure 2.1, an offshore arrangement comprises
an offshore wind turbine (formed by a turbine rotor, a transmission gearbox, an
electrical generator and a power converter), a wind farm collector where the power
delivered from other wind turbines is gathered and a transmission to shore system
that consists of an offshore substation, the HVAC or HVDC cable transmission
layout and an onshore substation from where the power is introduced into the
national grid.
Figure 2.1 Offshore wind turbine arrangement and transmission to shore
10
Although the final wind turbine layout depends on the manufacturer, four
different standard types exist. They can operate at either fixed or variable speed
and they can be controlled by either stall or pitch mechanisms:
a) Stall regulated fixed speed wind turbines

The constant speed at which they operate is established by the frequency
of the grid. They have induction generators which are directly connected
to the grid. The aerodynamic design of the blades makes the turbine stall
in high winds reducing the power and the thrust. The only control
options to manage the power output are to use a shaft brake and to
connect or disconnect the machine from the power system [4]. Some of
the most representative models of this type of wind turbines are:
 Nordex N62 1300 kW [5]
 Ecotecnia 62 1300 kW [6]
 REPower 48 600 kW [7]
b) Pitch regulated fixed speed wind turbines

These machines control their torque by pitching the blades. They usually
employ pitching mechanisms to start up and above rated wind speed
power control to maintain a constant rotor speed. The blades pitch to
increase rotor velocity up to operating speed. At this moment, the
electrical machine is connected and kept a constant speed. The Nordex
N60 (1.3 MW nominal power) is a typical example of a pitch regulated
fixed speed wind turbine [5].
c) Pitch regulated variable speed wind turbines

Pitch regulated variable speed turbines use pitch control above rated to
regulate the rotor speed and utilises torque control for the generator over
the operating range of the wind turbine to permit variable speed
operation of the machine. The V90 3 MW Vestas model and N80 2.5
11
MW Nordex model are examples of pitch regulated variable speed wind

turbines [8][5].
2.3 Wind turbine drivetrains
Drivetrains are the area of biggest variation in wind turbine design. Conventional
wind turbines are usually formed by a gearbox, a medium or high speed generator
and a power converter. A gearbox is a mechanical system which transmits the
rotational power of the wind turbine rotor to the electrical machine and increases the
rotational speed to match the one of the electrical generator. A generator is a device
which converts the mechanical power into electrical power. Generally speaking, two
distinct types of electrical generators exist: induction/asynchronous generators and
synchronous generators. Induction machines are alternating current (AC) electrical
generators that operate by mechanically turning their rotors quicker than the
synchronous speed, whereas synchronous machines operate turning their rotors at the
synchronous speed. The synchronous speed corresponds to the frequency of the
supply current. During operation, the rotor magnetic flux cuts the stator coils
producing an active current in the stator coils. Finally, a power converter is an
electro-mechanical device that converts electric energy by adjusting its voltage and
frequency.
Gearboxes have a limited lifetime that in most cases does not reach the expected
wind turbine lifetime. It is normal to replace them at least once in the course of the
wind turbine lifetime [9]. It is considered that the higher the number of gearbox
stages, the higher the gearbox failure rate [10]. Consequently, manufacturers are
studying new methods to replace or just to eliminate the gearbox by introducing the
direct drive concept.
A common layout of a fixed speed wind turbine with a squirrel cage induction
electrical machine is shown in Figure 2.2. Squirrel cage generators are induction
machines that have a rotor made of a ring of conducting bars, short-circuited at both
ends by rings forming a squirrel cage. One of the main components within the wind
turbine configuration is the soft-starter unit, which is composed by 6 thyristors, and
12
which function is to control how the magnetic flux is created to decrease current
ripple during the generator energising stage. Once the electrical machine has been
energised, the soft-starter takes the drivetrain to its operational speed by slowly
ramping up the voltage at the generator terminals until it is at the network voltage
[11]. This reduces the mechanical stress on the shaft and on the generator, as well as
the electrodynamic stresses on the attached power cables and electrical network,
extending the lifespan of the system.
Figure 2.2 Layout of a conventional fixed speed wind turbine [11]
If variable speed turbines are compared with fixed speed ones a considerable
reduction in loads is achieved. DFIG (doubly fed induction generator) wind turbines
use wound-rotor generators with slip rings. Wound-rotor machines are induction
generators where the rotor windings are connected through slip rings to external
resistances. By adjusting the resistance, the speed/torque characteristic of the
generator can be controlled. A variable frequency converter feeds the rotor winding.
It commonly consists of 2 AC/DC + DC/AC IGBT based back-to-back voltage
source converters linked by a DC bus [11]. Since with the power electronics
converter, the network electrical frequency is decoupled from the rotor mechanical
frequency variable speed operation becomes possible.
If the rotational velocity of the wind turbine is low, the rotor of a DFIG turbine can
absorb power from the grid through the power converter. On the contrary, if the
electrical machine works above synchronous speed, power can be sent to the power
13
network in the opposite direction through the power converter. Figure 2.3 displays
the typical configuration of a DFIG wind turbine.
Figure 2.3 Layout of a conventional DFIG wind turbine [11]
As it can be seen, another element used for protection, named crowbar, is present. It
is employed to protect the power converter. Over-currents and sudden increases in
voltage are typical faults that the crowbar system can cope with.
Figure 2.4 Layout of fully rated converter wind turbine [11]
On the other side, there are fully rated converter wind turbines. These can deliver a
higher level of control by making use of power converters. The fully rated converter
wind turbine can have various different layouts. A broad spectrum of electrical
machines can be used and might or might not have gearbox.
14
Advantages and disadvantages of a brushless synchronous generator with a gearbox

and a full converter are detailed in Table 2.1.
Table 2.1 Permanent magnet synchronous generator system advantages and disadvantages
Advantages Disadvantages
Less complicated grid fault ride-through A larger and more expensive converter is
competences needed
More simple control techniques used Higher losses in the power converter
(fully rated)
Used in both 50 & 60 Hz grids
Minimized maintenance. Eye-catching
for offshore purposes
With the drop in the power electronics cost a decade ago, this design became eye-
catching to manufacturers since they improve grid fault ride-through effectiveness.
This type of generators can be used in both 50 and 60 Hz grids meaning that they are
also suitable for the North American market, where the frequency of the grid is 60
Hz. The lack of brushes and slip rings minimises its maintenance and as the cost of
transporting and lifting equipment in deep waters is very high, this system has
become very attractive for offshore purposes.
As the power converter works as the interface between the grid and the generator, the
electrical machine is decoupled from the network. As a result, the generator’s
electrical frequency varies with wind speed variations, whereas the frequency of the
grid remains stable, allowing the machine to work at a variable speed [11].
The generator control scheme and power flow is picked in accordance to the sort of
power converter layout utilised. The side of the power electronics converter facing
the network (inverter) can be configured to keep a constant voltage in the DC bus by
applying torque to the generator that is regulated by the side of the converter facing
the generator (rectifier). This torque is also controllable by the network side of the
converter. Any of the banks can independently absorb or produce reactive power
[11].
15
2.4 Electrical machines
Motors are electrical machines which convert electrical power into mechanical
power, whereas generators are electrical machines which convert mechanical power
into electrical power. Figure 2.5 illustrates a simple electrical machine with a pole
pair (north and south) which creates an electromagnetic field. The wire loop is a
rotating armature. When current flows a force acts in an upwards direction on the
right hand side of the wire and a downwards direction on the left hand side of the
wire. Therefore, a torque is induced and the electrical machine works as a motor. If
no current is flowing in the wire but the armature rotates relative to the field, then an
induced voltage (EMF) would be induced according to Faraday’s Law. At this point,
there is no torque resisting the rotation. If some resistive electrical load is connected
to the terminals of the wire loop, current will start to flow. A torque will also be
developed meaning that work is used to rotate the wire loop. In this case the
electrical machine would work as a generator.
Figure 2.5 Illustration of a simple electrical machine [4]
Electrical generators can be synchronous or asynchronous machines. Synchronous

generators are alternating current, ‘AC’, machines where rotational speed is
dependent on the current frequency in the stator and the number of pole pairs. The
magnetic fields created by the current on the stator and on the rotor both rotate at the
same speed (called synchronous speed). Asynchronous generators are machines in
which the magnetic fields of the stator and the rotor do not rotate at the same speed.
16
By changing the magnitude of the rotor field current, the induced EMF can be altered
which means that the reactive power produced by a synchronous machine can be
controlled. This is an important ability that makes this type of electrical machines
suitable to maintain power system stability.
Figure 2.6 Illustration of an AC synchronous generator [11]
Figure 2.6 shows the rotor and the stator of a synchronous machine. The latter is
made up of a laminated steel core and a 3 phase winding that at the same time is
composed by coils.
2.6 Wind turbine generators
Until 1998, the vast majority of the manufacturers constructed constant speed wind
turbines with a 3 stage gearbox, a standard squirrel cage induction generator directly
coupled to the grid and a power output below 1.5 MW. Since then, manufacturers
started to look at more complex variable speed wind turbines in an attempt to
optimise the design of the turbine and its efficiency. Around 2005, wind turbine
generators began to be improved by manufacturers so as to meet the grid fault ride-
through requirements [12].
17
In [12], Polinder presents an overview of the available generators for wind turbines
and their trends. A good insight of the types of wind turbine generators used in the
present day as well as their advantages and disadvantages is given in this
comprehensive study. Till recent times, it was thought that constant speed wind
turbines were lighter, cheaper and have a more compact design than direct drive
turbines that were seen as heavy, large and expensive systems. Nevertheless,
companies, for instance Siemens, have found that enhanced direct drive designs have
a similar weight to conventional geared converters.
2.6.1 Electrical machines for direct drive wind turbines
As explained, direct drive wind turbines present a number of advantages that make
them suitable for offshore purposes. Among them, the removal of the gearbox stands
out. With it expensive gearbox matters can be avoided. Moreover, the decrease in the
number of moving parts, such as bearings, represents a significant reduction in
downtime periods, for example for oil replacement. These features have drawn the
attention of some manufacturers and nowadays 20 % of the wind turbines sold
worldwide are directly driven.
2.6.2 Excitation techniques
In order to excite an AC synchronous generator different ways exist: Electrical

excitation and permanent magnet excitation. Switch reluctance generators are
machines in which only the stator is electrically excited.
2.6.2.1 Electrically excited direct drive generators
In electrically excited generators, a DC source magnetizes the rotor poles. This DC

source is commonly given through brushes and slip rings. There are two types of
rotor poles: salient or cylindrical. Cylindrical poles are the most employed in the
wind industry although rotors with salient poles are more common when speaking
about direct drive machines. In this type of generators, the rotor poles must be large
18
enough to generate appropriate room for the excitation windings. The stator of an
EEDD is very similar to that of a 3 phase distributed winding introduced into a
slotted laminated iron core induction generator. The generator is connected to the
grid through a power electronics converter that permits full control of active and
reactive power as well as voltage in case of grid fault. In addition, the electrical
machine speed can be completely regulated over a broad range of wind speeds.
These machines are relatively easy to manufacture and for large power outputs they
provide better power factor and efficiency compared to induction machines.
Nevertheless, the constant injection of DC current leads to If2R losses on the field
decreasing the overall efficiency. However, since the external electrical excitation
can be tuned in accordance to the wind conditions, the voltage can be varied to
reduce If at lower outputs to minimise losses.
EEDD is currently considered a mature low speed direct drive technology widely
used in the wind energy market. The main supplier of EEDD is Enercon, serving 15
% of the whole market and 75 % of direct drive systems installed. As for instance,
the prototype E-126, shown in Figure 2.7, that has a 127 m rotor diameter and a 12 m
diameter generator. The turbine can generate up to 7.5 MW. The company MTorres
also manufactures EEDD turbines with outputs of 2.5 MW [13].
Figure 2.7 7 MW Enercon E-126 [14]
19
2.6.2.2 Permanent magnet excited direct drive generators
Among the drawbacks of electrically excited systems, efficiency reduction due to

If2R losses and maintenance issues because of the use of brushes stand up. Permanent
magnet generators can be considered superior to electrically excited machines
because of their lower weight, improved efficiency and compactness. These
generators do not need external electric excitation since their rotor poles are made of
permanent magnet material. The overall efficiency of the machine and the energy
capture rate then increases as there are no If2R losses. Besides, the lack of slip rings
improves the machine’s reliability.
On the side, permanent magnets are costly and difficult to handle during the
manufacturing stage. The stationary structure of a PMDD is similar to that of an
EEDD although different designs have been proposed [15][16]. For the stator’s
connection to the grid a fully rated converter is required to convert from variable
voltage and frequency to fixed voltage and frequency.
Permanent magnets are made of rare earth materials, e.g. samarium cobalt (SmCo) or
neodymium iron boron (NdFeB) which gives high magnetic densities in small
geometries and volume. SmCo permanent magnets are typically utilised in high-
temperature approaches. In [17], Vilsboll et al. states that NdFeB magnets are the
most suitable as they generate a larger remnant flux density (1.2 T) and can decrease
mass and price of the generator.
At the beginning, the high prices of PM materials discouraged manufacturers from

using them for this type of generator. However, the cost of rare earth materials
dropped between 1995 and 2005 by a factor of 10. In 2011, the prices went up again
due to a number of factors including a rise in demand and the sensitivity to
speculation and politics of the supply (95 % of the rare earth material is located in
China). Nowadays, prices have stabilized and although the mentioned issues
generated a great uncertainty regarding PM use, the future of PMDD generators is
promising since rare earths are being found in many other places.
With permanent magnet excitation a simpler and robust electrical machine can be
produced. Higher efficiencies and torque densities as well as limited life cost are
20
some of the advantages that have drawn the attention of big developers such as GE
that acquired Scanwind in an attempt to expand into this market [18]. For offshore
applications GE has offered a 4 MW design as seen in Figure 2.8.
If a comparison between a PMDD radial flux machines and a conventional generator,

(such as a constant speed geared induction machine) is made it can be seen that the
PM machine is less efficient at rated power but has greater average efficiency, e.g.
2.3 % higher for a 500 kW machine and 1.6 % higher for a 3 MW machine [19].
Grauers found a PM equivalent to the electrically excited synchronous machine of
Enercon with a 94 % smaller diameter and 50 % lower rotor volume. In [20], Bianchi
and Lorenzi showed that wound rotor designs are less efficient than permanent
magnet designs. On the other side, in [21], the authors found that electrically excited
machines have greater mass than PM excited machines whereas in [22], Dubois
noted that permanent magnet excitation is more profitable for pole pitches shorter
than 100 mm. In [17], Vilsboll et al. found that the efficiency of their permanent
magnet machines goes up with reducing load as far down as quarter load.
Figure 2.8 4 MW GE (former Scanwind) wind turbine [23]
Other companies entering into this MW scale market are: Siemens with its 3 and 7
MW design, Goldwind (1.5-2.5 MW), STX Windpower (1.5-2.0 MW), Emergya
21
Wind Technologies ‘EWT’ (0.5-2.0 MW), Vensys (1.5-2.5 MW), Leitwind (1.5-3.0
MW) and XEMC Darwind (5.0 MW) [24][25][26][27][28][29][30].
With the quick increase of commercial and military applications using PM over the
last decade, the industrial capacity for PM machines has grown and this is even
highlighted with the rise in the number of manufacturers establishing PM
synchronous generators as their first choice for their direct drive wind turbines [31].
2.6 Permanent magnet topologies
Permanent magnet generators are often characterized by the orientation of the

magnetic flux as it goes across the air gap, as follows (See Figure 2.9):
 Radial flux
 Axial flux
 Transverse flux
(a) (b) (c)

Figure 2.9 PMDD generator topologies; (a) Radial flux; (b) Axial flux; (c) Transverse flux [32][33]
The electrical machine can be slotted or slotless according to the design of the stator.
Permanent magnet direct drive generators can also be characterized depending on the
22
absence or the presence of iron in the core of the stator. As result, they can be either
iron cored or air cored machines.
2.6.1 Radial flux PM generators
These machines are so called because of the orientation of flux as it goes across the
airgap. They are very similar to wound rotor synchronous generators which are also
constructed as radial flux machines. Figure 2.10 displays the configuration of radial
flux PM machine.
Figure 2.10 PMDD radial flux machine and its components [34]
Some commercial models are out there since the number of manufacturers getting
interested in permanent magnet machines has been increasing over the last decade.
Most of them have a radial flux configuration [18] [28]. Figure 2.11 shows the layout
of the Zephyros/Harakosan Z72 wind turbine with accounts with 1.5 MW power
output. The electrical machine is rated at 18.5 rpm and owns a 4 m diameter structure
which weighs 47.2 tonnes [35].
23
Figure 2.11 Cutaway of the Zephyros/Harakosan Europe Z72 wind turbine [35]
2.6.2 Uncommon radial flux configurations
Aiming for a more compact lightweight design, Sway has developed a model as it is
shown in Figure 2.12. With a significant decrease in weight achieved by its novel
structural design, considerable tower and foundations optimisations can be made. By
having a large diameter ironless stator winding machine with a spoked rotor, the
system is 25 to 50 % lighter than conventional designs [36]. Low airgap flux
densities and shear stress are likely to be obtained with ironless stators and in spite of
higher eddy current and aerodynamic losses are higher than for conventional
turbines, developers say that the design’s efficiency is about 94 %. Therefore, an
overall decrease in cost of electricity per kWh is forecasted over existing machines.
Figure 2.12 ST10 10 MW offshore wind turbine developed by Sway [36]
24
The C-GEN concept developed by researchers at the University of Edinburgh is also

an air cored generator with a modular rotor consisting of C-core modules 50 %
lighter than a conventional iron cored PM direct drive generator. Its ease of
manufacturing and lightweight design makes it attractive not only for wind energy
converters but also for any type of marine power take off [37].
2.6.3 Further variations
An extra variation in permanent magnet direct drive generators is the option of

having an inner or an outer rotor. While the outer rotor may lead to a more compact
design and the centrifugal forces help magnets seat onto the rotor, the inner rotor
facilitates the stator cooling (naturally cooled by external air flow). A reduction in
the outer diameter can be achieved by using an outer rotor according to Vensys [28].
Figure 2.13 Inner and outer rotor generator variants [13]
2.6.4 Surface mounted or buried
In terms of rotor design, a further variation can be introduced. The permanent

magnets can be either surface mounted onto or buried into the rotor structure. High
energy magnets such as NdFeB are commonly used in surface mounted layouts.
They have a remnant flux density that goes above the wanted airgap flux density.
Although they are expensive a lightweight design can be produced. When these are
mounted on the rotor, they must be mechanically protected and coated since they are
prone to corrosion [38].
25
Buried ferrite magnets can be employed with flux concentration. They are lower
energy magnets and are much cheaper (a tenth of the cost per unit mass [39]).
Nevertheless, more magnets need to be used (5 times as much magnetic material
compared to an equivalent surface mounted machine [38]) and their assembly is
more complex and expensive.
According to Vilsboll et al. NdFeB magnets are more suitable than ferrite magnets
due to a reduction in overall mass, dimensions and price of a 20 kW generator for a
wind turbine can be obtained [17].
Several rotor designs for radial flux permanent magnet machines were studied by
Lampola such as curved, rectangular and rectangular equipped with shoes PM’s
surface mounted. The analysis demonstrated that the highest torque to cost of active
materials ratio was acquired with curved surface mounted permanent magnets [40].
2.6.3 Axial flux generators
An overview of axial flux machines, focused on axial flux permanent magnet

machines is given in [41]. The following features make this type of machines
different from conventional ones:
 Airgap is in the axial direction and conductors are radially lined up

 Rotor and stator are discs
 Discs under rotation work as fans
These electrical machines have been suggested for a diverse range of purposes, at
different speeds. Some examples include portable generator systems [42], direct
drive in-wheel motors [43], generator units in vehicles [44], for propulsion and ship’s
generators [45] and for aircraft drives [46].
2.6.4 Transverse flux generators
Transverse flux machines are similar to axial ones in that the flux track goes
perpendicular to the plane of rotor’s rotation. See Figure 2.14. Nevertheless, this type
of machines is very different to axial generators. The major difference between
26
transverse flux permanent magnet generators and RFPM and AFPM is that the
TFPM machine permits an increase of the space for the windings without reducing
the available space for the main flux. They can also be made with a very small pole
pitch compared to other machines. This results in higher force densities that make
this type of topology attractive for direct drive purposes.
Figure 2.14 A single-sided surface-mounted TFPM machine [47]
2.6.5 PM configurations comparison
An overview of the PM configurations available for wind turbine generators is

provided by Dubois in [48]. By making use of documentation about samples, Dubois
compares these technologies in accordance to torque density and cost per unit [49]. A
comparison of PM layouts for a wide range of turbines with power outputs no larger
than 200 kW was carried out by Chen et al. [50]. The parameters to make
comparison were torque density, mass of the active material, outer radius, total
length, overall volume and efficiency. The main conclusions were that axial flux
permanent magnet machines have simple winding, low cogging torque and noise,
short axial length and higher torque/volume ratio than radial flux machines
permanent machines. However, they also have lower torque/mass ratio, larger outer
diameter, large amount of permanent magnets and structural instability and
difficulties keeping airgap integrity and producing stator cores if compared with the
radial machines. Transverse flux machines present higher force densities than the
27
other 2 configurations, considerably low copper losses and simple winding but it is
rather complex to construct which makes it less attractive.
These comparison brings a problem with itself due to it is done by taken into
consideration only the active mass (magnets, iron in rotor and stator and copper in
windings). Nevertheless, the mass of the inactive material in direct drive machines
supposes the majority of the machine’s mass. In [21], Hartkopf et al. noted that about
2/3 of the total mass corresponds to the supporting structure. If the most suitable
permanent magnet topology is to be found then the mass of the inactive material also
needs to be considered.
In this thesis, a radial flux surface mounted permanent magnet machine has been
assumed. The methodology followed for its analysis can be applied to other types of
machines.
2.7 Structural Analysis, Modelling and Design of Direct Drive

Generators
Direct drive permanent magnet machines are attractive, however, these electrical
machines are currently very large and heavy and therefore expensive. Nevertheless,
they are also part of the turbine’s structure and withstand significant loads. This
section looks in detail how this type of electrical machine have been analysed
(considering the loads applied to its structure and the layout of the drivetrain),
modelled (simplified models such as disc or armed structures can be used) and
finally designed. It is thought that by employing an integrated approach which takes
into account the interactions between electrical, thermal and mechanical design
aspects of the machine, a reduction in weight could be achieved.
2.7.1 Dimensions of direct drive generators
So as to understand why these generators are large it is essential to consider the

torque rating T. If the output power of a generator is calculated as follows,
28
𝑃 = 𝑇𝜔 (2.1)
where 𝑇 is the torque produced by the electrical machine and 𝜔 is its angular
velocity, a large torque must be generated so as to produce high power. The torque
produced by a generator can be estimated by considering the generator as a cylinder
with a surface shear stress. See Figure 2.15. This can represent the rotor or stator of
the radial flux machine.
Figure 2.15 Cylinder model of torque produced by a generator [34]
Then, the equation to obtain the torque is,
𝑇 = 2𝜋𝜎𝑅 2 𝑙 (2.2)
where 𝜎 is the electromagnetic shear stress, 𝑅 is the radius of the airgap and 𝑙 is the
length in the axial direction. Since there are practical limits to magnetic and electrical
loading, a maximum shear stress exists. The typical shear stress for PM machines
used by designers is within the range between 25 and 50 kN/m2 [13].
2.7.2 Forces and moments acting on electrical machines
To keep the airgap clearance between the stator and rotor is vital from the design
point of view. The integrity of the entire machine relies on the capacity of the
supporting structure to maintain the airgap open and stable. Large direct drive
electrical generators are more demanding structurally speaking than conventional
generators due to their large surface areas. Larger moments are generated since
29
forces act at larger distances from the nacelle mounting location. The airgap
clearance is about 0.1 % of the airgap diameter for typical designs and a typical
maximum permitted deflection is about 10 to 20 % of that clearance. If this threshold
is breached, the airgap flux density will vary significantly and hence the loads will
also increase. The analysis stage is about recognising the forces acting on the
structure and their magnitude. On a radial flux electrical generator different loads
such shear, normal, gravitational and thermal, as well as centripetal forces and wind
turbine loading, are present.
The shear stress/torque transmission is generated in the area near the airgap where
mechanical energy is transformed into electrical energy. In the steady state, the shear
force on the rotor is met by an equal but opposite shear force on the stator. That shear
force on the stator comes about as current in the slots and the associated magnetic
field interact with the rotor magnetic field. See Figure 2.16.
Figure 2.16 Shear loading [34]
The normal stress, also named Maxwell stress, is produced by the effect of attraction
that the magnets mounted on the rotor generate between the moving and the
stationary parts of the machine. It is the largest load (in order of 200-400 kPa in
typical machines) and makes the reduction of machine’s weight a difficult task for
designers as a stiff and robust structure is needed to withstand it. See Figure 2.17.
Figure 2.17 Magnetic attraction of the moving and the stationary components of the generator [34]
30
A key issue to be considered during assembly, transportation and installation is

gravity. Generator’s weight must be taken into account. In [51] , Stander et al. states
that segmentation of substructures greater than 4 m (EU) and 5m (USA) can ease
manufacturability and transportability. The structural stiffening required to resist
gravitational loading can be acquired by arranging the structure geometry rather than
utilising stiffer materials. Different examples of stator/rotor yoke structures support
are provided: cantilever/Z-profile on one single side, E-profile in the middle or H-
profile on both sides. See Figure 2.18.
Figure 2.18 Gravitational loading [34]
Since large amounts of heat are generated during electrical machines operation,
thermal expansion or contraction of generator’s parts has to be taken into
consideration. The stator is typically is hotter than the rotor ΔTs > ΔTr. Figure 2.19
illustrates how the deformation caused by thermal expansion or contraction of
components can produce significant changes in flux density and therefore in the
forces acting on the generator structure.
Figure 2.19 Thermal expansion of the generator structure [34]
31
On the other side, designers need to be aware of centrifugal forces (~3 kPa). These
loads are fairly small at low speeds compared to the others.
Figure 2.20 Generator’s structure under centrifugal forces [34]
2.7.3 Integrating a direct drive generator within the wind turbine
The location and integration of the direct drive generator within the wind turbine and
drivetrain are key factors when calculating the loads generated by the rotor blades,
rotor blades weight, vertical and horizontal wind shear, yaw error and inertial effects.
Next figures show the different existing types of PMDD drivetrain configurations.
Figure 2.21 shows how the generator has been mounted upwind of the tower with a
single bearing, for example, Zephyros/Harakosan Z72 wind turbine. This type of
layout is popular within the market since a single bearing arrangement is allowed
leading to a reduction in cost. The generator hangs clear of the turbine, whereas the
bearings, which resist enormous loads, can rest on the nacelle’s structure. However,
it becomes an important structural load path of the wind turbine (its particular conical
supporting structure is able to effectively deal with radial and axial loads) and when
the electrical machine needs to be either repaired or replaced it is impossible to do it
without also removing the rotor
32
Figure 2.21 Zephyros/Harakosan Z72 wind turbine configuration [13]
An alternative design was proposed by Bywaters et al. [31] in the Northern Power
Systems WindPACT Drive Train Alternative Design Study Report. The main bearing
inner face is attached onto a spindle that carries the stator as well while the outer race
is connected to the rotor hub and the rotor of the generator. The spindle is anchored
to the turret providing the load path mentioned before. As seen in Figures 2.22 and
2.23, the generator is an integrated component which gives the possibility of
shipping an entirely assembled and tested machine to the site where it can be
installed onto the turret in one manoeuvre. With the capability of locking the
generator rotor to the stator frame, the main bearing can be accessed for maintenance
without taking out the generator. In addition, the seals of the bearing can also be
repaired or simply replaced without extracting the bearing.
33
Figure 2.22 PMDD Drivetrain [31]
Figure 2.23 PMDD Generator [31]
Figure 2.24 shows the drivetrain configuration of an MTorres 1.5 MW PMDD wind
turbine. The generator is located right onto the top of the tower and between
bearings.
34
Figure 2.24 MTorres 1.5MW PMDD wind turbine [13]
In Figure 2.25 a GE ScanWind 4.1-113 wind turbine is shown in detail. The

generator is located downwind of the tower and the two bearings sit down on
nacelle’s structure as in the case with the MTorres design. Again for this design, the
electrical machine becomes an integral part of the turbine’s structure acting as a load
path, although major rotor loads have been dealt with by the bearing before reaching
the generator. A more robust (in structural terms) and therefore heavier machine is
necessary.
35
Figure 2.25 GE ScanWind 4.1-113 wind turbine [13]
2.7.4 Design external loads for HAWT: International standards
In [51], Stander et al. differentiate between external and internal loads. Hub
(including thrust) and gravitational loads are considered external loads whereas
internal loads are electromagnetically induced and thermal. External loads have a
considerable influence on internal loads. For instance, in case of shaft misalignment
due to forces acting on the wind turbine rotor, a substantial increase in the attraction
forces within the airgap would be seen. The degree of influence of external loads on
electromagnetically induced stresses depends on how well integrated the generator is
in the wind turbine structure. In [52], the design loads for horizontal-axis wind
turbines are defined according to two different coordinate systems, one respect to the
blade and the other respect to the hub.
A more conscious differentiation of the types of stresses is given in the IEC 61400-1,
the Germanischer Lloyd rules for certification and the Danish Standard DS
472[53][54][55]. All the loads to be considered when designing a wind turbine
component, methods of analysis, material strengths, fatigue properties and the
36
corresponding safety limits are described in detail in the said international standards,
which are used for certification.
2.7.5 Configurations
Different bearing configurations have been described in Section 2.7.3. Looking at the
entire drivetrain, Alstom’s Pure Torque® drive system corresponds to an innovative
design, which separates torque transmission from load support and transfers all the
bending and gravitational loads directly to the tower. Two bearings connect the hub
to a cast iron structure, while the main shaft is attached to the electrical machine
through an elastic coupler, which allows certain degree of misalignment [56].
This system can be also used for conventional geared wind turbines. By making use
of this arrangement potential harmful loads for the gearbox or the generator are
eliminated elevating so the reliability of the turbine.
Other bearing layouts have been proposed with the main aim of obtaining a more
compact lightweight design. In [31], Bywaters et al. evaluated a number of bearing
configurations on the base of weight, cost, risk, shipping, assembly and
serviceability. An integrated configuration suitable for geared and direct drive
machines that removes the main shaft has been developed in [57]. For this system,
the gearbox or the generator are directly mounted on the outer side of the bearing
rotating with the same speed as the hub.
For this investigation, it was assumed that a system such as Alstom’s Pure Torque®
was utilised as the design of the generator supporting structure was carried out
considering only the major internal loads present during the machine operation, in
other words, isolating the generator from the rest of the wind turbine.
2.7.6 Integrated design of direct drive machines
In order to design PMDD machines different approaches can be followed. In [13],

McDonald proposed an integrated design that considers the interactions between
electrical, thermal and mechanical aspects of the generator.
37
Figure 2.26 Interactions between electrical, thermal and mechanical design aspects of the generator
[13]
A multidisciplinary approach is needed to design and manufacture this type of

machines due to their size, complexity and cost. A suitable compromise between the
priorities of each discipline is required for a proper design. Figure 2.27 shows the
flowchart of a conventional process starting from the specifications and giving a
privileged position to the aspects of the electrical design followed by mechanical and
cooling design aspects to end with the outline.
Figure 2.27 Early design stages – traditional approach [13]
38
However, as said, the idea behind the integrated design is to create a scheme that
equally considers the priorities of all the disciplines. Figure 2.28 shows the flowchart
of an integrated approach as proposed by McDonald [13].
Figure 2.28 Early design stages – integrated approach [13]
Figure 2.29 shows what the priorities of an electrical engineer are so as to have an
efficient and resilient generator. The electrical engineer would try to maximize
machine’s performance (efficiency and power output) employing as little amount of
material such as copper and permanent magnet as possible. This characteristic may
lead to a reduction in the physical clearance between rotor and stator and a large
radius. The type of material and its form is vital (high fill factors of copper and
correct design of steel laminations are desired). Electrical insulation is crucial for the
electrical engineer too. It is worth pointing out that the electrical design of a 3 MW
PMDD generator supposes 52 % of its total cost [13].
39
Figure 2.29 Electrical design perspective [13]
On the other side, the mechanical engineer would focus on the physical robustness of
the electrical machine, in terms of strain, stress, strength and fatigue.
Suitable safety factors can be achieved utilising low-cost and lightweight assemblies
if the mechanical scheme is good. However, the search of robustness might lead to
the utilisation of non-optimal materials electromagnetically speaking. The
mechanical design can be improved by optimising some electromagnetic parameters
which reduce torque ripple and fault torques
Looking at current designs of 5 MW generators is easy to recognize that around 55 %

of their total mass comes directly from the mechanical design.
40
Figure 2.30 Mechanical design perspective [13]
A special set of necessities can be identified for cooling the generator. The specialist
in heat transfer would add to and maximize thermal conduction and convection
tracks in order to reduce winding temperatures due to some electrical components are
temperature dependent. For instance, the resistance of copper is proportional to
temperature. Therefore, I2R losses are proportional to winding temperature as well.
Higher efficiencies and power densities can be acquired by decreasing temperature.
At the same time, the magnet BH curve is dependent of temperature. High
temperatures can cause a significant reduction in magnetic loading as well as to
elevate the risk of demagnetization in case of faults.
41
Figure 2.31 Thermal design perspective [13]
From the manufacturing point of view, a considerable cost reduction can be obtained
if the design is simplified. Cheaper generators could be constructed if concepts such
as modularisation (lower operational costs and labour are achieved as smaller field
crews are needed, project timelines are shorter and the use of material is more
efficient) and less strict tolerances were introduced. From the logistics viewpoint is
also an attractive option since only 4 m diameter structures can be transported by
road within the EU.
42
Figure 2.32 Manufacturing design perspective [13]
At this point, it is important to mention that might be an optimum electrical design

(in terms of cost, mass and efficiency) but not perhaps a global optimum.
The concept of an integrated design was successfully put in practice with the first 15
kW C-GEN machine prototype. Electromagnetic, structural and thermal issues were
considered by making use of a genetic algorithm that tries to maximise the energy
yield while minimising material and manufacturing costs [37].
43
2.7.7 Modelling methods
2.7.7.1 Analytical approaches for structural modelling
Although many different types of structures can be assumed to characterize

permanent magnet machines, see Figure 2.33, simple disc and arm structural models,
such as shown in Figure 2.34, are more common due to their relative simplicity.
Figure 2.33 Typical rotor structures [51]
Figure 2.34 a) Zephyros/Harakosan Europe Z72, b) MTorres 1.5 MW
44
McDonald [58] used these models to link the mechanical and the electromagnetic
design in radial flux machines and to estimate radial, axial and tangential deflections
in the structure.
Figure 2.35 Radial flux models: a) radial, b) axial and c) tangential deflection [13]
Axial flux machines can be easily modeled employing disc structures. Structural
deflection and dimensions, as well as magnetic forces, can be linked with the model
in question.
Figure 2.36 Axial flux model [13]
The stress shown here would take the form,
1 2 2 (2.3)
𝑞= 𝐵̂ 𝑠𝑖𝑛 (𝑛𝜃)
2𝜇0 𝑔
where q is the stress, 𝐵̂𝑔 is the peak airgap flux density, n is the deflection mode and θ
corresponds to the pitch angle. In reality, there are more forces at play in the machine
45
but it is considered appropriate to neglect these in an early stage design study as the
normal component of the Maxwell stress is the largest force by far.
2.7.7.2 Numerical approaches for structural modelling
The behavior of the generator structure can also be modelled using more advanced
methods, such as computational finite element techniques. For instance, ANSYS is a
widely used commercial piece of software that gives the designer the opportunity to
carry out dynamic behavior and fatigue studies. These are considered standard tools
whose outcome can be utilized with confidence.
On the other hand, they are in fact computationally expensive so their use is limited.
Fine meshes for large volumes own a large number of degrees of freedom which
requires more computational power in order to solve the model in a reasonable time.
2.7.7.3 Structural optimization
Making use of an analytical and finite element analysis optimization method, A.

Zavvos, A.S. McDonald and M. Mueller tried to minimize the structural mass of a
permanent magnet direct drive generator [59]. Three different iron cored generator
configurations rated at 5 MW were optimized concluding that a specific transverse
flux direct drive topology is the most suitable as its electromagnetic layout helps the
structural design. The said topologies were a radial flux and 2 transverse flux PMDD
generators, see Figure 2.37, that were simplified using a disc rotor structure and an
armed stator structure.
A constant amount of copper per unit of airgap surface area and iron with infinite
permeability was assumed so that only the airgap region was modelled [57]. The
mass of permanent magnets on the rotor of the machine was calculated as shown:
𝑏𝑚 (2.4)
𝑀𝑎𝑠𝑠𝑃𝑀 = 2𝜋𝑅𝑙 ( )ℎ 𝜌
𝜏𝑝 𝑚 𝑃𝑀
46
where 𝑅 is the radius of the rotor, 𝑙 is the rotor axial length, 𝑏𝑚 is the width of the
magnet, 𝜏𝑝 is the pole pitch, ℎ𝑚 is the height of the magnet and 𝜌𝑃𝑀 is the permanent
magnets’ density.
Figure 2.37 Illustrations of the three tested generator topologies: a) radial flux, b) transverse flux No.
1, c) transverse flux No. 2 [59]
(a) (b)
(c) (d)
Figure 2.38 Illustration of the variables that were utilised for the optimisation of the generator
structures; (a) The variables that describe a structure with arms; (b) The variables that describe arms
sub structure; (c) The variables that describe a rotor with discs; (d) The variables that describe the
electromagnetic model [59]
47
Figure 2.38 shows the parameters that were optimised in [59]. The effect on the
structural mass of the generator for various structural dimensions was calculated whilst
maintaining the structure’s aspect ratio (Krad = l/2R). The models successfully allowed
the designs to be structurally optimised and concluded that out of the three tested, the
transverse flux PMDD generator was the lightest option.
2.7.7.4 Modes of deflection
So far the discussion has assumed that local and global deflections are one and the
same. This one dimensional model is only correct in a limited case. Actually, the
deformation can be different at different parts of rotor and stator. In [60], the authors
noted that localized contact can eventually happen because of:
Mode 0: Relative radial expansion of the rotor or radial compression of the stator.
Mode 1: Relative displacement of the rotor and stator.
Mode 2: Distortion of either or both of the circular surfaces into ellipses (known as
ovalising).
Mode n: Distortion with ripples, n peaks around the circumference.
In general, the total change in air-gap clearance, δ (= δr + δs) can be expressed as a

function of circumferential angle around the machine’s axis, θ, by equation (2.5),
𝑛
𝛿(𝜃) = ∑ 𝛿𝑛 sin 𝑛(𝜃 − 𝜑𝑛 ) (2.5)

0
where δ(θ) is the change in airgap clearance at angle θ, δn is the amplitude of

component n, φn is the phase angle of component n, and n is the number of peaks,
hence:
n = 0 for deformation of mode 0 (Figure 2.39(a));
n = 1 for mode 1 (Figure 2.39(b));
n = 2 for mode 2 (Figure 2.39(c));
n ≥ 3 for mode 3 (Figure 2.39(d))and higher.
48
Often the air-gap deformation is dominated by a uniform mode 0 component with

amplitude, 𝛿 ̅, and a higher order component with amplitude 𝛿∆ and hence equation
(2.6) can be modified:
𝛿(𝜃) = 𝛿 ̅ + 𝛿∆ sin 𝑛(𝜃 − 𝜑𝑛 ) (2.6)
(a) (b)
(c) (d)
Figure 2.39 A rotor deforming into the airgap towards a stator
(a) Mode 0, uniform deflection, 𝜹(𝜽) = 𝜹̅

̅
(b) Mode 1, eccentricity, 𝜹(𝜽) = 𝜹 + 𝜹∆ 𝒔𝒊𝒏 (𝜽 − 𝝋)
(c) Mode 2, ovalisation, 𝜹(𝜽) = 𝜹̅ + 𝜹∆ 𝒔𝒊𝒏𝟐(𝜽 − 𝝋)
(d) Mode 3, 𝜹(𝜽) = 𝜹 ̅ + 𝜹∆ 𝒔𝒊𝒏𝟑(𝜽 − 𝝋)
49
Figure 2.40 is a plot of the airgap clearance for a prototype generator for a Northern
Power 1.5MW direct-drive wind turbine [61], showing both the designed and
measured air gap clearance, varying with angle. The deflection can be calculated and
is shown in Figure 2.41. The machine shows that the actual deflection exceeds the
designed limits. In this case the parameters in equation (2.6) can be approximated as
n = 2, 𝛿 ̅ = 9.3 mm and 𝛿∆ = 6.6 mm.
In order to understand why different machine designs lead to different mode shapes it
is necessary to understand the magnetic air-gap and structural stiffness properties.
Figure 2.40 Airgap clearance (in mm) for the Northern Power 1.5MW prototype [61]. The values
shown here are the mean of the airgap clearance at the upwind and downwind ends of the machine.
Clearance is plotted for different angles as seen from the upwind end of the machine
50
Figure 2.41 Total deflection from designed air-gap clearance (in mm) for the Northern Power 1.5MW
prototype [61]
2.8 The Concept of Direct Drive Generator Supporting Structures
The idea of designing a direct drive generator supporting structure utilising

composite materials was patented by Siemens Aktiengesellschaft in 2010. Figure
2.42 depicts a cutaway of a direct drive wind turbine with composite material
structures for a hollow main shaft (9) and the electrical machine, comprising the
rotor (18) and the stator (19) arrangements [62].
51
Figure 2.42 Siemens Direct Drive Wind Turbine Cutaway [62]
With the release of this patent, the foundations for the development of an idea that
has been widely proposed but not thoroughgoing studied have been set. In this thesis,
different structural configurations have been analysed and a potential way of
modelling and optimising simplified composite structures for direct drive generators
considering mechanical and electromagnetic issues has been proposed.
52
2.9 References
[1] I. Ency, “Key World Energy Statistics,” 2012.
[2] E. Hau, Wind Energy: Fundamentals, Technologies, Applications,

Economics; 2nd Edition, Berlin: Springer, 2006.
[3] Global Energy Council Report, “http://www.gwec.net/wp-

content/uploads/vip/GWEC-Global-Wind-2015-Report_April-
2016_22_04.pdf”. [Online]. [Accessed 11 September 2016]
[4] J. Manwell, J. McGowan and A. Rogers, Wind Energy Explained: theory,

design and application, 2nd Edition, John Wiley & Sons Ltd, 2009.
[5] Nordex AG, www.nordex-online.com, [Online] [Accessed 10 March 2017].
[6] Ecotècnia, GE Wind, “www.ge.com/wind,” [Online] [Accessed 30 September

2016].
[7] REpower Systems AG, www.repower.de, [Online] [Accessed 10 March

2017].
[8] Vestas Wind Systems A/S, www.vestas.com, [Online] [Accessed 10 March

2017].
[9] S. Alshibani and V. Agelidis, “Issues regarding cost estimation of permanent

magnet synchronous generators for MW level wind turbines,” in IEEE Int.
Elect. Mach. Drives Conf; pp. 1629-1634, May 2011.
[10] K. Smolders, H. Long, Y. Feng and P. Tavner, “Reliability analysis and

prediction of wind turbine gearboxes,” in EWEC, 2010.
[11] O. Anaya-Lara, N. Jenkins and J. Ekanayake, Wind Energy Generation:

Modelling and Control, 2011.
[12] H. Polinder, “Overview of the trends in wind turbine generator systems,” in

IEEE Power Energy Soc. Gen. Meet., pp.1-8, July 2011.
[13] M. Mueller and A. Zavvos, “Electrical drive technology,” in Electrical drives

for direct drive renewable energy systems, Woodhead Publishing Limited,
2013, pp. 3-30.
[14] “www.enercon.de/en-en/Windenergieanlagen.htm,” [Online]. [Accessed 30

September 2016].
[15] D. Bang, “Design of transverse flux permanent magnet machines for large
direct-drive wind turbines,” Ph.D. thesis, Delft University of Technology,
53
Delft, The Netherlands, 2010.
[16] E. Spooner and B. Chalmers, “TORUS: A slotless, toroidal-stator, permanent

magnet geneator,” in International Conference on Electrical Machines,
pp.1053-8, Cambridge, 1990.
[17] N. Vilsboll, A. Pinegin, D. Goussarov and J. Bugge, “The experience of

designing and testing a 20kW multi pole permanent magnet generator for wind
turbines,” DEWI Magazine, vol. no. 9, pp. 74-83, August 1996.
[18] Renewable energy world, http://www.renewableenergyworld.com, [Online].

[Accessed 10 March 2017].
[19] A. Grauers, “Design of direct-driven permanent magnet generators for wind

turbines,” Ph.D. dissertation, Dept. of Electric Power Engineering, Chalmers
University of echnology, Goteburg, Sweden, 1996.
[20] N. Bianchi and A. Lorenzoni, “Permanent magnet generators for wind power
industry: an overall comparison with traditional generators, pp. 49-54,” in IEE
Conf. on Opportunities and Advances in International Power Generation ,
Durham, UK, Mar. 1996.
[21] T. Hartkopf, M. Hofmann and S. Jockel, “Direct-drive generators for MW

wind turbines, pp. 668-671,” in European Wind Energy Conference, Dublin,
Ireland , 1997.
[22] M. Dubois, “Optimized permanent magnet generator topologies for direct-

drive wind turbines,” Ph.D. dissertation, Dept. of Electrical Engineering, Delft
University of technology, Delft, The Netherlands, 2004.
[23] General Electric GE, “www.ge.com/wind,” [Online]. [Accessed 30 September

2016].
[24] Siemens, “www.siemens.com/wind,” [Online]. [Accessed 30 September

2016].
[25] Goldwind, “www.goldwindglobal.com,” [Online]. [Accessed 30 September

2016].
[26] STX WindPower, “www.stxwind.com,” [Online]. [Accessed 30 September

2016].
[27] Emergya Wind Technologies, “www.ewinternational.com,” [Online].

[Accessed 30 September 2016].
[28] Vensys, “www.vensys.de,” [Online]. [Accessed 30 September 2016].
[29] Leitwind, “www.en.leitwind.com,” [Online]. [Accessed 30 September 2016].
54
[30] XEMC Darwind, www.xemc-darwind.com, [Online]. [Accessed 10 March

2017].
[31] G. Bywaters, V. John, J. Lynch, P. Mattila, G. Norton, J. Stowell, M. Salata,

O. Labath, A. Chertok and D. Hablanian, “Northern Power Systems
WindPACT drive train alternative design study report,” Report NREL/SR -
500-35524, 2004.
[32] M. A. Mueller, A. S. McDonald and D. E. Macpherson, “Structural analysis of

low-speed axial-flux permanent-magnet machines”, IEE Proceedings of
Electric Power Applications, Volume 152, No 6, pp. 1417-26, November
2005.
[33] G. Henneberger and M. Bork, “Development of a new transverse flux motor”,

in Proceedings of IEE Colloquium on New Topologies for PM Machines,
Volume 1, pp. 1-6, 1997.
[34] A. McDonald , M. Mueller and H. Polinder, “Structural mass in direct drive

permanent magnet electrical generator,” IET Renewable Power Generation
Special Issue- Selected Papers from EWEC 2007, vol. 2, pp. no. 1, 3-15,
March 2008.
[35] C. Versteegh, “Low speed direct drive PM generator for application in the
Zephyros Z72 wind turbine,” IEE Seminar on Electr. Aspects of Offshore
Renewable Energy Systems, NaREC, Blyth, Nothumberland, UK, 2004.
[36] http://www.windpowerengineering.com/design/sway-turbine-reveals-details-
10mw-offshore-design/, [Online]. [Accessed 10 March 2017].
[37] O. Keysan et al., “C-Gen, A Lightweigth Direct Drive Generator For Marine
Energy Converters,” in 5th IET PEMD International Conference, Brighton
(UK), 2010.
[38] E. Spooner, A. C. Williamson and G. Catto, “Modular design of permanent-

magnet generators for wind turbines”, IEE Proc. Electr. Power Appl.,
1996,143, (5), pp. 388-395.
[39] O. Danielsson, K. Thorburn, E. Sjöstedt, M. Eriksson and M. Leijon, M,

“Permanent magnet fixation concepts for linear generator”. Proc. of the Fifth
European Wave Energy Conf., Cork, Ireland, Sep 17-20, 2003.
[40] P. Lampola, “Directly driven, low-speed permanent-magnet generators for

wind power applications”, Ph.D. dissertation, Helsinki University of
Technology, Helsinki, 2000.
[41] A. Al-Badi, A. Gastli,H. Bourdoucen and J. Jervase, ‘Evolution of axial-field

electrical machines’. Science and Technology, Special Review (2000), Sultan
55
Qaboos University, pp. 227-245
[42] E. Spooner and B.J.Chalmers, “TORUS”: A slotless, toroidal-stator,

permanent magnet generator’, IEE Proc. B, 1992, 139, (6), pp. 497-506.
[43] F. Caricchi, F. Crescimbini, F. Mezzetti and E. Santini, ‘Multistage axial-flux

PM machine for wheel direct drive’, IEEE Trans. on Ind. Appl., 1996, 32, (4),
pp. 882-888.
[44] N. Brown, L. Haydock and J.R. Bumby, ‘Foresight vehicle: a toroidal, axial
flux generator for hybrid IC engine/battery electric vehicle applications’. Proc.
SAE Conf. paper 2002-01-0829, Detroit, USA, March 2002.
[45] F. Caricchi, F. Crescimbini and O. Honorati, ‘Modular axial-flux permanent-

magnet motor for ship propulsion drives’, IEEE Trans. on Energy Conv.,
1999, 14, (3), pp. 673-679.
[46] R. J. Hill-Cottingham et al.,‘Plastic structure multi-disc axial flux PM motor’.

Conf. Rec. of the 2002 IEEE Ind. Appl. Conf. 37th IAS Annual Meeting.
Pittsburgh, PA, USA, 13-18 Oct. 2002, 2, pp. 1274-1280.
[47] D. Bang, “Design of Transverse Flux Permanent Magnet Machines for Large
Direct-Drive Wind Turbines,” Ph.D. dissertation, Technical University of
Delft, Delft, The Netherlands, 2010.
[48] M. Dubois, “Review of electromechanical conversion in wind turbines, Final

literature review,” Technical University of Delft, Electrical Power Processing
Group, Delft, The Netherlands, 2000.
[49] M. Dubois, H. Polinder and J. Ferreira, “Comparison of generator topologies

for direct drive wind turbines,” in 2000 IEEE Nordic Workshop on Power and
Industrial Electronics, pp. 22-26, Aalborg, Denmark, June 2000.
[50] Y. Chen, P. Pillay and A. Khan, “PM wind generator comparison of differet
topologies,” in Conf. Rec. of the 2004 IEEE Ind. Appl. Conf. 39th IAS Annual
Meeting, Seattle, WA, USA, 2004.
[51] J. N. Stander, G. Venter, and M. J. Kamper, “Review of direct-drive radial

flux wind turbine generator mechanical design,” 2011.
[52] T. Burton, D. Sharpe, N. Jenkins, and E. Bossanyi, Wind Energy Handbook.

John Wiley & Sons, Ltd, 2001.
[53] IEC 61400 International Standard, Third Edition 2005-2008, www.iec.ch

[Online] [Accessed 13 March 2017].
56
[54] Germanischer Lloyd Industrial Services GmbH, Rules and Guidelines

Industrial Services IV, Renewables Certification, 1st July 2010. www.gl-
group.com/GLRenewables [Online] [Accessed 13 March 2017].
[55] Danish Standard DS 472, The Danush Energy Agency’s Approval Scheme for
Wind Turbines, www.wt-certification.dk [Online] [Accessed 13 March 2017].
[56] Alstom brochures,

www.alstom.com/Global/Power/Resources/Documents/Brochures/wind-
powersolutions.pdf, [Online] [Accessed 10 March 2017].
[57] S. Yagi and N. Ninoyu, “Technical trends in wind turbine bearings”,

Technical Report Review No. 76, NTN Corporation, Osaka, Japan, 2008.

generators for direct drive renewable energy converters,” Ph.D. dissertation,
[59] M. Mueller, A. Zavvos, and A. S. Mcdonald, “Optimization tools for large

permanent magnet generators for direct drive wind turbines,” in IET
Renewable Power Generation, vol. 7, no.2, pp. 163-171, Mar. 2013.
[60] P. J. Tavner and E. Spooner, “Light Structures for Large Low-Speed Machines
for Direct-Drive Applications.” in Proc. International Conference on
Electrical Machines, Chania, Greece, 2006.
[61] G. Bywaters, P. Mattila, D. Costin, J. Stowell, V. John, S. Hoskins, J. Lynch,

T. Cole, A. Cate, C. Badger, and B. Freeman, “Northern Power NW 1500
direct-drive generator.” Northern Power Systems Inc., Waitsfield, VT,
National Renewable Energy Laboratory, Subcontract Rep. NREL/SR-500-
40177, Oct. 2007.
[62] S. O. Lind and H. Stiesdal, “Wind Turbine,” Patent WO 2013/083386 A2,

June 2013.
57
Chapter 3
Magnetic Stiffness Modelling of

Wind Turbine Electrical Generators
3.1 Introduction
Direct drive electrical generators are low speed high torque machines whose robust
and stiff supporting structures are designed to withstand the significant loads present
during the assembly (gravitational and attraction forces) and operation stages. The
key load to be considered when designing this type of devices is a large force across
the air gap which is because of the normal component of the Maxwell Stress. Several
approaches were described by McDonald and Mueller in [1] for estimating the mass
58
of machine structures dealing with uniformly distributed force and deflection, also
known as Mode 0. By assuming the rotor and the stator structures were made up of
disc and arm sub structures, McDonald linked the electromagnetic and the
mechanical design to model radial, tangential and axial deformations in radial flux
machines [2]. In [3], Tavner and Spooner introduced a method which describes the
challenge in terms of stiffness, focusing the attention on Mode 1 deflection of rotor
and stator structures. Following that path, Jaen-Sola and McDonald presented in [4]
an electromagnetic model that can be utilised to calculate the required airgap
stiffness and therefore the stiffness of the generator structure. Going a step beyond,
the authors coupled this model with a parametric structural model, which was
produced by making use of finite element methods.
This chapter concentrates on the study of two different electrical machines under
distinct deflection modes from a static point of view. An overview of how to model
the different modes of generator structural deflection is given. The introduction and
use of the magnetic stiffness concept to characterize the behaviour of electrical
machines supporting structures is one of the main contributions made in this chapter
and the thesis. After this, in the main body of the chapter, the reader will be able to
see the development of an analytical model that links the magnetic and the
mechanical sides of the design of a conventional wound rotor synchronous generator
and a surface mounted permanent magnet generator, which has been developed
continuing the work presented in [4].
The analytical parametric model (magnetic) first assumes a deflection value, ‘δ’, that
is distributed along the outer surface of the rotor and the inner surface of the stator in
order to calculate the ultimate airgap closing force for deflection modes ranging from
0 to 4. The assumed deflection ‘δ’ is composed by a mean deflection, ‘𝛿 ̅’, and a
variable deflection, ‘δΔ’. Using the said deflection and the resultant force, the airgap
stiffness (magnetic), kM, is found dependent of δ. Finite element analysis of a two
pole model and full machine model are used afterwards to validate the analytical
models for airgap closing force and stiffness. With the analytical models
corroborated, a structural finite element model of the electrical machine is generated.
Making use of the airgap closing force computed with the analytical magnetic
59
approach, a deflection, ‘δStructural’, is obtained and used to calculate the structural

stiffness that will be compared to the already known magnetic stiffness so that the
designer can understand if the airgap of the proposed structure will remain stable and
what the margin of deformation is. See Figure 3.1.
Figure 3.1 Coupling of magnetic and structural models
As said, two different types of generators (surface mounted permanent magnet and
wound rotor) were analysed for distinct deflection modes. The achieved outcomes
revealed that a stiffer supporting structure is needed to maintain the airgap of a
wound rotor machine open and stable. In addition, it was identified that deflection
Mode 4 is the most damaging for any of the assumed structures. The conclusions
drawn from the obtained results are presented in the last section of this chapter.
3.2 Introduction to the Stiffness Concept
3.2.1 Mechanical Stiffness
In general terms, stiffness is a measure of the resistance offered by an elastic body to

a force deforming the body. The stiffness is defined as k = F/δ where F is the force
and δ is the displacement and it can be used to relate any F and δ, whereas a finite
element model of a structure only gives δ for one set of F. This concept of stiffness
can be expressed in terms of stress (the force per unit area, σ = F/A), strain (the
60
change in length divided by the original length, ε = δ/l) and major dimensions of the
body, thus,
𝐹 𝜎𝐴 (3.1)
𝑘= = .
𝛿 𝜀𝑙
Normally, one is interested in the strain response to the application of stress. A

positive value of stiffness means that as a positive force is applied, the change in
length is also positive. Occasionally, the stress itself depends on the strain. This is the
case for the magnetic forces in the airgap. Here as the airgap clearance reduces in
size (i.e. δ is negative) the magnitude of the forces become larger. Conversely, as the
airgap clearance increases in size, the magnitude of the force trying to close the
airgap reduces. In this case the stiffness is negative.
In the steady-state and with no external forces applied, stability is achieved and a
system is “stiff enough” when the sum of all the values of stiffness is equal to 0
(system is balanced). More stiffness is needed when other forces are introduced. A
detailed explanation on this statement is given in Section 3.2.3.
Most systems are made up of multiple bodies each with their own value of stiffness.
Two bodies with stiffness kA and kB can be combined into an equivalent stiffness
depending on whether they are in series (and hence experience the same force but
have different displacements),
𝑘A 𝑘B (3.2)
𝑘eq = ,
𝑘A + 𝑘B
or in parallel (and hence experience the same displacement but different applied
forces),
𝑘eq = 𝑘A + 𝑘B , (3.3)
or mixture of these two cases.
A cross section of a generator structure with a simplified structure for a direct drive
wind turbine is shown in Figure 3.2. A radial flux generator is formed by four main
61
components that in terms of stiffness are as follows: the bearing, kb, the structure of
the rotor, ks,r, the magnetic airgap stiffness, kM, and the structure of the stator, ks,s.
Combining the bearing and the rotor structure in series gives an equivalent stiffness,
𝑘s,r 𝑘b (3.4)
𝑘eq,r = .
𝑘s,r + 𝑘b
(a) (b)
Figure 3.2 (a) Generator structure (b) Shown as stiffness [4]
The magnetic attracting force acting on the rotor and stator surface also acts to
deform the rotor and stator structures. These structures have values of stiffness,
which are constant for elastic materials below the elastic limit. Equations (3.5a) and
(3.5b) express the common force in terms of stiffness and deflection,
𝐹c = 𝑘eq,r 𝛿r , (3.5a)
𝐹c = 𝑘s,s 𝛿s . (3.5b)
As they are connected to one another at the generator mounting point, and as they
have the same force applied to them both, one can consider them as two bodies with
stiffness in series, and so they can be expressed as an equivalent structural stiffness,
𝑘eq,r 𝑘s,s
𝑘s = 𝑘 . At one end of this composite structure, the force leads to rotor
eq,r +𝑘s,s
deflection into the airgap and at the other end the force leads to stator deflection,
𝐹 = 𝑘s (𝛿s + 𝛿r ) (3.6)
62
3.2.2 Magnetic Stiffness
With a force, F, caused by the normal component of the Maxwell stress, acting on
the rotor and stator surfaces, the airgap tends to close. The airgap closing force, ‘Fc’,
can be expressed in terms of a magnetic stiffness, ‘kM’, assuming a combination of a
radial mean deflection, ‘𝛿 ̅’, and a variable deflection, ‘𝛿∆ ’, which changes with angle
‘θ’, that alters the airgap clearance;
𝐹c = 𝑘M (𝛿 ̅ + 𝛿∆ sin (𝑛𝜃)) (3.7)
where n corresponds to the deflection mode and θ to the pitch angle. Figure 3.3
illustrates the airgap closing force calculated with equation (3.7) for 𝛿 ̅ = 1 mm and
𝛿∆ = 0.5 mm, at different angles for each deflection mode for kM = ks.
230
225
220
215
Force (kN)
210
205
200
195
190
185
180
175
0 20 40 60 80 100 120 140 160 180 200 220 240 260 280 300 320 340 360
θ (degrees)
Mode 0 Mode 1 Mode 2 Mode 3 Mode 4
Figure 3.3 Airgap Closing Force vs. Theta
Note that in the figure, the area of the rim is apportioned into 36 parts and that the
force for each span of β = 10 degrees is shown. Table 3.1 illustrates the
characteristics of the machine used in the analysis in this chapter. It is based on the
direct drive permanent magnet machine in [4].
63
Table 3.1 Case study generator data
Generator data
Axial length, ‘ls’ (m) 1.2
Rotor radius, ‘R’ (m) 2.5
Design airgap size, ‘𝑔’ (m) 0.005
Rotor yoke height, ‘hry’ (m) 0.05
Aspect ratio (proportional relationship between 0.6
width and height)
Magnet height, ‘hm’ (m) 0.01

Magnet width, ‘wm’ (m) 0.15
Number of pole pairs, ‘p’ 88
Pole pitch, ‘τp’ (m) 0.18
The force caused by the normal component of Maxwell stress will be calculated
analytically in Section 3.3. So as to develop this model, it was considered the
effective magnetic airgap clearance. This means that the changes in stiffness are
correlated to the alterations in the size of the airgap. Therefore, expressions
describing the airgap behaviour of electrically excited wound rotor machines and
permanent magnet generators needed to be produced. Equation (3.8) is suitable to
compute the airgap stiffness for electrically excited generators,
𝐹 (3.8)
𝑘WR =
𝑔−𝛿
where 𝑔 is the airgap size, whereas equation (3.9) should be used in the case of
having a permanent magnet machine (with surface mounted magnets),
𝐹 (3.9)
𝑘PM =
ℎ
𝑔 + 𝜇m − 𝛿
r
where ℎm is the height of the magnet and 𝜇r is the relative permeability of the
magnetic material. By introducing these two parameters into the equation, the fact of
having surface mounted permanent magnets can be considered. To evaluate the
64
stiffness of both types of machines for different deflection modes, δ can be

substituted by 𝛿 ̅ + 𝛿∆ sin (𝑛𝜃) , which would give us the following
𝐹
𝑘WR =
𝑔 − 𝛿 ̅ − 𝛿∆ sin (𝑛𝜃) (3.8a)
𝐹 (3.9a)
𝑘PM =
ℎ
𝑔 + 𝜇m − 𝛿 ̅ − 𝛿∆ sin (𝑛𝜃)
r
3.2.3 Overall Stiffness
As said, in order to keep the integrity of the electrical machine, the airgap must
remain open and stable. For this, it is necessary that the magnetic force, ‘FM’ and the
structural force, ‘Fs’, are equal and opposite. No external forces has been considered.
If equation (3.7) is manipulated, in equation (3.11), it can be seen that the equivalent
structural stiffness of the system must be equal (and opposite in sign) to the airgap
magnetic stiffness:
𝐹M + 𝐹s = 0 → 𝑘M (𝛿s + 𝛿r ) + 𝑘s (𝛿s + 𝛿r ) = 0 , (3.10)
𝑘s = −𝑘M . (3.11)
Where the structural stiffness, ‘ks’, will be calculated using finite element techniques.
Further methods to estimate the structural stiffness of the electrical generator will be
described in the next chapter.
3.3 Magnetic Airgap Stiffness
The concept of magnetic airgap stiffness was introduced in Section 3.2. The need for
a versatile quick model that can accurately predict the required magnetic stiffness in
several dimensions for different types of machines has led the author to create a 2
dimensional parametric model that can be used for optimization purposes. Equations
65
will be developed here for the airgap closing force per unit area as a function of
deflection and angle for both wound rotor and surface mounted permanent magnet
machines. Finally the formulations for the magnetic airgap stiffness will be
developed.
As explained in Chapter 2, deflection can be different at distinct zones of rotor and

stator [3]. Airgap collapse can take place due to:
Mode 0: Radial expansion of the rotor or radial compression of the stator.
Mode 1: Rotor eccentricity (localized deflection).
Mode 2: Distortion of either or both of the circular surfaces into ellipses.
Mode 3: Distortion with ripples around the circumferences.
The magnetic airgap stiffness expressions for both the wound rotor and the surface
mounted permanent magnet machines will be derived attempting to address all of
these scenarios.
3.3.1 Airgap closing force per unit area
Magnetic airgap stiffness arises because of the influence of the airgap clearance on
the airgap permeance and hence airgap flux density. This in turn affects the airgap
closing force. In the case of the airgap closing, the flux density increases and the
force per unit area increases. This airgap closing force can be found from the normal
component of Maxwell stress σ with equation (3.12), where B is the airgap flux
density,
𝐵2 (3.12)
𝜎= ,
2𝜇0
and μ0 is the permeability of free space.
The flux density distribution, B, in the airgap can be found as follows,
𝑃(𝜃) (3.13)
𝐵(𝜃) = F (𝜃) ,
𝐴
66
where F (θ) is the MMF set up by the rotor field (winding or magnets) and armature
windings current and P(θ)/A is the magnetic permeance per unit area. This flux
density distribution can be found for a generic machine. Having a pole number of 2p
the main airgap MMF is assumed to be sinusoidally distributed,
̂ cos(𝑝𝜃 − 𝜑).
F (𝜃) = F (3.14)
Although the airgap flux density distribution for a surface mounted permanent
magnet machine is often more akin to a square or quasi-square wave, equation (3.14)
̂ PM = 𝐵r ℎm 4 sin (𝜋 𝑤m ), where hm is the

is normally a good approximation with F 𝜇 𝜇 𝜋 2 𝜏0 r p
magnet height, wm is the magnet height, τp is the pole pitch and μr is relative
permeability. It should be noted that because the surface-mounted permanent magnet
machine has a larger airgap permeance than conventional salient pole synchronous
machines, the MMF per pole will be higher to produce the same flux density
(assuming the same number of poles, rating and airgap geometry). Indeed equation
(3.13) suggests that for the same airgap flux density, the ratio of MMFs approximates
̂ PM
F 𝑃
to ≈ 𝑃 . This is because the final part of equation (3.13) is the magnetic
F̂ PM
permeance of the airgap. This can be defined in general terms as,
𝑃 𝜇0 (3.15)
= ,
𝐴 𝑙
where A and l are the cross sectional area and length of the region in question.
Assuming that the iron in the magnetic circuit is infinitely permeable and ignoring
slots then the magnetic permeance reduces to the permeance of the airgap, and l = 𝑔.
Before any deflection occurs, the ratio of magnetic permeance of the airgap of the
𝑃PM 𝑔
two machines would be ≈ ℎ .
𝑃 𝑔+ m
𝜇r
As the deflection occurs the local airgap changes with the circumferential angle, θ,
according to,
𝑔(𝜃) = 𝑔 − 𝛿(𝜃) = 𝑔 − 𝛿 ̅ − 𝛿∆ sin(𝑛𝜃 − 𝜑), (3.16a)
67
ℎm ℎm (3.16b)
𝑔eff,PM (𝜃) = 𝑔 + − 𝛿(𝜃) = 𝑔 + − 𝛿 ̅ − 𝛿∆ sin(𝑛𝜃 − 𝜑),
𝜇r 𝜇r
where 𝑔 and 𝑔eff,PM are the nominal airgap clearance. The permeance per unit area
can be approximated as,
𝑃 (𝜃) (3.17)
≈ 𝑃̅ + 𝑃∆ sin(𝑛𝜃 − 𝜑) ,
𝐴
where 𝑃̅ is the mean value of airgap permeance per unit area and 𝑃∆ is the amplitude
of variation of the airgap permeance per unit area. If 𝛿Δ2 terms are neglected, then
𝜇0 𝜇0
𝑃̅ ≈ 𝑔−𝛿̅ and 𝑃∆ ≈ 𝛿 .
̅ )2 ∆
For a surface mounted permanent magnet machine, the
(𝑔−𝛿
magnetic airgap and airgap clearance are no longer one and the same; the mean and
𝜇0
amplitude permeance per unit area terms become 𝑃̅ ≈ and 𝑃∆ ≈
𝑔+ℎ𝜇m −𝛿
̅
r
𝜇0
2 𝛿∆ .
(𝑔+ℎ𝜇m −𝛿
̅)
r
Figure 3.4 shows a comparison between the outcomes achieved for the magnetic
permeance per unit area calculated using equations (3.15) and (3.17) for a permanent
magnet machine.
Permeance per unit area (H/m2)
9.80E-05
9.60E-05
9.40E-05
9.20E-05
9.00E-05
8.80E-05
0 20 40 60 80 100 120 140 160 180 200 220 240 260 280 300 320 340 360
θ (degrees)
Magnetic permeance Magnetic permeance approximation
Figure 3.4 Magnetic permeance per unit area comparison. Magnetic permeance per unit area vs.
Magnetic permeance per unit area approximation assuming infinite permeability for the back iron and
ignoring slots.
68
So as to evaluate the goodness of fit of the magnetic permeance approximation to the

magnetic permeance, the average root-mean-square error (RMSE) was calculated
giving a value of 7.77562×10-8. The RMSE corresponds to the sample standard
deviation of the differences between the approximated values acquired with the
model and the observed values and it is considered a good measure of accuracy
statistically speaking. The normalized root-mean-square-error was also computed
revealing a value of 1.13 % of residual variance. Thus, the permeance
approximation calculated using equation (3.17) was treated as valid.
Substituting equations (3.14) and (3.17) into (3.13) and assuming that φ is changed
so that peak deflection is at θ = π/2, then it is found that
B(𝜃) = F̂ cos(𝑝𝜃)(𝑃̅ + 𝑃∆ sin(𝑛𝜃)) (3.18a)
Equation (3.18b) is the corresponding equation but for the permanent

magnet machine.
B(𝜃) = F̂ PM cos(𝑝𝜃)(𝑃̅ + 𝑃∆ sin(𝑛𝜃)) (3.18b)
Here there are two spatial frequencies; a high frequency, p, corresponding to the pole
pairs and a lower frequency, n, corresponding to the mode of deflection. By
substituting (3.18a) into (3.12) and rearranging and noting that in the case of many
pole pairs, the variation in force distribution due to poles (i.e. the compared cos(pθ)),
becomes less significant for structural deflections, then the mean value of cos2(pθ) is
½ and so the stress distribution can be simplified as
̂ 2 cos2 (𝑝𝜃)𝜇0
F 2𝛿∆ sin(𝑛𝜃) 𝛿∆2 sin2 (𝑛𝜃) ̂ 2 𝜇0
F (3.19a)
𝜎(𝜃, 𝛿 ̅, 𝛿∆ ) = ̅ )2
[1 + 𝑔−𝛿 ̅ + ̅ )2
] ≈ ̅ )2
[1 +
2(𝑔−𝛿 (𝑔−𝛿 4(𝑔−𝛿
2𝛿∆ sin(𝑛𝜃) 𝛿∆2 sin2 (𝑛𝜃)
̅ + ̅ )2
] .
𝑔−𝛿 (𝑔−𝛿
For Mode 0 (n = 0), equation (3.19) becomes (3.20),
̂ 2 cos 2 (𝑝𝜃)𝜇0
F ̂ 2 𝜇0
F (3.19b)
𝜎𝑛=0 (𝜃, 𝛿 ̅) = 2 ≈ 2
2(𝑔 − 𝛿 ̅) 4(𝑔 − 𝛿 ̅)
69
For a surface mounted permanent magnet machine, the equivalent of equation

(3.19a) becomes 3.20a,whereas 3.19b becomes 3.20b for Mode 0,
̂ PM 2 cos2 (𝑝𝜃)𝜇0
F 2𝛿∆ sin(𝑛𝜃) 𝛿∆2 sin2 (𝑛𝜃)
𝜎PM (𝜃, 𝛿 ̅, 𝛿∆ ) = 1+ + (3.20a)
ℎm 2 ℎ 2
̅
2 (𝑔 + 𝜇 − 𝛿 ) [ 𝑔 + 𝜇m − 𝛿 ̅ (𝑔 + ℎm − 𝛿 ̅)
r r 𝜇r ]
̂ PM 2
F 2𝛿∆ sin(𝑛𝜃) 𝛿∆2 sin2 (𝑛𝜃)
≈ 1+ +
ℎm 2 ℎ 2
̅
4 (𝑔 + 𝜇 − 𝛿 ) [ 𝑔 + 𝜇m − 𝛿 ̅ (𝑔 + ℎm − 𝛿 ̅)
r r 𝜇r ]
̂ PM 2 cos2 (𝑝𝜃)𝜇0
F ̂ PM 2 1⁄ 𝜇0
F
𝜎PM (𝜃, 𝛿 ̅, 𝛿∆ ) = = 2
2 2
ℎ ℎ (3.20b)
2 (𝑔 + 𝜇m − 𝛿 ̅) 2 (𝑔 + 𝜇m − 𝛿 ̅)
r r
̂ PM 2 𝜇0
F
≈ 2
ℎ
4 (𝑔 + 𝜇m − 𝛿 ̅)
r
Figure 3.5 illustrates how the magnetic stress varies with angle for different
deflection modes in a permanent magnet electrical machine.
5.40E+05
5.20E+05
Magnetic stress (Pa)
5.00E+05
4.80E+05
4.60E+05
4.40E+05
4.20E+05
0 20 40 60 80 100 120 140 160 180 200 220 240 260 280 300 320 340 360
θ (degrees)
Figure 3.5 Magnetic stress vs. Theta for different deflection modes
70
To find the force on the rotor or stator surface closing the airgap, equation (3.19a)
can be integrated over the axial length of the machine, ls, and over any angle, ‘β’. To
find the force over for an angle β, we can integrate half an angle either side of the
value of θ. For a wound rotor machine, the radial force on an arc of span β centred at
angle θ for Modes 1, 2, 3 and 4 is as follows,
𝜃+𝛽
2 ̂ WR 2 𝑙s 𝑅𝜇0
F (3.21a)
𝐹WR = 𝑙s 𝑅 ∫ 𝜎(𝜃, 𝛿 ̅, 𝛿∆ , 𝛽)𝑑𝜃 = 4 𝛼WR
𝜃−𝛽
2 4(𝑔 − 𝛿 ̅)
with 𝛼WR equal to
1 𝛿∆ 𝑛 𝑛 (3.21b)
𝛼WR = [2𝛿∆ (𝑔 − 𝛿 ̅ − sin ( (𝛽 − 2𝜃))) cos ( (𝛽 − 2𝜃))
𝑛 4 2 2
𝛿∆ 𝑛 𝑛
− 2𝛿∆ (𝑔 − 𝛿 ̅ + sin ( (𝛽 + 2𝜃))) cos ( (𝛽 + 2𝜃))
4 2 2
2
𝛿∆2
+ 𝛽𝑛 ( + 𝑔 − 𝛿 ̅) ].
2
While for Modes 0 is
̂ WR 2 𝑙s 𝑅𝜇0
F (3.21c)
𝐹WR = 2 𝛽
4(𝑔 − 𝛿 ̅)
For a surface mounted PM machine, the radial force on an arc of span β centred at
angle θ for Modes 1, 2, 3 and 4 is
𝜃+𝛽
2
(3.22a)
𝐹 = 𝑙s 𝑅 ∫ 𝜎(𝜃, 𝛿 ̅, 𝛿∆ , 𝛽)𝑑𝜃
𝜃−𝛽
2
̂ PM 2 𝑙s 𝑅𝜇0
F 𝜃+𝛽
2 2𝛿∆ sin(𝑛𝜃)
= 2∫ 𝛽 1+
ℎ ℎ
4 (𝑔 + 𝜇m − 𝛿 ̅) 𝜃− 2 𝑔 + 𝜇m − 𝛿 ̅
r [ r
𝛿∆2 sin2 (𝑛𝜃) ̂ PM 2 𝑙s 𝑅𝜇0

F
+ 2 𝑑𝜃 = 4 𝛼PM
ℎm ̅ ℎm ̅
(𝑔 + 𝜇 − 𝛿 ) ] 4 (𝑔 + 𝜇 − 𝛿 )
r r
71
with 𝛼PM being
1 ℎm 𝛿∆ 𝑛 𝑛 (3.22b)
𝛼PM = [2𝛿∆ (𝑔 + − 𝛿 ̅ − sin ( (𝛽 − 2𝜃))) cos ( (𝛽 − 2𝜃))
𝑛 𝜇r 4 2 2
ℎm
− 2𝛿∆ (𝑔 + − 𝛿̅
𝜇r
𝛿∆ 𝑛 𝑛
+ sin ( (𝛽 + 2𝜃))) cos ( (𝛽 + 2𝜃))
4 2 2
2
𝛿∆2 ℎm
+ 𝛽𝑛 ( + 𝑔 + − 𝛿 ̅) ].
2 𝜇r
Whereas for Mode 0 is
𝜃+𝛽
2
(3.22c)
𝜃+𝛽 2 2
2 ̂ PM 𝜇0
F ̂ PM 𝑙s 𝑅𝜇0
F
𝐹 = 𝑙s 𝑅 ∫ 2 𝑑𝜃 = 2𝜃
𝜃−𝛽 ℎ ℎm
2 4 (𝑔 + 𝜇m − 𝛿 ̅) ̅
[4 (𝑔 + 𝜇r − 𝛿 ) ]𝜃−𝛽
r
2
̂ PM 2 𝑙s 𝑅𝜇0
F
= 2 𝛽
ℎ
4 (𝑔 + 𝜇m − 𝛿 ̅)
r
The magnetic stiffness of the same arc for Modes 1, 2, 3 and 4 of a wound rotor
machine can be calculated using the following equation
𝐹 ̂ WR 2 𝑙s 𝑅𝜇0
F (3.23a)
𝑘WR = = 𝛼WR.
𝑔 − (𝛿 ̅ + 𝛿∆ sin(𝑛𝜃)) 4
4(𝑔 − 𝛿 ̅) (𝑔 − 𝛿 ̅ − 𝛿∆ sin(𝑛𝜃))
For Mode 0, the stiffness would be computed using equation (3.23b)
𝐹 ̂ WR 2 𝑙s 𝑅𝜇0 𝛽
F (3.23b)
𝑘WR = 𝛿 = ̅ )3
.
4(𝑔−𝛿
The magnetic stiffness of the said arc is for Modes 1, 2, 3 and 4 of a PM machine
equal to,
72
𝐹 (3.24a)
𝑘PM =
𝑔 − (𝛿 ̅ + 𝛿∆ sin(𝑛𝜃))
F̂ PM 2 𝑙s 𝑅𝜇0
= 4 𝛼PM
ℎ
4 (𝑔 + 𝜇m − 𝛿 ̅) (𝑔 − 𝛿 ̅ − 𝛿∆ sin(𝑛𝜃))
r
whereas for Mode 0 is
𝐹 ̂ PM 2 𝑙s 𝑅𝜇0 𝛽
F (3.24b)
𝑘PM = = 3
𝛿 ℎ
4 (𝑔 + 𝜇m − 𝛿 ̅) .
r
3.3.2 Validation using finite element code
Finite element analysis of a two pole model was used to validate the analytical
models for airgap closing force and stiffness. Applying correct periodic boundaries
this type of model can represent the magnetic flux in the magnetic circuit in a more
realistic manner than the simplified lumped parameter approach of the analytical
solution. For instance, the analytical solution effectively ignores MMF drops in the
iron yokes and teeth and neglects leakage paths but these can be captured in the finite
element model.
In this case, a 2D code was utilized (FEMM[5]) and so axial end effects are
neglected. To make the model more versatile, the two pole model has been
geometrically linearized so that radial lines and arcs are mapped onto vertical and
horizontal lines respectively. By changing the airgap clearance by a deflection δ a
number of magnetostatic runs were processed and the results were interrogated to
find the airgap closing force. Using a force via weighted stress tensor approach the
obtained results are shown in Figure 3.6(a)-(c) where the force is plotted against
deflection δ, the resulting physical airgap clearance (𝑔 − 𝛿), and the resulting
ℎm
effective magnetic airgap clearance (𝑔 + − 𝛿) respectively.
𝜇r
73
250
Airgap closing force on one pole pair (kN)

200
150
100
No slots, linear BH
Slots, linear BH
50 Slots, non-linear BH
Analytical model
Analytical with 3rd harmonic
Analytical with 3rd harmonic and Carter factor
0
0 0.001 0.002 0.003 0.004 0.005
Airgap clearance (m)
(a)
250
200
150
100
No slots, linear BH
Slots, linear BH
Slots, non-linear BH
50
Analytical model
0
0 0.001 0.002 0.003 0.004 0.005
Change in airgap clearance (m)
(b)
74
250

200
150
No slots, linear BH
100
Slots, linear BH
Slots, non-linear BH
50 Analytical model
0
0.0000 0.0020 0.0040 0.0060 0.0080 0.0100 0.0120 0.0140 0.0160
Magnetic effective airgap (m)
(c)
Figure 3.6 Magnetic airgap stiffness for a pole pair of a PM generator for a direct drive wind turbine,
based on [3]. (a) Airgap closing force on one pole pair vs. Airgap clearance (b) Airgap closing force
on one pole pair vs. Change in airgap clearance (c) Airgap closing force on one pole pair vs. Magnetic
effective airgap
There are 3 FE cases: (i) where the materials are assumed to have linear BH curves
and the stator has no slots (similar to the analytical model); (ii) where slotting is
introduced but the materials have liner BH curves and (iii) where slotting is present
and non-linear BH curves are used.
Also on Figure 3.6(a)-(c) one can see the analytical solutions for the same
dimensions and materials using equation (3.22a) for 𝛿∆ = 0 and 𝛿 ̅ = 𝛿 and one pole
2𝜋
pair (i.e. 𝛽 = ).
𝑝
The analytical model clearly underestimated the force; this can be seen when
comparing the results with those of the idealized FE model (i). This suggested that
using only the fundamental MMF as an input to the analytical solution is incorrect as
it leads to the model neglecting higher order airgap flux density spatial harmonics
75
and the resulting force contributions. The MMF of a surface mounted permanent
magnet of width 𝑤m can be written as:
∞
𝐵r ℎm 4 𝑚𝜋 𝑤m (3.25)
F m (𝜃h ) = ∑ sin ( ) sin(𝑚𝜃h )
𝜇0 𝜇r 𝑚 2 𝜋 2 𝜏p
𝑚=1,2,3,…
where m is the harmonic order. Including m = 1 and m = 3 in equations (3.21)-(3.22)

leads to amended force equation of
2 2
̂ +F
(F ̂ 3 )𝑙s 𝑅𝜇0 (3.26)
F≈ 1
ℎ ̅ )2
𝛼PM .
4(𝑔+ m −𝛿
𝜇r
This has been plotted in Figure 3.6(a)-(c). This shows better agreement with the
idealized FE results (i). When slotting is introduced (FE models (ii) and (iii)) there is
a noticeable reduction in force. The analytical model ignores the reduction in
permeance due to slotting. This can be taken into account by applying the Carter
factor 𝑘cr =1.18, to the effective magnetic airgap, in other words,
2 2
̂ +F
(F ̂ 3 )𝑙s 𝑅𝜇0 (3.27)
F≈ 1
2 (𝑔+ℎm −𝛿
̅ )2
𝛼PM .
4𝑘cr
𝜇r
The Carter factor was calculated using equation 3.28,
𝜏s (3.28)
𝑘cr =
𝜏s − 𝑔1 𝛾
ℎm
where the slot pitch, 𝜏s = 0.06 m, the slot width, 𝑤s = 0.03 m, 𝑔1 = 𝑔 + and
𝜇r
4 𝑤 𝑤 𝑤 2
𝛾 = 𝜋 (2𝑔s 𝑎𝑟𝑐𝑡𝑎𝑛 (2𝑔s ) − 𝑙𝑛√1 + (2𝑔s ) ).
1 1 1
The results for equation (3.27) are plotted in Figure 3.6(a)-(c) and show good
agreement with the FE models (ii) and (iii). The difference between linear and non-
linear materials is relatively modest, if the magnetic circuit is designed to avoid
saturation in the default state.
76
Making use of the same data, a full model of the permanent magnet machine was
generated in FEMM for validation. Assuming a linear B-H relationship for the
NdFeB magnets and the M19 Steel parts, five different studies were carried out
under Mode 1 deflection. As seen no coils have been included in the analysis as it
was assumed that the stator electrical connection does not have any influence in the
magnetic stiffness result. By looking at Figure 3.7(a), one can see how the north and
south poles are clearly differentiated by green arrows pointing up and down that
specify the direction of the magnetic flux. An adaptive mesh focused on the airgap
area with 514,225 nodes was created for the evaluation.
(a)
77
(b)
Figure 3.7 PM generator full model validation; (a) Detailed view of the triangular mesh; (b) Density
plot showing the behaviour of the magnetic flux at a particular moment in time
As said, the analysis comprised studies under Mode 1 deflection. To simulate the
eccentricity inherent to Mode 1, the rotor structure was shifted horizontally to the left
so that the airgap clearance is reduced by a deflection δ. Again, a force via weighted
stress tensor approach was utilized. The results achieved are displayed in Figure 3.8,
where the airgap closing force is plotted against deflection δ. The analytical solutions
can be obtained for the same dimensions and materials using equation (3.27) with
𝛿∆ = 𝛿 and 𝛿 ̅ = 0.
78
0
0 1 2 3 4 5
-200
Airgap closing force (kN)

-400
-600
-800
-1000
-1200
-1400
Deflection (mm)
Analytical model FEMM model
Figure 3.8 Comparison between analytical model and FEMM model (Mode 1 deflection)
As observed, the analytical model slightly overestimates the airgap closing force.
Nevertheless, the level of agreement is considered good overall, which means that
the analytical model can be used with high level of confidence.
3.4 Case study generator
Magnetic stiffness and structural stiffness are brought together in this section in order
to study a 3 MW wind turbine generator. By assuming a deflection, the closing force
acting on the airgap can be estimated making use of the magnetic model and utilized
to calculate the structural deflection through a structural model. The characteristics
of the electrical machine used in this analysis are displayed in Table 3.1 located in
Section 3.2.2.
As explained, the required stiffness for the generator structure can be computed in
different ways. In this case a structural finite element model of the generator was
created in SolidWorks.
For the FE model, the rotor and the stator structures were loaded with radial stresses
which were calculated as explained in the previous sub section using the data
presented in Table 3.1 and a mean deflection, ‘𝛿 ̅’, of 0.001 m and variable deflection,
79
‘𝛿∆ ’, of 0.0005 m. The cylindrical sub structures of both the rotor and the stator were
apportioned into 36 parts so that the appropriate forces corresponding to Modes 0, 1,
2, 3 and 4 could be applied. With this, the deflection experienced by the structure
was found allowing the structural stiffness of the generator to be evaluated.
The structural radial deflection of each 10 degrees part was measured making use of
deflection sensors located on the outer face of each part in the case of the rotor and
on the inner face in the case of the stator. See Figure 3.9.
Figure 3.9 Measured structural deflection of rotor
The data retrieved from the structural FE analysis are presented in Figure 3.10.
80
4.50E-01
4.00E-01
3.50E-01
Deflection (mm)
3.00E-01
2.50E-01
2.00E-01
1.50E-01
1.00E-01
5.00E-02
0.00E+00
0 20 40 60 80 100 120 140 160 180 200 220 240 260 280 300 320 340 360
θ (degrees)
Figure 3.10 Rotor structural deflection (m) vs. Theta (degrees)
It can be appreciated how the structural deformation of the rotor varies according to
the mode number (Figure 3.11). If the deflections obtained are plotted against the
applied forces, it can be seen that the same gradient (structural stiffness) is
maintained throughout the whole range although larger spread is found as the mode
number goes up. This increasing might be attributed to the increase in the structural
shear stresses with the number of ripples. See Figure 3.11(a)-(e).
81
0.5 0.5
Deflection (mm)
Deflection (mm)
0.4 0.4
0.3 0.3
0.2 0.2
0.1 0.1
0 0
190000 200000 210000 220000 230000 190000 200000 210000 220000 230000
Force (N) Force (N)
(a) (b)
0.5
0.5
Deflection (mm)
Deflection (mm)
0.4 0.4
0.3 0.3
0.2 0.2
0.1 0.1
0 0
190000 200000 210000 220000 230000 190000 200000 210000 220000 230000
Force (N) Force (N)
(c) (d)
0.5
Deflection (mm)
0.4
0.3
0.2
0.1
0
190000 200000 210000 220000 230000
Force (N)
(e)
Figure 3.11 Rotor structure deflection vs. Applied Force; (a) Mode 0; (b) Mode 1; (c) Mode 2;
(d) Mode 3; (e) Mode 4.
The typical oval shapes shown in the deflection diagrams above were obtained for
the corresponding sinusoidal stress applied to the rotor structure according to the
deflection mode. As only for Mode 0, a unique uniform load is applied, 36 different
finite element studies using 36 distinct loads had to be carried out in order to acquire
the straight line displayed in Figure 3.11(a) that lets us know the stiffness. For the
82
rest of the modes only one study was needed. The loads utilized in the study are as
illustrated in the figure above and depend on the deflection mode. So as to know if
the structure will be able to resist the load, the absolute value of the magnetic
stiffness estimated using the analytical model must be equal or smaller than the
structural stiffness calculated with the finite element study. This means that 𝑘s (𝜃) ≥
|𝑘M (𝜃)|. Combining the stiffnesses of the bearing, the rotor and the stator in series
as explained is Section 3.2.1, the total stiffness of the generator structure is assessed.
The bearing stiffness is assumed constant with a value of 3×109 N/m. In order to
calculate the equivalent stiffness for each 10 degrees part, the bearing has been
modelled as a finite number of radial stiffnesses set in parallel as shown in Figure
3.12, where kr corresponds to the radial stiffness, γ is the angle between stiffnesses
and N is the total number of radial stiffnesses, in our case 36.
Figure 3.12 Bearing model showed as stiffness
Paying special attention to Mode 1, where a force, F, is applied to the top of the
structure generating a deflection δ, which gives a stiffness kb = F/ δ, the bearing has
been split into a top structure and a bottom structure as depicted in Figure 3.13(a)-
(b).
83
(a) (b)
Figure 3.13 Bearing structure split into top and bottom parts; (a) top part; (b) bottom part
The total stiffness of the top structure, 𝑘T , can be calculated using equation 3.29a,
where the vertical components of all the radial stiffnesses in the top structure, in this
case 18, are added up. Similarly, the stiffness of the bottom structure, 𝑘B , can be
estimated making used of equation 3.29b. It is assumed that under Mode 1
deflection, the top bearing structure is under tension, whereas the bottom part works
under compression.
𝑖=𝑁⁄2
𝑘T = ∑ 𝑘r sin(𝑖γ) (3.29a)
𝑖=1
𝑖=𝑁
𝑘B = ∑ 𝑘r sin(𝑖γ − π)
(3.29b)
𝑖=𝑁⁄2+1
Thus, the total stiffness, ‘kb’, can be computed as follows,
𝑖=𝑁/2 𝑖=𝑁
𝑘b = 𝑘T − 𝑘B = ∑ 𝑘r sin(𝑖γ) − ∑ 𝑘r sin(𝑖γ − 𝜋) (3.30)

𝑖=1 𝑖=𝑁⁄2+1
If we know that,
sin(𝑖γ − 𝜋) = − sin(−𝑖γ + 𝜋) = − sin(−𝑖γ) = −sin(𝑖γ) (3.31)
84
then
𝑁
𝑖= 𝑖=𝑁
2
𝑘b = 𝑘T − (−𝑘B ) = ∑ 𝑘r sin(𝑖γ) + ∑ 𝑘r sin(𝑖γ − 𝜋) (3.32)

𝑖=1 𝑖=𝑁⁄2+1
𝑖=𝑁
= 𝑘𝑟 ∑ sin(𝑖γ)
𝑖=1
Finally, considering the identity
γ 1
𝑖=𝑁 cos (2) − cos ((𝑁 + 2) γ)
∑ sin(𝑖γ) = (3.33)
γ
𝑖=1 2sin (2)
and rearranging equation 3.32, it can be obtained that the stiffness for each 10
degrees section is
𝜋
𝑘b sin (𝑁)
𝑘r = (3.34)
𝜋 𝑁+1
cos (𝑁) − cos (( 𝑁 ) 𝜋)
In this case, with N equal to 36, kr shows the value of 1.31 × 108 N/m. Figure 3.14
and 3.15 present the results acquired for the rotor and the stator structures utilizing
the FE model.
8.00E+08
Rotor structural stiffness (N/m)
7.00E+08
6.00E+08
5.00E+08
4.00E+08
3.00E+08
2.00E+08
1.00E+08
0.00E+00
0 20 40 60 80 100 120 140 160 180 200 220 240 260 280 300 320 340 360
θ (degrees)
Figure 3.14 Rotor structural stiffness for deflection modes ranging from 0 to 4 vs. Theta
85
6.00E+08
Stator structural stiffness (N/m)

5.00E+08
4.00E+08
3.00E+08
2.00E+08
1.00E+08
0.00E+00
0 20 40 60 80 100 120 140 160 180 200 220 240 260 280 300 320 340 360
θ (degrees)
Figure 3.15 Stator structural stiffness vs. Theta
As seen, not all the waves are sinusoidal although they are periodic. Distortions are
more pronounced as the mode number increases. This might be again attributed to
the alterations in the structural shear stresses with the number of ripples. It can also
be appreciated how the rotor structure is stiffer than the stator structure. This is
because the structural geometry of the rotor, with the disc acting as a radial support,
allows coping with the radial deflection more effectively than that of the stator,
where the discs are placed at the edges of rim. Figure 3.16 gives the results of
combining the structures in series as mentioned. If these data are compared to those
achieved for the necessary magnetic stiffness introduced in Figure 3.17, it can be
seen that the structural stiffness is higher than the magnetic stiffness.
86
9.20E+07
Generator Structural Stiffness (N/m)

9.00E+07
8.80E+07
8.60E+07
8.40E+07
8.20E+07
8.00E+07
7.80E+07
0 20 40 60 80 100 120 140 160 180 200 220 240 260 280 300 320 340 360
θ (degrees)
Figure 3.16 Generator structural stiffness vs. Theta
The results achieved for the generator structural stiffness vary about 8.7×107 N/m,
whereas the magnetic stiffness for each 10 degrees β section is always around -2×107
N/m. As observed in Figure 3.17, the sinusoidal waves describing the magnetic
stiffness for each mode are asymmetric. This means that the straight line representing
Mode 0 does not go through the inflexion points of the rest of the modes and it is
because of the extra factor, ‘𝛿 ̅ + 𝛿∆ sin(𝑛𝜃)’, added to the denominator of the
stiffness equation for the rest of the modes.
0.00E+00
0 20 40 60 80 100 120 140 160 180 200 220 240 260 280 300 320 340 360
Magnetic Stiffness (N/m)
-5.00E+06
-1.00E+07
-1.50E+07
-2.00E+07
-2.50E+07
-3.00E+07
θ (degrees)
Figure 3.17 Stiffness on beta degree section
87
In Figure 3.19 the combination of magnetic stiffness and structural stiffness is plotted
against the angle for each deflection mode. Four distinct scenarios have been
analysed in order to identify the mode giving the lowest value. With 𝛿 ̅ going from 1
mm up to 4 mm in steps of 1 mm and 𝛿∆ taking 0.5 mm, 0.75 mm and 1 mm values,
all the options have been addressed. See Table 3.2. Since the structure geometry and
material are the same, it was assumed, for simplicity, that the generator structural
stiffness remains invariant with angle, ‘θ’, and mode, ‘n’, for all scenarios, although
as it is shown in Figure 3.18, it does depend on the angle and on the deflection mode.
As the variation is relatively small and it would only represent a slight alteration in
the final result, the assumption was considered valid. Figure 3.18 illustrates the
structural stiffness for all the modes and cases 1 to 4. As seen, an average value for
the structural stiffness is about 6.2×108 N/m for all cases. Nevertheless, it is also
important to highlight the fact that the stiffness varies with angle, ‘θ’, and that the
higher the mode, the higher the amplitudes obtained. Mode 4 shows the most
unstable behaviour with the lowest stiffness at 4.73×108 N/m, as it can be observed
in Table 3.2, where the lowest stiffnesses acquired for each mode and case are
illustrated. Peaks obtained in Figure 3.18(a)-(e) are because of approximation.
7.00E+08
Structural stiffness (N/m)
6.00E+08
5.00E+08
4.00E+08 Case 1
3.00E+08 Case 2
Case 3
2.00E+08
Case 4
1.00E+08
0.00E+00
0 60 120 180 240 300 360
θ (degrees)
(a)
88
9.00E+08

8.00E+08
7.00E+08
6.00E+08
5.00E+08 Case 1
4.00E+08 Case 2
3.00E+08 Case 3
2.00E+08
Case 4
1.00E+08
0.00E+00
0 60 120 180 240 300 360
θ (degrees)
(b)
9.00E+08
8.00E+08
7.00E+08
6.00E+08
5.00E+08 Case 1
4.00E+08 Case 2
3.00E+08
Case 3
2.00E+08
1.00E+08 Case 4
0.00E+00
0 60 120 180 240 300 360
θ (degrees)
(c)
1.00E+09
9.00E+08
8.00E+08
7.00E+08
6.00E+08 Case 1
5.00E+08
4.00E+08 Case 2
3.00E+08 Case 3
2.00E+08
1.00E+08 Case 4
0.00E+00
0 60 120 180 240 300 360
θ (degrees)
(d)
89
1.20E+09

1.00E+09
8.00E+08
Case 1
6.00E+08
Case 2
4.00E+08 Case 3
2.00E+08 Case 4
0.00E+00
0 60 120 180 240 300 360
θ (degrees)
(e)
Figure 3.18 Rotor structural stiffness; (a) Mode 0; (b) Mode 1; (c) Mode 2; (d) Mode 3; (e) Mode 4
Table 3.2 gives the minimum structural stiffness per case (different 𝛿 ̅ and 𝛿∆ utilized)
and the mode number. If cases 1 and 3 are compared, it can be seen that an increment
of 1 mm in variable deflection 𝛿∆ corresponds to a drop in the minimum stiffness of
about 16 %. If cases 2 and 4 are contrasted, it can be observed that an increase of 3
mm in mean deflection represents a decrease in the minimum stiffness of about 5 %.
This demonstrates that both deflections exert an influence of different weight over
the stiffness of the generator structure.
Table 3.2 Minimum structural stiffness per case and mode
Case 1 Case 2 Case 3 Case 4

Mode 𝛿̅=1mm 𝛿∆ =0.5mm 𝛿̅=1mm 𝛿∆ =1mm 𝛿̅=1mm 𝛿∆ =1.5mm 𝛿̅=4mm 𝛿∆ =1mm
1 5.92×108 N/m 5.65×108 N/m 5.43×108 N/m 5.52×108 N/m

2 5.82×108 N/m 5.49×108 N/m 5.21×108 N/m 5.32×108 N/m
3 5.7×108 N/m 5.28×108 N/m 4.95×108 N/m 5.09×108 N/m
4 5.58×108 N/m 5.1×108 N/m 4.73×108 N/m 4.87×108 N/m
By looking at Figure 3.19, one can appreciate that for the worst case scenario, which
corresponds to the collapse of the airgap with the stator structure physically touching
90
the magnets, the mode presenting the worst performance is Mode 4, having the
lowest stiffness value at 5.79×107 N/m.
8.00E+07
Magnetic stiff. + Structural stiff. (N/m)

7.00E+07
6.00E+07
5.00E+07
4.00E+07
3.00E+07
2.00E+07
1.00E+07
0.00E+00
0 20 40 60 80 100 120 140 160 180 200 220 240 260 280 300 320 340 360
θ (degrees)
Figure 3.19 Magnetic stiffness + Structural stiffness vs. Theta for the worst case scenario
̅ = 0.004 m; 𝜹∆ = 0.001 m)
(𝜹
It was observed that the structure selected for the study was very stiff and it is rather
difficult to appreciate the overall impact of the magnetic stiffness even in the worst
case scenario. By carrying out a considerable reduction in the thickness of both disc
and rim sub structures of the rotor and the stator a more compliant structure was
generated for its study. With higher magnetic stiffnesses, lower overall stiffnesses are
achieved and it is expected to see that at some point the total stiffness reach zero
values. The thicknesses used for both analyses are given in mm in Table 3.3.
Table 3.3 Thicknesses for stiff and compliant structures
Stiff structure Compliant structure

Rotor Stator Rotor Stator
Cylinder Disc Cylinder Disc Cylinder Disc Cylinder Disc
thickness thickness thickness thickness thickness thickness thickness thickness
(mm) (mm) (mm) (mm) (mm) (mm) (mm) (mm)
40 56 25 56 30 40 15 40
91
A drop in the rotor and the stator masses of 27 % and 31 % respectively was
obtained. This corresponds to a total mass reduction of 30 %, which corresponds in
this case to an overall stiffness reduction of 21 %. Figure 3.20 displays the results
acquired for the rotor and the stator compliant structures. As seen, the equivalent
rotor stiffness average is about 4.6×108 N/m, whereas the stator stiffness is around
2.6×108 N/m.
6.00E+08
Compliant rotor structure stiffness
5.00E+08
4.00E+08
(N/m)
3.00E+08
2.00E+08
1.00E+08
0.00E+00
0 20 40 60 80 100120140160180200220240260280300320340360
θ (degrees)
(a)
3.50E+08
Compliant stator structure stiffness
3.00E+08
2.50E+08
2.00E+08
(N/m)
1.50E+08
1.00E+08
5.00E+07
0.00E+00
0 20 40 60 80 100120140160180200220240260280300320340360
θ (degrees)
(b)
Figure 3.20 Compliant structure stiffness vs. Theta; (a) Rotor; (b) Stator
92
The difference between the rotor and stator stiffnesses for both structures stiff and
compliant is about 2×108 N/m, which brings down the overall stiffness (magnetic +
structural). See Figure 3.21.
8.00E+07
Generator Structural Stiffness (N/m)
7.80E+07
7.60E+07
7.40E+07
7.20E+07
7.00E+07
6.80E+07
6.60E+07
6.40E+07
0 20 40 60 80 100 120 140 160 180 200 220 240 260 280 300 320 340 360
θ (degrees)
(a)
7.00E+07
Magnetic stiff. + Structural stiff.
Compliant structure (N/m)
6.00E+07
5.00E+07
4.00E+07
3.00E+07
2.00E+07
1.00E+07
0.00E+00
0 20 40 60 80 100120140160180200220240260280300320340360
θ (degrees)
(b)
Figure 3.21 Compliant structure stiffness vs. Theta; (a) Generator structural stiffness; (b) Magnetic +
Structural stiffness
93
In Figure 3.22, a comparison between the stiffnesses for each mode and for each
structure is displayed. As observed, a drop in the minimum value of the overall
stiffness of more than 1×107 N/m is achieved.
8.00E+07
7.50E+07
7.00E+07
Stiff structure (N/m)
6.50E+07
6.00E+07
5.50E+07
5.00E+07
4.50E+07
4.00E+07
0 20 40 60 80 100120140160180200220240260280300320340360
θ (degrees)
Stiff Mode 0 Stiff Mode 1 Stiff Mode 2

Stiff Mode 3 Stiff Mode 4
(a)
6.50E+07
Compliant structure (N/m)
6.00E+07
5.50E+07
5.00E+07
4.50E+07
4.00E+07
0 20 40 60 80 100120140160180200220240260280300320340360
θ (degrees)
Compliant Mode 0 Compliant Mode 1 Compliant Mode 2

Compliant Mode 3 Compliant Mode 4
(b)
Figure 3.22 Overall stiffnesses comparison; (a) Stiff structure; (b) Compliant structure
94
The reduction in thickness of each sub structure forming the machine, revealed a
substantial drop of stiffness for this permanent magnet generator. This shows the
effect of compliant (lighter) structures and how the approach can be used in the
design process.
With both structures fully described and analysed for the PM machine, the stiff
structure was studied under the demanding loading conditions of a wound rotor
machine. Figure 3.23 illustrates the forces applied on each β degree section. Since
the effective airgap size of a wound rotor machine is smaller than that of a PM
generator due to the lack of magnets attached to the rotor surface, the forces are one
order of magnitude larger in all cases and for all the modes.
2000
1800
1600
Airgap closing force (kN)
1400
1200
1000
800
600
400
200
0
0 20 40 60 80 100 120 140 160 180 200 220 240 260 280 300 320 340 360
θ (degrees)
Figure 3.23 Wound rotor machine airgap closing force vs. Theta
Figure 3.24 depicts the magnetic stiffness of the wound rotor generator. Again, if it is
compared to its PM counterpart it can be observed that it is one order of magnitude
larger, and so it will have a profound impact on the overall stiffness.
95
0.00E+00
0 20 40 60 80 100 120 140 160 180 200 220 240 260 280 300 320 340 360
-1.00E+08
Magnetic stiffness (N/m)

-2.00E+08
-3.00E+08
-4.00E+08
-5.00E+08
-6.00E+08
θ (degrees)
Figure 3.24 Wound rotor machine magnetic stiffness vs. Theta
As predicted, the effect of the large increase in magnetic stiffness produced a very
significant drop in the overall stiffness, as seen in Figure 3.25. The negative values
for the minimum stiffness means that the structure is not stiff enough to resist the
loads and the airgap would close causing the collapse of the overall structure.
0.00E+00
Magnetic stiff. + structural stiff. (N/m)
0 20 40 60 80 100 120 140 160 180 200 220 240 260 280 300 320 340 360
-5.00E+07
-1.00E+08
Wound Rotor Generator
-1.50E+08
-2.00E+08
-2.50E+08
-3.00E+08
-3.50E+08
-4.00E+08
-4.50E+08
-5.00E+08
θ (degrees)
Figure 3.25 Wound rotor overall stiffness vs. Theta
96
3.5 Discussion
Various approaches exist in order to design a supporting structure for a wind turbine
electrical generator capable of withstanding the loads. In this chapter, a stiffness
model joining the magnetic and the mechanical designs has been developed. The 2D
magnetic model assumes a uniform radial deflection, ‘𝛿 ̅’, and a variable deflection,
‘𝛿∆ ’, that changes with angle, to estimate the resulting airgap closing force under
different modes of deflection. The assumed deflection and the obtained force is then
utilized to calculate the airgap stiffness. At this point, a structural model was created
and making use of the computed loads a set of finite element analyses was run for a 3
MW machine with a simplified steel structure made with discs. With the deflection,
the structural stiffness of the machine could be approximated. A comparison between
the airgap stiffness acquired with the magnetic model and the structural stiffness
determined if the generator would resist the input loading and what the stiffness
margin would be.
3.5.1 Magnetic stiffness model
The magnetic model for the airgap closing force and stiffness of a PM machine was
validated using a 2D finite element code. A two pole model with periodic
boundaries, neglecting axial effects and geometrically linearized so that radial lines
and arcs are mapped onto vertical and horizontal lines, was produced to carry out the
task. Three finite element cases were generated: (i) the stator has no slots and the
materials have a linear BH behaviour; (ii) the stator has slots and the materials have
linear BH curves and (iii) the stator has slots and the materials have no linear BH
curves. Comparing the analytical model with the idealised FE model (i), it could be
observed how the analytical model underestimated the force. It was proven that the
use of the fundamental MMF only, leads to neglecting higher order airgap flux
density spatial harmonics and the resulting force contributions. With that, the
analytical model was amended to incorporate the 3rd harmonic achieving better
results. Nevertheless, the model did not take into consideration the slotting that
according to the FE models (ii) and (iii) significantly reduced the forces. So as to
97
replicate this behaviour, the Carter factor was introduced into the analytical model.
An excellent agreement was achieved, hence the magnetic model was considered
valid. A FE magnetic model of a full PM machine was also produced for validation
purposes. The retrieved data showed a very good agreement with the analytical
results which reinforces the usefulness of the model. The FE magnetic studies did not
include coils as it was assumed that the stator electrical connection did not have any
influence in the magnetic stiffness final result. It is also necessary to highlight that
the effect of the armature reaction which can weaken or strengthen the airgap flux
density affecting the airgap closing force was not considered either.
3.5.2 Permanent magnet and wound rotor machines
Two distinct types of generators were analysed: a permanent magnet machine and a
wound rotor machine. The main reason to make this differentiation was that a
permanent magnet generator has a certain number of magnets attached to the rotor
ℎm
surface that increases the effective size of the airgap (𝑔 + ), whereas a wound
𝜇r
rotor machine is electrically excited and thus the effective airgap size is equal to the
physical airgap clearance ‘𝑔’. This means that the existing attractive forces in the
airgap of a wound rotor machine are larger than the forces acting on the airgap of a
permanent magnet generator. Having this in mind, it can be said that a stiffer and
more robust structure is necessary for a wound rotor machine which in turn leads us
to have a heavier generator. On the other hand, it is also essential to highlight the fact
that the magnetic design of a PM machine can be designed to avoid saturation in the
default state, while when a wound rotor machine airgap closes the iron parts of the
magnetic circuit go deep into saturation bringing about significant reluctance.
3.5.3 Structural stiffness model
As mentioned, the magnetic study of both machines was completed first assuming a
uniform radial deflection and variable deflection which changes with angle, ‘θ’.
Looking at the results obtained for both cases, it could be understood their effect on
the magnetic, structural and overall stiffnesses. The increase of 𝛿 ̅ supposes a
98
noticeable decrease in the absolute values of the maximum and minimum magnetic
stiffness, whereas the increment in force as the airgap closes causes a substantial
reduction of the structural stiffness that puts down the overall stiffness. If 𝛿∆ is
augmented, the absolute value of the maximum magnetic stiffness increases while
the minimum stays at the same level. The structural stiffness diminishes and the
overall stiffness affected by the increase in the magnetic stiffness and the drop in the
structural stiffness goes down although it never reaches a zero value, which means
that the structure is eventually very stiff and it would easily support the imposed
loading conditions. A more compliant structure was also looked at. A total reduction
in mass of 30 % was accomplished. That meant a decrease of 27 % in rotor mass and
31 % in stator mass. After the analysis, it could be observed that the drop in mass
corresponded to a decrease in the overall stiffness of 21 %. This gives a clear picture
of the trade-off process that should be made during an optimization study. For these
analyses it was assumed, for simplicity, that the structural stiffness was totally
invariant although it has been demonstrated that it changes with deflection mode,
deformation and angle. Nonetheless, as the variation is very small and it would only
represent a slight alteration in the final result, the assumption was considered valid.
3.5.4 Modes of deflection
Mode number was another factor that had a significant impact on the distinct
stiffnesses. In case of the magnetic stiffness, it could be seen that despite of having
different frequencies, all the modes but Mode 0 shared the same maximum value.
Mode 0 appeared as a straight line (due to its constant uniform load applied)
coinciding with the inflexion points of the curves corresponding to the rest of the
modes. Similar behaviour was noticed for the structural and overall stiffnesses,
although in the case of the structural stiffness, Mode 4 stood out showing the worst
performance with the minimum stiffness at 8.29×107 N/m for the stiff structure and
6.94×107 N/m for the compliant structure. The overall stiffness presented the same
type of shape as the magnetic stiffness, with all the modes but Mode 0 having the
same minimum value, which takes us to think that high order modes have to drive
the design of any type of electrical machine.
99
However, it is thought that the model can be utilized not only during the design stage
but also after manufacturing and during operation as part of an online airgap
condition monitoring system. Manufacturing defects can make the machine more
prone to deform following certain patterns, which would turn into high order
deflection modes. The model would be able to predict the airgap behaviour and its
suitability for operation. As part of a condition monitoring system, the model would
be capable of evaluating the time-varying output data and assessing the structural
integrity of the machine.
3.6 Conclusion
A stiffness model coupling the electromagnetic and mechanical designs of wound

rotor generators and surface mounted permanent magnet machines has been
developed. The results of the validation with the FE model showed the usefulness
and accuracy of the magnetic analytical tool. In addition, the parametric nature of
this analytical model makes it easy to use helping the designer to carry out quick
estimations for any deflection mode or deformation, in the early stages of the design
or after the manufacturing process. It could also be very handy for optimization
purposes or as part of an online condition monitoring system as it could assessed the
structural integrity of the machine at any time. On the other hand, it is important to
point out that this is a linearized 2D model, which neglects the axial end effects. The
dynamic behaviour of the machine is not captured either by the model and no
external forces have been considered.
With a view to improve the accuracy of the model, it is thought that all of these
features should be included.
Regarding the mechanical model, which is fed by the finite element analyses
retrieved data, it can be concluded that it is suitable to calculate the generator
structural stiffness. It was assumed that the structural stiffness was invariant although
it could be seen that it changes with deformation,‘𝛿 ̅’and‘𝛿∆ ’, deflection mode, ‘n’,
and angle, ‘θ’. However, it was observed that the variations were very small and
therefore only a slight alteration in the total stiffness result would be obtained. On the
100
other side, it is thought that the assumption of a constant stiffness for the bearing
introduces inaccuracies into the tool. Having said that, the author believes that an
accurate stiffness model of the bearing should also be generated and introduced.
A comparison between a very stiff structure and a more compliant structure was
made in order to understand the effect of mass reduction in the overall stiffness. It
could be seen how for a PM machine both structures were capable of withstanding
the loads although it was observed that the compliant structure had more difficulties.
In the wound rotor machine case, it could be seen how even the stiff structure was
not able to successfully support the loads. Taking this into consideration, the use of
the stiffer structure could be associated to a wound rotor machine, while the use of
the more compliant structure can be related to a permanent magnet generator due to
the inherent characteristics of the wound rotor generator make it more structurally
demanding than a permanent magnet electrical machine subject to the same
deflection. Hence, it can be concluded that a wound rotor electrical machine is
heavier than a permanent magnet generator.
Magnetic and structural stiffnesses were combined together and plotted against the
angle in search of the most dangerous mode. By varying the mean and the variable
deflection values, twelve different scenarios were assessed obtaining that Mode 4 is
the most damaging. Bearing this in mind, designers can tailor the structure with a
view to resist this mode in the lightest manner, by just introducing stiffeners in the
corresponding direction, by modifying the geometry of the structure or by utilizing
distinct structural materials with higher Young’s modulus to density ratios.
101
3.7 References
[1] A. S. McDonald , M. Mueller and H. Polinder, “Structural mass in direct drive

permanent magnet electrical generator,” IET Renewable Power Generation
Special Issue- Selected Papers from EWEC 2007, vol. 2, no. 1, pp. 3-15, March
2008.

generators for direct drive renewable energy converters,” PhD thesis,
Electrical Machines, Chania (Greece), 2006.
[4] P. Jaen-Sola and A. S. McDonald, “Structural Analysis and Characterization of

Radial Flux PM Generators for Direct-Drive Wind Turbines”, 3rd Renewable
Power Generation International Conference (RPG 2014), Naples (Italy), Sept.
2014.
[5] FEMM ‘Finite Element Method Magnetics’, www.femm.info/, [Online] Last

Accessed 28/02/2017.
102
Comparison of Methods for Estimating Generator Structural Stiffness
Chapter 4
Comparison of Methods for

Estimating Generator Structural
Stiffness
4.1 Introduction
Minimising the structural mass of low speed multi MW electrical machines for
renewable energy purposes have become an important object of study as with the
reduction in mass a substantial decrease in the machine capital cost can be achieved
[1]. In [2], Grauers introduced a procedure which estimated the final cost of a
103
generator based on its diameter and length, compared to dimensions of reference

structures. The excessive weight of the generator structure was highlighted by
Hartkopf et al. who claimed that 2/3 of a direct drive radial flux electrical machine
mass corresponded to the inactive material [3]. Mueller, McDonald and MacPherson
claimed in [4] that at multimegawatt ratings the inactive mass of a direct drive axial
flux machine it is almost 90 % of the total mass. Several studies have been written
on this regard presenting different approaches that can be utilised when designing
this type of machines in order to minimize their structural mass [5][6]. Some of them
have been already mentioned in Chapters 2 and 3. In these studies, analytical and
numerical analysis techniques were used with the main aim of finding the minimum
required stiffness so that the machine can withstand the loads. As known, the
structural stiffness is the ratio between the force applied on a body and the deflection
produced by the force along the same degree of freedom. By having the minimum
stiffness, the designers are able to estimate the minimum structural mass in a low
cost and fast manner. On the other side, there are studies which suggest the use of
magnetic and other innovative types of bearings to reduce the mass [7][8][9]. New
lightweight concepts have been also proposed, such as the NewGen one, where the
bearings are placed adjacent to the airgap so as to resist the loads and reduce the
structural stiffness demand in the rotor, the stator and the shaft [10]. Although these
are very promising options, the use of any of these elements have not been included
in the analyses presented in this thesis.
The main objective of this chapter is to explain and present the results obtained from
analyses carried out using three different approaches to estimate the minimum
machine structural stiffness: finite element, analytical and hybrid method. This is a
new concept that consists of combination of the data retrieved from finite element
studies and the outcomes acquired from dimensional homogeneity analyses. In
Section 4.2, a hierarchy of the methods will be given according to their suitability,
reliability and speed to accurately estimate the generator stiffness. A description of
the techniques and how and where they can be applied is also included in this
section. The results obtained from the analyses carried out with these methods over
three different structural layouts are presented in Section 4.3. A disc structure
arrangement offering the best structural performance under certain loading
104
conditions will be also tested under distinct modes of deflection. Finally, a further
optimisation of the said structure will be made employing finite element tools. A
detailed discussion about the data achieved will be introduced in Section 4.4,
whereas the drawn conclusions will be presented in Section 4.5.
4.2 Estimating minimum generator structural stiffness
4.2.1 Finding Structural Stiffness through the Combination of Sub Structures

Stiffness
The structural stiffness either of the rotor or the stator can be calculated by putting
their sub structures together, in series or in parallel as it has been described in
Chapter 3. When looking at a rotor or stator structure one can identify coherent
structural elements that lend themselves to separate evaluation of stiffness. For
instance, the stiffness of a rotor disc structure could be estimated combining in series
the stiffness of the disc and the stiffness of the cylinder as follows,
𝑘s,d 𝑘s,c (4.1)

𝑘s,r =
𝑘s,d + 𝑘s,c
and the total structural stiffness of the generator must be either equal to or higher to
kM (magnetic airgap stiffness) as explained in [11]. A graphical representation of
how the rotor structural stiffness can be combined is displayed in Figure 4.1.
Figure 4.1 Rotor structure split into disc and cylinder models [9]
Bearing this in mind and looking at the entire wind turbine, a complete structural
stiffness model was defined in the previous chapter employing equations (4.2a) and
(4.2b),
105
𝑘s,r 𝑘b
𝑘eq,r =
𝑘s,r + 𝑘b (4.2a)
𝑘eq,r 𝑘s,s
𝑘s =
𝑘eq,r + 𝑘s,s (4.2b)
where keq,r and ks correspond to the equivalent rotor structure stiffness and the total
structural stiffness, respectively.
So as to satisfy equation (4.2b) with the minimum mass, it is necessary to find a way
of evaluating the structural stiffness and mass of rotor and stator structures.
There exist different procedures to calculate the structural stiffness needed by an

electrical machine. A hierarchy should be defined considering the advantages and
drawbacks of each one. Among the three techniques described in this section, the
most sophisticated is the FE method as it is capable of capturing geometric, loading
and material features, as well as giving very accurate data. However, this approach is
computationally expensive and time consuming and is more suited for final design
analysis, rather than early stage optimisation.
The analytical methods presented in this section are able of giving accurate results
for Mode 0 and Mode 1 deflection of sub structures, such as arms, in a much quicker
manner. In order to estimate the stiffness of advanced structures it is often necessary
to combine the stiffness of different sub structures in series. Typically, these
analytical tools are precise when the geometry is simple and the loading corresponds
to Mode 0. No analytical tools have been proposed for higher modes.
The hybrid procedure combines the data retrieved from dimensional homogeneity
studies and a limited number of FE results and fits functions to the outcomes. It is
also much quicker than the FE approach and it is able to generate precise results for
Mode 0 and Mode 1 deflection of sub structures. This technique is more suited for
individual elements as additional independent variables make the function fitting
much more challenging although high precision results can also be obtained for
complete structures.
106
4.2.2 A Case Study Generator
For this investigation, a 3 MW machine made up of steel of 4 m diameter and 1.2 m

axial length, rotating at 12.7 rpm, as shown in Figure 4.2, has been utilised. In order
to estimate the stiffness, dimensions tc and td in the case of the disc structures and tc,a
and tarm for the armed structures have been varied. Afterwards, FE studies were
carried out so that the radial deflection could be obtained.
Figure 4.2 Rotor structures with thickness dimensions as altered in this analysis (a) Disc structure (b)
Arm structure [12]
Mass of the rotor and stator disc structures was found out with equations (4.3a) and
(4.3b), whereas mass of the rotor and stator armed structures was computed using
equations (4.4a) and (4.4b).
𝑚s,r = 𝜌[𝜋((𝑅 + 𝑡c )2 − 𝑅 2 )𝑙 + 𝜋(𝑅 2 − 𝑟 2 )𝑡d ] (4.3a)

2
𝑚s,s = 𝜌[𝜋 ((𝑅 + 𝑡s,c ) − 𝑅 2 ) 𝑙 + 2(𝜋(𝑅 2 − 𝑟 2 )𝑡s,d )] (4.3b)
2
𝑚s,ar = 𝜌[𝜋 ((𝑅 + 𝑡c,a ) − 𝑅 2 ) 𝑙 + 𝑛arms ((4𝑡arm 𝑤 − 4𝑡arm
2 )𝑙
arm )]
(4.4a)
2
𝑚s,as = 𝜌[𝜋 ((𝑅 + 𝑡c,as ) − 𝑅 2 ) 𝑙 + 2(𝑛arms ((4𝑡arm 𝑤 − 4𝑡arm
2 )𝑙
arm ))]
(4.4b)
Where r is the radius of the shaft and 𝑤 is the width of the arm structure.
107
4.2.3 Analytical characterisation of disc and arm structures
Different approaches are available to characterize disc and armed structures.

Whereas the use of FE techniques bring on certain difficulties that have been
explained in Section 4.2.1, analytical approaches allow the user to find the structural
stiffness of the components by introducing their dimensions and their material
characteristics into the equation. This technique has been already employed by other
authors. For instance, in [5], McDonald derived and validated (using finite element
methods) a series of equations for rotor and stator structures made with arms or
discs, that can be utilised to accurately calculate their radial deflection under Mode 0
loading and their axial deflection due to gravity for a wide range of dimensions and
materials. In this sub section, the author has focussed his attention on the calculation
of the stiffness of sub structures, such as discs and arms, under distinct modes of
deflection.
Rotor disc model: central hole and loaded boundary

The approach proposed by Benham et al. has been used in this paper to examine the
stiffness for the disc structure. In [13], the authors assumed a disc structure with a
central hole and unloaded boundaries that rotates at a constant velocity and therefore
is subjected to stresses induced by centripetal acceleration. Since the approach
presented in this paper considered a disc structure with a central hole subject to an
expansion load uniformly distributed along its edge and ω → 0, Benham’s model
was to some extent modified to match these features and help corroborate the
models. The stress-strain relationship is,
𝜎r 𝑣𝜎θ (4.5)
𝜀𝑟 = −
𝐸 𝐸
with 𝜎r and 𝜎θ being the radial stress and the angular stress respectively. If no
motion is considered, the radial stress can be calculated either
𝑌 (4.6a)
𝜎r = 𝑋 −
𝑅2
𝑌 (4.6b)
𝜎r = 𝑋 −
𝑟2
108
Any of the following equations can also be utilised to find the angular stress, 𝜎θ =
𝑌 𝑌
𝑋 + 𝑟 2 or 𝜎θ = 𝑋 + 𝑅2.
With no load acting on the central hole

𝑌 (4.7)
𝑋=
𝑟2
Replacing this into equation (4.6a) and rearranging then

𝜎r (4.8)
𝑌=
1 1
( 2 − 2)
𝑟 𝑅
Radial strain displacement for axial symmetry gives

𝑑𝑢 ∆𝑅 (4.9)
𝜀r = =
𝑑𝑟 𝑅
Rearranging equation (4.9) and substituting 𝜀r in equation (4.5) the following results
can be achieved,
1 1 Y (4.10)
Y ( 2 − 2 ) 𝑣 (X + 2 )
∆𝑅 = ( 𝑟 𝑅 − 𝑅 )𝑅
𝐸 𝐸
𝜎r 2𝜋𝑡d 𝐸
𝑘d = (4.11)
𝜎r
(𝑅 2 + 𝑟 2) ( 1 )
−1
𝑟2 𝑅2
𝜎r − 𝑣
𝑟 2𝑅2
( ( ))
A comparison between the results acquired from FE analyses and those achieved
with the equation is shown in Section 4.3.2.
4.2.4 Modelling structural stiffness: Finite Element Analysis
Complete rotor and stator structure models were studied utilising finite element
methods. For the analyses, the model was constrained at the shaft and evaluated for
Mode 1. Considering the variations of the flux density within the electromagnetic
circuit, a maximum normal stress of 411 kPa was located on the top of the structure
109
while the minimum normal stress (335 kPa) was placed at the bottom with the stress
varying sinusoidally. To apply the loads correctly, the rim sub structure was divided
into 36 parts (of 10 degrees each). This helped us later to deal with different
deflection modes without making major changes into the initial model. Note that the
maximum structural deflection was limited to a 10 % of the airgap size in any
direction. The thicknesses of the cylinders, tc, the thicknesses of the discs in the case
of the disc structures, td, and the thicknesses of the arms in the case of arm structures,
tarm, were changed and the obtained FE values plotted. For the armed structures, 10
ties of width, w = 0.35 m, were employed. In order to carry out the optimisation, a
fine tetrahedral mesh with an element size as suggested by the software was used.
a) Rotor analysis
The structure was studied as explained above. Figure 4.3 illustrates a disc rotor
structure model constrained at the shaft with a radial load acting on it at the left side
and an arm rotor structure at the right side. The material characteristics of this
structure made up of steel are: Young’s modulus, 𝐸 = 2.1 × 1011 Pa, Poisson’s
𝑘𝑔
ratio, v = 0.3 and density, ρ = 7850 𝑚3 .
Figure 4.3 Rotor model; Disc structure showing loading conditions and constraints (left side); Arm
structure (right side)
110
b) Stator analysis
For the stator structures the same methodology was put in practice. A sinusoidally
distributed compression load acts on the inner face of the stator cylinder, as it is
shown in Figure 4.4 (left side). Both discs and arms assemblies are constrained at the
shaft. The material properties of these structures are the same as those of the rotor
structure.
Figure 4.4 Stator model; Disc structure showing loading conditions and constraints (left side); Arm
structure (right side)
c) Conical rotor analysis
Although the main purpose of this section is to look at the mentioned type of
structures, conical rotor structures have been also proposed in this sub section.
Conical rotor structures have been already suggested by other authors, such as
Stander in [14], due to the superior axial stiffness that the geometry introduces, as
well as, its inherent radial stiffness provided by the cone sub structure. Nonetheless,
it is worth highlighting that the optimisation of a rotor cone structure includes other
variables that make it more tedious to study. These variables are the position of the
cone and the angle of the cone. See Figures 4.5 and 4.6. A rotor cone structure of 4
m diameter and 1.2 m of axial length made of steel was modelled in SolidWorks for
its optimisation. Then the structure was constrained at the shaft and the same loading
conditions as applied to the disc structure utilised. The element size suggested by the
111
piece of software for the tetrahedral mesh was accepted. The optimisation process
and the results achieved are described and presented in Section 4.3.
(1) (2) (3) (4) (5)
Figure 4.5 View of the different positions for the cone sub structure
Figure 4.6 illustrates the rotor conical structure showing the dimensions as altered in
the study. The angle ‘ψ’ is shown as well for clarification.
Figure 4.6 Rotor conical structure as altered in the study
112
4.2.5 Modelling structural stiffness using a hybrid technique: Rotor disc model
Looking at the rotor components separately is possible to find physically meaningful

algebraic equations that describe with precision their structural behaviour. In this
thesis, analytical techniques based on the principle of dimensional homogeneity have
been used [15]. In the case of the disc structure, if it is assumed that its stiffness
depends on the Young’s modulus, E, thickness, td, Poisson’s ratio, v, and R-r, where
r is the radius of the shaft, we can proceed as follows. Let
[𝑘d ]=[𝐸 a 𝑡db (𝑅 − 𝑟)c ] (4.12)
Replacing the dimensional combination for each factor in terms of [𝐹], force, and
[𝐿], length, it can be obtained that
[𝐹𝐿−1]=[𝐹 a 𝐿−2a 𝐿b 𝐿c ] (4.13)
Note that the Poisson’s ratio is a dimensionless variable.
Equating powers it is obtained that a=1 and -1=-2a+b+c. It can be observed that the
analysis is unable to tell the powers of the thickness and the length. Nevertheless,
they can be estimated by looking at how they vary with the stiffness, which was
approximated using the FE data that were validated with Benham’s model. The
implementation of a constant was necessary so that the equation could be finally
balanced.
𝐶1 𝑡d 2 𝐸(1 + 𝑣 2 ) (4.14)
𝑘d =
(𝐶2 𝑡d + 𝐿)𝛾
where C1 = 4160 and C2 = 400. Note that a dimensionless variable, 𝛾, which depends
on the mode of deflection, has also been introduced. With it, the stiffness of the rotor
components can be calculated taking into consideration the deflection mode. 𝛾 is
𝜎radial,max
equal to .
𝜎radial,min
A comparison between the results for Mode 0 obtained with equation (4.14) and data
retrieved from the FE simulation studies is shown in Figure 4.7. As it can be seen, a
113
good agreement was achieved over the whole range. The relationship acquired from
the linear regression was y = x, with 𝑅 2 = 0.9945.
Equation Disc Stiffness(N/m)

11
4 x 10
0.5 1 1.5 2 2.5 3 3.5 4

FE Disc Stiffness(N/m) 11
x 10
Figure 4.7 Equation vs. FE disc stiffness [11]
4.2.6 Structural optimisation
A detailed description of the different methods that can be used to estimate the
minimum required stiffness and structural mass of an electrical machine has been
given. Nevertheless, other ways of further reducing the mass of the structure exist. In
this section, the shape optimisation add-on tool available in ANSYS Workbench© is
presented. Both small and large scale generator structures have been looked at in this
chapter. An optimisation study of a small scale generator structure made of steel was
run first in order to verify the usefulness of the said tool. Moreover, the creation and
analysis of a small scale model at first instance was considered good practice and it
could be utilised for other purposes in the future. A 100 kW electrical machine with
0.42 m radius, 0.21 m axial length, 2.08 mm airgap, 140 rpm rotor speed and 6.8
kNm torque was assumed. With a radial expansion load of 400 kPa and a tangential
load of 30 kPa applied on the outer face of the rim structure and a gravitational load
applied globally according to the Y axis, the shape optimisation study was made.
After constraining the structure at the shaft, a fine tetrahedral mesh was produced,
and as it can be seen, the red elements, which are mostly placed within the disc sub
structure, are the ones that can be removed. See Figure 4.8.
114
(a)
(b)
Figure 4.8 Rotor structure shape optimisation; (a) Rotor structure highlighting the elements to be
eliminated; (b) Cutouts of the optimised rotor structure (dimensions shown in mm)
Linking the result to a CAD model in SolidWorks with the same dimensions and
material characteristics a handmade removal of material was carried out always
taking into consideration the deflection limit, which in this case corresponds to 0.208
mm. An elimination of 6 mm of material could be done for the disc. The same
115
procedure was followed for the stator acquiring similar data. Having a 6 mm cylinder
thickness and 12 mm thickness discs, a removal of 7.5 mm of material from each
disc, as it was done with the rotor, could be accomplished. The total mass for a steel
generator structure being able to withstand the mentioned loads without deforming
more than as stated, is 313.22 kg with 114.76 kg for the rotor and 198.46 kg for the
stator. After further optimisation using the ANSYS shape optimisation tool an
overall reduction in mass of about 15 % can be achieved. The final structural mass
would be 266.7 kg with 91.1 kg for the rotor and 175.6 kg for the stator.
CAD Model
Shape optimisation
study in ANSYS
Approximated CAD model of the

obtained shape in SolidWorks
Structural optimisation
Using Design Explorer

By hand
Optimised structure
Figure 4.9 Flowchart of the structural topology optimisation process
116
Figure 4.9 shows a flowchart of the structural topology optimisation process as it

was followed. Two different ways of approaching the structural optimisation can be
tracked: by hand or using the Design Explorer ANSYS tool coupled to the model in
SolidWorks. The first procedure was used to optimise the small scale model,
whereas the second was utilised to optimise the large scale one. See Section 4.3.5.
Models with relative simple shapes can be easily optimised by hand. However, when
the shapes are more complex, the use of an instrument that standardizes the process
as Design Explorer does is necessary.
4.3 Results
In this section, outcomes obtained by the three distinct types of methods are given.
The FE technique is shown first and the data are utilised as a benchmark to validate
the other approaches.
4.3.1 Finite element approach
Figure 4.10 (1), (2), (3) and (4) are contour plots for the disc and arm rotor and stator
structures displaying their stiffness for different sub structures thicknesses. In the
generator disc structure case, the variables are the thicknesses of the disc and the
cylinder sub structures that have been altered as seen in the figure. With the disc
thickness in the Y axis and the cylinder thickness in the horizontal axis, the coloured
lines plotted represent stiffness. In the case of the generator with arm rotor and stator
structures, the variables are the arm sub structure thickness and again the cylinder
thickness. As said in Section 4.2.4, the following graphs represent the results
obtained from analysing the structures under Mode 1 (localized deformation due to
eccentricity) deflection.
117
Figure 4.10 2D optimisation for 3 MW rotor and stator disc and arm structures with structural
stiffness criterion; (1) Rotor disc structure; (2) Stator disc structure; (3) Arm rotor structure; (4)
Arm stator structure
118
By making use of this tool, a quick estimation of either the dimensions or the
stiffness needed by the electrical machine structure can be made. These contour plots
would result very useful during the early stage of the design.
Different methods for estimating the minimum structural stiffness required have
been described and explained in Section 4.2. However, the structures are more prone
to deform following certain patterns depending on their geometric configuration.
With this in mind the optimised rotor and stator disc structures have been tested
under distinct modes of deflection. The characteristics of the electrical machine are
the following,
- Axial length, ‘l’ = 1.2 m

- Rotor radius, ‘R’ = 2 m
- Airgap, ‘g’ = 0.005 m
- Rotor yoke height, ‘ℎry ’ = 0.04 m
- Aspect ratio = 0.6
- Magnet height, ‘lm’ = 0.017 m
- Magnet width, ‘bp’ = 0.1 m
- Flux density, ‘Br’ = 1.02 T
The stiffness of the said components was approximated for deflection modes going
from 0 to 4 by using the maximum pressure, which was calculated making use of the
data presented above. Its average value corresponded to approximately 518,404 Pa.
The rotor structure formed by a disc sub structure with 56 mm of thickness and a
cylinder sub structure with a thickness of 40 mm was analysed employing FE
techniques as described in the previous section obtaining the results presented in
Figure 4.11. As seen, the rotor structure stiffness drops with the mode of deflection.
Mode 4 is shown as the most hazardous with a stiffness below 5.6×108 N/m and
according to the hierarchy of approaches given in this chapter, only finite element
methods are able to accurately predict the minimum stiffness needed to comply with
the structural requirements for this mode.
119
6.30E+08
6.20E+08
6.10E+08
Stiffness (N/m)
6.00E+08
5.90E+08
5.80E+08
5.70E+08
5.60E+08
5.50E+08
0 1 2 3 4 5
Deflection modes
Figure 4.11 Structural Stiffness vs. Deflection Modes (Disc Rotor Structure)
The stator structure is composed by two discs of 40 mm thickness each and a

cylinder with a thickness equal to 25 mm. The optimised structure was analysed
tracking the procedure defined in Section 4.2.4 and under the same loading
conditions as used for the rotor achieving similar results. The structural stiffness
goes down again with the deflection mode. As observed, Mode 4 is the most
damaging for the structure with a stiffness barely overtaking 4×108 N/m. See Figure
4.12.
4.35E+08
4.30E+08
4.25E+08
Stiffness (N/m)
4.20E+08
4.15E+08
4.10E+08
4.05E+08
4.00E+08
0 1 2 3 4 5
Deflection modes
Figure 4.12 Structural Stiffness vs. Deflection Modes (Disc Stator Structure)
120
4.3.2 Analytical approach
a) Rotor disc sub structure
Figure 4.13 illustrates the comparison between the modified rotor disc sub structure
model and the FE results. An excellent agreement is achieved as Benham’s model
stiffness is equal to the FE stiffness with R2 = 1.
Benham's Model for Disc Stiffness(N/m)
11
4.5 x 10
4
3.5
3
2.5
2
1.5
1
0.5
0
0 0.5 1 1.5 2 2.5 3 3.5 4
FE Disc Stiffness(N/m) 11
x 10
Figure 4.13 Comparison of stiffness calculated from FE and Benham model [11]
However, this approach only works for Mode 0. Hence, an approach that accurately
predicts the structural stiffness of this rotor component for different modes of
deflection needs to be found.
b) Arms sub structure

The arms are generator sub structures that connect the external cylinder to the main
shaft in the case of the rotor. For the stator, these arms, also called ties, are attached
to the turret. The aim of the arms is to stiff the generator structure in order to
withstand the large loads present during operation as well as during the
transportation and installation stages. The modulus of elasticity is,
𝐹
𝜎arm 𝐴arm (4.15)
𝐸= =
𝜀arm 𝛿
𝑙arm
121
where F is the force applied to the structure, 𝐴𝑎𝑟𝑚 is the cross sectional area of the
arm, 𝛿 is the deflection in the longitudinal direction and 𝑙𝑎𝑟𝑚 is the length of the
arm. Reordering the equation
𝐹𝑙arm (4.16)
𝐸=
𝐴arm 𝛿
𝐹 𝑘arm 𝑙arm
With 𝑘 = 𝛿 it is obtained that 𝐸 = . Rearranging
𝐴arm
𝐸𝐴arm (4.17)
𝑘arm = .
𝑙arm
If it is considered that the ties are hollow square structures of width, w, and
thickness, 𝑡arm , then the cross sectional area
𝐴arm = 𝑤 2 − (𝑤 − 2𝑡arm )2 (4.18)
Expanding the polynomial expression and rearranging the equation
2
𝐴arm = 4𝑡arm 𝑤 − 4𝑡arm (4.19)
Substituting 𝐴arm into equation (4.17) the following expression for the arms
structural stiffness can be found,
4𝐸𝑡arm (𝑤 − 𝑡arm ) (4.20)
𝑘=
𝑙arm
Due to the relative simplicity of this sub structure, equation (4.20) is supposed to be
valid to calculate the required stiffness of the arms for Mode 0 and Mode 1. Figure
4.14 shows a comparison between the data obtained from FE analyses and the values
acquired with equation (4.20).
122
8
x 10
10
Arms Stiffness Equation (N/m)

7
0
0 1 2 3 4 5 6 7 8 9 10
Arms Stiffness FE (N/m) x 10
8
Figure 4.14 Comparison of stiffness estimated with analytical model and FE [6]
As it is shown, a very good agreement has been achieved over the whole range.
Fitting a straight line of gradient 1 passing through the origin it is obtained that
R2=0.9829.
4.3.3 Hybrid approach
Using the hybrid approach, the structural stiffness of several components has been
calculated. Equations for the rotor and stator cylinders of the structures made with
discs have been obtained. Arm structures have been also looked at acquiring
equations for their components. Moreover, equalities for complete rotor and stator
structures made with discs and arms have been tried achieving good results in most
of the cases.
a) Rotor cylinder model as found from FE results

A similar methodology was tracked for the rotor cylinder. Assuming, 𝑘c =
𝑓(𝐸, 𝑡c , 𝑙, 𝑅), the dimensional analysis was carried out. As it happened with the disc,
the study could not predict all the powers of the variables so they had to be found by
looking at their variation with the stiffness. Equation (4.21), which accurately
describes the behaviour of the cylinder structure, was found after the analysis,
123
[𝐶3 𝐸𝑡c2 + 𝐶4 (𝑅 + 𝑙)](1 + 𝑣 2 ) (4.21)

𝑘c =
(𝑅 + 𝑙)𝛾
where C3 = 82.3, C4 = 8.23 × 109 and l is the cylinder axial length. Again a
comparison between the equation results and the data from the FE analyses for mode
0 is presented in Figure 4.15. High accuracy was also achieved according to the
relationship obtained from the linear regression, Equation Cylinder Stiffness = FE
Cylinder Stiffness with 𝑅 2 = 0.9721.
10
18 x 10
Equation Cylinder Stiffness (N/m)
16
14
12
10
8
6
4
2
0
0 2 4 6 8 10 12 14 16
FE Cylinder Stiffness (N/m) 10
x 10
Figure 4.15 Equation vs. FE cylinder stiffness [11]
b) Stator Cylinder Sub Structure Model

As illustrated in Section 4.2.5, a dimensional analysis of the component in question
was made. In this particular case it was assumed that the stiffness of the cylinder
depends on the Young’s Modulus, E, thickness, tc,s, length, lc,s, radius, Rc,s and
Poisson’s ratio, v. As expected, the study could not predict all the powers of the
variables present in the equation, therefore they had to be determined by analysing
the variation of each parameter with stiffness. A constant had to be introduced to
balance the equation,
2
𝐶6 𝐸𝑡c,s 𝑙c,s (1 + 𝑣 2 ) (4.22)
𝑘c = 𝐶5 + [ 2 𝛾
]
𝑅c,s
124
where C5 = 3.82 × 1010 and C6 = 442.14. Comparing the equation retrieved data and
the FE studies results for Mode 0, it can be observed that a reasonable level of
accuracy was obtained, as a straight line of gradient 1, passing through the origin,
fits the data with an R2 equal to 0.9225. See Figure 4.16.
10
x 10
6
Stator Cylinder Stiffness Equation (N/m)
0
0 1 2 3 4 5 6
Stator Cylinder Stiffness FE (N/m) 10
x 10
Figure 4.16 Stator cylinder Equation vs. FE [6]
c) Cylinder Sub Structure Model for Armed Rotor

In this case, it was assumed that 𝑘c,a = 𝑓(𝐸, 𝑡c,a , 𝑙c,a , 𝑅c,a , 𝑣). Once the dimensional
analysis and sensitivity analysis was completed and the retrieved data were analysed,
equation (4.23) was found.
2 2
𝐶8 𝐸𝑡c,a (1 + 𝑙c,a )(1 + 𝑣 2 ) (4.23)
𝑘c,a = 𝐶7 + [ 2 𝑙 𝛾
]
𝑅c,a c,a
where C7 = 1 × 1010 and C8 = 86.35. As seen in Figure 4.17, a fair precision was
achieved. Nevertheless, higher volatility can be appreciated for models
corresponding to cylinders with very large thicknesses (150 mm). The equality
presented the highest accuracy within a range between 5 and 7 metres diameter. The
results have a R2 value of 0.9454 regarding a straight line of gradient 1 passing
through the origin.
125
Cylinder Armed Rotor Structures Stiffness Equation (N/m)

10
x 10
14
12
10
0
0 2 4 6 8 10 12
Cylinder Armed Rotor Structures Stiffness FE (N/m) 10
x 10
Figure 4.17 Armed rotor cylinder Equation vs. FE [6]
d) Cylinder Sub Structure Model for Armed Stator

Using the same arguments as for Section 4.2.5, the final equation to find the cylinder
stiffness is as follows,
2
𝐶10 𝐸𝑡as,c 𝑙as,c (1 + 𝑣 2 )
𝑘as,c = 𝐶9 + [ 2 𝛾
] (4.23)
𝑅as,c
with C9 = 1.19 × 1010 and C10 = 128.44. A comparison between the results obtained
from the equation for the cylinder under Mode 0 deflection and the data acquired
from the FE simulation studies was made. As it can be seen in Figure 4.18, a good
agreement was achieved again as the data has an R2 = 0.9455.
126
Cylinder Armed Stator Structures Stiffness Equation (N/m)

10
x 10
12
10
0
0 2 4 6 8 10 12
Cylinder Armed Stator Structures Stiffness FE (N/m) 10
x 10
Figure 4.18 Armed stator cylinder Equation vs. FE [6]
e) Rotor and stator structural stiffness model

In order to estimate the structural stiffness of the whole disc rotor, the data for the
disc and the cylinder sub structures were introduced into equation (4.1). An overall
good agreement was achieved with a R2 = 0.9828 for a gradient 1 straight line
passing through the origin fitting. The component dimensions are shown in Table
4.1. They are the same as the ones that will utilised later on in this chapter to
optimise the generator structural mass using the FE approach.
The stator’s structural stiffness was predicted by putting together the cylinder, 𝑘c,s ,
and the discs sub structures. Because the two discs are in parallel, they are added
together to give an equivalent stiffness, 𝑘eq,d = 𝑘d1 + 𝑘d2 .
The same approach that was used to find out the equation of the rotor was utilised in
the case of the stator. As the stator discs are constrained in the same way as it was
done with the rotor disc, equation (4.14) was considered valid. However, a new
formula for the stiffness of the cylinder was needed.
127
Table 4.1 Disc and arm generator structures data
Disc structure Arm structure

General data Rotor Stator General data Rotor Stator
tc (m) ts,d (m) tc,a (m) tc,as (m)
R=2m 0.023 0.054 R=2m 0.03475 0.038
r = 0.625 m 0.026 0.046 r = 0.625 m 0.0365 0.039
l = 1.2 m 0.029 0.039 l = 1.2 m 0.0395 0.04
v = 0.3 0.032 0.036 v = 0.3 0.042 0.0415
E = 2.1×1011 Pa 0.035 0.034 E = 2.1×1011 Pa 0.045 0.044
Rotor 0.038 0.032 larm = 1.375 m 0.047 0.049
td = 0.04 m
Stator 0.041 0.03 w = 0.35 m 0.05 0.0565
ts,d = 0.02 m
0.043 0.029 Rotor 0.053 0.0615
tarm = 0.08 m
Stator
ta,s = 0.03 m
After introducing the results acquired with equations (4.14) and (4.22) into the
2𝑘d 𝑘c
stator’s equation, 𝑘s,s = , a low agreement with the FE studies was achieved
2𝑘d +𝑘c
with a R2 = 0.3647. This is because the equation predicts a higher contribution of the
discs to the overall stiffness.
For the armed structures case, having the equivalent stiffness of the arms, 𝑘a,eq =
𝑛𝑎𝑟𝑚𝑠 × 𝑘𝑎 , and the stiffness of the cylinder, 𝑘c,a, the stiffness of the rotor can be
calculated. As it can be seen in Figure 4.19, the data obtained for arm rotor structures
shows a good agreement with the FE results having a R2 = 0.9091 for a straight line
of gradient 1 intercepting the origin fitting.
The stator’s cylinder stiffness was also estimated using the method described in
above, but this time considering that there are two sets of 10 arms. A decent
agreement was acquired this time as seen in the figure. The data has a R2 = 0.8358
for a linear fitting intercepting the origin. As understood, the equalities tend to
overestimate the overall stiffness with the stator structures giving the lowest
agreements.
128
Figure 4.19 Equation Stiffness vs. FE Stiffness for complete structures; (a) Disc rotor structure; (b)
Arm rotor structure; (c) Disc stator structure; (d) Arm stator structure
4.3.4 2D optimisation of simplified structures
Going a step beyond in structural optimisation, disc structures mass can be

minimized using plots showed in Figure 4.20. For instance, if it is assumed a
constant bearing stiffness 𝑘b = 3 × 109 N/m and having that the total needed
stiffness 𝑘s = 1 × 109 N/m (acquired from equation (4.2b)), a comprehensive table
gathering all the relevant data can be easily created. As it can be observed, it was
identified that the disc of the rotor must have at least 40 mm of thickness whereas the
thickness of the stator discs must be over 20 mm due to torque requirements (green
lines for rotors and red lines for stators determine the minimum required stiffness in
the tangential direction). This necessary stiffness to withstand torque loads was
calculated according to the deflections obtained from applying a torque of 2,250
kNm to the structures. This value was retrieved from a simulation study carried out
with Bladed over a 3 MW direct drive wind turbine with a PM generator at 12 m/s of
wind speed.
129
Taking all of these features into consideration, the minimum mass of the rotor
structure can be calculated for the whole range of structural stiffnesses by looking at
the plot displayed in Figure 4.20(a), where the red line states the minimum stiffness
in the normal direction. Replacing ks,r and kb into equation (4.2a), the equivalent
rotor stiffness can be computed. Then, substituting keq,r and ks into equation (4.2b)
and reordering, ks,s can be achieved. With the stator discs thickness and the structural
stiffness of the stator known, its mass can be found by entering into the plot
displayed in Figure 4.20(b), where the black line determines the minimum required
stiffness in the normal direction. The total mass of the generator is estimated by
adding up the mass of the rotor structure, ms,r, and the mass of the stator structure,
ms,s.
80
Structural stiffness(N/m)
Structural mass(kg)
2.2x10 9
2.8x10
2.6x10
Rotor disc thickness(mm)
Tangential stiffness(N/m)
13
70
11
00
Normal stiffness(N/m)
00
0
9
9
0
10
90
60
2.1x109
00
00
12
2.4x10
3.2x10
0
3x10
00
50
9
9
9
3.4x10
3.6x1
1 6
80
2 3 4 5 7 8
60
70
40
00
9
2.3x10
00
00
0
9
9
30
20 25 30 35 40 45
Rotor cylinder thickness(mm)
(a)
130
80
2.2x10
3.1x10
4.9x
4x10
200
Structural mass(kg)
Stator discs thickness(mm) 70 Tangential stiffness(N/m)
10
0
9
2.6x10
0
5.4
9
9
60 9
x10
3.6x10
9
9
120
160
50
2.1x109
140
00
00
100
180
00
800
00
40
00
0
4.5
30
x10
9
8 7 6 5 4 3 2 1
20 2.3x109
600
0
10
20 25 30 35 40 45 50 55
Stator cylinder thickness(mm)
(b)
Figure 4.20 2D optimization for 3MW rotor and stator disc structures with structural stiffness
criterion [6]
For this example, it can be seen that the minimum mass of the electrical machine
structure, ms, is 19,260 kg. See Figure 4.21. The models utilised to develop this case
study has been highlighted with red dots and their corresponding numbers on the
optimization graphs.
4
2.5 x 10
Stator mass(kg)
Rotor mass(kg)
Minimum mass Total mass(kg)
Structural mass(kg)
1.5
0.5
1 2 3 4 5 6 7 8
Models
Figure 4.21 Mass optimisation result for disc structures [6]
131
b) Armed Structure
The same methodology was followed for the arm structures case. See Figure 4.22.
110
Structural mass(kg)
9
1.9x10
2.2x10
Rotor arms thickness(mm)
100
2.1x109 150
9
2.5x10
19
16
17
18
90 00
00
00
00
0
0
0
0
3.7x10
3x109
1 2 3 4 5 6
0
7 8
80
0
9
2.3x109
3.2x109
3.5x109
2.7x10
13
14
70
00
0 00
0
9
60
30 35 40 45 50 55
Rotor cylinder thickness(mm)
(a)
80
4.2x10
Structural mass(kg)
70
Stator arms thickness(mm)
3.2x10
2.1x10
9
200
32
60
00
0
3.7x10
26
0
0
9
9
230
29
00
50
00
0
00
2.6x10 7000
5.3x10
0
9
40
4.8x10
9
9
1
8 7 6 5 4 3 2 1
30
2.3x109
9
140
20
00
10
30 35 40 45 50 55 60 65
Stator cylinder thickness(mm)
(b)
Figure 4.22 2D optimization for 3MW rotor and stator armed structures with structural stiffness
criterion [6]
With 𝑘b = 3 × 109 N/m and the total needed stiffness 𝑘s = 1 × 109 N/m, the
minimum generator mass was 35,500 kg as it can be observed in Figure 4.23.
132
4
4.5 x 10
Rotor mass(kg)
Stator mass(kg)
4
Total mass(kg)
Structural mass(kg)
3.5
3 Minimum mass
2.5
1.5
1
1 2 3 4 5 6 7 8
Models
Figure 4.23 Mass optimisation result for armed structures [6]
After a detailed analysis of both types of structures, it could be observed that armed
structures were not capable of resisting torque loads as disc structures do unless the
thickness of the hollow arms is considerably increased with the consequent rise in
mass. It was also noticed that arm structures are slightly weaker than disc structures
in the radial direction, as it could be seen from their behaviour under the normal
component of the Maxwell stress. If a comparison between these two types of
structures supporting the same loads is made, a difference in mass of about 17,000
kg is achieved.
4.3.5 Structural topology optimisation
In this section, the data obtained from the optimisation study of the proposed conical
rotor structure are presented, as well as the results achieved from the shape
optimization carried out over the large scale model of the generator structure.
133
a) Conical rotor optimisation
The optimisation study of the rotor structure was carried out as follows,
1. Initial values were given to the thicknesses of the cone and the cylinder (40 mm
for the cone and 22 mm for the cylinder).
2. The most suitable position for the cone sub structure was found. By altering the
cone angle, 9 different models with five dissimilar cone positions as shown in Figure
4.5 were analysed obtaining the results displayed in Table 4.2.
Table 4.2 Cone structure optimisation results
Position
1 2 3 4 5
Angle (º) Mass (kg) δ(m) δ(m) δ(m) δ(m) δ(m)
×10-4 ×10-4 ×10-4 ×10-4 ×10-4
30 6873 4.8 4.4 4.8 4.8 4.4
35 7080 4.9 4.6 4.9 4.9 4.4
40 7384 4.7 4.5 4.9 4.9 4.4
45 7775 4.8 4.5 5 5 4.5
50 8281 4.9 4.7 5.1 5 4.6
55 8958 5 4.9 5.3 5.2 4.7
60 9881 5.2 5.1 5.7 5.4 4.9
65 11200 5.7 5.5 5.9 5.7 5.2
70 13205 7.1 6.6 6.7 6.6 5.9
3. As seen, position 5 has the best radial stiffness. In addition it was observed that
the overall structural mass decreases with the cone angle. The thickness of each sub
structure was increased by 5 mm in order to see which one was introducing more
mass into the overall system. After a comparison, it was seen that the cylinder sub
structure put about 151 kg more than the cone in any case.
4. With the variables to be targeted (cone angle and cylinder thickness), the
optimisation study was defined in SolidWorks. The test settings include the
deflection constraint of 0.5 mm in any direction and the goal of minimizing the mass
of the whole structure. A complete batch of scenarios was looked at. Having the cone
placed at position 5, the angle of the sub structure was varied from 35 to 5 degrees in
steps of 5. Each step was study with the cone thickness fixed at 40 mm and the
cylinder thickness varying from 22 mm to 14 mm in steps of 1 mm. The best
134
outcome was acquired for an angle of 10 degrees and a cylinder thickness equal to
20 mm. The overall mass was 6,104 kg and the radial deflection 0.497 mm.
5. So as to complete the rotor conical structure optimisation, it was checked that at

that point, the variation of the cone thickness did not introduce any further
improvement. For that, the cylinder thickness was diminished to 18 mm while the
cone thickness was kept fixed giving a radial deflection of 0.54 mm and a mass of
5,861 kg. Then, the thickness of the cone sub structure was pushed up by 5 mm. The
study showed that the deflection had gone down to 0.536 mm whereas the mass had
gone up to 6,320 kg. Bearing in mind this result, the author terminated the
optimisation process.
b) Large scale disc structure optimisation
Having demonstrated the utility of this ANSYS instrument in Section 4.2.6, it was
again utilised to further optimise the large scale model. In this case, different
outcomes were obtained for the rotor and the stator structures compared to those of
the small scale model. With the structure loaded as specified in Section 4.2.6, the
shape optimisation analysis was made revealing that the elements to be removed are
not only within the disc sub structure but also in the outer surfaces. In addition, they
tracked a certain pattern, as observed in Figure 4.24 that can be utilised to eliminate
the material in a standard way making the structure easier to manufacture.
Figure 4.24 Large Scale Rotor Structure Shape Optimisation
135
Considering the data retrieved from the shape optimisation study, a model in
SolidWorks was modified as seen in Figure 4.25(a) trying to standardize the shape of
the clusters of elements to be removed. Then, by linking this model to the Design
Explorer tool of ANSYS Workbench, the dimensions and the number of those
shapes could be altered. Keeping in mind the deflection limit (in this case it is 0.5
mm), the optimum number of gaps was found to be 9. The size of the shapes was
maximised so that the maximum amount of material could be taken out. A total
number of 57 iterations were necessary to find out the optimum profiles for the rotor
structure. Figure 4.25(b) shows the variables as changed in the Design Explorer
study with lr = 1,265 mm, Rs = 135 mm, Rm = 2,345 mm and Rl =1,240 mm.
(a) (b)
Figure 4.25 Design Explorer Optimisation (Large Rotor Structure)
For the stator structure a similar methodology was followed although a different
shape was acquired from the study, as seen in Figure 4.26. The shape was
approximated as shown in Figure 4.27(a) with the optimum found at Ds = 1,400 mm
and ds = 1,150 mm with 5 circular holes. A total number of 23 iterations were needed
so as to figure out the optimum values of the two variables.
136
Figure 4.26 Large Scale Stator Structure Shape Optimisation
(a) (b)
Figure 4.27 Design Explorer Optimisation (Large Stator Structure)
If the mass of the resultant generator structure is compared with a solid disc structure
(overall mass of 19,260 kg with 9,809 kg for the rotor and 9,451 kg for the stator)
capable of supporting the already said loads, a difference of almost 38 % is achieved.
With a total mass of 12,000 kg with the rotor accounting for 5,694 kg and the stator
for 6,306 kg a substantial drop in mass was achieved.
137
4.4 Discussion
There exist different procedures to calculate the structural stiffness needed by an

electrical machine. A hierarchy have been defined considering the advantages and
drawbacks of each one. Among the three techniques described here, the most reliable
is the FE method due to the high precision data that can be obtained for any type of
structure under any deflection mode. Nevertheless, this approach it is considered
computationally expensive and time consuming. Two distinct structural
configurations, arm and disc, were analysed using this method and it could be
observed the arm structures were not capable of resisting torque loads as disc
structures did unless the thickness of the hollow arms was considerably increased
with the consequent rise in mass. It was also noticed that arm structures were slightly
weaker than disc structures in the radial direction, as it could be seen from their
behaviour under the normal component of the Maxwell stress. After a comparison
between these two types of structures supporting the same loads, a difference in mass
of about 17,000 kg was achieved, with 19,260 kg for the disc structure and 35,500
for the armed structure.
On the other hand, the analytical method presented here was capable of producing
accurate results for Mode 0 deflection of disc and arm sub structures in a much faster
way with R2 not going below 0.9829 in either case. For the disc structure, the
approach was developed by modifying an existing model of a rotating disc with a
central hole and unloaded boundaries, whereas for the case of the arm, the final
expression was obtained by playing with the physical dimensions describing the
structure. Last but not least, the hybrid procedure, which was produced by examining
the results from the FE studies and utilising the data achieved from dimensional
analyses, was also much quicker than the FE approach and it was able to generate
precise results for Mode 0 and Mode 1 deflection of disc and arm structures.
Excellent agreements between FE data and the results achieved with the expressions
were obtained for all the analysed sub structures with R2 never going below 0.9225.
However, when looking at complete structures, the equations did not show that
outstanding performance. The lowest agreements between FE data and the results
acquired from the equations were achieved for the stator structures, with the disc
138
arrangement giving the worst outcome (R2 = 0.3647). It could be understood that the
equations tend to overestimate the overall stiffness by giving too much weight to the
stiffness of the discs. Considering the applicability and reliability of this method for
sub structures, it is thought that better outcomes could be achieved if a more detailed
research on the calculation of the stiffness of complete structures is carried out.
Rotor and stator disc structures were tested under different modes of deflection in
order to identify which one is the most dangerous and how it would affect the
structures. As the data revealed, the stiffness drops with each deflection mode
showing its lowest value at Mode 4 in both cases. By taking into account this
important feature, the design can be tuned so that the structures can withstand this
deflection mode in the most reliable, lightweight and cost effective manner. On the
other hand, if the designer is interested in varying the most damaging deflection
mode (due to different requirements), it can be done by altering the thicknesses of
the sub structures. This is a trade-off process in which the stiffness of the overall
structure can be increased for some deflection modes at expense of the others.
Finally, with the optimum disc structure fully described, a further optimisation was
carried out using finite element tools. The ANSYS Shape Optimisation© add-on was
utilised to identify the areas not contributing to carry the loads. The Design Explorer
ANSYS tool was linked to a CAD model in SolidWorks which was produced by
looking at the data retrieved from the shape optimisation add-on so that the
dimensions of the areas to be removed could be maximised in order to diminish the
overall structural mass. Two models were studied. One at a small scale which
showed a 15 % reduction in mass whether compared with a solid disc structure
capable of supporting the same loading conditions and one at a large scale which
gave a drop of 38 % if compared to the disc structure model obtained in Section
4.3.4.
A rotor conical structure at a large scale was also proposed for study due to the
excellent radial and axial characteristics that this type of geometry has. After the
completion of an optimisation study where the variables to be targeted were the cone
angle and the cylinder thickness, it was obtained that the best arrangement
compromised a cone sub structure with 40 mm of thickness and 10 degrees of slope
139
angle and a cylinder thickness of 20 mm. This gave an overall mass of 6104 kg and a
radial deflection of 0.497 mm. If a comparison with the optimised disc structure
presented in Section 4.3.4 is made a mass reduction of about 37.7 % is achieved.
4.5 Conclusions
Three different techniques have been proposed and verified in this chapter.
According to their reliability, suitability and speed to determine the structural
stiffness of the components forming the machine it can be said that the first approach
to be used would be the hybrid one when studying disc and arm sub structures and
complete structures for Mode 0 and Mode 1, whereas the analytical one would be
utilised in first instance to analyse Mode 0 deformation of disc and arm sub
structures. These two methods give high design freedom as the stiffness can be found
by simply introducing the required values describing the structure. They are meant to
be used in early design stages, while the FE alternative could be employed either
when none of these approaches apply or when data validation is needed.
Nonetheless, special attention should be payed when estimating the stiffness of
complete rotor and stator structures using the hybrid technique as the equations are
prone to overestimate the influence of the discs or the arms into the overall equation.
More research is necessary on the area of stiffness calculation of complete rotor and
stator structures.
On the other side, the performance of two types of structures arm and disc have been
analysed. Due to the high torques present during machine’s operation, a high
tangential stiffness is needed. Arm structures composed by a cylinder and hollow
arms requires high thickness values for the arms which makes the structure heavier
than its counterpart made with discs. For this reason it can be concluded that disc
structures are more suitable for this type of applications. With the optimum layout
identified, its structural performance was tested under different modes of deflection.
The studies revealed that the most harmful mode corresponded to mode 4. This data
can be considered of vital importance due to the engineers can make a design that
avoids this mode by altering the dimensions of the machine or by introducing
additional features such as stiffeners.
140
Further optimisation of the rotor and stator disc structures was accomplished using
finite element techniques. A large drop in mass was achieved by removing material
following a certain pattern that facilitates the manufacturing of the structure.
Looking at the results obtained and the relatively low complex shapes that can be
acquired tracking this methodology, the author recommends its use for additional
structural optimisations.
The optimisation of another type of structure as the rotor conical one has been
completed using the design study optimisation tool. This analysis has opened the
door to the exploration of other types of layouts. No further optimisation of this
structure was made although as it could be seen, the arrangement showed great
potential for structural mass savings. More progress should be made on the research
of this structure. Taking into account the results achieved, it can be concluded that
rotor conical structures are arrangements that should be considered when designing
weight sensitive rotating machinery supporting structures.
141
4.6 References
[1] A. C. J. A. Versteegh, P. Nl, and H. Nl, “Design of the Zephyros Z72 wind
turbine with emphasis on the direct drive PM generator .,” 2004, vol. 31, no.
0, pp. 1–7.
[2] A. Grauers, “Design of direct-driven permanent magnet generators for wind
turbines,” Chalmers University of Technology, Goteborg, Sweden, 1996.
[3] T. Hartkopf and S. Hofmann, M. and Jockel, “Direct-drive generators for MW
wind turbines,” in European Wind Energy Conference, 1997, pp. 668–671.
[4] M. Mueller, A. S. McDonald and D. MacPherson, “Structural analysis of low-
speed axial-flux permanent-magnet machines,” IEE Proc. Electr. Power.
Appl., 2005, 152, (6), pp. 1417-1426.
[5] A. S. Mcdonald, “Structural analysis of low speed, high torque electrical
generators for direct drive renewable energy converters,” University of
Edinburgh, Edinburgh, 2008.
[6] P. Jaen-Sola, A.S. McDonald, “A Comparative Study of Methods for
Modelling the Structural Stiffness of Generator Components”, IET Power
Electronics, Machines & Drives (PEMD), Glasgow (UK), Apr. 2016.
[7] G. Shrestha, H. Polinder, D. Bang, and J. A. Ferreira, “Structural Flexibility :
A Solution for Weight Reduction of Large Direct-Drive,” in IEEE
Transactions on Energy Conversion, 2010, vol. 25, no. 3, pp. 732–740.
[8] G. Shrestha, H. Polinder, D. J. Bang, and J. A. Ferreira, “Direct Drive Wind
Turbine Generator with Magnetic Bearing,” pp. 1–10.
[9] D. Bang, H. Polinder, G. Shrestha, and J. A. Ferreira, “Possible Solutions to
Overcome Drawbacks of Direct-Drive Generator for Large Wind Turbines,”
pp. 1–10.
[10] D. Bang, H. Polinder, G. Shrestha, and J. A. Ferreira, “Promising Direct-
Drive Generator System for Large Wind Turbines I . Introduction II .
Comparison of Different Generator Concepts.”
[11] P. Jaen-Sola and A. S. McDonald, “Structural Analysis and Characterization
of Radial Flux PM Generators for Direct-Drive Wind Turbines,” 3rd Renew.
Power Gener. Conf. (RPG 2014), Naples (Italy), 2014.
[12] A. S. McDonald, P. Jaen-Sola, “A New Method for Coupling Structural and
Magnetic Models for the Design and Optimization of Radial Flux PM
Generators for Direct-Drive Renewable Energy Applications”, IET Renewable
Power Generation Journal, Jan. 2016.
[13] P. Benham, R. Crawford, and C. Armstrong, Mechanics of Engineering
Materials, Second Edi. Prentice Hall, 1996.
142
[15] R. Pankhurst, Dimensional Analysis and Scale Factors. London: The Institute
of Physics and The Physical Society, 1964.
143
Lightweight materials in generator structures
Chapter 5
Lightweight materials in generator

structures
5.1 Introduction
Lightweight materials such as composites have been considered for applications

where structures need to comply with demanding requirements at the lowest possible
cost and weight, for example aircrafts, automobiles and ships [1]. In wind turbines,
drivetrain weight is highly correlated with machine’s capital cost and therefore to its
levelised cost too. With less weight on top of the tower, the structural requirements
for tower and foundations are less demanding. The reduction in mass means a
reduction in the cost of material and manufacturing process, as well as time savings.
144
In addition, the combination of an adequate design with the decrease in the amplitude
of fatigue loads achieved by the drop in mass contributes to enlarge the wind turbine
lifespan. In [2], Shrestha et al. suggest the use of light structural material, such as
aluminium alloy and composites as a method to reduce the weight of direct-drive
machines. In 2010, Siemens patented a wind turbine design where both the rotor and
the stator structures are made of composite materials [3].
The main aim of this chapter is to produce a lightweight design, using composite
materials, that meets all the structural requirements as described in previous chapters in
an economic manner. For this, different ways of designing generator structures must be
considered. Once the most suitable technique is identified, a rigorous process of
optimisation and evaluation have to be completed. So as to check weight savings, a
comparison with typical steel structures can be made. At the end of this chapter, a mass
comparison between a steel structure, an optimised steel structure, a composite structure
modelled by using conventional approaches and a composite structure modelled by using
a more advanced methodology will be given. The results will be presented for rotor and
stator structures.
Distinct approaches can be utilised to design an advanced composite material

structure. In Section 5.2 the most relevant factors to be taken into consideration when
designing a composite structure are outlined. Section 5.3 describes in detail the
procedure tracked to design the structure and Section 5.5 shows all the results
obtained. The last two sections, 5.6 and 5.7, correspond to the discussion of the
acquired outcomes and the conclusions reached respectively.
5.2 Composite Materials
5.2.1 What are composite materials and how do we form them?
Composites consist of a bulk material (the ‘matrix’) and some sort of reinforcement,
which is typically in the form of fibres, although particles and flakes are also
available. The main task of the fibres is to carry loads, as well as to augment the
strength and the stiffness of the matrix. They are much bigger in length than in
diameter. This makes them stronger since almost no room for defects is allowed, in
145
other words, with very small diameters the chances of having an imperfection in the
fibre are minimised. Whether an accident occurs and the fibre gets damaged
somehow, the effects would be easily noticeable permitting the engineer to take the
appropriate measures. In order to have a better binding between the fibres and the
matrix, the fibres are treated with chemicals, also known as interface. Figure 5.1
displays the different stages that must be followed when designing advanced
composite structures.
Figure 5.1 Advanced composite structure design stages [4]
The first step in the design of a composite material is to pick a suitable type of fibre.
Factors to be considered are the strength and the Young’s Modulus. Cost is also a
major issue but material properties can be balanced against the cost. Then, according
to the project’s demands, the type of matrix is chosen. At this point, plies or layers
can be formed by combining the fibres and the matrix as shown in Figure 5.1. The
higher the fibre volume fraction, the better the mechanical properties of the
composite, such as strength and stiffness. At a microscopic level, the distribution and
the orientation of the fibres, as well as their properties and the ones of the matrix
determine the properties of the composite material. As said, the main objective of the
fibres is to carry the loads, hence the higher the volume fraction of the fibres the
146
better the composite mechanical properties. Nevertheless, it is crucial not to

underestimate the importance of the matrix task. In reality, a minimum matrix
volume fraction of 30 % is needed, giving a maximum of 70 % for the fibres
according to the ‘rule of mixtures’. Theoretically, almost 91 % fibre volume fraction
can be acquired if the fibres are hexagonally ‘close packed’ with the fibres touching
each other [5].
The main goal of any manufacturing method is to properly wet the fibres with the
selected resin and consolidate the laminate in a cost effective and reliable manner.
This is achieved by a combination of elevated pressure and temperature.
Temperature is needed to initiate and sustain the chemical reactions, while pressure
is necessary to consolidate the fibres into the matrix and obtain the maximum volume
fractions. Time, also called cure cycle, is a key feature too as it determines the
production rate of parts. Cure is a transformation of uncured (liquid stage) or partly
cured polymer composites into c-stage (solid stage). Proper cure or cooking must be
achieved within the shortest time.
The most common composites currently produced can be split into three different
groups:
- Polymer matrix composites (PMC’s); these are the most common and are
made of a polymer-based resin (plastic) matrix, while the reinforcement is
created using glass, carbon or aramid fibres.
- Metal matrix composites (MMC’s); MMC’s are widely used in the
automotive industry and consist of a metal matrix (aluminium) and
reinforcement made of either particles or fibres typically of silicon carbide.
- Ceramic matrix composites (CMC’s); suitable for very high temperature
environments. These are made of a ceramic matrix with a reinforcement
usually composed by short fibres or whiskers of silicon carbide and boron
nitride [6],[7].
Advanced composite materials can be defined as composites composed by high
performance reinforcements of a thin diameter embedded into an epoxy or
aluminium matrix. Some examples are graphite/epoxy, Kevlar/epoxy and
boron/aluminium composites, which are widely used in the aerospace and naval
147
aircraft industries (for example F/A-18, AV-8B, SH-60B and CH-53E aircraft) due to
their superior strength and stiffness properties [8].
When designing an advanced composite structure many factors need to be considered

so that the most suitable material can be chosen.
5.2.2 Carbon fibre and epoxy
As said, reinforcement can be in the form of fibres, flakes and particles and it
provides strength and stiffness. There is a plethora of reinforcement types: Glass,
Boron, Carbon, Graphite (Carbon + Graphite), Silicon Carbide, Ceramic and Kevlar.
In this investigation, carbon fibre has been selected as the type reinforcement to be
used in the design of the advanced composite structure forming the generator.
Carbon fibre modulus ranges from 40 to 100 million psi and it depends on the
manufacturing techniques. The diameter of a fibre is about 7 to 8 μm. Their tensile
strength is 13 times higher than that of aluminium. They also have very high stiffness
and are electrically conductive. The microstructure of carbon fibres follows an A-B-
A packing sequence and the atoms within the layer are held by strong covalent
bonds. Fibre density is about 1.8 g/cm3. This is 65 % of that of aluminium which is
much heavier. It is important to notice that carbon fibres are nearly pure carbon and
their Young’s modulus correlates to the carbonization temperature.
Between the diverse types of matrices, epoxy offers the best properties. This
thermoset is an amorphous polymer with the molecules randomly distributed. They
have 3 dimensional crosslinks between the molecules. These links are very strong
and difficult to break even applying heat. Epoxy is also highly resistant to
environmental and solvent attack and has excellent adhesion. It shrinks very little
during cure (manufacturing process) and it can be made tougher by embedding
thermoplastic interlayers into its structure, also called toughened matrix.
Nonetheless, it is an expensive polymer which needs to be cured over long periods at
high temperatures. This makes them not suitable for mass production.
By wetting carbon fibres with liquid epoxy and applying the right combination of
elevated temperature and pressure, laminates get consolidated. Uniform high
148
pressure and high temperature contribute to more uniform structure. High quality
structures must have the fibres uniformly dispersed over the matrix due to resin rich
areas are more prone to microcracking. Nevertheless, if not enough resin is applied
(resin starved areas) shear transfer cannot take place.
Composite materials properties can be tailored by combining different percentages of

0, 45, -45 and 90 degrees plies. When weight, strength and stiffness are critical
design factors carbon/epoxy outperforms steel and aluminium. If a weight
comparison between these three materials is made it can be obtained that steel is 3
times heavier than aluminium and at the same time, aluminium is 2 times heavier
than carbon/epoxy. In addition, specific strength of composites (not only
carbon/epoxy) is much better that steel and aluminium. To tailor carbon/epoxy
structure’s elastic constants Figure 5.2 can be used as a reference. It displays how the
4 independent constants that are needed to analyse a 2D orthotropic material behave
against orientation of the plies. Note that for defining a 3D orthotropic material 9
independent elastic constants are necessary.
Figure 5.2 Laminate elastic constants for high modulus carbon/epoxy [9]
149
5.2.3 Manufacturing processes
Amongst the advantages of manufacturing composites are reduced part count,

reduced machining and assembly time and minimum number of threaded fasteners
(nuts, bolts, etc.,). Different manufacturing processes exist: hand layup, automated
layup, resin transfer moulding and vacuum-assisted resin transfer moulding.
Hand layup is a flexible and low capital investment composite manufacturing method
involving safety issues and labour intensity as plies are produced by hand, whereas
for the automated layup process the fibres are placed using a numerical control
machine with 6 degrees of freedom.
On the other hand, resin transfer moulding (RTM) is a very precise and efficient
method for which two match-dies are required. In the first instance, dry fibres are
placed into the mould. Liquid resin is injected afterwards generating pressure.
Finally, the product is heated up. It is considered as a costly and stiff process.
Nonetheless, significant savings of money and time can be made since as the high
precision offered by this technique helps minimizing trimming and finishing after
curing.
Vacuum-assisted resin transfer moulding is easier to carry out and cheaper although
it implies a certain loss of quality and mechanical properties. The main difference
between this process and common resin transfer moulding is that the resin is applied
from the top to the bottom through a hole to be vacuumed later on. This can be
produced an uneven distribution of the resin [10].
One example of a composite structure produced by hand layup methods, among

others, are wind turbine blades [11]. Automated layup techniques are taking a key
role in the production of composite frames of commercial jets [12], whereas resin
transfer moulding technologies are applied in the process of truck panels, boat hulls,
wind turbine blades again, aerospace and automobile parts, medical composites,
bathroom fixtures, car body, helmets and so on [13]. Lastly, the vacuum assisted
resin transfer moulding method is typically utilised for aircraft fuselage sections,
aircraft landing gear doors, large composite panels, wind turbine blades, high fibre
content parts, low void content parts and carbon fibre [14].
150
Laminates are created by stacking these plies following a certain sequence. The
stacking sequence will define the properties of the laminate that will be utilised in the
final stage of the advanced structure design.
5.2.4 Advantages and drawbacks of using composites
Composites can meet some of the more advanced material requirements demanded
by high performance technologies in a more lightweight manner than metals and
alloys. Metals are isotropic and homogenous (they have the same properties in all
directions) whereas composites are anisotropic and inhomogeneous (they have the
same properties at all points in the body), with properties 400 times better in some
directions.
Other advantages for composites are enhanced fatigue and impact resistance,
strength, stiffness, corrosion resistance and thermal conductivity [10]. Components
made with these materials last longer and need less maintenance and thus, fewer
inspections. Larger elements can be designed leading to quicker assemblies.
Significant cost savings in manufacturing, machining and assembly processes can be
achieved by utilising composites. It is estimated that if well-designed the reduction in
the cost of the parts forming an aircraft and the labour wages can be around 25-30 %
[10]. Furthermore, as the Young’s Modulus to density ratio is very high substantial
mass reductions of up to 30 % can be achieved by making use of this type of
materials. In the case of airplanes, it is also more efficient to use composites since by
reducing their mass the fuel consumption sees itself decreased too [15].
However, composites also present a set of drawbacks that must be taken into
account. Fabrication cost is a critical concern. Repair of composite materials is a
complex process that requires very skilled labours to detect flaws and cracks. In [16],
Su presented a review on the state of the art of Lamb wave-based damage
identification approaches for composite structures, showing the latest advances in
this technique. Moreover, mechanical characterization of composite structures is
more complicated than that of structures made of metal, as composites properties are
not equal in all directions.
151
Composite materials are brittle, therefore higher stresses can be handled with little
changes in strain. If a comparison of the structural behaviour between a specimen
made of a ductile material, for example steel, and a specimen made of a brittle
material, such as composites, both loaded with tension forces is made, ultimate
failure of composites occurs without going through a plastic stage as it is shown in
Figure 5.3. Energy is absorbed causing internal delamination (separation of layers)
that causes the sudden collapse of the structure. This particular characteristic can be
considered advantageous in our case as very little deformations are allowed ensuring
so that the airgap deflection will be maintained within the limits. However, the fact
of not having a plastic stage makes hard to identify any structural damage that can
lead to a sudden collapse. In a breakdown event, replacement of the whole machine
would be probably required as the composite features make them extremely difficult
to repair.
Figure 5.3 Stress-strain curves for Ductile (left) and Brittle (right) materials
5.2.5 Other factors to be considered in design
The thickness of laminates is an issue that needs to be addressed since with thinner
laminates (10 to 16 plies for example), the energy from loading deforms the laminate
and what is left delaminates the laminate. For thicker laminates, it works the other
way around. In other words, the energy is spent on delamination purposes and the
remaining is left for the deformation of the laminate itself. The thickness of the
laminates is a design choice usually made according to the loads that the structure
must withstand. The larger the loads, the thicker the laminates. As very large torque
152
and radial loads are expected for direct drive machines, the use of thick laminates
will be necessary. Bearing all of this in mind, a suitable factor of safety must be
taken into account. Figure 5.4 displays an example of delamination acquired during
the stator structure design stage.
Figure 5.4 Composite structure delamination [17]
Finally, corrosion resistance of composites needs to be considered. Plastics do not

corrode even from salt water. Nevertheless, when aluminium, magnesium, cadmium
or steel are used with composites a galvanic type of corrosion can happen, as these
materials own different electrical properties. To avoid galvanic corrosion, titanium
fasteners are required. They are more expensive and they will add more weight.
5.2.6 Establishing the design conditions: loading
The design process starts by looking at the project’s requirements. Different types of
loads can act on the assembly and the structure must be ready to resist them without
being damaged in any way. There are four key loads that every single structure must
be able to withstand: tension, compression, shear and flexure.
153
a) Tension
A composite under tensile loads can respond in many different ways depending on
the tensile stiffness and strength of the reinforcement fibres (stiffness and strength
properties are higher than the resin based matrix on its own). Figure 5.5 illustrates
the four typical loads acting on a composite.
(a) (b)
(c) (d)
Figure 5.5 Composite loading: (a) Tensile; (b) Compressive; (c) Shear and (d) Flexural [6]
b) Compression
Under compressive loading the adhesive and stiffness properties of the resin system
(matrix) are critical in order to maintain the fibres of the reinforcement straight at all
times and protect them against buckling.
c) Shear
Under a shear load, the matrix transmits the stresses across the composite. Excellent
mechanical properties are desirable, as well as high adhesion to the reinforcement
fibres, for the composite to withstand this load, which tries to slide nearby layers of
fibres over each other. In a laminate, the interlaminar shear strength (ILSS) is
employed to designate this property. This limiting property must be taken in to
consideration when approaching a composite structure design. Nonetheless, it is also
154
convenient to remember that the use of different geometric configurations for the
structure and fasteners will affect this property.
In the case of having no matrix, if a fibre breaks the whole structure would become
useless. However, with the use of a matrix, shear stress transfer is achieved. This,
also called crack arresting system, allows the structure to keep working even if a
fibre breaks in that the fibres located within a short distance of the broken fibre are
able to recover the stress. This is an indirect contribution of the matrix [18].
d) Flexure
Under flexural loads the structure experiences a combination of tensile load (lower
face), compression load (upper face) and shear load (central section of the laminate)
as can be seen in Figure 5.5 d).
Fatigue behaviour of composites is a remarkable point to be considered when

designing composite structures. With metals, after 10 million cycles only 20 % of
their strength remains whereas with composites 95 % of the properties are still usable
[10]. Composites work better under tension than under compression. Despite this,
their structural behaviour is still better than metals in both cases. In order to design
composite structures, stress concentration must be analysed.
Typical understanding of fracture mechanics cannot be employed when designing

composite structures because fracture toughness is not a material constant. For
composites, a zone of damage can be noticed while for metals the most crucial
damage is a sharp crack. Composite fracture mechanics is described by the Mar-Lin
mathematical model which is widely used in industry. This model is for predicting
the strength of composites with open holes and cut outs. To calculate the mentioned
residual strength of the structure, ‘𝜎N∞ ’, the Mar-Lin formula is as follows [10],
𝐻C (5.1)
𝜎N∞ =
2𝐿𝑚
155
where 𝐻C is the composite fracture toughness having units of stress, 𝐿 is the half
notch length and the exponent 𝑚 is related to the stress singularity at the crack tip of
the bimaterial interface. The order of singularity depends on the ratio of the shear
moduli of the matrix and the fibre, and the Poisson’s ratio [10].
5.2.7 Mechanically analysing a composite structure
Since composite materials are composed by two or more elements, their analysis is
completely different from that of metals.
Looking at the micromechanics of a lamina, the average properties of the ply can be
found by observing the properties of the said ply individual components. The
following assumptions when considering the fibres and the matrix are usually made:
- Both fibres and matrix are linearly elastic

- The fibres are infinitely long
- The fibres are spaced periodically either in square-packed or hexagonal
packed arrays
Three distinct techniques are available so as to calculate the elastic constants for the
composite material founded on micromechanics:
- The use of numerical methods (FE)

- The use of models grounded on the theory of elasticity
- The use of the rule of mixture models based on strength of materials method
If the structure needs to be analysed looking at the macromechanics of a laminate the
use of failure theories, such as the Classical lamination theory, and the development
of stress-strain relationships are required.
5.2.8 Classical lamination theory
In finding the properties of a laminate, the use of the Classical lamination theory is a
must. Superimposing the properties of each ply is not an option. This theory is
utilised to estimate internal stress state, dimensional stability of laminated
composites and stiffness. It couples extensional, shear, bending and torsional loads
156
with strain and curvatures. By introducing an environmental load analogy, residual

strains and warpage caused by differential shrinkage and swelling of plies in a
laminate can be included in lamination theory as well. The ABD matrix shown below
describes the united influence of diverse types of loads and moments on laminated
plate response. In the equality, N corresponds to loads, M are moments, ε are strains
and κ are curvatures. Aij are extensional and shear stiffnesses, Bij are extension-
bending coupling stiffnesses and Dij are bending and torsional stiffnesses. For further
information on where ABD elements come from see [9].
Nx A11 A12 A16 B11 B12 B16 εx

Ny A12 A22 A26 B12 B22 B26 εy (5.2)
Nxy A16 A26 A66 B16 B26 B66 εxy
=
MX B11 B12 B16 D11 D12 D16 κx
My B12 B22 B26 D12 D22 D26 κy
κ
[Mxy ] [B16 B26 B66 D16 D26 D66 ] [ xy ]
Three distinct factors determine the mechanical behaviour of laminates: symmetry,

balanced vs. unbalanced and stacking sequence.
A laminate is considered symmetric whether for every ply to one side of the laminate
reference plane with specific material properties, a specific thickness and specific
fibre orientation, exists another ply at the same distance on the opposite side of the
plane with the same material properties, thickness and fibre orientation. If this is not
the case, the laminate is referred to as an unsymmetric laminate. Unsymmetric
laminates can have up to six different types of deformation caused by a single
applied load. When studying symmetric laminates, all the extension-bending
coupling stiffnesses (B elements) are zero. It is important that the elements of the B
matrix tend to zero in order to avoid laminate curvature.
A balanced laminate is one which for every ply with specific material properties, a
specific thickness and specific fibre orientation, there is another ply somewhere in
the laminate with the same specific material properties, specific thickness and
157
opposite fibre orientation. A16 and A26 components are always zero for balanced
laminates. Therefore, no shear will take place.
Symmetric and balanced laminates are desirable so that neither warpage nor shear
can occur. Then, the laminate will nearly behave as an isotropic material.
Matrices B and D are dependent of stacking sequence. So as to reduce deformation,

D matrix elements should be as big as possible.
5.3 Composite Structure Modelling
In order to model the supporting composite structure for the electrical machine of a
direct drive wind turbine, a disc structure was selected and a finite element approach
adopted. As Abdul-Aziz and Uddin propose for flywheels in [19] and [20]
respectively, the ANSYS Parametric Design Language (APDL) was employed to
carry out this task as it allows easy parametric changes in material, geometry, mesh,
loads and boundary conditions.
The material selected was carbon-epoxy due to its excellent strength-to-weight ratio.
Table 5.1 shows the material properties used for study,
158
Table 5.1 Material properties for a filament wound disc [20]
EX 2.03×1011 Pa
EY 1.12×1010 Pa
EZ 1.12×1010 Pa
vXY 0.33
vYZ 0.45
vXZ 0.33
GXY 8×109 Pa
GYZ 3.8×109 Pa
GXZ 4.2×109 Pa
ρc 1,600 kg/m3
δ 5×10-6 m
θ 1.96×10-11 m2
where E corresponds to the Young’s modulus, G is the shear modulus, v, is the

Poisson’s ratio, ρc is the density, δ is the distance between fibres and θ is the cross-
section area of a single fibre. Data for the Shear modulus in the YZ and XZ planes
where estimated using equation 5.3:
𝐸 (5.3)
𝐺=
2(1 + 𝑣)
Assuming a perfect bonding between fibres and matrix, the distance between fibres,
δ, and the cross-section area of a single fibre, θ, were picked so that the highest
possible fibre volume fraction can be obtained. The fibre volume fraction utilised in
these analyses was 90%.
The type of element utilised in the analysis was Shell281. As seen in Figure 5.6, it
owns eight nodes (I, J, K, L, M, N, O and P) with six degrees of freedom at each
node: translations in the X, Y and Z axes, and rotations about the X, Y and Z axes. It
is suitable for studying thin to moderate-thick shell structures. The element accounts
159
for consequential effects of distributed pressures [21]. Figure 5.6 displays the
geometry, node locations and the element coordinate system.
Figure 5.6 Shell281 Geometry [21]
The model of the structure was created using the GUI tool available in ANSYS
APDL. It was constrained at the shaft and three different sorts of loads were applied
to the generator structure:
- Radial expansion load acting on the rim sub-structure of 400 kPa

corresponding to the Maxwell stress. This corresponds to an airgap flux
density of 1 T.
- Tangential load again acting on the rim of 30 kPa corresponding to the shear
stress. This loading was divided by four and applied in the form of forces of
120kN for the large scale model and 5.5 kN for the one at a small scale, to the
keypoints located at the cylinder midspan and equally spaced. This can be
seen in Figure 6.6.
- Gravitational loading in the Y axis of 9.81 m.s-2. Gravity was also tested in
the Z direction in order to consider the weight effect during the generator’s
transportation stage.
Figure 5.7 depicts in detail all the loads acting on a generator rotor disc structure
model.
160
Cylinder sub structure
Disc sub structure
Figure 5.7 Loads acting on the generator rotor structure
Two distinct methods were used when creating the model of the disc that would be
merged with the outer cylinder sub structure lately: the conventional approach and
the mosaic pattern disc approach as stated by Morozov in [22].
In the conventional approach (CA), the disc is modelled as a laminated circular plate
consisting of a certain number of plies; see Figure 5.10, whereas the mosaic pattern
disc is composed by four plies of different thicknesses and orientations depending on
the corresponding area. For this model, the disc area is divided into sub areas as it
can be seen in Figure 5.11.
Electrical machines structures for direct driven wind turbines were modelled
according to at both small and large scales. The benefits of composites are likely to
be different at these scales, and so two case studies were performed. At the small
scale, a 100 kW buoyant airborne wind turbine, also known as BAT, created by
Altaeros Ltd. [23] was investigated. This device shares much of the wind turbine
technology with its grounded cousins and also introduces additional constraints. One
such limitation is the requirement of the turbine equipment to be lightweight. With
161
this is mind, the design of the supporting structure for a generator of 0.42 m radius
and 0.21 m axial length rotating at 140 rpm and with a torque of 6.8 kNm was started
first following the conventional approach to then continue with the more complex
one. In the early stages, it was thought that fibre reinforcement of certain plies could
enhance the stiffness characteristics of the machine modelled using the more
complicated configuration. This option was tried, as shown in Figure 5.12, and
discarded as it was observed that it did not introduce any further noticeable
improvement to the design. The mesh for the small scale model was created
following the same procedure as stated in the following section. However, it is
important to highlight that the size for the quads of the inner/outer areas was 0.05m,
whereas for the rest was of 0.005 m. The elements in the areas at the edges had to be
refined twice.
Having the small scale structure modelled and tested, it was proceeded to draft a 3
MW large scale case study model. The same approaches were tracked to design this
structure of 2 m radius and 1.2 m axial length again with the purpose of minimising
its mass.
With the composite structures finished, a mass comparison with models made of
steel (AISI 304) and designed to comply with the same structural requirements, this
means, as already mentioned, that deflection in all directions must remain below the
stipulated limit of 10 % of the airgap size. For the small scale model that is equal to
0.208 mm, whereas for the large scale it is equal to 0.5 mm.
Two electrical machine supporting structures made with discs have been modelled
following the methodology described in this section, one at small scale and one at
large scale corresponding to 100 kW and 3 MW wind turbines, respectively.
So as to come up with a design that can fulfil the structural requirements stated a
factorial design of experiments was developed for both the small scale and large
scale models. For the models with conventional configurations a design of
experiments process was completed for each section (disc and cylinder). For the
more mosaic pattern composite structure a design of experiments process was carried
out for every area forming the disc structure and for the cylinder.
162
After those procedures were terminated it was obtained that a stacking sequence of
90/0/45/0/90 was the one presenting less deflection for both the disc and the cylinder
of the CA structures no matter the size. With this in mind, an optimisation process of
the thicknesses of the plies was made. It was observed that the influence of the 90
degrees and the 45 degrees plies in the cylinder was negative and elevated the
deflection in the sub-structure for both the small and the large scale models.
Therefore, they were removed. The deflection for each attempted stacking sequence
of the laminates forming the CA structure with 5 mm plies is given in Table 5.2.
Figure 5.8 Stator structure according to the conventional method
5.4 Investigation
5.4.1 Conventional Approach
For this model, the construction of a mesh of an adequate density was determined by
an independence study which gave a result of 0.05 m quad element size. A mapped
mesh was used as depicted in Figure 5.8 for the generator with a CA disc sub
structure. The orientation of all the fibres within this model was set according to the
original Cartesian coordinate framework and it was assigned to each element through
the creation of two different sections, one for the disc sub-structure and one for the
cylinder sub-structure. In order to find the best stacking sequence for either the disc
or the cylinder a factorial experimental design was carried out [24]. Three different
163
thicknesses (0.005 m, 0.01 m and 0.02 m) and three different fibre orientations of 45,
0 and 90 degrees were assumed. All the possible combinations for the stacking
sequence (keeping symmetry), see Table 5.2, and thicknesses were tried as stated by
the factorial design. For each stacking sequence, the thicknesses of two orientations
were fixed to 0.005 m, whereas the one remaining was changed.
After attempting the 54 possible combinations, it was found that 90/0/45/0/90 was
the most suitable for the disc and the cylinder sub-structures. The deflection results
obtained for the six stacking sequences with all the plies with 0.005 m thickness are
given in Table 5.2.
Table 5.2 CA Composite Structure Stacking Sequence
Stacking Outer Inner Mid Inner Outer Deflection (mm)

Sequence Ply (1) Ply (1) Ply Ply (2) Ply (2) for 0.005 m ply
thickness
1 90 0 45 0 90 0.139
2 0 90 45 90 0 0.14
3 45 0 90 0 45 0.14
4 45 90 0 90 45 0.141
5 90 45 0 45 90 0.14
6 0 45 90 45 0 0.14
The last step was to optimise the thickness of plies taking into consideration the
deflection limit. In the cylinder case, it was observed that the 45 and 90 degrees plies
did not introduce any improvement in the stiffness of the overall generator structure,
hence they were removed, leaving the 0 degrees ply alone, which corresponds to the
hoop direction. Figure 5.9 depicts a flowchart describing the process, while in Figure
5.10, the composite material fibre orientations utilised in this investigation can be
recognised.
164
Figure 5.9 CA composite structure design process used in this chapter
165
(a) (b)
(c)
Figure 5.10 Example of Composite Fibre Orientations & Stacking Sequences for the Generators
Components; (a) Following a Cartesian Coordinate Frame; (b) Following a Cylindrical Coordinate
Frame; (c) For the Cylinder Sub Structure Following a Cylindrical Coordinate System
5.4.2. Mosaic Pattern Approach
For the mosaic pattern disc sub-structure, 5 cylindrical coordinate systems were
created so that the orientation of the fibres in the different areas could be assigned.
Eleven unlike sections were generated for the disc, while for the cylinder only one
was needed. Afterwards, the sections were applied to the areas. In the first instance, a
quarter of the rotor was developed and then reflected in the X direction with the YZ
plane acting as a mirror. Finally, the resultant structure was again mirrored in the Z
axis direction. The same procedure was utilised to model the stator structure. See
Figure 5.11.
166
Figure 5.11 View of the areas forming the mosaic pattern composite disc sub structure
A more concise mesh control had to be used for this model due to the sharp edges of
the outer and inner areas of the disc. As a result, the element size for the mentioned
areas was first set to 0.1 m. A free quad mesh was employed, although a triangle one
could also be used especially taking into account the shape of these zones obtaining
good outcomes. Then, the elements generated were refined twice using the minimal
level of refinement available, or level 1, which produces an element of half the
length of the original element [21]. The rest of the areas were meshed utilising a free
quad mesh of 0.05 m element size and no further refinement was necessary. The
main aim to use a more refined mesh in the inner/outer zones was to control the
shape and the size of the disc elements merging with the cylinder ones and to avoid
errors in the grid creation. See Figure 5.12.
The models with a more advanced structure, also known as mosaic pattern composite
structures, were designed following the guidelines specified above. However, for the
first attempt (the small scale model), fibre reinforcement and only one fibre
167
orientation per area as stated by Sayem in [20] were utilised as observed in Figure
5.12, where the fibre orientations have been highlighted with coloured arrows
matching the colour of the origin of the cylindrical coordinate system that they track.
The fibre reinforcement option available in ANSYS APDL allows the user to
simulate a cluster of fibres placed according to the structure needs in order to
enhance its mechanical properties. Certain key zones were reinforced using smear
layers of the same material so that the stiffness of the structure could be improved. A
secure bond between the reinforcing fibres and the base element is assumed by ANSYS
and the motion of the reinforcing fibres is determined solely by the motion of the base
element as the relative movement between the mentioned components is not allowed
[21].
Despite the potential of this tool no influence in the properties of the structures was
noticed. Moreover, the lack of flexibility in the design made extremely difficult to reduce
the mass and comply with the imposed specifications at the same time. Thus, this option
was discarded. More flexibility was given to the design as requested in search of decent
results that would set the path to track. Following the preceding established with the
CA generator structure, a factorial design of experiments was carried out utilising the
same thicknesses and fibre orientations but in this occasion for each area. It was
checked that the most suitable fibre orientation for the rim was still 0 degrees and the
addition of any ply of any thickness or fibre orientation makes the deflection go up.
Once the most favourable orientations were found (for either the disc areas or the
rim), the thicknesses were optimised so that the overall weight of the structure could
be minimised. Thickness of areas with small sizes was increased as much as possible
so the thickness of the largest areas could be decreased. This trade off process of
optimisation went on until the deflection limit was reached.
168
11
15 12
11
11
13
13 15 11
11 12 11
11 11
14
11
11
13
12
Figure 5.12 Detailed view of the composite rotor disc structure based on Morozov’s model
highlighting variable fibre orientation (small green bits) and reinforcement in purple [25]
169
5.4.3 Conventional Approach vs. Mosaic Pattern Approach
A comparison between the stress plots for a CA and mosaic pattern composite
structures is depicted in Figure 5.13. Stress concentrations in the disc sub structure
have been cancelled in the mosaic pattern structure due to the effect of the fibres
interlacing.
(a)
170
(b)
Figure 5.13 Contour plots highlighting stress in the radial direction (a) CA composite rotor structure
(b) Mosaic pattern composite rotor structure
The contour plot illustrated in Figure 5.13(b) shows a result of -0.472×107 Pa of

tensional stress across the disc, with certain areas working again under tension at
even less stress (-215170 Pa) for the more advanced model, while the CA model
shows a combination of stress loading of distinct order of magnitude (tension and
compression of -0.564×107 and 0.156×108 Pa respectively) that will have a big
negative influence in the life of the structure.
Table 5.3 shows the effect of reducing the thickness of the critical areas of the large
scale model mosaic pattern disc sub-structure on the deflection and by definition on
the stress. Areas 2 and 3 were not at play as they produced either no noticeable
improvement or unacceptable deflections with large stress concentrations.
171
Table 5.3 Effect Caused by the Reduction of Critical Areas Thickness
Areas Thickness Mass (kg) Total Deflection

Reduction (mm) (mm)
1 4 8 0.2
4/5 2 4 0.1
8/7 2 8 0.2
9 2 8 ~ 0.1
10/11 2 9 0.4
As understood, the combination of areas 10 and 11, which correspond to midspan

areas of the disc sub structure as Figure 5.11 illustrates, are the ones giving the
highest deflection, although they are also taking away the highest amount of material
with 9 kg per 0.2 mm of thickness reduction. The combination of areas 4 and 5 give
the lowest deflection but with the lowest mass reduction, whereas area 1 and the
combination of areas 8 and 7 offer fair results with 8 kg per 0.2 mm of deflection.
The best outcomes were obtained by minimising the thickness of area 9. As seen, 8
kg of mass drop was achieved per 0.2 mm of thickness reduction which gives ~ 0.1
mm of deflection.
5.5 Results
Different materials and design approaches for electrical machine supporting

structures have been presented in this investigation. Suitable disc steel structures
have been proposed for both the small and big scale models. As presented in Chapter
4, the total mass (includes rotor and stator structures) of the small scale model made
of steel was 313.2 kg with 114.76 kg for the rotor and 198.46 kg for the stator. A
further optimisation of this structure was carried out obtaining a total mass of 266.7
kg, with a mass of 91.1 kg for the rotor and 175.6 kg for the stator. This means a
drop of about 15 %.
172
With the lightest steel structure configuration identified for the small scale model, the
use of composite materials in the structure design was attempted tracking two
distinct methods named conventional approach, ‘CA’, and mosaic pattern approach.
They have been already described in Sections 5.3 and 5.4.
In the case of the CA small structure, the rotor mass was 34.1 kg, while the stator
mass went up to 41.4 kg giving a total of 75.5 kg. If a comparison with the optimised
steel structure is made, it can be appreciated that a significant reduction of nearly 72
% was achieved.
For the mosaic pattern fibre orientation generator structures, it was estimated that the
mass for the small scale rotor structure is 16.96 kg. The stator structural mass was
33.25 kg. This makes a sum of 50.21 kg, which gives us a difference of about 25.3
kg with the conventional configuration model. This means an overall reduction in
mass of about 33.5 %. The results obtained for the small scale model structures have
been plotted in Figure 5.14 for comparison.
350
300
250
Mass (kg)
198.46
200
175.6
150
100
50 114.76 91.1 41.4

33.25
34.1 16.96
0
Steel Optimised steel CA Composite Mosaic Composite
Type of Structure
Rotor Stator
Figure 5.14 Small Scale Model (Mass vs. Type of Structure)
Large scale models corresponding to a 3 MW electrical machine supporting structure

were developed following the same procedure. A steel structure was modelled, as
described in Chapter 4, achieving a result of 9809 kg for the rotor structure and 9451
173
kg for the stator. Adding up these values, it was obtained an overall mass 19260 kg.
After further optimisation, a decrease of 38 % was acquired. The total mass for the
optimised model corresponded to 12000 kg with the rotor accounting for 5694 kg.
The mass of the stator structure was 6306 kg and as it can be seen, its optimisation
contributed less to the weight reduction than the rotor one.
Regarding the composite structures, in the case of the CA large scale model, the rotor
structure had a mass of 2488.5 kg, whereas the stator one was of 4346 kg. The total
mass of the generator structure was 6834.5 kg, which means a drop of 43 % if
compared to its optimised counterpart made of steel.
For the mosaic pattern approach large scale model, the rotor had an estimated mass
of about 2418 kg. The mass of the stator structure was about 4550 kg giving us an
overall value of 6968 kg. The difference between this model and the one with the
conventional layout was 133.5 kg and it can be neglected. A tiny saving in mass of
70.5 kg could be made for the rotor, while for the stator a negative result of 204 kg
was understood. Figure 5.15 shows the outcomes achieved for the large scale
structures.
20000
18000
16000
14000 9451
Mass (kg)
12000
10000
6306
8000
6000
9809 4346 4550
4000
5694
2000
2488.5 2418
0
Steel Optimised steel CA Composite Mosaic
Composite
Type of Structure
Rotor Stator
Figure 5.15 Large Scale Model (Mass vs. Type of Structure)
Figure 5.16 shows the result achieved for the final design of the mosaic pattern
approach stator structure. The concentration of deflection highlighted in green on the
174
cylinder corresponds to the location of the tangential forces, which were placed like
that as the available ANSYS license had a limited number of nodes that could be
loaded. With a load uniformly distributed along the rim, it is thought that less
deflection would be obtained allowing for further optimisation of the structural mass.
Figure 5.16 Contour plot displaying maximum deflection in the hoop direction of the mosaic pattern
approach composite stator structure
Having said that, further improvements might be achieved by considering designs

with larger number of areas as proposed by Sayem in [20] that would provide with
the necessary flexibility to the design.
5.6 Discussion
Structures made of steel have been assumed and analysed with the main of reducing
their mass. Under radial, torque and gravitational loading conditions and constraints
as described in Section 5.3, the minimum mass for a conventional steel structure was
of 313.2 kg. However, if the steel structure is optimised a substantial difference can
175
be obtained. The total mass for an optimised steel generator structure would be 266.7
kg, which means that an overall mass drop of 15 % can be achieved.
On the other hand, two different ways of designing electrical generator supporting
structures made with discs for direct-drive wind turbines using low density materials,
such as composites, have been also presented. First, a small scale model was created
following the conventional approach and tested under the typical loads acting on the
drivetrain of a direct driven machine of 0.42 m radius, 0.21 m axial length, 2.08 mm
airgap, 140.1 rpm rotor speed and 6,817 Nm torque corresponding to a 100 kW wind
turbine. Then, a mosaic pattern configuration was modelled and tested under the
same parameters. After a comparison, a difference in mass between the two models
of 25.3 kg was achieved. This is equivalent to a reduction of up to 33.5 %.
Furthermore, whether a comparison with the optimised steel model is made an
overall mass reduction of 72 % in the case of the conventional model and 82 % in the
case of the more advanced mosaic pattern design have been reached.
It could be observed that higher mass savings can be acquired for the rotor structure
than for the stator structure. For instance, in the mosaic pattern approach model case,
a decrease of almost 50 % was achieved for the rotor, while a 20 % was reached for
the stator. This difference can be explained by looking at the sort of loading acting
on the structures in question. The rotor structure is subject to a radial expansion load
uniformly distributed along the outer surface of the rim, besides the tangential and
gravitational loads. On the contrary, the stator is subject to a radial load acting on the
inner surface of the rim and pointing inwards besides the tangential and gravity
loads. This means that the rotor structure is mainly working under tension when the
stator is working under compression. Performance of composites decreases when
they work under compression. Typically, mechanical properties of composites
working under compression are about 30 % lower than when subject to tensile
stresses. Due to this important factor, the thicknesses of the largest areas that form
the disc sub-structures of the stator had to be increased.
In addition, the configuration of the structures is completely different with the rim of
the rotor being supported by a disc concentrically located and the cylinder of the
stator being maintained by two discs placed at its edges. The stator layout can be
176
considered much weaker as the area of the rim not being supported is larger than that
of the rotor.
Generally speaking, it can be said that the mosaic pattern model without
reinforcements and using multiple fibre orientations gave better results if compared
with the conventional model because of the addition of more flexibility to the design.
The use of fibre reinforcement and a fixed fibre orientation for each area was seen as
a constraint during the early stage of the design as mass optimisation could not be
carried out if the structural requirements wanted to be fulfilled. For this reason it was
discarded.
By dividing the disc sub structure area into smaller zones that can have different
layups, the designer introduces more variables to play with to this parametric model.
The use of several plies with unlike fibre orientations that follow five distinct
coordinate systems allows the designer to manage the stress by spreading it or simply
concentrating it in purpose into smaller areas. With this, minimisation of thickness of
large areas can be achieved. Special care has to be taken regarding the nature of the
stress. When designing a structure like this, the avoidance of cyclic loading of large
amplitudes is a must. Taking into account the brittle behaviour of this sort of
materials, the variable thickness of the structure will play a crucial role in its working
life.
The design of the composite structure for the large scale model with a radius of 2 m
and an axial length of 1.2 m was made following the same procedure as for the small
scale one. Very significant mass reductions were acquired for both models, the
conventional one and the mosaic pattern one if compared with a structure made of
steel. As explained in Chapter 4, the steel structure was optimised by using the
Design Explorer tool of ANSYS Workbench giving a result for the mass of the
whole structure of 12000 kg. The rotor accounts for 5694 kg while the stator weights
6306 kg. In the worst case scenario, a mass drop of 6968 kg could be achieved by
utilising composite structures, which means a decrease of nearly 42 %. Bearing this
mind, it can be said that the use of this type of lightweight materials is a viable
option whether a significant structural mass reduction is pursued.
177
Nevertheless, it is essential to point out that the mosaic pattern design was not able to
make any further improvement to the mass reduction. Yet, the structural performance
of both the rotor and the stator were enhanced as the cyclic loading acting on the
discs and on the rim individually was eliminated.
It is considered that further mass reductions could be achieved if the designer goes a
step beyond and introduces even more flexibility to the mosaic pattern model.
Models (a) and (b) displayed in Figure 5.17 have been analysed in this research. Due
to time constraints, designs (c) and (d) could not be explored. However, the author
thinks that they would introduce the required resilience to approach more ambitious
optimisations. As suggested in Section 5.4, the use of more variables would add
more complexity to the design. For this, the development of a script that automates
the process is recommended. A powerful computer would be necessary to handle and
process the amount of data that these simulations would generate.
Looking at the structural design of the machine from the manufacturing perspective,
it can be easily asserted that the more complexity added to the design, the more
expensive will be to manufacture the structure. Therefore, the cost of manufacturing,
which would include the cost of material, wages of skilled labours, the use of special
tooling and handling equipment amongst others, should be estimated before pursuing
such investigation. Nevertheless, if the amount of material to be removed is large
enough the overall cost could be balanced.
The structural performance has been enhanced, as it has been demonstrated, by

utilising the mosaic pattern model. Working life of the structure is another important
input to be included when carrying out the risk analysis of the project.
Last but not least, certain simplifications and assumptions were made during the
accomplishment of this study. For example, thermal aspects were no considered. The
fibres utilised to form composite materials own very low coefficients of thermal
expansion. Nonetheless, the sensitivity of the composite structure thermal expansion
coefficient to variations from fibre orientations can be significant as thermal response
of the composite materials can be tailored directionally as desired by placing the
laminates in the appropriate manner [26]. Due to the complexity and the variety of
178
approaches that could be followed, it was thought that a more concise analysis was
required. For this reason, a deep investigation of this field and its effects on the
design of any composite structure proposed is recommended.
On the other hand, it was considered that steel back iron does not add any strength or
stiffness to the structure. By taking into account this feature, more accurate structural
optimisations might be achieved.
Finally, this research has no concerns about the connection of any composite
structure presented here to the shaft, bearings, turret, bed plate and rotor and stator
back iron that can be very challenging. Taking the aircraft industry as an example,
titanium fasteners are usually employed to join metallic parts with composites, in
order to avoid galvanic corrosion. This would add an extra weight that would be
necessary to include. In addition, the cost of these hardware devices would need to
be taken into consideration.
179
(a) (b)
(c) (d)
Figure 5.17 Potential disc sub-structure designs (a) Conventional model (b) mosaic pattern approach
disc (c) 6-unit disc (d) 8-unit disc [20]
5.7 Conclusions
In this chapter, the possibility of using low density materials in order to minimise the
mass of the overall structure has been explored. Two different approaches have been
utilised: a conventional method and a more advanced approach, which divided the
full area of the disc sub-structure into smaller zones. Both techniques were employed
to optimise the mass of the rotor and the stator of two models one at a small scale
and another at a large scale. A comparison between the small scale model created
180
following the conventional approach and an optimised steel structure under the same
loads and constraints showed that the composite structure was 72 % lighter than the
steel one.
Afterwards, an evaluation of the mass of the CA structure and the structure modelled
tracking the mosaic pattern approach revealed a difference of 33.5 % in favour of the
mosaic pattern design, as well as a better structural performance. The mass of this
model was also contrasted with the one of a model made of steel working under the
same circumstances, giving a mass drop of about 82 %.
A similar procedure was chased for the large scale model. The masses of the models
designed following the conventional and the mosaic pattern approaches were
contrasted with their optimised counterpart made of steel revealing a drop in mass of
about 60 %. Later on, the masses and performances of the two composite structures
were compared. No significant difference in mass was observed, although the more
advanced mosaic pattern model showed a better structural performance. Further
improvements might be acquired if more flexibility is added to the design.
Looking at the results obtained, it can be concluded that the use of materials with
higher Young’s Modulus to density ratios, such as composites, is without doubt, an
option to be taken into consideration when attempting the design of lightweight
electrical generator structures or any other type of weight sensitive rotating
machinery working in a similar environment.
The high cost of these materials and the complexity that they introduce into the
design could make the engineers reject its use. During the feasibility study that needs
to be accomplished before giving the go-ahead to the project, the advantages brought
to the design by the composite materials cannot be ignored. As explained in Section
5.2.4, larger components can be produced. Less number of parts will be then joint
and less number of fasteners will be needed saving money, weight and time. The
decrease in the amplitude of fatigue loads and cancellation of stress concentrations
achieved with a concise design would contribute to extend the lifespan not only of
the generator but also of the entire wind turbine. Moreover, if well designed, thermal
and corrosion issues might be taken out of the equation. The opinion of highly skilled
181
engineers with capacity to assess correctly the economic impact of all of these factors
would be needed at this early phase.
182
5.8 References
[1] A. American Composites Manufacturers, “Where are composites used?”,

www.acmanet.org/composite/where-are-composites-used, 2016. Last
Accessed [05/05/2016].
[2] G. Shrestha, H. Polinder, D. Bang, and J. A. Ferreira, “Structural Flexibility :
A Solution for Weight Reduction of Large Direct-Drive,” in IEEE
Transactions on Energy Conversion, 2010, vol. 25, no. 3, pp. 732–740.
[3] H. Lind, S. O. Stiesdal, “Wind Turbine,” Patent, WO 2013/083386 A2.
[4] ANSYS, “ANSYS Webinar on advanced composite structures analysis,” 2015.
[5] University of Plymouth, “Laminate Thickness and Vf,” Laminate thickness
and Vf presentation, University of Plymouth. www.tech.plym.ac.uk/. Last
Accessed [01/09/2016].
[6] Gurit, Gurit - Wind Energy Handbook (Guide to composites), GTC-3rd–0509
ed. www.gurit.com.
[7] Ronald F. Gibson, Principles of composite material mechanics. McGraw-Hill
Inc, 1994.
[8] Aviation Structural Mechanic (H&S) 3&2, “How airplanes are built and how
to maintain them,” http://navyaviation.tpub.com/14018/css/Composite-
Material-56.htm. Last Accessed [01/09/2016].
[9] US Dept. of Defense Handbook, "Composite Materials Handbook. Polymer
Matrix Composites", vol. 3, June (2002).
[10] University of WashingtonX, “Composite materials overview for engineers
AA432x,” 2015.
[11] S. C. Nolet,TPI Composites, “Composite Wind Blade Engineering and
Manufacturing,”
http://web.mit.edu/windenergy/windweek/Presentations/Nolet_Blades.pdf,
2011.
[12] C. World, “ATL and AFP Defining the Megatrends in Composite
Aerostructures,” http://www.compositesworld.com/articles/atl-and-afp-
defining-the-megatrends-in-composite-aerostructures. 2016. Last Accessed
[01/09/2016].
[13] G. P. Thomas, “Resin Transfer Moulding,”
http://www.azom.com/article.aspx?ArticleID=8620. Last Accessed
[01/09/2016].
[14] Advanced Process Technology Inc., “Vaccum Assisted Resin Transfer
Moulding,” http://www.advancedprocess.com/vartm-vacuum-assisted-resin-
transfer-m, 2015. Last Accessed [01/09/2016].
[15] V. V. V. and E. V. Morozov, “Optimal Composite Structures,” in Advanced
183
Mechanics of Composite Materials, 2nd Edition, Elsevier Ltd, 2007, pp. 437–
480.
[16] Z. Su, L. Y. Ã, and Y. Lu, “Guided Lamb waves for identification of damage
in composite structures : A review,” J. Sound Vib., vol. 295, pp. 753–780,
2006.
[17] I. Dresden, “Composite Delamination,” http://www.ima-dresden.de/, 2008.
Last Accessed [01/09/2016].
[18] J. G. (NASA. Goree, “Preliminary Investigation of Crack Arrest in Composite
Laminates Containing Buffer Strips,” 1978.
[19] A. Abdul-aziz, G. Baaklini, and J. Trudell, “Structural Analysis of Composite
Flywheels : An Integrated NDE and FEM Approach,” Cleveland, Ohio, 2001.
[20] M. Sayem, E. V. Morozov and K. Shankar, “The effect of filament winding
mosaic pattern on the stress state of filament wound composite flywheel disk,”
Compos. Struct., vol. 107, pp. 260–275, Jan. 2014.
[21] ANSYS Apdl, Help topics. ANSYS, 2015.
[22] M. Sayem, E. V. Morozov, and K. Shakar, “The effect of filament-winding
mosaic patterns on the strength of thin-walled composite shells,” Compos.
Struct., vol. 76, pp. 123–129, 2006.
[23] Altaeros Energies, “www.altaerosenergies.com/,” 2014. Last Accessed
[01/09/2016].
[24] The Pennsylvania State University,
“https://methodology.psu.edu/ra/most/factorial,” Introduction to Factorial
Experimental Design, 2016. Last Accessed [01/09/2016].
[25] P. Jaen-Sola, A. S. Mcdonald, and E. Oterkus, “A Lightweight Approach for
Airborne Wind Turbine Drivetrains,” in European Wind Energy Association
International Conference, Paris (France), Nov. 2015, pp. 1–9.
[26] R. R. Johnson, M. H. Kural, and G. B. Mackey, “Thermal Expansion
Properties of Composite Materials”, Lockhead Missiles and Space Company,
Inc., NASA Contractor Report 165632, Sunnyvale, California, 1981.
184
Dynamics of a direct drive generator
Chapter 6
6.1 Introduction
In previous chapters, the structural integrity of the electrical generator has been
analysed by carrying out static studies using FE tools. In this chapter, the machine
will be studied from a dynamic point of view. As a rotating piece of machinery, the
generator vibrates when its natural frequencies are excited introducing potentially
large amplitude oscillations into the forces acting on it that could cause structural
fatigue, noise and, in the worst case scenario, the sudden collapse of the structure.
The main aim of this chapter is to develop suitable approaches and to investigate a
185
set of possible designs that would ensure the dynamic integrity of the machine
without significantly increasing its weight.
When designing a wind turbine system special consideration must be taken when
looking at these machines from a dynamic viewpoint as they inherently vibrate
during operation. This vibration produced by the excitation of the structure natural
frequencies can increase the magnitude of fatigue forces leading to an early
breakdown. Following the guidelines and specifications stated in [1] and [2],
Bywaters et al. presented results for various sets of modal analyses accomplished for
wind turbines with different rated powers and drivetrain configurations [3]. Figure
6.1 displays the Campbell diagram for a 3 MW machine operating between 8.5 and
17 rpm.
Figure 6.1 Campbell diagram for a 3 MW machine [3]
A Campbell diagram, named after Wilfred Campbell who first introduced the
concept, also called interference diagram, represents the frequency in Hertzs versus
the rotational speed of the rotating piece of machinery in rpm, in this case, the wind
turbine rotor. The development of the natural frequencies corresponding to the mode
shapes are presented in function of the rotational speed of the rotor as horizontal
lines. To excite the structure, any of the excitation frequencies, which in this graph
186
corresponds to the 1P, 3P and 6P inclined lines; have to intercept any of the natural
frequencies. The most dangerous natural frequencies to be excited are typically the
lowest ones as they have more possibilities of coinciding with a larger number of
excitation frequencies [4].
As seen, Figure 6.1 shows the 1st mode for the tower at around 0.3 Hz, whereas the
1st modes flapwise and edgewise of the wind turbine rotor take place at around 1.375
and 1.55 Hz respectively. Excitation frequencies are calculated according to the wind
turbine rotor rotational speed, ‘P’. This technique has been utilised, as it will be
described in the next section, to find out whether the optimised generator disc
structure is at risk dynamically speaking and if so, what the frequencies to be avoided
are.
The analysis and optimisation of the mechanical and structural design of a radial flux
direct drive generator rated between 0.75-3 MW for land based and fixed bottom
offshore wind turbine applications were carried out in [5] and [6]. Authors in [6]
claimed no vibration issues with the generator according to the results achieved from
a collection of experimental studies on a 1.5 MW machine allowing an eccentricity
of up to 50 % under extreme loads.
A permanent magnet synchronous generator was dynamically evaluated and the

effects on the turbine operational speed range caused by the harmonics of the
cogging torque and torque ripple quantified in [7], [8] and [9]. In [10] and [11], the
bearing working life and the structural suitability of a permanent magnet generator
with an outer rotor were surveyed. Hub and nacelle design were suggested too. Large
scale direct drive generators feasibility (10 MW) is under investigation as seen in
[12].
In [13], Sethuraman characterised the dynamic behaviour of the direct drive power
train of a floating wind turbine following the well-known two step de-coupled
approach proposed by Xing et al. [14], [15]. This method, originally created for
analysing conventional geared wind turbines, utilises the data obtained from an aero-
elastic simulation code, on global motion response and loads as inputs for a detailed
drive train model created in a Multi-Body Simulation piece of software which
187
enables kinematic and dynamic analyses of mechanical systems. However, as the

main purpose of this research was to study a direct drive machine, the author also
needed to address how significant the dynamic effects and feedback forces from the
drivetrain were.
A new concept for the design of a 5 MW direct drive generator is presented by

Shrestha et al. in [16]. The main goal was to reduce the weight of a large diameter
machine by introducing axial flexibility in the generator support structure. The data
retrieved from modal analyses are given highlighting the fact that the use of
transverse plate stiffeners for a hollow rotor structure not only provides axial
stiffness and limits the radial deflection but also helps keep the first bending mode
frequency in a safe region without significantly increasing machine´s weight. See
Figure 6.2 for a detailed view of the proposed structure.
Figure 6.2 Wire frame view of the rotor structure with transverse plate stiffeners [16]
Zavvos shows in [17] the outcomes acquired from the modal analyses he made over
three different 5 MW direct drive electrical machines: radial, axial and transverse
flux. The first was also studied having distinct structural configurations, such as disc
and armed. He found that axial and transverse flux machines and radial flux
188
machines made with discs have a moderate risk of experiencing vibrations, with the
axial flux machine being the most vulnerable to fatigue. On the opposite side were
the conventional radial flux machines made with hollow arms, which was the least
susceptible to fatigue type of generator.
In [18], Kirschneck produced a two way model, which links mechanical and
magnetic systems, that identifies the dynamics of a specific electrical machine.
Going through a collection of all the relevant modelling techniques and formulae, he
derived a set of equations that can be used to optimize the structural mass of the
generator and to determine the effect of the excitation forces.
Chapter 6 is composed of 6 sub sections: Section 6.1 reviews the state of the art in
the technology. Although several studies on the dynamics of the wind turbine
drivetrain have been published, only a few concentrate on the direct drive generator
structure. In Section 6.2, the distinct options explored to improve the dynamic
behaviour of the machine are explained in detail. All the outcomes obtained from the
said research are presented in Section 6.3. These results are analysed and discussed
in Section 6.4, leading to conclusions and recommendations for further design
enhancements in Section 6.5.
6.2 Methodology
6.2.1 Evaluating structural natural frequencies
A finite element modelling package, such as SolidWorks, was employed in this

chapter so as to estimate the lowest mode shapes of the generator structures. The
theory behind the modal analyses carried out with this piece of software leans on the
structural linear vibrations principle.
Starting with basics, if we want to explain why a body vibrates, it is essential to

recall the energy conservation principle. Assuming a mass ‘m’ suspended on a spring
of stiffness ‘k’, as shown in Figure 6.3, consider that the mass is pulled up from its
static equilibrium position by ymax and then released.
189
Figure 6.3 Spring-mass arrangement [19]
When the spring deforms some energy gets stored in it. This energy is known as the
potential energy or ‘Vmax’ and represents the maximum energy transmitted to the
spring by deforming it by ymax. After realising the mass, the spring will start
oscillating up and down until recovering its original position. The energy gained with
the motion of the mass is called kinetic energy and can be described by the following
equality,
1 (6.1)
𝐸k = 𝑚𝑦̇ 2
2
with 𝑦̇ being the mass linear velocity, which corresponds to the mass displacement
time derivative. The potential energy stored in the spring at a midway position y is
1 2 (6.2)
𝑉= 𝑘𝑦
2
Assuming an ideal spring, the sum of the kinetic energy gained with the mass motion
and the potential energy acquired by the spring at any moment in time must be equal
to Vmax. Therefore,
1 1 (6.3)
𝑉𝑚𝑎𝑥 = 𝑚𝑦̇ 2 + 𝑘𝑦 2
2 2
190
It can be observed that an energy transformation occurs in a spring-mass system,

from potential to kinetic and vice versa while Vmax always remains constant. It
happens that when y = ±ymax, V is equal to Vmax and 𝑦̇ = 0. However, when y = 0,
which represents a static displacement, ẏ = 𝑦̇𝑚𝑎𝑥 . As already mentioned, this
periodic move from extreme positions is named oscillation and can be characterized
by the next two parameters: the period of oscillation, which is the time elapsed
between two consecutive maximum deformations, and the amplitude of oscillation,
which is the maximum displacement of the mass from its static equilibrium position
or ‘ymax’.
Knowing that the mass moves in a periodic fashion with an amplitude of ymax, the
process can be described by the following the function,
𝑡 (6.4)
𝑦 = 𝑦max 𝑠𝑖𝑛 (2𝜋 + 𝛼)
𝑇
where t is the time, T is the oscillation period and α is a constant which is defined by
the requirement that at t = 0, y = ymax. As seen, ymax is known leaving T as the only
unknown (besides α) in the equation. After derivations, y and 𝑦̇ can be substituted in
expression (6.3), obtaining
𝑚(2𝜋)2 2 𝑡 (6.5)
[ − 𝑘] 𝑦max 𝑐𝑜𝑠 2 (2𝜋 𝑇 + 𝛼) = 0.
𝑇2
For equation (6.5) to be true at any moment in time, the term in brackets must be
equal to zero,
𝑚(2𝜋)2 (6.6)
[ − 𝑘] = 0
𝑇2
hence the period of oscillation is defined as a function of mass and stiffness

according to equation (6.7):
𝑚 (6.7)
𝑇 = 2𝜋√
𝑘
191
Note that the oscillation period does not depend on the initial disturbance. The
number of oscillations per unit time is called oscillation frequency (typically
expressed in oscillations per second or Hertz) and it corresponds to the inverse of T
[19]. Then,
1 𝑘 (6.8)
𝑓 = 2𝜋 √𝑚 .
The evaluation of resonant frequencies, also known as natural frequencies, of simple

or composite structures is based on the concept explained above. This evaluation
would require the use of more complex equations or stress analysis packages
depending on the complexity of the structural configuration, which is the case of
rotor and stator structures. The example provided owns only one degree of freedom,
whereas more complex layouts have various. This implies having several modes of
deflection and therefore, several excitation modes and natural frequencies.
As seen from equation (6.8), it is possible to avoid resonance by design influencing

the structural stiffness or mass or both. Moreover, by having full control of the
system you may be able to avoid dynamic loading at these frequencies.
In Chapter 4 was found that generator disc structures are much lighter than armed
structures, as the hollow arms cannot cope with the tangential loading as well as the
disc sub-structures do unless the thickness of the arms is considerably increase with
the subsequent rise in mass. With the optimum configuration identified, the first 5
natural frequencies of the rotor and stator structures have been assessed, by
performing modal analyses, and plotted in a Campbell diagram. For that, the
SolidWorks Simulation© add-on package was used. The frequency analysis study
option was selected among the distinct types of studies available. Then, structural
steel was assigned to all the components forming the structures as their material. Its
characteristics are the same as the ones employed in chapter 4: Young’s modulus,
𝑘𝑔
𝐸 = 2.1 × 1011 Pa, Poisson’s ratio, v = 0.3 and density, ρ = 7850 𝑚3 . The next step
was to constrain the structures at the shaft and to create an appropriate mesh. In
SolidWorks, it is possible to adjust the mesh element size manually by making use of
a slider which goes from a finer mesh to a coarser one. The piece of software also
192
suggests an element size by default that is supposed to give accurate results in a

reasonable period of time. In addition, two types of adaptive meshes are available
too: h-type and p-type. The h-adaptive mesh diminishes the size of the elements in an
iterative manner, locally or globally depending on the settings specified by the user,
until convergence is accomplished, while the p-adaptive mesh increases the element
order (linear, quadratic, cubic, etc,.), again locally or globally depending on the
specifications, in an iterative way until convergence is achieved. See Figure 6.4 for
further clarifications on element order. Similar results should be obtained by using
any of these two techniques which are very useful instruments especially at the time
of finding stress singularities. For instance, when looking at sharp corners, which do
not exist in the real world, 3D finite element tools, tend to overestimate the stress no
matter how fine the mesh is. This phenomenon is known as a stress singularity. The
stress results go higher and higher as the user refines the mesh without levelling off
or converging. So as to make sure that good outcomes have been acquired the trend
tracker option included in the adaptive mesh tree can show the user if a levelling off
have been reached and when.
Figure 6.4 (a) First order (linear tetrahedral solid element with no nodes at the midpoints); (b) Second
order for a higher quality mesh (parabolic tetrahedral solid element with nodes at the midpoints) [20]
For these studies the mesh density suggested by the FE package was utilised. A
linear tetrahedral high quality mesh with 7,682 elements and 15,640 nodes was
created. The element size was 148.5 mm.
The estimation of the first five mode shapes of the rotor and the stator structures was
performed with no loads applied at any time.
193
6.2.2 Dynamically designing a direct drive generator supporting structure
Few studies have been published on the design of large diameter direct drive
generator structures. Most of them concentrate all the attention on the effects of the
external loads acting on the whole drivetrain but not many look at how these forces
affect the components forming the electrical machine structure individually. As
mentioned, the natural frequencies of the structure to be designed must be either
avoided or passed as quickly as possible. The potential excitation frequencies that
could activate the structure natural frequencies are as follows [17],
 The frequency of the wind turbine rotational speed (1P).

 The fundamental electrical frequencies (pP, with p being the number of pole
pairs per stage).
 The frequency of the rotor blades passing in front of the tower (3P and 6P).
In this regard, designers of this type of devices often follow an empirical rule that
natural frequencies between 70 % and 130 % of the exciting frequency must be
avoided [16].
The lowest natural frequencies of both the rotor and the stator structures can be
estimated utilising FE techniques. Different excitation frequencies are able to trigger
the cited natural frequencies. This research has been carried out considering the three
types of frequencies defined above. With this, distinct options were explored so as to
increase the natural frequencies of the structures without adding significant weight.
Figure 6.4 displays the power spectrum of rotor speed of a variable speed
conventional geared wind turbine, with the rotational velocity, expressed in rad/s, in
the x axis and the amplitude of the loads in the y axis. The first step when
dynamically studying a wind turbine or a wind turbine component is to analyse its
power spectrum of rotor speed so that the most dangerous modes can be recognised
and avoided later on in the design. As seen in the said picture, the first drive train
mode is the most dominant. In our case, as the drive train of a direct drive wind
turbine is very different (no gearbox), the amplitude of this mode and its frequency
will be distinct. In fact, for a direct drive machine it is expected to have a lower
amplitude for the first drive train mode and lower frequencies as the rotor speed is
significantly lower. Nonetheless, the geared wind turbine example can be utilised as
194
a reference due to this sort of highly resonant modes are also present during direct
drive machines operation and are the ones to be avoided in order to maintain the
integrity of the machine.
Figure 6.5 Rotor speed power spectrum of a VS conventional geared wind turbine [21]
The rotation of the wind fields generate deterministic and stochastic peaks in
frequencies such as 1P, 2P, 3P and so on for the blades of a 3 bladed rotor, while for
the drive train and tower the stochastic peaks take place at 3P, 6P and all its multiples
and the deterministic peaks happen at 1P, 2P, 4P and all its multiples [21]. As stated
by Kirschneck in [18], the dynamic of the structure is dominant while the dynamic of
the magnetic field has a rather limited influence. Hence, it is necessary to concentrate
the attention on the structural side and to analyse the power spectrum of a direct
drive rotor speed to find the most harmful modes for the structure of the electrical
machine, to then plot them in a Campbell diagram with the structure natural
frequencies and encounter the best way to evade them. As already said, in absence of
such information, the author employed the data provided by Zavvos in [17].
Considering a number of pole pairs equal to 60, the excitation frequencies for the
drivetrain were encountered for a wide range of rotational speeds going from 0 to 20
195
rpm. The outcomes were plotted in the frequency interference diagram shown in
Figure 6.6 for comparison.
Figure 6.6 Campbell diagram of the system
Looking at the Campbell diagram, it can be seen that the weakest component is the
rotor with a frequency for the first mode of 13.403 Hz, whereas the frequency of the
first mode for the stator is 16.976 Hz. From Figure 6.6, it can be understood that for a
3 MW direct drive machine optimised structure made with discs the suitable range of
operation is between 7.6 and 18.5 rpm. The need to avoid the lowest resonant
frequencies forces the designer to sacrifice the collection of energy at the lowest and
at the highest wind speeds for the good of the structure. Nevertheless, it would be
also possible to step through frequencies using a control system.
Different ways of increasing the natural frequencies of the disc structures have been
analysed in this chapter. In addition, another type of structure layout, such as a rotor
cone structure, has been proposed and studied. The results acquired for these tests are
illustrated in detail in Section 6.3.
196
6.2.3 Techniques for elevating structure’s natural frequencies
There exist 4 distinct ways of elevating the natural frequency of a design [22]:
 Alter the geometry.

 Strategically locate mass elements.
 Change materials (resonant frequencies are directly proportional to the elastic
modulus of the materials).
 Modify the features of the shock isolators.
The first three ways have been looked at in this investigation. Since the structural
design of the electrical machine has been approached from a general perspective, the
fourth manner of pushing up the frequencies has not been studied.
Looking at the structure mode shapes obtained from modal analyses, the sort of
technique that might apply to push the natural frequencies up can be identified.
6.2.3.1 Dimensional alteration of structures made with discs
As stated, a 3 MW machine of 4m diameter and 1.2 m axial length made up of steel,

as shown in Figure 6.7, has been used. Dimensions tc and td have been varied and
modal analyses carried out so that the effect caused by those alterations in the natural
frequencies could be studied and quantified.
Figure 6.7 Rotor structure as changed in the analysis
197
The thicknesses, measured in millimetres, of both sub structures were given the
following values,
- Cylinder thickness, ‘tc’: 20, 40, 60, 80, 100, 120 and 140 mm
- Disc thickness, ‘td’: 40, 60, 80, 100, 120, 140 and 160 mm
These dimensions were picked looking at the values presented in Chapter 4, in which
the contour plots are utilised to optimise the generator structure with a view to
minimise its mass. They are all within a realistic range which varies around the
optimum obtained results. A total number of 49 modal studies were made. The
outcomes achieved were plotted and given in Section 6.3.
The following corresponds to a simplified example of a rotor structure that will help
to understand the idea presented in this sub section. If a rotor structure is simplified
by assuming that the disc sub structure is equal to a cantilever beam and the outer
cylinder is an attachment located at the free end of the beam, as seen in Figure 6.7, a
modal analysis can be made so that the effect that each sub structure has in the
dynamic behaviour of the overall structure can be appreciated.
Figure 6.8 Simplified Rotor Structure
As observed in Figure 6.9, the first resonant mode takes place in the Y direction as
the area moment of inertia (I = bh3/12), with b being the width and h being the
height, is the lowest in that direction. This can be simply explained by saying that the
Y direction is the weakest due to it is the thinnest. The characteristics of the model
are shown in Table 6.1.
198
Table 6.1 Simplified rotor model features
b (width in X direction) 0.1 m

h (height in Y direction) 0.035 m
L (length in Z direction) 1m
m (beam mass per unit length) 28 kg/m
M (attachment mass) 85.8 kg
E (Young’s modulus) 2×1011 Pa
𝒈 (gravity in Y direction) 9.81 m/s2
The weakest parts of the design show themselves as low frequency modes and as it
can be observed in the mentioned picture, the beam is so weak in the Y direction that
3 resonant modes occur in that direction for the beam alone, whereas when the
attachment is included 2 modes are obtained.
199
Mode 1 → f = 27 Hz, Shape: Primary Y Direction; f = 18 Hz, Shape: Primary Y Direction
Mode 2 → f = 78 Hz, Shape: Primary X Direction; f = 51 Hz, Shape: Primary X Direction
Mode 3 → f = 172 Hz, Shape: Secondary Y Direction; f = 132 Hz, Shape: Secondary Y Direction
Mode 4 → f = 447 Hz; Shape: Primary Z Direction; f = 248 Hz, Shape: Primary Z Direction
(Torsional)
Mode 5 → f = 469 Hz; Shape: Tertiary Y Direction; f = 362 Hz, Shape: Secondary X Direction
Figure 6.9 Simplified Rotor Mode Shapes
200
In order to validate these outcomes, the resonant modes in the weakest direction of
the cantilever beam have been calculated using empirical equations [23]. Figure 6.10
shows the first mode whose shape corresponds to the one acquired with the finite
element tool.
Figure 6.10 First mode shape in Y direction
By employing equation 6.9,
3.52 𝐸𝐼
𝑓= √ (6.9)
2𝜋𝐿2 𝑚
where L is the beam length in meters, E is the Young modulus in Pa, I is the area
moment of inertia in m4 and m is the mass per unit length in kg/m, it was found that
the first natural frequency was 27.7 Hz. The result obtained using FE techniques was
27.65 Hz.
For the second mode, equation 6.10 was utilised. The mode shape is displayed in
Figure 6.11 and again it can be seen that it fairly matches the form acquired by the
FE outcome for the cantilever beam.
Figure 6.11 Second mode shape in Y direction
201
As it can be observed, the parameters in equation 6.10 are the same as those in
equation 6.11. The only difference between these two equalities is the constant in the
first fraction which varies with modes.
22 𝐸𝐼 (6.10)
f = 2𝜋𝐿2 √ 𝑚
In this case, the result was 173.4 Hz while the data retrieved form the software was
172.32 Hz.
Finally, mode 3 was calculated utilising equation 6.11.
61.7 𝐸𝐼
f = 2𝜋𝐿2 √ 𝑚 (6.11)
Its result was 485.6 Hz whereas SolidWorks gave an outcome of 469.75 Hz. Its
shape is shown Figure 6.12.
Figure 6.12 Third mode shape in Y direction
For a cantilevered beam with a mass attached to its free end, Young and Budynas
[23] derived an empirical equation that is capable of estimating its first natural
frequency. Assuming a beam negligible mass, equation 6.12, where 𝑔 corresponds to
the gravity in m/s2 and M is the mass of the attachment in kg, gave a result of 24.38
Hz. So as to carry out this study in SolidWorks, a material density of 1 kg/m 3 was
assumed. The result obtained from the software was 23.8 Hz.
1.732 𝐸𝐼𝑔
f= √ (6.12)
2𝜋 𝑀𝐿3
202
If a comparison between the FE data and the outcomes achieved with the equations is
made, it can be observed that the largest error obtained was lower than ±5 %. See
Table 6.2.
Table 6.2 Analytical vs. FE comparison
Mode Frequency (Hz) Error (%)

Equation FE
1st without attachment 27.7 27.65 0.2
nd
2 without attachment 173.4 172.32 0.6
3rd without attachment 485.6 469.75 3.3
st
1 with attachment 24.38 23.8 2.4
Therefore, it can be said that the results acquired using the software are accurate
enough. Once the data was validated, the attention was again focussed on the FE
study results. By reviewing the resonant modes acquired with and without the
attachment, it is easy to understand that the addition of mass at the free edge of the
beam makes the structure even weaker in all directions. A considerable reduction in
frequency happens for all the modes with the 5th one experiencing the highest drop
from 469 to 362 Hz and changing the mode shape from the Y to the X direction and
it would have been worse if a heavier attachment had been used. This is the type of
behaviour that can be expected from rotor and stator structures. The results given in
Section 6.3 will corroborate this statement.
6.2.3.2 Use of stiffeners
Another way of manipulating the natural frequencies of the structures is through the
introduction of other structural components, named stiffeners, which provide the
structure with the necessary directional stiffness. When finding the structure’s natural
frequencies, special attention must be taken to the mode shapes. The different forms
adopted by the structure give the designer a clear picture of type of stiffener required
and their location. With this in mind, it is important to remember that the main aim
of this investigation is to reduce the mass of the machine while meeting structural
203
requirements. Therefore, the utilisation of stiffeners will be suitable only when a

substantial increase in frequency is observed without adding too much mass.
In this research, an isosceles triangular type of stiffener with a thickness of 60 mm

joining the cylinder and the disc sub structures has been proposed. Located in the
axial direction, its equal length sides are joined to the inner face of the rim and the
axial surface of the disc and its dimensions vary with the disc thickness. As shown in
Figure 6.13, a finite number of stiffeners equally spaced have been employed. First
on one side of the disc only and then on both sides.
Figure 6.13 Rotor structures with axial stiffeners
A different number of stiffeners, as well as the use of a distinct material with lower
density (alloy steel) in their design were also tried with the main aim of minimising
the overall mass. Again, all the outcomes obtained from this part of the investigation
are given in Section 6.3 of this chapter.
6.2.3.3 Rotor conical structure
In the last sub section, the techniques utilised for incrementing the natural
frequencies of an optimised disc structure have been presented. However, other types
of simplified structural configurations exist as stated by Stander in [24].
Rotor conical structures provide an axial stiffness that a disc structure misses, while
the cone sub structure acts as a disc sub structure bringing radial stiffness to the
204
model. In this chapter, a rotor conical structure has been optimised according to the
deflection criterion as applied to the disc structure and under the same loading
conditions but also considering its natural frequencies.
Last but not least, the most dangerous modes of deflection for the rotor conical
structure were found. By applying the same loads and constraints as for the disc
structure in Chapter 4, the optimised cone structure was tested and the retrieved data
analysed.
6.3 Results
As explained in Section 6.2, the natural frequencies of the generator structure must
be avoided in order to avoid resonance. Different techniques can be used to increase
the said natural frequencies. In this section, the data retrieved from the analyses
carried out over 3 distinct types of structural configurations are presented.
As with Chapter 4, one can plot the mass of a disc structure on a 2D plot, with disc
thicknesses on the X-axis, cylinder thicknesses on the Y-axis and contours
representing rotor or stator structural mass. Another set of contours can be used to
show the natural frequency. See Figure 6.14. The frequencies presented in these plots
correspond to the first mode shapes. Plots showing the variation of frequency with
dimensions for the rest of the modes have been included in Appendix A.
205
(a)
(b)
Figure 6.14 2D Optimisation for 3 MW rotor and stator disc structures with 1 st mode natural
frequencies criterion
206
After optimising a disc generator structure by looking at it from a static viewpoint,

one can assess its dynamic behaviour and estimate the structure natural frequencies
by entering its dimensions in the contour plots given. With the information provided
by this reliable and fast method and the resonant frequencies given by the power
spectrum of rotor speed a Campbell diagram can be created and the best ways to
maintain the machine’s integrity identified.
At this point, it is important to mention that the rotor and the stator structures have
been studied individually and the potential interactions between them have not been
taken into consideration. An inner rotor type was utilised in the studies. The results
shown in this section do not apply to machines with outer rotor configurations as the
change in geometry will produce different outcomes for the same loading conditions.
Figure 6.15 displays the shapes for the first 5 modes of a typical rotor disc structure.
In the first mode, the cylinder sub structure stretches in the horizontal direction
making the disc sub structure twist about the vertical axis (in red) as it is illustrated
in the figure below. A similar shape is acquired for mode 2 although in this case the
cylinder sub structure stretches in the vertical direction making the disc twist about
the X axis. The excitation of the third natural frequency would produce a lateral
displacement in the Z direction of both the cylinder and disc sub structures. As the
disc is fixed at the shaft, the displacement in the Z direction would induce the disc
sub structure to acquire a conical shape. In the last two modes presented here, the
cylinder sub structure shows large deflection at its edges taking place periodically
that produce ripples. The disc sub structure deforms accordingly with the cylinder
shape. Although, these modes seem to produce random shapes, it is not difficult to
recognise a deformation pattern. The lowest deflection is located on the disc and it
follows a square shape that changes in direction with the mode as it can be
appreciated.
These modes obtained due to the excitation of the structure natural frequencies create
an uneven deformation across the structure that would introduce such instability in
the airgap that could cause the collapse of the machine. Two options can be taken
into account to avoid this situation:
207
- To make use of the techniques given in Section 6.2.
- To design a structure capable of withstanding the stress without significantly

deforming.
Note that the latter would imply a substantial increase in the structural mass.
Figure 6.15 First 5 mode shapes of a typical rotor disc structure (1 st Mode: top left; 2nd Mode: Middle
top; 3rd Mode: top right; 4th Mode: bottom left; 5th Mode: bottom right)
The evaluation of this behaviour helps the design engineers to make decisions on
how to alter the natural frequencies of the structures. For instance, in this case where
the disc deforms axially due to radial alterations in the cylinder sub structure, the use
of stiffeners providing some extra axial stiffness might help. A finite number of
stiffeners of isosceles triangular shape with 60 mm of thickness, equally spaced,
were implemented in the optimum steel structure, with a mass of 19,260 kg, as
described in the previous section, in order to quantify their effect in the structure
natural frequencies.
208
15.5
15.0
Frequency (Hz)
14.5
14.0
13.5
13.0
4 5 6 7 8 9 10
One side 13.9 14.1 14.3 14.5 14.8 15.0 15.2
Two sides 13.8 14.1 14.3 14.6 14.9 15.1 15.3
Cast alloy steel 13.9 14.1 14.3 14.6 14.8 15.1 15.3
No. of Stiffeners
One side Two sides Cast alloy steel
Figure 6.16 Quantification of the effect produced by the implementation of stiffeners in a rotor disc
structure
The investigation was started with a modal study of the structure without any
stiffener. The result for the first mode shape was 13.403 Hz. Then, 2 stiffeners were
introduced on one side of the disc verifying that their effect on the structure natural
frequencies was almost negligible, although 154 kg were added to the overall mass.
An unacceptable outcome was again obtained for 3 stiffeners; hence the data began
to be taken into consideration after them. As shown in Figure 6.16, a different
number of stiffeners located on one side first and on both sides second were tried. In
addition, a less dense material such as cast alloy steel, with a density of 7,300 kg/m3
was also assumed for the stiffeners with the main objective of reducing their mass.
It was understood that the structure natural frequency increases linearly with the
number of stiffeners placed on one side of the disc. The same behaviour was
observed when using stiffeners on both sides with slightly higher frequencies. The
utilisation of stiffeners made of cast alloy steel on one side of the disc slightly
enhanced the dynamic performance of the structure but the reduction in mass
achieved was not very significant as each stiffener weighted 71.65 kg for 77 kg of a
stiffener made of conventional structural steel. The data retrieved from the analyses
showed that the introduction of stiffeners, either on one side or on both, pushed up
209
the natural frequencies less than 2 Hz (with 10 stiffeners). However, a mass of 770
kg had to be included for that.
Having the disc structure deeply analysed, another type of structural configuration
was looked at. As explained in Section 6.2, rotor conical structures include extra
axial stiffness that disc structures miss, while keeping excellent radial stiffness
characteristics. A rotor cone structure suitable for the proposed direct drive machine
has been optimised this time not only considering the deflection criterion as in
Chapter 4, where a complete set of static analyses was carried out, but also the
structure natural frequencies that were found utilising finite element methods. The
methodology tracked was:
First a cone structure was CAD modelled in SolidWorks as in Chapter 4. Random

dimensions were picked for both the thicknesses of the cone and the cylinder sub
structures. Their values were 40 and 22 mm respectively. After that, a set of static
and dynamic analyses were run for each position and angle of the cone with the main
aim of obtaining the mass, the maximum deflection and the first mode shape of the
structures. See Table 6.3.
210
Table 6.3 Cone Structure Optimisation Results
Position
1 2 3 4 5
Angle Mass δ(m) f(Hz) δ(m) f(Hz) δ(m) f(Hz) δ(m) f(Hz) δ(m) f(Hz)
(kg) x10-4 x10-4 x10-4 x10-4 x10-4
(º)
30 6873 4.8 19.1 4.4 25.7 4.8 31.3 4.8 33.5 4.4 29.8
35 7080 4.9 20.4 4.6 26.8 4.9 32.8 4.9 36 4.4 33
40 7384 4.7 21.1 4.5 27.4 4.9 33.5 4.9 37.5 4.4 35.3
45 7775 4.8 21.1 4.5 27.5 5 33.4 5 38 4.5 37.6
50 8281 4.9 21 4.7 26.9 5.1 32.4 5 37.1 4.6 38.1
55 8958 5 21 4.9 25.4 5.3 30.3 5.2 34.7 4.7 36.7
60 9881 5.2 19.4 5.1 23.2 5.7 27.3 5.4 31 4.9 33.1
65 11200 5.7 17.3 5.5 20.1 5.9 23.3 5.7 26 5.2 27.6
70 13205 7.1 14.5 6.6 16.3 6.7 18.3 6.6 20 5.9 21
As seen, the lowest deflections are acquired when the cone is placed in position 5. In
addition, the highest natural frequencies for the first mode shape are also achieved in
this position. By looking at the second column of position 5, it can be observed that
the frequency increases with the cone angle until reaching 50º. After that, it starts
going down again. Since the breakeven point took place at 50º of position 5, this
arrangement was taken as the optimum.
Then, it was checked which sub structure put more weight into the overall structure
when adding the same thickness (5 mm). It could be understood that the cone sub
structure included 87 kg more than the cylinder. With the targeted sub structure
found, its thickness was diminished while the thickness of the cylinder was kept
constant achieving a final result of 17 mm thickness for the cone and 22 mm for the
cylinder. That gave a total mass of 5,062 kg and a radial deflection of 4.827×10-4 m.
Tangential and axial deflections were also checked observing that they were still
within the limits. The natural frequencies for the first five mode shapes were as
follows,
211
 1st Mode shape → 26.92 Hz

 2nd Mode shape → 27.23 Hz
 3rd Mode shape → 27.64 Hz
 4th Mode shape → 27.66 Hz
 5th Mode shape → 36.18 Hz
Once the optimum thicknesses were found, a sanity check was carried out obtaining
that 50º cone angle at position 5 was still the best layout.
With the finest configuration encountered, the dimensions of the rotor, ‘tcon’ and
‘tcyl’, were altered and modal analyses run in order to quantify the effect of
increasing the thicknesses of the sub structures in the overall frequency. The obtained
results showed that the natural frequencies of the rotor structure rise with the
thickness of the cone in a linear manner. However, as the cylinder thickness goes up,
the natural frequencies go down following a linear trend again. See Figure 6.17. The
plots for the rest of the modes can be seen in Appendix A.
70
60
50
Frequency (Hz)
40
f = 291.99tcon + 18.473
30 R² = 0.9971
20
10
0
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18
Cone thickness (m)
(a)
212
35
30
25
Frequency (Hz)
20 f = -46.891tcyl + 30.28
R² = 0.6875
15
10
0
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16
Cylinder thickness (m)
(b)
Figure 6.17 Frequency variation as dimensions are altered; (a) With cylinder thickness maintained at
0.02 m; (b) with cone thickness kept at 0.04 m
Finally, with the conical structure optimised and studied, the most dangerous modes
of deflection were identified. Figure 6.18 shows the outcomes achieved for the
different existing modes of deflection starting from 0 and ending with 4. The
deflection was calculated using the same loads as for the disc structure in Chapter 4
that are eventually higher than the ones used in the optimisation of the structure.
Only radial loads were applied to obtain these data, whereas tangential and
gravitational were also utilised for the optimisation. Hence, this information must be
seen only as a reference when designing conical structures. From the graph, it is easy
to see that the highest deflection is obtained for mode 3 and therefore, this mode can
be considered as the most damaging for this type of structures. So as to withstand the
demanding loading conditions the use of thicker sub structures or the introduction of
stiffeners are the most straightforward solutions that can be adopted.
213
8.00E-03
7.00E-03
6.00E-03
Deflection (m) 5.00E-03
4.00E-03
3.00E-03
2.00E-03
1.00E-03
0.00E+00
0 1 2 3 4 5
Deflection modes
Figure 6.18 Deflection vs. Modes of Deflection (Cone Structure)
6.4 Discussion
Three distinct types of structural arrangements have been analysed in this chapter. A
useful tool developed for disc structures to help the engineers during the early stage
of their dynamic design has been presented. It consists of contour plots where the
natural frequencies and the structural mass vary with thicknesses of the sub
structures forming the generator components. Alternatively, the use of stiffeners has
been introduced as a potential solution to push up the natural frequencies of the
structures without sensibly augmenting their mass. A series of modal analyses were
carried out over the optimised structure having stiffeners equally spaced. Firstly, they
were placed on one side of the disc sub structure and then on both sides. It was
understood that the utilisation of this sort of stiffeners puts too much weight (77 kg
per stiffener) on the structure while the first natural frequency only increases by
around 0.2 Hz per stiffener. However, it was observed that the use of stiffeners on
only one side was much more efficient than on both sides in terms of frequency and
mass. For this reason, it was thought that a less dense material, such as cast alloy
steel, might be employed to further reduce the mass of such layout. The results
obtained from the modal studies revealed that the addition of structural mass was still
high. Nonetheless, the author believes that a distinct shape type of stiffener made of a
214
lower density material might be employed. At this point, it is important to highlight

that the techniques suggested in this chapter have been only tried on rotor structures.
It was considered good practice to make an attempt to find suitable methods for more
simple structures before proposing them for stator structures.
On the other hand, rotor conical structures were assumed. Attracted by their good
axial stiffness features, a conical structure was optimised with a view of minimising
its mass considering deflection limitations and the frequencies of the lowest mode
shapes. It was found that a rotor structure with a cone angle of 50º placed at the 5 th
position with 17 mm thickness for the cone and 22 mm for the cylinder presented the
best outcomes. That gave a total rotor mass of 5,062 kg. Whether this data is
compared to the one of the optimised disc structure, it can be observed a drop in
mass of more than 48 % and 33 % if it is contrasted with the further optimised steel
structure, although it is still heavier than any of the composite models. Moreover, the
natural frequencies of the first five mode shapes of the rotor conical structure were
all between 27 and 37 Hz, which means an increase of 58 % in the worst case
scenario if compared with the disc structure. However, this increment in the natural
frequencies of the first mode shapes can be interpreted as an issue too. This increase
means that for the example given in this chapter, the electrical resonant frequency
would cross all the natural frequencies of the rotor at higher rotation speeds. This
would force the engineer to sacrifice a lot of energy collection in order to maintain
the integrity of the machine unless the structure is designed to withstand the stress or
the thickness of the cone is significantly elevated so that the resonant frequency is
avoided. The structure would see its mass considerably augmented anyway.
The last step of this investigation on rotor cone structures included the data acquired
from a set of conventional static studies made over the optimised conical structure
which showed that the highest deformation takes place under mode of deflection 3. It
is thought that this is a key factor that must be taken into account when designing
this kind of supporting structures for electrical generators.
Chapters 3, 4 and 5 look at the design of the machine structure from a static point of
view. The loads were estimated using a parametric model and applied to rotor and
stator structures with different characteristics and made of distinct materials. Yet, no
215
external loads (except torque) or influence were included during the optimisation
process. Rotation of the rotor and transient loads (wind induced forces, short
circuit,..) should be taken into account in future models. Exploration of the rotor and
stator interference and bearing location are other features that have not been
considered in these analyses. By checking the Campbell diagram presented in this
chapter, it can be seen that the highest rotor modes and lowest stator modes are fairly
close (this combination could easily lead to a structural collapse). In order to avoid
them, it would be necessary to have the electrical machine working in a range
between 7.6 and 18.5 rpm. This means that the wind energy that could be harvested
below 7.6 rpm and over 18.5 rpm is lost. A potential solution would be to increase
the thickness of the disc sub structure with the consequent rise in mass. Another
solution would be to make use of rotor conical structures. However, it was
understood that the use of this particular geometry did not push up the frequency far
enough, thus an elevation of the cone sub structure thickness would be again needed.
The last solution that can be noticed by having a look at the said diagram goes
through altering the number of pole pairs so that the slope of the electrical frequency
line can be changed. With a higher number of poles, the inclination of the line would
go down interfering with the lowest rotor and stator modes at higher rotational
speeds which would force us to narrow down the machine operating range. In
addition, saturation problems could arise. But, if the number of poles is decreased,
the slope of the line would go up giving us more room for machine operation.
Nonetheless, it is important to notice that by doing so the highest stator modes could
get excited at lower rotational speeds. Also, with lower number of poles thicker rotor
and stator yokes would be necessary which would make the machine heavier.
During a short circuit event, the magnetic field could react causing rotor eccentricity,
transferring the entire load from the generator structure to the bearing. Moreover, the
excitation of the structural modes can generate changes in the inductance of the coils
introducing higher harmonics in the currents. Therefore, a more accurate model of
the bearing that replicates their characteristic non-linear behaviour would be
required.
216
In Chapter 6, the dynamic behaviour of the structure has been studied and tried to
improve. Several techniques have been described and tested here. The results
achieved were filtered until finding the appropriate data that could be useful for a
design engineer looking at this sort of devices. Again, no external influence was
considered in these studies.
6.5 Conclusions
In this chapter, a useful instrument for the dynamic design of disc structures has been
presented. The use of the provided contour plots allows the engineer to make
accurate estimations of the weight and the natural frequencies of the structure to be
produced. Later on, the introduction of stiffeners was proposed with the main aim of
elevating the structure natural frequencies without adding too much weight. The best
results were obtained by placing the stiffeners on one side of the disc sub structure
with the frequency of the overall structure increasing linearly with their number.
Nevertheless, as the studies demonstrated, the augment in frequency was not
significant enough for the amount of material that is put on and for this reason their
use was discarded.
Due to its potentially good radial and axial stiffness characteristics, a rotor conical
structure was proposed. It was observed that its performance is superior to that of a
disc structure with less mass needed to support the same loads and natural
frequencies almost 2 times higher. Having said that, the author believes that conical
structures should be taken into consideration when approaching the design of a
generator structure due to their outstanding static and dynamic characteristics. More
research would be necessary on this field as with the use of the already described
methods and materials the mass of these structures might be further optimised. For
instance, a composite material conical structure would present fewer issues than a
disc structure at the time of joining the sub structures as the cone slope creates a
larger surface of attachment that would facilitate the transition of the composite
fibres from one sub structure to another. With this, a considerable reduction in the
number of fasteners and in the manufacturing time would be achieved. It would be
217
even possible to think that the structure might be constructed as a whole and not in
two pieces (cone and cylinder) eliminating the need for fasteners and the challenging
process of attachment. The introduction of external loads and influence would be
indispensable in order to carry out an accurate optimisation study.
218
6.6 References
[1] IEC 61400-1. Wind Turbine Generator Systems, Part 1: Safety Requirements.
2nd edition. 1999.
[2] Germanischer Lloyd. Non-Marine Technology, Part 1: Regulations for the

Certification of Wind Energy Conversion Systems. Rules and Regulations, IV.
1999.
[3] G. Bywaters, V. John, J. Lynch, P. Mattila, G. Norton, J. Stowell, M. Salata,

O.Labath, A. Chertok, D. Hablanian,Northern Power Systems WindPACT
Drive Train Alternative Design Study Report April 12, 2001 to January 31,
2005.
[4]
http://www.energy.kth.se/compedu/webcompedu/S5_Aeroelasticity/B1_Introd
uction_to_Aeroelasticity/C3_Bladed-
Disk_Vibrations/S5B1C3_files/Campbell_diagram.htm,[Online] Last
Accessed 19/10/2016.
[5] R.Poore, T. Lettenmaier,Alternative Design Study Report: WindPACT
Advanced Wind Turbine Drive-train Designs Study. NREL/SR-500-33196.,
August 2003.
[6] G. Bywaters, P. Mattila, D. Costin, J. Stowell, V. John, S. Hoskins, J. Lynch,

T. Cole, A. Cate, C. Badger, B. Freeman, Northern Power NW 1500 Direct-
Drive Generator. NREL/SR-500-40177, October 2007.
[7] J. Heikkinen, J. Sopanen, V. Ruuskanen, J. Nerg,Dynamic analysis of a direct-

driven permanent magnet generator drive train including flexible turbine
blades, ASME International Design Engineering Technical Conferences
Computers and Information in Engineering Conference, IDETC/CIE
2011,Washington, DC, USA, Aug. 2011.
[8] J. Islam , D. Svechkarenko , R. Chin , A. Szucs , J. Mantere , R. Sakki,

Cogging torque and vibration analysis of a direct-driven PM wind generator
with concentrated and distributed windings, 15th International Conference on
Electrical Machines and Systems (ICEMS), Sapporo, 21-24 Oct. 2012.
[9] J. Sopanen, V. Ruuskanen, J. Nerg, J. Pyrhonen, Dynamic Torque Analysis of

a Wind Turbine Drive Train Including a Direct-Driven Permanent-Magnet
Generator, IEEE Transactions on Industrial Electronics, (2011).58, (9), pp.
3859-3867.
[10] E.B.Smith, Nacelle Design for a 10MW reference wind turbine, M.S thesis.
Thesis, NTNU, 2012.
219
[11] S. S. Klair, Design of Nacelle and Rotor Hub for NOWITECH 10MW
Reference Turbine, MS Thesis, NTNU, 2013.
[12] L. Frøyd, O.G. Dahlhaug, Rotor Design for a 10 MW Offshore Wind Turbine,
Proceedings of the Twenty-first (2011) International Offshore and Polar
Engineering Conference, Maui, Hawaii, USA, 19-24 June 2011.
[13] L. Sethuraman, “Hydrodynamics and drive-train dynamics of a direct-drive

floating wind turbine”, PhD thesis, University of Edinburgh, Edinburgh, 2014.
[14] Y. Xing, M. Karimirad, T. Moan, Effect of Spar-Type Floating Wind Turbine
Nacelle Motion on Drivetrain Dynamics, In Proceedings of EWEA 2012
Annual Event, Copenhagen, Denmark,16-19 April 2012.
[15] Y. Xing, M. Karimirad, T. Moan, Modelling and analysis of floating spar-type

wind turbine drivetrain, Wind Energy, (2012).(DOI: 10.1002/we.1590), pp.1-
23.
[16] Shrestha et al., “Structural Flexibility: A Solution for Weight Reduction of

Large Direct-Drive Wind-Turbine Generators”, IEEE Transactions on Energy
Conversion, Vol. 25, No. 3, Sept. 2010.
[17] A. Zavvos, “Structural Optimisation of Permanent Magnet Direct Drive

Generators for 5MW Wind Turbines”, PhD thesis, University of Edinburgh,
Edinburgh, 2013.
[18] M. Kirschneck, “Mastering Electro-Mechanical Dynamics of Large Off-Shore

Direct-Drive Wind Turbine Generators”, PhD thesis, TU Delft, Delft (The
Netherlands), 2016.
[19] https://sites.ualberta.ca/~yousefim/bfiles/Ebooks/Vinogradov%20-
%20Fundamentals%20of%20kinematics%20and%20dynamic%20of%20mach
ines%20and%20mechanisms.pdf, [Online] [Accessed 10 March 2017].
[20] http://www.prismengblog.com/2016/05/31/faster-solidworks-simulations-
with-draft-quality-mesh, [Online] Last Accessed 19/10/2016.
[21] W. Leithead, Wind Turbine Technology 2, Wind CDT material course,

University of Strathclyde, Glasgow, 2012.
[22] SolidWorks, http://www.solidworks.com/sw/products/simulation/frequency-
analysis.htm, [Online] Last Accessed 19/10/2016.
[23] R. J. Roark and W. C.Young, ‘Roark's Formulas for Stress and Strain’,
McGraw-Hill International Editions, Singapore, 1989, 6th Edition.

220
Discussions and Conclusions
Chapter 7
7.1 Chapter summaries
This thesis addresses the structural analysis and the different ways of efficiently
designing low speed, high torque radial flux permanent magnet electrical generators
for direct drive wind energy converters with a view to minimise their mass. In this
chapter, a summary of each chapter is presented. The outcomes obtained from this
investigation will be discussed accompanied by the conclusions drawn.
Chapter 2 introduces the reader to renewables and the diverse types of wind energy
converters. Looking more in detail, two alternatives for the drivetrain are available:
221
one which utilises a gearbox to step up the low speed of the wind turbine rotor so that
it can be connected to a conventional high speed electrical generator and one in
which the electrical generator is directly connected to the turbine rotor. The latter,
also known as a direct drive system, removes the gearbox from the arrangement
minimizing the number of mechanical moving parts making the device potentially
more reliable and efficient and less noisy.
However, with the layout where the gearbox is eliminated, the electrical machine
copes with a very large torque and forces, hence it needs to be bigger, heavier and
more robust than conventional high speed generators. Whereas the common way to
approach the design of such machines was to focus all the attention on the mass of
the active material and give the highest priority to the electrical aspects, in this
investigation the author has concentrated on the amount of material required to
maintain the structural integrity of the generator taking into account mechanical
matters.
In Chapter 3, the distinct existing types of simplified radial flux generator structures
are shown as described in [1] and the concepts of structural stiffness and magnetic
stiffness introduced [2]. Due to the variety of forces acting on the direct drive
generator, the structure can deform in many different ways which can be
characterized by a mode number, ‘n’. These modes of deflection can affect the
magnetic flux density in the generator airgap inducing forces that try to close it. The
most significant load is known as the normal component of Maxwell stress and the
necessary stiffness to resist it has been calculated employing a parametric model,
which links magnetic and structural design, previously proposed by Tavner and
Spooner in [3]. This method has been corrected, enhanced and further developed.
Equalities capable of computing the required structural stiffness for conventional
synchronous machines and permanent magnet machines have been generated. A
permanent magnet machine has been assessed using this model and FEMM for
validation.
222
Chapter 4 presents three techniques that have been used to estimate the minimum
stiffness and mass that the generator structure needs to withstand static loads without
sensibly deforming. The mentioned three methods, which correspond to finite
element, analytical and hybrid, have been categorised according to their flexibility
and suitability to produce accurate results against the benchmark of finite element
analysis. The hybrid approach consists of combination of the data acquired from
dimensional homogeneity studies and the results achieved from finite element
analysis. This versatile method is capable of producing accurate results for Mode 0
and Mode 1 deflection much quicker than the FE method. The analytical technique is
also much faster than the finite element approach although it is only suitable to
predict the stiffness of certain types of structures which are under Mode 0 deflection.
However, although much slower, the FE method is able to estimate the stiffness of
any type of structure under any sort of deflection mode. Moreover by using the
retrieved data from FE analyses, 2D structural contour maps, where the cylinder and
the disc thicknesses are set as the independent and dependent variables respectively,
can be generated and utilized as a mean for finding the minimum structural mass of
the generator.
Structural stiffness modelling of disc and armed structures is also presented in

Chapter 4. The outcomes obtained from the analyses carried out using the said three
techniques are shown and evaluated. A comparison between disc and armed
structures is presented revealing that structures with hollow arms are not able to
resist high torque loads as well as structures made with discs do unless the thickness
of the arms is considerably increased. This leads to an increase in mass. Thereafter, a
disc structure suitable for a 3 MW direct drive generator was optimised with the
main aim of minimising its weight considering normal, tangential and gravitational
loads. For the mass minimisation, as well as the loads, a stiffness and mass trade-off
study of the rotor and stator structures was carried out. By setting the minimum
stiffness of each structure forming the generator as a threshold, and considering the
minimum stiffness needed by the entire machine, its mass can be diminished by
targeting the heaviest structure that typically corresponds to the stator. Other ways of
reducing the mass of electrical machines disc structures are suggested and put in
223
practice and a distinct structural configuration, such as rotor conical structures,

proposed and analysed. Both disc and conical structures are studied under several
deflection modes.
Chapters 2, 3 and 4 provide the reader with the necessary information to understand
the theory and the issues behind the static design of a supporting structure for a multi
MW electrical machine. Different design tools, which can be employed during the
early stages of the design process for the estimation of the required minimum
structural stiffness, are developed and described.
Chapter 5 introduces the reader to composite materials. An overview of the theory

behind and the work done in the past in related fields using these materials is given.
A small scale disc structure corresponding to a 100 kW generator was first modelled
following what it is called the conventional approach, in which the plies forming the
stacking sequence are placed according to the Cartesian coordinate frame. Then, the
disc structure was again modelled but this time following an innovative approach
utilised for the design of flywheels, in which the fibres are placed following 5
different cylindrical coordinate systems. The disc sub structure was divided into 11
distinct areas where the plies formed by the said fibres were located. An optimisation
of this type of structure was finally made obtaining a difference in mass with its
optimised counterpart made of steel of 82% and 33.5% if compared with the
composite structure modelled following the conventional approach.
A model of the larger direct drive generator corresponding to a 3 MW machine was
also created utilising composite materials employing the conventional and the more
advanced approach. In this case, the difference in mass is about 60%. However, the
more advanced structure resulted in a slightly heavier than the conventional approach
one due to a lack of adaptability in the design. Diverse ways of implementing the
needed adaptability to pursue more ambitious optimisations are proposed.
In Chapter 6 the optimised steel structure is studied from a dynamic point of view
making use of modal analyses. The main aim was to work out the natural frequencies
of the structure and the distinct ways of changing them so that the excitation
224
frequencies do not match them. The thickness of both the disc and the outer rim sub
structures were varied revealing that thicker cylinders reduce the natural frequencies
of the overall structure, whereas thicker discs have the opposite effect. Different
numbers of stiffeners, first located on one side of the disc sub structure and then
placed on both sides, were tried too. It was observed that the structural mass was
increased, while the natural frequencies did not increase as expected. A distinct
material with lower density was tried in order to see how the reduction in weight of
the stiffeners would affect the overall mass. The result achieved showed that the
increase in mass was still high for the small increase in the natural frequencies of the
structure. With all of these options exploded, it was thought of another kind of
structural configuration such as a conical structure where the cone contributes with a
higher axial stiffness. The location of the cone sub structure was altered – with
respect to the cylinder – until the ideal position was found. The structure was then
optimised following the same parameters as with the disc structure, obtaining an
optimum cone angle of 50 degrees, and analysed using FE tools. The studies revealed
a superior performance of the conical structure if compared to the disc structure with
less mass needed to withstand the same loads and natural frequencies almost 2 times
higher. The advantages and drawbacks of utilising rotor conical structures are
discussed.
7.2 Discussion
In this section, the gained insights from the chapters above are outlined and
discussed. The broad approach of this thesis includes the next ideas:
1. The use of simple generator structures that can reproduce the types and
character of airgap deflection and mass features of real generators used in
industry.
2. Development of design tools and knowledge that can be used in the early
phase of the design process to estimate the required minimum structural
stiffness at the lowest mass that allows fast calculations and straightforward
optimisations.
225
3. To find a simplified structure made with conventional materials that can meet
major structural requirements in a lightweight manner.
4. To explore the possibility of employing low density materials, such as
composites, in the supporting structure design in order to minimise its overall
mass.
5. Whether stiffness through static modelling is sufficient or if dynamic
modelling is also necessary.
7.2.1 Evaluating stiffness
Simplified structural generator models have been analysed using different

approaches. A method coupling the magnetic and mechanical design has been
developed and described in Chapter 3. It is a versatile 2D magnetic model which by
assuming a deflection computes the airgap closing force and the airgap stiffness of
two different types of electrical machines under distinct modes of deflection. Then,
by making a comparison with the results obtained from the mechanical model fed
with the outcomes of a finite element structural model of the machines, the structure
in question can be considered suitable or not to successfully carry the imposed loads.
The given data corroborated the applicability and accuracy of this technique to
estimate the required stiffness that can be utilized during the design stage, after the
manufacturing process and as part of a condition monitoring system. In Chapter 4,
another three distinct methods capable of calculating the necessary stiffness and
structural mass of the entire machine with high accuracy were presented and
evaluated. As proved, they can be used to optimize the structure so that it can
withstand the major internal stresses present during operation at the lowest mass.
All of the mentioned techniques are very easy to use. In addition, their high
versatility make them very useful especially at the early stages of the design process.
They would accelerate the calculations and give the designer the possibility of
carrying out a quick estimation of the needed stiffness and mass.
Further structural optimizations could also be made by using topology optimization

packages. Their use after the studies made with the tools mentioned above is
226
recommended by the author as additional mass savings can be achieved through the
elimination of parts which do not contribute to carry the loads.
On the other side, it is important to highlight that these methods have been developed
taking into account the major internal loads that can act on the generator structure.
This means that the electrical machine is completely isolated from the rest of the
turbine and that no external loads, except torque, have been considered. Generator
dynamics have not been included either. Therefore, these techniques only apply to
the generator supporting structure design from a static viewpoint.
7.2.2 Lightweight materials for generator supporting structures
The use of low density materials, such as composites, has been considered in the
design of the generator structure with the main aim of reducing its overall mass. The
anisotropic nature of composite materials makes them harder to model and
introduces additional complexity to the structure design. Nevertheless, the graphical
interface of the software and its flexibility to change parameters makes the modelling
and optimisation process more straightforward. The results obtained showed that an
advanced composite material structure can deal with loads as well as a structure
made of steel does. By tailoring its structural properties, it was possible to find an
arrangement capable of spreading out the stress across the structure and withstand
the loads keeping the deflection limits within the specified range. In addition, the
better fatigue properties of composite materials help to extend the lifespan of the
entire wind turbine and their higher Young’s modulus to density ratios reduces the
overall generator mass by about 60 %. It is important to note that the plies utilised for
modelling the structure were assumed perfectly manufactured. This means that no
defects and ideal alignment of fibres were considered. In reality, it is not possible to
perfectly align the fibres and although very small, the possibility of having voids or
any other kind of imperfection is something to take into account. These factors can
affect the structural weight reduction achieved although not heavily.
During the machine modelling and optimisation stage, thermal considerations were
not assumed. Wind turbine electrical generators operate for long periods of time at
really high temperatures. Thermal expansion properties of composite structures can
227
also be tailored to meet the needs of the design potentially introducing further
enhancements and that is why the author believes that further research must be done
in this area. The inherent dynamic features of the electrical machine and the
influence of external loads on its composite structure have not been investigated
here. However, the characteristics of the structure with tailored high stiffness and
low mass suggests that it is possible to achieve satisfactory dynamic features and still
have a lighter machine. Extensive investigation on further mass optimisation of the
mosaic pattern composite structure is required as with the introduction of higher
design adaptability additional mass savings can be achieved. The interaction between
the rotor and the stator yokes, made of iron, with the composite structure should also
be studied. In the aerospace industry, the attachment of composite structures
typically implies the utilisation of titanium fasteners which are lighter than its steel
counterparts and that not only introduces the necessary stiffness but also impedes
galvanic corrosion. The fact of employing carbon/epoxy for the structure and
titanium fasteners will considerably increase the cost of the machine. Its
manufacturing process will be more expensive than that of a conventional generator
steel structure. Different options can be studied in order to diminish the cost of the
machine structure and its manufacturing process such as using rotor conical
structures. The cone angle can benefit the integration of the cone sub structure fibres
into the outer casing, opening the door to manufacture the structure as only one piece
eliminating so the need for fasteners and reducing the weight and the production
time. Bearing all of this in mind and once the suggested research has been carried
out, the author would propose to build up a model of the structure and test it before
undertaking large scale projects. A detailed economic study considering the
advantages and drawbacks derived from the use of composites to produce a generator
structure is needed before giving the go-ahead or rejecting the task.
7.2.3 Direct drive electrical generator dynamics
Several tools have been developed and evaluated for the static design of simplified
direct drive electrical machine structures. However the inherent dynamic nature of
228
the wind turbine led the author to look at the generator from a dynamic perspective in
Chapter 6.
Considering an ideal bearing configuration in which the external loads coming from
the hub are directly transmitted to the tower, only the major internal loads acting on
the generator structure have been assumed for its design and analysis. Nevertheless,
the external loads have an important influence on the internal loads that must be
taken into account. This influence strongly depends on how well integrated in the
wind turbine system the generator is. The structure has to be capable of dealing with
internal stresses with larger amplitudes and variable frequencies produced by the
induced effect of the external loads without sensibly deforming. In order to make this
possible it is estimated that an extra 10 % of stiffness as composed to static case
needs to be added to the structure in every direction.
Resonance is another issue that needs to be addressed so as to maintain the integrity

of the machine. For that, a range of possibilities has been explored in Chapter 6. To
increase the structure natural frequencies, the use of stiffeners introducing more
support into sensitive zones is a solution widely adopted by manufacturers. Other
options include the utilisation of low density materials in the structure design or a
change in geometry. By looking at the Campbell diagram of the example provided in
Chapter 6, the operation range of the machine could be identified. In addition, it was
observed that the said operational range narrows down as the structure natural
frequencies increase due to the match with the fundamental electrical frequency. Due
to this, other possibilities were proposed but this time the attention was focused on
the electrical side. In order to expand the machine range of operation so that the
energy capture can be maximized the number of pole pairs can be modified. An
increase would mean that the electrical frequency would cross the lowest modes of
the rotor and the stator at higher rotational speeds reducing the operational range
even further. But if the number of pole pairs is diminished the slope of the line would
increment. With this, it would be possible to avoid the lowest modes of both the rotor
and the stator structures although it would require having thicker rotor and stator
yokes, which would elevate the mass of the generator. In any case, the author
229
believes that it is necessary to go through this trade-off process so that the optimum
integrated design is achieved according to the requirements of the project.
7.3 Revisiting the thesis research question
In Chapter 1, the research question is stated as:
“Can electromagnetic and structural stiffness models be used effectively to minimise

the mass of electrical generators for direct drive wind turbines?”
In order to find an answer, it is necessary to know the minimum needed stiffness so

that all the loads acting on the generator can be resisted without deforming to find the
minimum required mass.
In this project, a number of models were generated. The way in which they can be
used during the design and optimisation stages was also described and explained. In
Chapter 3, an electromagnetic-mechanical stiffness model was developed. The
retrieved results showed how versatile and accurate the model is so as to calculate
the minimum required structural stiffness of two different types of electrical
machines. Chapter 4 illustrates the distinct techniques available and establishes a
hierarchy according to their suitability and accuracy to estimate the minimum
stiffness and mass of electrical machine structures. In Chapter 5, models of generator
structures made of low density materials are analysed revealing a high positive
impact on the effort to reduce the machine’s supporting structure mass. As the
mentioned chapters covered the design of the generator from a static perspective, in
Chapter 6, the generator structures were studied from a dynamic point of view. The
outcomes helped to understand the influence of the dynamics in the electrical
machine design. It was estimated that an extra 10 % of stiffness, as composed to the
static case, in every direction is necessary to deal with the internal stresses with
larger amplitudes and variable frequencies produced by the induced effect of the
external loadings although this figure might vary as the influence of the external
loads on the generator structure strongly depends on how well the electrical machine
is integrated in the wind turbine. It could be said that the answer to the research
230
question is affirmative as the models presented in this thesis were proved to be

effective when they were used to reduce the structural mass of the machine.
7.4 Contribution to knowledge
This thesis has contributed to knowledge in a number of ways:
 A consistent set of equations coupling electromagnetic and mechanical

designs of different types of electrical machines, which is used to model the
minimum required stiffness, has been derived and validated for different
modes using finite element tools. A case study illustrating the suitability of
the magneto-mechanical model has been introduced (Chapter 3).
 Distinct structural design tools have been developed, validated and
categorized according to their flexibility and suitability to estimate the
necessary stiffness of sub structures and complete structures (Chapter 4).
 Diverse structural configurations have been analysed and compared in order
to identify the arrangement that is capable of coping with the loads with the
minimum mass. The optimum disc model, found by using 2D contour maps
of stiffness and mass of rotor and stator structures, was studied under several
deflection modes and further optimised using a commercial piece of software.
A rotor conical structure was proposed and analysed under different
deflection modes (Chapter 4).
 A new concept for the design of electrical machine disc structures using low
density materials has been introduced and verified for small and large scale
generators. The structural mass reduction achieved with this design is
investigated (Chapter 5).
 The options available to alter the natural frequencies of a structure have been
presented and analysed through modal studies. A dynamic study of the
optimum disc structure has been conducted, establishing structural excitation
sources and how to interact with them. A new structural configuration (rotor
conical) has been dynamically studied and optimised (Chapter 6).
The work in this thesis has contributed to a number of other publications:
231
Structural stiffness analysis of radial flux permanent magnet direct drive generators
in Chapter 3 in [2]; A comparative study of methods for modelling the structural
stiffness of generator components in Chapter 4 in [4] and A New Method for
Coupling Structural and Magnetic Models for the Design and Optimization of Radial
Flux PM Generators for Direct-Drive Renewable Energy Applications in [5]; A
lightweight approach for airborne wind turbine drivetrains in Chapter 5 in [6].
7.5 Further work
This thesis covers a number of topics that are of vital importance in the design of
electrical generators structures. Due to time limitations, it did not go deeper into the
fields it addresses. In order to improve the techniques introduced in this dissertation
further research is necessary. The following points need to be investigated in more
detail:
 The model coupling magnetic and mechanical designs it is a static

representation that neglects axial end effects and does not consider any
external influence. Further investigation should be carried out so that these
points can be covered.
 The optimised generator models produced in this thesis do not consider
thermal issues or any other external load coming from the wind turbine rotor.
External loads compromise any load coming from the wind turbine rotor
(including thrust). External loads have a significant influence on internal
loads. As said, the degree of influence depends on how well integrated the
generator is in the wind turbine structure. The electrical machine is directly
connected to the hub through a shaft. Misalignments and other factors that
can affect the normal operation of the generator and the integrity of its
structure should be also analysed.
 The majority of the simplified structures proposed by Stander in [1] have
been studied. However, supporting structures made with spokes or star shapes
are still missing. Further optimisations might be possible with the use of such
layouts.
232
 The mosaic pattern structural model produced for the 3 MW direct drive
machine using composite materials in Chapter 5 was not able to reduce the
weight of the structure more than the conventional approach model because
of the lack of flexibility in its design. Additional optimisations could be made
by making use of mosaic pattern structures with more sub divisions in the
disc sub structure. This would increase considerably its cost of
manufacturing; hence a detailed economic study should be made before
approaching this task.
 Avoidance of structure’s natural frequencies is a key matter that must be
addressed with the highest precision possible. Extensive research should be
carried out on this field with the main aim of maximising the machine
operation range with the minimum mass.
 A deeper investigation on the distinct methods for minimising the mass of
conical structures is desirable.
233
7.6 References

[2] P. Jaen-Sola and A. S. McDonald, “Structural Analysis and Characterization
of Radial Flux PM Generators for Direct-Drive Wind Turbines,” 3rd Renew.
Power Gener. Conf. (RPG 2014), Naples (Italy) Sept. 2014.
Electrical Machines, Chania, Greece, 2006.
[4] P. Jaen-Sola and A. S. McDonald, “A Comparative Study of Methods for
Modelling the Structural Stiffness of Generator Components”, IET Power
Electronics, Machines & Drives (PEMD), Glasgow (UK), Apr. 2016.
[5] A. S. McDonald and P. Jaen-Sola, “A New Method for Coupling Structural
and Magnetic Models for the Design and Optimization of Radial Flux PM
Generators for Direct-Drive Renewable Energy Applications”, IET Renewable
Power Generation Journal, Dec. 2016.
[6] P. Jaen-Sola and A.S. McDonald, “A Lightweight Approach for Airborne
Wind Turbine Drivetrains”, European Wind Energy Association International
Conference (EWEA), Paris (France), Nov. 2015.
234
Appendix
Appendix
Natural frequencies: Disc rotor structure
60
50
f = 349.58td - 3.5856
R² = 0.9998
Frequency (Hz)
40
30
20
10
0
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18
Disc thickness (m)
(a)
70
60 f = 366.1td + 0.0431
R² = 0.9999
50
Frequency (Hz)
40
30
20
10
0
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18
Disc thickness (m)
(b)
235
Appendix
70
60
f = 394.03td + 0.1858
50 R² = 0.9995
Frequency (Hz)
40
30
20
10
0
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18
Disc thickness (m)
(c)
70
60
f = 394.2td + 0.1881
50 R² = 0.9995
Frequency (Hz)
40
30
20
10
0
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18
Disc thickness (m)
(d)
Figure A.1 Frequency variation as dimensions are altered with rotor cylinder sub structure thickness maintained
at 0.02 m; (a) 2nd mode; (b) 3rd mode; (c) 4th mode; (d) 5th mode
236
Appendix
20
18
16
14
Frequency (Hz)
12
10
8
f = 462.71tc2 - 141.26tc + 19.747
6 R² = 0.9969
4
2
0
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16
(a)
25
20
Frequency (Hz)
15
f = -72.263tc + 23.102
R² = 0.9858
10
0
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16
(b)
237
Appendix
50
45
f = 1025.5tc2 + 13.01tc + 22.569
40
R² = 0.9961
35
Frequency (Hz)
30
25
20
15
10
5
0
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16
(c)
50
45
f = 1026.4tc2 + 12.857tc + 22.578
40 R² = 0.9961
35
Frequency (Hz)
30
25
20
15
10
5
0
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16
(d)
Figure A.2 Frequency variation as dimensions are altered with rotor disc sub structure thickness maintained at
0.06 m; (a) 2nd mode; (b) 3rd mode; (c) 4th mode; (d) 5th mode
238
Appendix
Natural frequencies: Disc stator structure
120
100
Frequency (Hz)
80
f = -4296ts,d2 + 1009.6ts,d + 36.9

60
R² = 0.9716
40
20
0
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16
Disc thickness (m)
(a)
120
100
Frequency (Hz)
80
f = -4092.4ts,d2 + 970.33ts,d + 38.681
60 R² = 0.9745
40
20
0
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16
Disc thickness (m)
(b)
239
Appendix
200
180
160
140
Frequency (Hz)
120
100 f = -9559.1ts,d2 + 2620.1ts,d + 7.3029
R² = 0.9993
80
60
40
20
0
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16
Disc thickness (m)
(c)
200
180
160
140
Frequency (Hz)
120
100 f = -9908ts,d2 + 2661ts,d + 7.8964
R² = 0.9997
80
60
40
20
0
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16
Disc thickness (m)
(d)
Figure A.3 Frequency variation as dimensions are altered with stator cylinder sub structure thickness
maintained at 0.02 m; (a) 2nd mode; (b) 3rd mode; (c) 4th mode; (d) 5th mode
240
Appendix
55
54
Frequency (Hz) 53
52
51 f = -812.68ts,c2 + 104.35ts,c + 51.079

R² = 0.9452
50
49
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16
(a)
57
56
55
Frequency (Hz)
54
53
52
f = -854.29ts,c2 + 101.33ts,c + 52.558
51 R² = 0.9468
50
49
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16
(b)
241
Appendix
62
61
Frequency (Hz) 60
59
f = 10108ts,c3 - 2995.7ts,c2 + 286.52ts,c + 51.934
R² = 0.9788
58
57
56
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16
(c)
63
62
61
Frequency (Hz)
60
f = 9427.1ts,c3 - 2812.3ts,c2 + 272.62ts,c + 52.906
59 R² = 0.9825
58
57
56
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16
(d)
Figure A.4 Frequency variation as dimensions are altered with stator disc sub structure thickness maintained at
0.02 m; (a) 2nd mode; (b) 3rd mode; (c) 4th mode; (d) 5th mode
242
Appendix
Natural frequencies: Rotor conical structure
70
60
50
Frequency (Hz)
40
30 f = 291.75tcon + 18.537
R² = 0.9972
20
10
0
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18
Cone thickness (m)
(a)
90
80
70
Frequency (Hz)
60
50
f = 274.91tcon + 36.015
40 R² = 0.9858
30
20
10
0
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18
Cone thickness (m)
(b)
243
Appendix
90
80
70
Frequency (Hz) 60
50
f = 274.83tcon + 36.063
40 R² = 0.9864
30
20
10
0
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18
Cone thickness (m)
(c)
100
90
80
70
Frequency (Hz)
60
50
f = 350.92tcon + 34.702
40 R² = 0.987
30
20
10
0
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18
Cone thickness (m)
(d)
Figure A.5 Frequency variation as dimensions are altered with cylinder thickness maintained at 0.02 m; (a) 2 nd
mode; (b) 3rd mode; (c) 4th mode; (d) 5th mode
244
Appendix
35
30
25
Frequency (Hz)
20
15 f = -47.152tcyl + 30.311
R² = 0.6906
10
0
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16
(a)
50
45
40
35
Frequency (Hz)
30
25
f = 3369.1tcyl2 - 542.53tcyl + 53.758
20 R² = 0.9571
15
10
5
0
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16
(b)
245
Appendix
50
45
40
35
Frequency (Hz)
30
25
f = 3383.2tcyl2 - 545.32tcyl + 53.885
20
R² = 0.9571
15
10
5
0
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16
(c)
90
80
70
Frequency (Hz)
60
50
f = -4456.9tcyl2 + 943.12tcyl + 25.127
40 R² = 0.8897
30
20
10
0
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16
(d)
Figure A.6 Frequency variation as dimensions are altered with cone thickness kept at 0.04 m; (a) 2nd mode; (b)
3rd mode; (c) 4th mode; (d) 5th mode
246

Advanced Structure Modeling ....

Uploaded by

Copyright:

Available Formats

Advanced Structure Modeling ....

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Advanced Structure Modeling ....

Uploaded by

Copyright:

Available Formats

UNIVERSITY OF STRATHCLYDE

Pablo Jaen Sola

A thesis presented in fulfilment of the requirements

Centre for Doctoral Training in Wind Energy Systems

Department of Electronic and Electrical Engineering

I am grateful to my second supervisor, Dr Erkan Oterkus for his support and

Finally, but most importantly, I would like to express my eternal gratitude to my

In this thesis a stiffness approach is introduced which combines electromagnetic

Chapter 1: Introduction ............................................................................................ 1

Chapter 2: Literature Review ................................................................................... 8

Chapter 3: Magnetic Stiffness Modelling of Wind Turbine Electrical

Chapter 4: Comparison of Methods for Estimating Generator Structural

Chapter 5: Lightweight Materials in Generator Structures .............................. 144

Chapter 6: Dynamics of a Direct Drive Generator ............................................. 185

Chapter 7: Discussions and Conclusions ............................................................. 221

Appendix ............................................................................................................. 235

Figure 3.1 Coupling of magnetic and structural models ...................................................................... 60

Figure 6.1 Campbell diagram for a 3 MW machine [3] .................................................................... 186

Table 3.1 Case study generator data .................................................................................................... 64

Table 6.1 Simplified rotor model features ......................................................................................... 199

Aarm Cross-section area of the arm, m2

Aij Extensional and shear stress stiffnesses, N/m

B Magnetic flux density, Wb/m2

Bg Peak flux density in the airgap, T

Bij Extension-bending coupling stiffnesses, N/m

Br Remanent flux density in the airgap, T

b Beam width in X direction, m

Dij Bending and torsional stiffnesses, N/m

Ds Distance between holes centres (stator further optimisation), m

ds Optimisation holes diameter (stator further optimisation), m

F Force caused by the normal component of the Maxwell stress, N

Fc Airgap closing force, N

F̂ Peak magnetomotive force, At

h Beam’s height in Y direction, m

hry Rotor yoke height, m

hsy Stator yoke height, m

i Finite number of bearing radial stiffness

𝑘a,eq Arms equivalent stiffness, N/m

𝑘B Stiffness of component B, N/m

𝑘c,s Stator cylinder stiffness for disc structures, N/m

𝑘d Disc stiffness, N/m

𝑘eq,d Equivalent rotor stiffness, N/m

𝑘eq,r Equivalent rotor stiffness, N/m

𝑘s,s Stator stiffness, N/m

ms,s mass of stator made with discs, kg

ms,as mass of stator made with arms, kg

P Frequency of the wind turbine rotational speed, Hz

ts,d Stator disc sub structure thickness, m

δ Distance between composite fibres, m

δStructural Structural deflection, m

𝛿∆ Deflection change respect to the mean, m

𝜀arm Arm strain

θ Angular displacement, degrees

θ Cross-section area of a single fibre, m2

μ0 Permeability of free space, 4̟×10-7 H/m

ρc Composite material density, kg/m3

σ Airgap shear stress, Pa

𝜎PM Stress during operation in permanent magnet machine, Pa

𝜎N∞ Residual stress from Mar-Lin formula, Pa