GUIDELINES For Design of Wind Turbines PDF

Guidelines for
Design
of
Wind Turbines
A publication from
DNV/Risø in technical co-operation
Guidelines for Design of Wind Turbines
1st Edition
 Det Norske Veritas, Copenhagen (Certification@dnv.com) and Wind Energy Department,
Risø National Laboratory (Certification@risoe.dk) 2001
All rights reserved. No part of this publication may be reproduced, stored in a retrieval sys-
tem, or transmitted, in any form or by any means, electronical, mechanical, photocopying,
recording and/or otherwise without the prior written permission of the publishers.
This book may not be lent, resold, hired out or otherwise disposed of by way of trade in any
form of binding or cover other than that in which it is published, without the prior consent of
the publishers.
The front-page picture is from Microsoft Clipart Gallery ver. 2.0.
Printed by DSI Grafisk Service, Denmark 2001
ISBN 87-550-2870-5
2
Guidelines for Design of Wind Turbines − DNV/Risø
Preface quirements that need to be fulfilled in order

to achieve certification of his wind turbine.
The guidelines for design of wind turbines The guidelines can be used by wind turbine
have been developed with an aim to compile manufacturers, certifying authorities, and
into one book much of the knowledge about wind turbine owners. The guidelines will
design and construction of wind turbines also be useful as an introduction and tutorial
that has been gained over the past few years. for new technical personnel and as a refer-
This applies to knowledge achieved from ence for experienced engineers.
research projects as well as to knowledge
resulting from practical design experience. The guidelines are available as a printed
In addition, the various rules and methods booklet in a handy format as well as elec-
required for type approval within the major tronically in pdf format on a CD-ROM.
markets for the wind turbine industry form a
basis for the guidelines, with emphasis on The development of the guidelines is the
the international standards for wind turbines result of a joint effort between Det Norske
given by the International Electrotechnical Veritas and Risø National Laboratory. The
Commission, IEC. development is based on funding from Dan-
ish Energy Agency, Det Norske Veritas and
The objective of the guidelines is to provide Risø National Laboratory.
a unified basis, which can be used for design
of different types of wind turbines in the These guidelines for design of wind turbines
future. The guidelines provide recommen- have been thoroughly validated and verified.
dations and guidance for design together However, no warranty, expressed or im-
with application-oriented solutions to complied, is made by Det Norske Veritas and
monly encountered design problems. The Risø National Laboratory, as to the accuracy
level of detail is such that the user of the or functionality of the guidelines, and no
guidelines should be able to meet the re- responsibility is assumed in connection
therewith.
Guidelines for Design of Wind Turbines

© 2001 Det Norske Veritas and Risø National Laboratory
Preface i
ii Preface
0. Contents 2.5.3 Normal operation .................. 27

2.5.4 Service, maintenance and repair
1. WIND TURBINE CONCEPTS..... 1 ....................................................... 27
2.6 CODES AND STANDARDS....... 28
1.1 INTRODUCTION .......................... 1 DS472............................................. 28
1.2 CONCEPTUAL ASPECTS ............ 1 IEC61400-1.................................... 28
1.2.1 General.................................... 1 NVN11400-0 .................................. 28
1.2.2 Vertical axis turbines .............. 2 DIBt RICHTLINIEN....................... 28
1.2.3 Horizontal axis turbines .......... 2 GL REGULATIONS ....................... 28
1.2.4 Number of rotor blades ........... 3 REFERENCES ................................... 28
1.2.5 Power control aspects ............. 3
1.3 ECONOMICAL ASPECTS............ 4
1.4 EFFICIENCY ASPECTS ............... 5 3. EXTERNAL CONDITIONS ............ 31
1.5 CONFIGURATIONS AND SIZES 6 3.1 WIND CONDITIONS .................. 31
1.6 FUTURE CONCEPTS.................... 7 3.1.1 10-minute mean wind speed .. 31
3.1.2 Standard deviation of wind
2. SAFETY AND RELIABILITY .......... 9 speed .............................................. 32
3.1.3 Turbulence intensity .............. 35
2.1 SAFETY PHILOSOPHY................ 9 3.1.4 Lateral and vertical turbulence
2.2 SYSTEM SAFETY AND ....................................................... 35
OPERATIONAL RELIABILITY....... 10 3.1.5 Stochastic turbulence models 36
2.2.1 Control System ...................... 10 3.1.6 Wind Shear............................ 38
2.2.2 Protection System.................. 11 3.1.7 Transient Wind Conditions ... 40
2.2.3 Brake system.......................... 13 3.1.8 Extreme winds – gusts........... 41
2.2.4 Failure mode and effects 3.1.9 Site assessment ...................... 42
analysis........................................... 14 3.2 OTHER EXTERNAL CONDITIONS
2.2.5 Fault Tree Analysis ............... 14 ............................................................ 45
2.3 STRUCTURAL SAFETY ............ 16 3.2.1 Temperatures ........................ 45
2.3.1 Limit states ............................ 16 3.2.2 Density of air......................... 46
2.3.2 Failure probability and other 3.2.3 Humidity................................ 47
measures of structural reliability ... 17 3.2.4 Radiation and Ultraviolet Light
2.3.3 Structural reliability methods 18 ....................................................... 47
2.3.4 Code format, characteristic 3.2.5 Ice.......................................... 47
values, and partial safety factors ... 18 3.2.6 Rain, Snow and Hail ............. 47
2.3.5 Code calibration.................... 19 3.2.7 Atmospheric Corrosion and
2.3.6 Example – Axially Loaded Steel Abrasion......................................... 48
Tower ............................................. 21 3.2.8 Earthquake ............................ 48
2.3.7 Example – Fatigue of FRP Blade 3.2.9 Lightning ............................... 49
Root in Bending.............................. 22 REFERENCES ................................... 50
2.3.8 Tests and calculations for
verification ..................................... 24
2.3.9 Inspection and inspection 4. LOADS ............................................... 53
intervals.......................................... 24 4.1 LOAD CASES.............................. 53
2.4 MECHANICAL SAFETY............ 25 4.1.1 Design situations................... 53
2.5 LABOUR SAFETY ...................... 26 4.1.2 Wind events ........................... 53
2.5.1 General.................................. 26 4.1.3 Design load cases.................. 53
2.5.2 Transportation, installation and 4.2 LOAD TYPES .............................. 56
commisioning ................................. 27
0 – Contents iii
4.2.1 Inertia and gravity loads ....... 56 5.2.1 Determination of Design Loads
4.2.2 Aerodynamic loads................ 57 ..................................................... 117
4.2.3 Functional loads.................... 58 5.2.2 Strength Analyses ................ 117
4.2.4 Other loads............................ 58 5.2.3 Analyses of Bolt Connections118
4.3 AEROELASTIC LOAD 5.2.4 Hub Enclosure..................... 118
CALCULATIONS.............................. 58 5.2.5 Materials ............................. 118
4.3.1 General.................................. 58 5.2.6 Standards ............................ 119
4.3.2 Model elements...................... 59 REFERENCES ................................. 119
4.3.3 Aeroelastic models for load
prediction ....................................... 68
6. NACELLE .................................. 121
4.3.4 Aerodynamic data assessment68
4.3.5 Special considerations........... 69 6.1 MAIN SHAFT ............................ 121
4.4 LOAD ANALYSIS AND 6.1.1 Determination of Design Loads
SYNTHESIS....................................... 75 ..................................................... 121
4.4.1 Fatigue loads......................... 75 6.1.2 Strength Analysis................. 121
4.4.2 Ultimate loads ....................... 83 6.1.3 Fatigue Strength.................. 122
4.5 SIMPLIFIED LOAD 6.1.4 Ultimate strength................. 126
CALCULATIONS.............................. 87 6.1.5 Main Shaft-Gear Connection127
4.5.1 Parametrized empirical models 6.1.6 Materials ............................. 128
........................................................ 87 6.1.7 Standards ............................ 128
4.5.2 The simple load basis ............ 87 6.2 .............................. MAIN BEARING
4.5.3 Quasi-static method .............. 88 .......................................................... 128
4.5.4 Peak factor approach for 6.2.1 General ............................... 128
extreme loads ................................. 89 6.2.2 Determination of Design Loads
4.5.5 Parametrized load spectra .... 89 ..................................................... 131
4.6 SITE-SPECIFIC DESIGN LOADS92 6.2.3 Selection of Bearing Types.. 131
4.7 LOADS FROM OTHER SOURCES 6.2.4 Operational and Environmental
THAN WIND ..................................... 93 Conditions .................................... 131
4.7.1 Wave loads ............................ 93 6.2.5 Seals, Lubrication and
4.7.2 Current loads ........................ 99 Temperatures ............................... 131
4.7.3 Ice loads ................................ 99 6.2.6 Rating Life Calculations ..... 133
4.7.4 Earthquake loads .................. 99 6.2.7 Connection to main shaft .... 134
4.8 LOAD COMBINATION .............. 99 6.2.8 Bearing Housing ................. 134
REFERENCES ................................. 100 6.2.9 Connection to Machine Frame
..................................................... 134
6.2.10 Standards .......................... 134
5. ROTOR ....................................... 103
6.3 MAIN GEAR.............................. 135
5.1 BLADES ..................................... 103 6.3.1 Gear types ........................... 135
5.1.1 Blade geometry.................... 103 Rolling bearings........................... 137
5.1.2 Design Loads....................... 104 6.3.2 Loads and Capacity ............ 138
5.1.3 Materials and manufacturing104 6.3.3 Codes and Standards........... 142
5.1.4 Strength analyses ................ 110 6.3.4 Lubrication.......................... 142
5.1.5 Tip deflections ..................... 113 6.3.5 Materials and testing........... 143
5.1.6 Lightning protection............ 114 6.4 COUPLINGS.......................... 146
5.1.7 Blade testing........................ 114 6.4.1 Flange Couplings................ 146
5.1.8 Maintenance........................ 116 6.4.2 Shrink Fit Couplings ........... 147
5.2 HUB ............................................ 116 6.4.3 Key Connections ................. 147
iv 0 – Contents
6.4.4 Torsionally Elastic Couplings. 6.10.8 .. Special design considerations

...................................................... 147 ..................................................... 167
6.4.5 Tooth Couplings .................. 147 REFERENCES ................................. 167
6.5 MECHANICAL BRAKE ........... 147
6.5.1 General................................ 147
7. TOWER............................................ 169
6.5.2 Types of brakes.................... 147
6.5.3 Brake discs and brake pads. 149 7.1 INTRODUCTION ...................... 169
6.5.4 Brake torque sequence ........ 149 7.2 LOAD CASES............................ 170
6.6 HYDRAULIC SYSTEMS .......... 149 7.3 DESIGN LOADS........................ 170
6.6.1 General................................ 149 7.3.1 Loads and responses ........... 171
6.6.2 Arrangement........................ 150 7.3.2 Extreme loads occurring during
6.6.3 Accumulators....................... 150 normal power production............. 171
6.6.4 Valves .................................. 150 7.3.3 Extreme loads – parked turbine
6.6.5 Application in safety systems150 ..................................................... 173
6.6.6 Additional provisions .......... 152 7.3.4 Fatigue Loading .................. 173
6.6.7 Codes and Standards........... 152 7.4 GENERAL VERIFICATIONS FOR
6.7 GENERATOR ............................ 152 TOWERS.......................................... 175
6.7.1 Types of generators ............. 152 7.4.1 Dynamic Response and
6.7.2 Climate aspects ................... 154 Resonance .................................... 175
6.7.3 Safety aspects ...................... 154 7.4.2 Critical Blade Deflection
6.7.4 Cooling and degree of sealing Analysis ........................................ 177
...................................................... 155 7.5 TUBULAR TOWERS ................ 178
6.7.5 Vibrations............................ 156 7.5.1 Loads and responses ........... 178
6.7.6 Overspeed............................ 156 7.5.2 Extreme loads...................... 179
6.7.7 Overloading ........................ 156 7.5.3 Fatigue loads....................... 179
6.7.8 Materials ............................. 156 7.5.4 Vortex Induced Vibrations .. 180
6.7.9 Generator braking............... 157 7.5.5 Welded joints....................... 180
6.7.10 Lifetime.............................. 158 7.5.6 Stress Concentrations near
6.7.11 Testing of generators......... 158 Hatches and Doors....................... 182
6.8 MACHINE SUPPORT FRAME. 158 7.5.7 Stability analysis ................. 182
6.9 ................. NACELLE ENCLOSURE 7.5.8 Flange connections ............. 184
.......................................................... 159 7.5.9 Corrosion protection........... 184
6.10 YAW SYSTEM ........................ 159 7.5.10 Tolerances and specifications
6.10.1 Determination of Design Loads ..................................................... 184
...................................................... 160 7.6 ACCESS AND WORKING
6.10.2..................................Yaw Drive ENVIRONMENT ............................. 185
...................................................... 162 REFERENCES ................................. 185
6.10.3................................... Yaw Ring
...................................................... 163
6.10.4................................. Yaw Brake 8. FOUNDATIONS.............................. 187
...................................................... 163 8.1 INTRODUCTION ...................... 187
6.10.5..............................Yaw Bearing 8.2 SOIL INVESTIGATIONS.......... 187
...................................................... 164 8.2.1 General ............................... 187
6.10.6............. Yaw Error and Control 8.2.2 Recommendations for gravity
...................................................... 167 base foundations .......................... 188
6.10.7 Cable twist......................... 167 8.2.3 Recommendations for pile
foundations................................... 189
0 – Contents v
8.3 GRAVITY-BASED C. FATIGUE CALCULATIONS ....... 229

FOUNDATIONS .............................. 189
C.1 METHODS FOR STRUCTURAL
8.3.1 General................................ 189
STEEL .............................................. 229
8.3.2 Bearing capacity formulas .. 190
C.2 FATIGUE IN WELDED
8.4 PILE-SUPPORTED
STRUCTURES................................. 231
FOUNDATIONS .............................. 193
C.3 CHARACTERISTIC S N CURVES
8.4.1 General................................ 193
FOR STRUCTURAL STEEL........... 232
8.4.2 Pile groups .......................... 194
C.4 CHARACTERISTIC S N CURVES
8.4.3 Axial pile resistance ............ 195
FOR FORGED OR ROLLED STEEL233
8.4.4 Laterally loaded piles.......... 197
C.5 S N CURVES FOR COMPOSITES
8.4.5 Soil resistance for embedded pile
.......................................................... 234
caps .............................................. 200
REFERENCES ................................. 234
8.5 FOUNDATION STIFFNESS ..... 200
8.6 SELECTED FOUNDATION
STRUCTURE CONCEPTS FOR D. FEM CALCULATIONS ................ 237
OFFSHORE APPLICATIONS......... 206
8.6.1 Introduction to concepts...... 206 D.1 GENERAL................................. 237
8.6.2 Monopile ............................. 206 D.1.1 The use of FEM analyses ... 237
D.1.2 What type of analysis.......... 237
8.6.3 Tripod.................................. 212
D.2 MODELLING............................ 238
REFERENCES ................................. 218
D.2.1 Model ................................. 238
D.2.2 Elements ............................. 239
9. ELECTRICAL INSTALLATIONS 221 D.2.3 Boundary conditions .......... 240
D.2.4 Loads.................................. 241
9.1 SAFETY SYSTEMS .................. 221
D.3 DOCUMENTATION................. 242
9.2 ELECTRICAL SAFETY ............ 221
D.3.1 Model ................................. 242
9.3 GRID CONNECTION................ 221
D.3.2 Results ................................ 243
9.3.1 Power factor correction ...... 221
9.3.2 Frequency converter ........... 222
9.3.3 Thyristor cut-in.................... 222 H. TABLES AND CONVERSIONS... 245
9.4 WIND TURBINE CONTROLLER
H.1 ENGLISH/METRIC CONVERSION
.......................................................... 222
.......................................................... 245
9.5 REMOTE CONTROL ................ 223
H.2 AIR DENSITY VS.
TEMPERATURE ............................. 245
10. MANUALS ..................................... 225 H.3 AIR DENSITY VS. HEIGHT.... 245
H.4 RAYLEIGH WIND
10.1 .............................USER MANUAL
DISTRIBUTION .............................. 245
.......................................................... 225
10.2 SERVICE AND MAINTENANCE
MANUAL......................................... 225
10.3 INSTALLATION MANUAL ... 225 INDEX .................................................. 247
REFERENCE.................................... 225
A. BOLT CONNECTIONS................. 227

A.1 BOLT QUALITIES ................... 227
A.2 S N CURVES........................... 227
vi 0 – Contents
1. Wind Turbine Concepts operating at constant rotor speed have been

dominating up to now, turbines with vari-
able rotor speed are becoming still more
1.1 Introduction common in an attempt to optimize the en-
Wind-powered ships, grain mills, water ergy capture, lower the loads, obtain better
pumps, and threshing machines all exem- power quality, and enable more advanced
plify that extraction of power from wind is power control aspects.
an ancient endeavour. With the evolution of
mechanical insight and technology, ma- 1.2 Conceptual aspects
chines to efficiently extract power from 1.2.1 General
wind have been developed, in particular
over the last few decades of the 20th cen- Some early wind turbine designs included
tury. ”Wind turbines” is now being used as a multiple-bladed concepts as the ones shown
generic term for machines with rotating in Figure 1-1.
blades that convert the kinetic energy of
wind into useful power.
Early wind turbine designs in the 20th cen-

tury were driven by three basic philosophies
for handling the loads: (1) withstanding the
loads, (2) shedding or avoiding the loads,
and (3) managing the loads mechanically,
electrically, or both. During this evolution,
many wind turbine designs have seen the
light of day. These include horizontal-axis
and vertical-axis turbines that spin about
horizontal and vertical axes, respectively,
and are equipped with one, two, three or
multiple blades.
Modern turbines evolved from the early

designs and can be classified as two- or
three-bladed turbines with horizontal axes
and upwind rotors. The choice between two
and three blades for wind turbines designed
today is merely a matter of trade-off be- Figure 1-1 Multiple-bladed wind turbines of
tween aerodynamic efficiency, complexity, various designs
cost, noise and aesthetics.
These turbines are all characterized by rotors
Other key turbine design considerations to- with high solidity, i.e., the exposed area of
day include wind climate, rotor type, gen- the blades is relatively large compared to the
erator type, load and noise minimisation, swept area of the rotor. A disadvantage of
and control approach. Current trends, driven such a high-solidity rotor is the excessive
by the operating regime and the market en- forces that it will attract during extreme
vironment, involve development of low- wind speeds such as in hurricanes. To limit
cost, megawatt-scale turbines and light- this undesirable effect of extreme winds and
weight turbine concepts. Whereas turbines increase efficiency, modern wind turbines
1 – Wind Turbine Concepts 1

are built with fewer, longer, and slender tical axis wind turbine as the one in Figure
blades, i.e., with a much smaller solidity. To 1-2 are that the generator and gearbox are
compensate for the slenderness of the placed on the ground and are easily accessi-
blades, modern turbines operate at high tip ble, and that no yaw mechanism is needed.
speeds. Among the disadvantages are an overall
slightly lower efficiency, and the facts that
1.2.2 Vertical axis turbines the turbine is not self-starting and needs
total dismantling just to replace the main
Vertical axis wind turbines (VAWTs) such
bearing.
as the one shown in Figure 1-2 with C-
shaped blades are among the turbine types
1.2.3 Horizontal axis turbines
that have seen the light of day during the
past century. Horizontal axis wind turbines (HAWTs)
such as the ones shown in Figure 1-3 form
the most common wind turbine type in use
today. In fact all grid-connected commercial
wind turbines today are designed with pro-
peller-type rotors mounted on a horizontal
axis on top of a vertical tower. In contrast to
the mode of operation of the vertical axis
turbines, the horizontal axis turbines need to
be aligned with the direction of the wind
such that the wind flows parallel to the axis
of rotation.
Figure 1-2 Eole C, a 4200 kW vertical axis

Darrieus wind turbine with 100 m rotor di-
ameter at Cap Chat, Québec, Canada. The
machine, which is the world's largest wind
turbine, is no longer operational, from
www.windpower.org (2000), © Danish Figure 1-3 Three-bladed upwind turbines
Wind Turbine Manufacturers Association being tested at Risø, August 1986
Classical water wheels let the water arrive For horizontal axis wind turbines, a distinc-
tangentially to the water wheel at a right tion is made between upwind and downwind
angle to the rotational axis of the wheel. rotors. Upwind rotors face the wind in front
Vertical axis wind turbines are designed to of the vertical tower and have the advantage
act correspondingly towards air. In princi- of very little or no wind shade effect from
ple, such a design could work with a hori- the presence of the tower. Upwind rotors
zontal axis as well, but it would hardly be need a yaw mechanism to keep the rotor axis
able to beat the efficiency of a propeller- aligned with the direction of the wind.
type turbine. The major advantages of a ver- Downwind rotors are placed on the lee side
2 1 – Wind Turbine Concepts

of the tower, where the wind shade of the turbine with upwind rotor, asynchronous
tower is significant. A disadvantage of this generator and active yaw system is usually
design is the fluctuation of the produced referred to as the Danish concept. This is a
power every time a rotor blade passes concept, which tends to be a standard
through the tower wake. This feature may against which other concepts are evaluated.
give rise to more fatigue loads in a down- Relative to the three-bladed concept, the
wind rotor than in an upwind rotor. Down- two- and one-bladed concepts have the ad-
wind rotors can theoretically be built with- vantage of possible savings in cost and
out a yaw mechanism, if the rotor and na- weight of the rotor. However, their use of
celle can be designed in such a way that the fewer rotor blades implies that a higher ro-
nacelle will follow the wind passively. tational speed or larger chord is needed to
However, when the rotor yaws passively in yield the same power output as a three-
the same direction for a long time, this may bladed turbine of similar size. The two- and
cause twist of the generator power cables. one-bladed concepts have usually so-called
For large wind turbines, the use of slip rings teetering hubs, which implies that they have
or mechanical collectors to circumvent this the rotor hinged to the main shaft. This de-
problem is very difficult. Upwind rotors sign allows the rotor to teeter in order to
need to be rather inflexible to keep the rotor eliminate some of the unbalanced loads.
blades clear of the tower. Downwind rotors One-bladed wind turbines are less wide-
can be made more flexible, which implies spread than two-bladed turbines. In addition
possible savings with respect to weight, and to a higher rotational speed, more noise and
which may contribute to reduce the loads on visual intrusion problems, one-bladed tur-
the tower. The vast majority of wind turbines need a counter-weight to balance the
bines in operation today have upwind rotors. rotor blade.
1.2.4 Number of rotor blades 1.2.5 Power control aspects

For horizontal axis wind turbines, a distinc- Wind turbines are designed to produce elec-
tion is also made between rotors according tricity cheapest possible. For this purpose,
to the number of rotor blades. Wind turbine wind turbines are in general designed to
concepts with one, two and three rotor yield a maximum power output at wind
blades have been developed, see Figure 1-4. speeds around 15 m/s. It would not pay to
design turbines to maximise their power
output at stronger winds, because such
stronger winds are usually too rare. How-
ever, in case of stronger winds, it is neces-
sary to waste part of the excess energy to
avoid damage of the wind turbine. The wind
turbine therefore needs some sort of power
control. The power control has two regimes,
viz.
Figure 1-4 Three-, two- and one-bladed • power optimisation for low wind speeds
wind turbine concepts, from • power limitation for high wind speeds
www.windpower.org (2000), © Danish which are separated by the wind speed at
Wind Turbine Manufacturers Association which the maximum power output is
achieved, typically about 15 m/sec.
The three-bladed concept is the most com-
mon concept for modern wind turbines. A

Basically, there are two approaches to power

control, viz. This is illustrated in Figure 1-5, which
• stall control shows iso-power curves for a wind turbine
• pitch control as a function of the blade angle and the
In addition comes active stall control, which mean wind speed. The ranges for pitch con-
can be viewed as a version of stall control. trol and active-stall control are separated at a
Stall-controlled wind turbines have their blade angle of 0° with the rotor plane. At
rotor blades bolted to the hub at a fixed an- low wind speeds, the optimal operation of
gle. The stall phenomenon is used to limit the wind turbine is achieved at a blade angle
the power output when the wind speed be- close to 0°. At higher wind speeds, the tur-
comes too high. This is achieved by design- bine will overproduce if the blade angle is
ing the geometry of the rotor blade in such a not adjusted. With pitch control, the blade is
way that flow separation is created on the pitched positively with its leading edge be-
downwind side of the blade when the wind ing turned towards the wind. With active-
speed exceeds some chosen critical value. stall control, the blade is pitched negatively
Stall control of wind turbines requires cor- with its trailing edge turned towards the
rect trimming of the rotor blades and correct wind. The power control, and in particular
setting of the blade angle relative to the rotor the power limitation at higher wind speeds,
plane. is indicated in idealised manner for both
control approaches by the dashed curves in
Pitch-controlled wind turbines have blades Figure 1-5. The dashed curves illustrate how
that can be pitched out of the wind. The the transition between operation with 0°
power output is monitored and whenever it blade angle at low wind speeds and power-
becomes too high, the blades will be pitched limiting operation along an isopower curve
slightly out of the wind to reduce the pro- at high wind speeds can be achieved for a
duced power. The blades will be pitched three-bladed rotor at a rated power of 400
back again once the wind speed drops. Pitch kW. For this example, the rated power is
control of wind turbines requires a design reached at a wind speed of about 12 m/sec.
that ensures the blades are pitched exactly
the amount required, in order to optimise the 1.3 Economical aspects
power output at all wind speeds. As of to-
day, pitch control of wind turbines is only The ideal wind turbine design is not dictated
used in conjunction with some kind of vari- by technology alone, but by a combination
able rotor speed. of technology and economy. Wind turbine
manufacturers wish to optimise their ma-
Active stall-controlled turbines resemble chines, so that they deliver electricity at the
pitch-controlled turbines by having pitchable lowest possible cost per unit of energy. In
blades. At low wind speeds, active stall tur- this context, it is not necessarily optimal to
bines will operate like pitch-controlled tur- maximise the annual energy production, if
bines. However, at high wind speeds, they that would require a very expensive wind
will pitch the blades in the opposite direc- turbine.
tion of what a pitch-controlled turbine would
do and force the blades into stall. This en- The choice of rotor size and generator size
ables a rather accurate control of the power depends much on the distribution of the
output, and makes it possible to run the tur- wind speed and the wind energy potential at
bine at the rated power at all high wind a prospective location. A large rotor fitted
speeds. with a small generator will produce electric-
ity during many hours of the year, but it will

only capture a small part of the wind energy at low wind speeds. Sometimes it will be
potential. A large generator will be very beneficial to fit a wind turbine with two or
efficient at high wind speeds, but inefficient more generators.
Figure 1-5 Iso-power curves for a wind turbine vs. blade angle and mean wind speed
1.4 Efficiency aspects in which Pwind=0.5ρAV03 is the power of the

wind without the presence of the wind tur-
The efficiency of a wind turbine is the frac-
tion of the wind energy, which is extracted bine. Here A=πR2 is the swept area of the
from the wind when it passes through the rotor, R is the rotor radius, ρ is the density of
rotor disc. It can be shown that the maxi- air, and V0 is the farfield wind speed in front
mum efficiency is 0.59 and is achieved of the wind turbine. The nominal efficiency
when the turbine reduces the wind speed to η is a function of v0 and typically reaches a
one-third the free wind speed. The extracted maximum when V0≈8-9 m/sec. Hence the
energy is converted to electrical energy by a electrical power produced by the wind tur-
generator in the wind turbine. The nominal bine can be expressed as
efficiency η is defined as the ratio between
the electrical power output and the maxi- Pturbine=η0.59⋅0.5⋅ρπR2V03
mum possible power input
The electrical power is thus proportional to
Pturbine the square of the rotor diameter and to the
η= cube of the wind speed, provided η is kept
0.59 ⋅ Pwind
unchanged. The electrical power typically

reaches the rated power of the turbine at a While wind energy is already economic in
wind speed V0≈14-15 m/sec. good onshore locations, it is currently about
to cross the economic frontier set by shore-
The efficiency of wind turbines generally lines: Offshore wind energy is becoming
improves with increasing tip speed ratio. competitive with other power-generating
However, the higher the speed of the blade technologies. Offshore wind energy is a
tip, the larger is the noise from the rotor. promising application of wind power, in
Today, tip speeds for turbines are limited to particular in countries with high population
about 70 m/sec due to concern about noise. density and thus with difficulties in finding
The tip speed increases with rotor speed and suitable sites on land. Construction costs are
radius. much higher at sea, but energy production is
also much higher. Currently, wind energy
1.5 Configurations and sizes from turbines erected on fixed foundations
in up to 15 m water depth is considered eco-
Wind turbines are erected as stand-alone nomically feasible. Figure 1-6 shows an
turbines, in clusters of multiple turbines, or example of an early offshore wind farm.
– on a larger scale – in park configurations.
Before the 1980’es, wind energy develop-
ment focused on the individual wind turbine.
By the late 1980’es, this perspective began
to change as attention shifted to collective
generation of electric power from an array
of many wind turbines located in the vicinity
of each other and commonly referred to as
wind parks or wind farms. In the early
1980’es the typical size of a wind turbine
was about 55 kW in terms of rated power,
whereas turbine sizes today have exceeded 2
MW. Table 1-1 gives examples of typical
combinations of rotor diameter and rated
power for a number of different tower
Figure 1-6 Vindeby Offshore Wind Farm,
heights.
from www.windpower.org (1997), © Bonus
Table 1-1 Typical wind turbine sizes
Tower Rotor Rated
height (m) diameter (m) power (kW)
22 21 55
31 30 225
35 35 450
35-40 41-44 500
44 43 600
50 48 750
50 54 1000
60 58 1500
64-80 72-76 2000

1.6 Future concepts The expected trend for the future is charac-
terised by more flexible wind turbine con-
cepts. One element in this is an expected
increase in the structural flexibility of wind
turbines, and Figure 1-7 shows one example
of how this can be conceived. Another ele-
ment is an expected increase in the flexibil-
ity of the drive train, e.g. in terms of gearless
designs and variable rotational speeds, and a
more extensive use of power electronics can
also be expected.
Also more flexible control systems can be

foreseen as still more computers and sensors
are incorporated to allow for adaptive op-
eration. In this context, it is possible that a
shift may take place from focusing on wind
Figure 1-7 A conception of the flexible turbine control to focusing on wind farm
wind turbine of tomorrow. During stand- control, and this will inevitably imply other
still, the blades deflect as a rush in the wind requirements to the individual wind turbine
in addition to the ones we see today.


2. Safety and Reliability fill its task. Non-redundant structural parts

of the protection system are therefore usu-
ally designed to high safety class.
2.1 Safety philosophy
A wind turbine should be designed, dimen- The prescribed safety or the choice of safety
sioned and manufactured in such a way that class can be different for different parts of
it, if correctly used and maintained over its the wind turbine. The rotor is usually de-
anticipated service life, can withstand the signed to at least normal safety class. Other
assumed loads with a prescribed safety and structural parts such as tower and foundation
will possess a sufficient durability and ro- are usually assigned safety classes according
bustness. Calculation or testing, or a combi- to the possible consequences of a failure.
nation thereof, can be used to demonstrate Since failure of the foundation will have
that the structural elements of a wind turbine consequences for the tower and the rotor,
have the prescribed safety. while failure of the rotor or the tower may
not necessarily have consequences for the
The prescribed safety for a structural ele- foundation, an attractive approach to choices
ment can be expressed in terms of a re- of safety classes for various structural parts
quirement to the failure probability and can could be to attempt a so-called fail-grace
be determined from risk acceptance criteria. sequence, i.e., a sequence of failures in
It depends on the type and consequence of which the foundation will be the last struc-
failure. The type of failure can be charac- tural part to fail. This is based on a purely
terized by the degree of ductility and the economical assessment of the failure conse-
amount of reserve capacity or structural re- quences with a view only to the economical
dundancy. The consequence of failure can consequences for the wind turbine structure
be characterized in terms of fatalities and and its foundation. However, the rotor is still
economical loss involved with the failure. usually designed to normal or high safety
The more severe the consequence and the class, which is not necessarily in accordance
less the reserve capacity, the smaller is the with the fail-grace philosophy. Require-
acceptable failure probability. The pre- ments to design a wind turbine rotor to nor-
scribed safety is standardized in terms of mal or high safety class come about from the
safety classes as decribed in more detail in hazards that the rotor poses on the sur-
Section 2.3. A distinction is made between roundings when the turbine runs away or the
low, normal and high safety class. The rotor fails. In such events, parts of the rotor
higher the safety class, the stronger is the may be shed in distances up to a kilometer
requirement to the safety, i.e., the smaller is or even more away from the turbine loca-
the acceptable failure probability. tion.
The wind turbine is equipped with a control Note that safety is usually the result of a
and protection system, which defines an trade-off with economy. In DS472, empha-
envelope of possible design situations that sis is placed on safety. However, for tur-
the wind turbine will experience. To keep bines to be installed offshore, it is inevitable
the turbine within this envelope, it is part of that economical aspects will become more
the safety philosophy that the protection predominant than they are onshore, and
system shall possess a sufficiently high reli- more emphasis will eventually be placed on
ability to render the joint probability negli- economy when it comes to safety issues and
gible that a failure occurs during an extreme determination of acceptable safety levels for
event and the protection system cannot ful- offshore turbines.
2 – Safety and Reliability 9

Limit state design is used to achieve the that some of the partial safety factors, which
prescribed safety. It is common to verify the are specified in standards, are not safety
safety of a wind turbine with respect to the factors in the true sense, but rather reduction
following limit states factors which account for degradation ef-
• ultimate limit state fects, scale effects, temperature effects etc.
• serviceability limit state and which happen to appear in the design
• accidental limit state expressions in exactly the same manner as
For this purpose, design loads are derived by true partial safety factors.
multiplying characteristic loads by one or
more partial safety factors, and design ca- With the structural safety as a major goal in
pacities are derived by dividing characteris- the design, it is important to make sure that
tic capacities by one or more other partial the characteristic values of load and material
safety factors. Verification of the structural quantities, which have been assumed for the
safety is achieved by making sure that the design, are achieved in practice. Non-
design load, or the combination of a set of destructive testing of completed structural
design loads, does not exceed the design parts plays a role in this context, and control
capacity. Note that for combination of loads of workmanship another. Material certifi-
it is common always to combine one ex- cates also come in handy in this context. In
treme load with one or more “normal” loads. general one may say that inspection is an
Combination of two or more extreme loads important part of the safety philosophy. It
is usually not done unless they have some will allow for verification of assumptions
correlation. made during the design and for taking reme-
dial actions if averse conditions are detected
Characteristic loads and characteristic ca- during the service life of the wind turbine.
pacities are important parameters in the de-
sign process. Characteristic loads for as- For details about structural safety and limit
sessment of the ultimate limit state are usu- state design, reference is made to Section
ally determined as load values with a 50- 2.3. For details about combinations of de-
year recurrence period, and they are there- sign situations and external conditions, as
fore often interpreted as the 98% quantile in well as definition of load cases, reference is
the distribution of the annual maximum made to Chapter 4.
load. This choice does not necessarily imply
that a design lifetime of exactly 50 years is
considered. It is more a matter of tradition 2.2 System safety and operational
and convenience. Nor does it imply that no reliability
failure for sure will take place until after 50 A wind turbine is to be equipped with con-
years have passed. For assessment of fa- trol and protection systems which are meant
tigue, a design life time is needed, and in to govern the safe operation of the wind
this context it is common to consider a 20- turbine and to protect the wind turbine from
year design life for wind turbines. Charac- ill conditions.
teristic capacities are usually chosen as low
quantiles in the associated capacity distribu- 2.2.1 Control System
tions. The partial safety factors that are ap-
plied in design account for the possible more Controls are used for the following functions
unfavourable realizations of the loads and • to enable automatic operation
capacities than those assumed by the choices • to keep the turbine aligned with the
of characteristic values. Note in this context wind
10 2 – Safety and Reliability

• to engage and disengage the generator turbine in the wind speed interval
• to govern the rotor speed [Vmin;Vmax], where Vmin denotes the start
• to protect the turbine from overspeed or wind speed for the turbine
damage caused by very strong winds • a maximum nominal power Pmax, which
• to sense malfunctions and warn opera- on average over 10 minutes may not be
tors of the need for maintenance or re- exceeded for a wind speed at hub height
pair of V10min,hub < Vmax
• a maximum operating frequency of ro-
The control system is meant to control the tation nr,max for the wind turbine
operation of the wind turbine by active or • a maximum transient frequency of rota-
passive means and keeps the operating pa- tion nmax for the wind turbine
rameters within their normal limits. Passive • a wind speed below which the wind
controls use their own sensing and are exer- turbine may be stopped.
cised by use of natural forces, e.g. centrifu- The wind turbine is kept within its normal
gal stalling or centrifugal feathering. Active operating range by means of the control
controls use electrical, mechanical, hydrau- system which activates and/or deactivates
lic or pneumatic means and require trans- the necessary controls, e.g.
ducers to sense the variables that will deter- • yaw (alignment to the wind)
mine the control action needed. Typical • blade angle regulation
variables and features to be monitored in • activation of the braking system
this respect include • power network connection
• rotor speed • power limitation
• wind speed • shutdown at loss of electrical network
• vibration or electrical load.
• external temperature
• generator temperature In addition, it must be possible to stop the
• voltage and frequency at mains connec- wind turbine, for example for the purpose of
tion inspection and repairs, or for emergencies.
• connection of the electrical load The monitoring of the control system and
• power output the control system functions must be
• cable twist adapted to the actual design of the wind tur-
• yaw error bine. Design of a wind turbine control sys-
• brake wear tem requires a background in servo theory,
i.e., theory for control of continuous sys-
The control system is meant to keep the tems.
wind turbine within its normal operating
range. As a minimum, the normal operating 2.2.2 Protection System
range should be characterized by the fol- The protection system is sometimes referred
lowing properties and requirements: to as the safety system. Mechanical, electri-
• a maximum 10-minute mean wind cal and aerodynamic protection systems are
speed at hub height, Vmax, i.e. the stop available. The protection system is to be
wind speed, below which the wind tur- activated when, as a result of control system
bine may be in operation failure or of the effects of some other failure
• a maximum long-term mean nominal event, the wind turbine is not kept within its
power Pnom, interpreted as the highest normal operation range. The protection sys-
power on the power curve of the wind tem shall then bring the wind turbine to a

safe condition and maintain the turbine in The protection system shall include one or
this safe condition. It is usually required that more systems (mechanical, electrical or
the protection system shall be capable of aerodynamic) capable of bringing the rotor
bringing the rotor to rest or to an idling state to rest or to an idling state from any operat-
from any operating condition, and that ing condition.
means shall be provided for bringing the
rotor to a complete stop from a hazardous An emergency stop button, which will over-
idling state in any wind speed less than the ride the automatic control system and result
10-minute mean wind speed with a 1-year in a machine shutdown, shall be provided at
return period. The activation levels for the every working place. In cases of conflict, the
protection system have to be set in such a protection function shall over-rule the con-
way that design limits are not exceeded. trol function.
Situations which call for activation of the In addition to what is stated above, the
protection system include, but are not neces- safety system is, as a minimum, to be sub-
sarily limited to ject to the following requirements:
• overspeed • The safety system must take precedence
• generator overload or fault over the control system.
• excessive vibration • The safety system must be fail-safe in
• failure to shut down following network the event of a failure of the power sup-
loss, disconnection from the network, or ply.
loss of electrical load • Structural components in mechanisms
• abnormal cable twist owing to nacelle of the safety system shall be designed to
rotation by yawing high safety class.
The protection system should therefore as a
• The safety system must be able to reg-
minimum cover monitoring of the following ister a fault and bring the wind turbine
• rotational speed or rotational frequency to a standstill or to controlled free-
• overload of a generator or other energy wheeling in all situations in which the
conversion system/load rotor frequency of rotation is less than
• extreme vibrations in the nacelle nmax. In this context, fault means
• safety-related functioning of the control changes in the operation of the wind
system. turbine, resulting in an inability to
Overspeed is by far the most critical error, maintain the safety laid down in the
so rotational speed monitors form a crucial standard (e.g. network failure).
element of the protection system. • The safety system must be tolerant
vis-a-vis a single fault in a sensor, in the
A protection system consists of electronic and electrical as well as the
• a registering unit hydraulic systems or in active mechani-
• an activating unit cal devices, i.e. an undetected fault in
• a braking unit the system must not prevent the system
It is usually required that at least two brak- from detecting a fault condition and car-
ing systems are included in the braking unit. rying out its function.
For more details about braking systems ref- • The reliability of the safety system must
erence is made to Section 2.3.3. be such as to make it possible to disre-
gard operational modes in which the
safety system fails, with the extreme

operating range being exceeded as a re- As an alternative to using blade tip brakes,
sult. braking by rotation of the entire blade can be
The reliability of the safety system may be adopted and is suitable for both stall- and
ensured by means of either (1) the entire pitch-controlled turbines, and spoilers can
safety system being of a fail-safe design, or also be used. A blade tip brake is shown in
(2) redundancy of the parts of the safety Figure 2-1.
system where it cannot be made fail-safe, or
(3) frequent inspections of the functioning of
the safety system, in which risk assessment
is used to determine the interval between
inspections. Figure 2-1 Tip brake, from
www.windpower.org (2000), © Danish
Control and protection systems are of par- Wind Turbine Manufacturers Association
ticularly great importance in areas where
weak grids are encountered. Weak grids can, The reliability of a brake system is of utmost
for example, be found in areas with a small importance to ensure that the system will
population, and the capacity of the grids in serve its purpose adequately. In this respect,
such areas can often be a limiting factor for it is important to be aware of possible de-
the exploitation of the wind resource. There pendencies between different brakes or dif-
are two problems in this context ferent brake components. For example, if all
• increase of the steady-state voltage level three blades are equipped with tip brakes,
of the grid above the limit when power some dependency between the three tip
consumption is low and wind power in- brakes can be expected, cf. the common-
put is high cause failures that can be foreseen for these
• voltage fluctuations above the flicker brakes. This will influence the overall reli-
limit may result from fluctuating wind ability against failure of the system of the
power input caused by fluctuating wind three tip brakes and needs to be taken into
and wind turbines cut-ins account.
The solution to this is a so-called power
control as part of the protection system. The Brakes or components of brake systems will
power control concept implies buffering the be subject to wear, so current monitoring
wind turbine power in periods where the and maintenance are required.
voltage limits may be violated and releasing
it when the voltage is lower. This is com- IEC61400-1 requires that the protection
bined with a smoothening of the power out- system shall include one or more systems
put, such that fluctuations are removed, in (i.e., mechanical, electrical, or aerodynamic
particular those that would exceed the brakes) capable of bringing the rotor to rest
flicker limit. or to an idling state from any operating con-
dition. At least one of these shall act on the
2.2.3 Brake system low-speed shaft or on the rotor of the wind
turbine. The idea behind this is to have a
A brake is a device capable of reducing the brake system such that a fault will not lead
rotor speed or stopping the rotation of the to a complete failure of the wind turbine.
rotor. Examples of brake systems are
• mechanical brake DS472 is more strict by requiring at least
• blade tip brake, i.e. aerodynamically two fail-safe brake systems. If the two sys-
operated brake unit tems are not independent, i.e. if they have

some parts in common, then the turbine shall ponents. If conducted at too detailed a level,
automatically be brought to a complete stop the analysis can be time-consuming and
or to controlled idling in the event of a fail- tedious, but will lead to great understanding
ure in the common parts. At least one brake of the system.
system is required to have an aerodynamic
brake unit. Table 2-1 Example of worksheet
COM- FAIL- FAIL- FAIL- FAIL- FRE- SE-
PO- URE URE URE URE QUEN- VERITY
2.2.4 Failure mode and effects analysis NENT MODE CAUSE EFFECT DETEC- CY RATING
TION RATING
A failure mode and effects analysis is a Valve Leak Deterio- Oil leak Visual Low Low
qualitative reliability technique for system- past
stem
rated
seal
by ROV
atical analysis of a mechanical or electrical Fails to Control Valve Flow Medium Low
close on system will not does not
system, such as a wind turbine safety sys- com- failure shut off shut off
tem. The analysis includes examination of mand flow
each individual component of the system for

determination of possible failure modes and The failure mode and effects analysis is
identification of their effects on the system. primarily a risk management tool. It is a
The analysis is based on a worksheet that stronghold of the failure modes and effects
systematically lists all components in the analysis that, if carried out correctly, it
system, and for each component includes identifies safety-critical components where a
• Component name. single failure would be critical for the sys-
• Function of component. tem. It is a weakness that it depends on the
• Possible failure modes. experience of the analyst and that it cannot
easily be applied to cover multiple failures.
• Causes of failure.
• How failures are detected.
2.2.5 Fault Tree Analysis
• Effects of failure on primary system
function. A fault tree is a logical representation of the
• Effects of failure on other components. many events and component failures that
• Necessary preventative/repair actions. may combine to cause one critical event
The failure mode and effects analysis can be such as a system failure. It uses “logic
supplemented by a criticality analysis, which gates” (mainly AND and OR gates) to show
is a procedure that rates the failure modes how “basic events” may combine to cause
according to their frequency or probability the critical “top event”.
of occurrence and according to their conse-
quences. The assigned ratings can be used to Application
rank the components with respect to their Fault tree analysis has several potential uses
criticality for the safety of the system. An for wind turbine safety systems:
example of a worksheet is given in Table 2- • In frequency analysis, it is commonly
1. used to quantify the probability of the
top event occurring, based on estimates
As a failure mode and effects analysis can of the failure rates of each component.
be conducted at various levels, it is impor- The top event may be an individual
tant to decide before commencing what level failure case, or a branch probability in a
will be adopted as otherwise some areas event tree.
may be examined in great detail while others • In risk presentation, it may also be used
will be examined at the system level only to show how the various risk contribu-
without examination of the individual com- tors combine to produce the overall risk.

• In hazard identification, it may be used Various standards for symbols are used −
qualitatively to identify combinations of typical ones are shown in Figure 2-2. An
basic events that are sufficient to cause example fault tree is shown in Figure 2-3.
the top event, known as “cut sets”.
Some types of events, for example a fire or
Construction of a Fault Tree power failure, may affect many components
Construction of a fault tree usually starts in the system at once. These are known as
with the top event, and works down towards “common-cause failures”, and may be repre-
the basic events. For each event, it considers sented by having the same basic event oc-
what conditions are necessary to produce the curring at each appropriate place in the fault
event, and represents these as events at the tree.
next level down. If any one of several events
may cause the higher event, they are joined Combination of Frequencies and Prob-
with an OR gate. If two or more events must abilities
occur in combination, they are joined with Both frequencies and probabilities can be
an AND gate. combined in a fault tree, providing the rules
in Table 2-2 are followed.
Table 2-2 Rules for Combining Frequencies

and Probabilities
Gate Inputs Outputs
OR Probability + Probability Probability
AND Frequency + Frequency Frequency
Frequency + Probability Not permitted
Probability × Probability Probability
Frequency × Frequency Not permitted
Frequency × Probability Frequency
Figure 2-2 Fault Tree Symbols
If quantification of the fault tree is the ob-

jective, downward development should stop
once all branches have been reduced to
events that can be quantified in terms of
probabilities or frequencies of occurrence. Figure 2-3 Example Fault Tree

Minimal Cut Set Analysis Strongholds and Weaknesses

Cut sets are combinations of events that are Preparation and execution of a fault tree
sufficient to cause the top event. Minimal analysis is advantageous in the sense that it
cut sets contain the minimum sets of events forces the wind turbine manufacturer to
necessary to cause the top event, after elimi- systematically examine the safety system of
nating any events that occur more than once his wind turbine. When probabilities can be
in the fault tree. For simple fault trees with assigned to events, the fault tree methodol-
each basic event only occurring once, the ogy can be applied to assess the overall reli-
minimal cut sets can be identified by inability of the wind turbine safety system. It
spection. For more complex trees, formal can also be used to determine which events
methods such as Boolean analysis are re- and which parts of the safety system are
quired. More commonly, computer pro- most critical.
grams are used to identify minimal cut sets.
Beware, however, that it is often hard to find
The minimal cut sets can be used in hazard out whether a fault tree analysis has been
identification to describe combinations of carried out properly. Fault tree analyses be-
events necessary to cause the top event. come complicated, time-consuming and
difficult to follow for large systems, and it
The minimal cut sets can also be used to becomes easy to overlook failure modes and
rank and screen hazards according to the common cause failures. It is a weakness that
number of events that must occur simultane- the diagrammatic format discourages ana-
ously. In principle, single event cut sets are lysts from stating explicitly the assumptions
of concern because only one failure can lead and conditional probabilities for each gate.
to the top event. In reality, larger cut sets This can be overcome by careful back-up
may have a higher frequency of occurrence. text documentation. It is a limitation that all
Nevertheless, the method can be useful for events are assumed to be independent. Fault
hazard screening, and for suggesting where tree analyses lose their clarity when applied
additional safeguards may be needed. to systems that do not fall into simple failed
or working states such as human error, ad-
Quantification of a Fault Tree verse weather etc.
Simple fault trees may be analysed using a
gate-by-gate approach to determine the top
event probability, provided that all events
are independent and there are no common-
cause failures. This gate-by-gate approach is
useful for QRA, because it quantifies all
intermediate events in the fault tree and pro-
vides a good insight into the main contribu-
tors to the top event and the effectiveness of 2.3 Structural safety
safeguards represented in the tree. However, 2.3.1 Limit states
because it cannot represent repeated events
or dependencies correctly, it is not usually During the lifetime of a structure, the struc-
used for formal reliability analysis. Reliabil- ture is subjected to loads or actions. The
ity analysis of more complex fault trees re- loads may cause a change of the condition or
quires minimal cut set analysis to remove state of the structure from an undamaged or
repeated events. intact state to a state of deterioration, dam-
age, or failure. Structural malfunction can

occur in a number of modes covering all on random realizations according to some

failure possibilities that can be imagined for probability distribution.
the structure. Although the transition from For the considered failure mode, the possi-
an intact state to a state of malfunction can ble realizations ofX can be separated in two
indeed be continuous, it is common to as- sets; namely the set for which the structural
sume that all states with respect to a par- component will be safe, and the set for
ticular mode of malfunction can be divided which it will fail. The surface between the
in two sets, namely the states that are failed safe set and the failure set in the space of
in this mode, and the states that are unfailed basic variables is denoted the limit state sur-
or safe in this mode. The boundary between face, and the reliability problem is conven-
the safe states and the failed states is re- iently described by a so-called limit state
ferred to as the set of limit states. The safety function g(X) which is defined such that
or reliability of a structure is concerned with
how likely it is that the structure will reach a ì > 0 for X in safe set
limit state and enter a state of failure. ï
g ( X)í= 0 for X on limit state surface
ï< 0 for X in failure set
There are several types of limit states. Two î
types are common, ultimate limit states and
serviceability limit states. Ultimate limit The limit state function is usually based on
states correspond to the limit of load- some mathematical engineering model for
carrying capacity of a structure or structural the considered limit state, based on the un-
component, e.g., plastic yield, brittle frac- derlying physics, and expressed in terms of
ture, fatigue fracture, instability, buckling, the governing load and resistance variables.
and overturning. Serviceability limit states
imply deformations in excess of tolerance The failure probability is the probability
without exceeding the load-carrying capac- content in the failure set
ity. Examples are cracks, wear, corrosion,
permanent deflections, and vibrations. Fa- PF = P[g ( X) ≤ 0] = ò f X ( x ) dx
tigue is sometimes treated as a separate type g ( X ) ≤0
of limit state. Other types of limit states are

possible, e.g., accidental limit states and where fX(x) is the joint probability density
progressive limit states. function for X and represents the uncertainty
and natural variability in the governing vari-
2.3.2 Failure probability and other ables X. The complement PS=1−PF is re-
measures of structural reliability ferred to as the reliability and is sometimes
also denoted the probability of survival. The
In structural design, the reliability of a
reliability may be expressed in terms of the
structural component is evaluated with re-
reliability index,
spect to one or more failure modes. One
such failure mode is assumed in the follow-
β=−Φ−1(PF),
ing. The structural component is described
by a set of stochastic basic variables
grouped into one vector X, including, e.g., where Φ is the standardized normal distri-
its strength, stiffness, geometry, and loading. bution function. The failure probability, the
Each of these variables is stochastic in the reliability, and the reliability index are all
suitable measures of structural safety.
sense that it − owing to natural variability
and possible other uncertainties − may take

For the simple example that X consists of 2.3.4 Code format, characteristic values,
two variables, the load L and the resistance and partial safety factors
R, and the limit state function can be speci-
A structural design code specifies design
fied as g(X)=R−L, the failure probability
rules that are to be fulfilled in the code
becomes a simple convolution integral
checks which are carried out during the de-
sign of a structure or structural component.
PF = P[R − L < 0] = òf R
(r ) f L (l )drdl The general layout of the design rules in a
R − L<0
design code is known as the code format.
∞ l
= òòf
−∞ − ∞
R
(r ) f L (l )drdl
The code format most frequently used in
∞ design codes today is a format which is ex-
= òF
−∞
R
(l ) f L (l )dl pressed in terms of design values of gov-
erning load and resistance variables. These
design values are defined as characteristic
where fR and fL are the probability density values of the load and resistance variables,
functions of R and L, respectively, and factored by partial safety factors. Such a
fR(r)=dFR(r)/dr, where FR is the cumulative form of a code format results from require-
distribution function of R. ments to an easy and yet economic design
and is known as a design value format. De-
2.3.3 Structural reliability methods sign according to a design value format is
sometimes referred to as load and resistance
The reliability index β can be solved in a factor design (LRFD).
structural reliability analysis by means of a
reliability method which can be any amongst In its simplest form, a code requirement can
several available methods, including nu- be expressed as a design rule in terms of an
merical integration, analytical first- and sec- inequality
ond-order reliability methods, and simula-
tion methods. Reference is made to Madsen LD<RD
et al. (1986). Some of these methods are
approximate methods, which will lead to in which LD is the design load effect and RD
approximate results for the reliability index. is the design resistance. The design load
Numerical integration is usually only feasi- effect is calculated as
ble when X consists of very few stochastic
variables such as in the example above.
LD=γfLC
Analytical first- and second-order solutions
to the failure probability are often suffi-
where LC is the characteristic load effect and
ciently accurate, and they are advantageous
γf is a load factor. Similarly, the design re-
to simulation results when the failure prob-
sistance is calculated as
abilities are small.
A useful byproduct of a structural reliability RC

R D=
analysis by these methods is the so-called γm
design point x*. This is a point on the limit
state surface and is the most likely realiza- where RC is the characteristic resistance and
tion of the stochastic variables X at failure. γm is a material factor.

Usually, a number of different load effects prescribed safety.

have to be combined into a resulting load
effect, and often the largest of several differ- Structural reliability analysis results, ob-
ent such load combinations is used for the tained as outlined above, play an important
load effect in the design rule. An example of role in codified practice and design. Their
such a load combination is the combination application to calibration of partial safety
of gravitational loads and wind loads. factors for use in structural design codes is
of particular interest. Structural reliability
The characteristic values are usually taken analysis directly capitalizes on the variabil-
as specific quantiles in the load and resis- ity and uncertainties in load and strength and
tance distributions, respectively. The char- therefore produces a reliability estimate,
acteristic values are often the mean value for which is a direct measure of the structural
dead load, the 98% quantile in the distribu- safety. Because of this, structural reliability
tion of the annual maxima for variable (en- analysis forms the rational basis for calibra-
vironmental) load, and the 2 or 5% quantile tion of the partial safety factors, which are
for strength. Reference is made to Figure 2-4 used in code checks during conventional
for an example. The code also specifies val- deterministic design. With structural reli-
ues of the partial safety factors γf and γm to ability method available, it is possible to
be used. determine sets of equivalent partial safety
factors which, when applied with design
fL, fR rules in structural design codes, will lead to
designs with a prescribed reliability.
L R As a first step, a target reliability index βT

must be selected. The choice for the target
L, R reliability index can be derived from a util-
ity-based feasibility assessment in a decision
LC RC analysis, or by requiring that the safety level
as resulting from the design by a structural
Figure 2-4 Probability density functions and reliability analysis shall be the same as that
characteristic values for load L and resis- resulting from current deterministic design
tance R practice. The latter approach is based on the
assumption that established design practice
The design equation is a special case of the is optimal with respect to safety and econ-
design rule obtained by turning the inequal- omy or, at least, leads to a safety level ac-
ity into an equality, i.e., RD=LD in the exam- ceptable by society.
ple.
When the target reliability index βT cannot
2.3.5 Code calibration be established by calibration against estab-
lished design practice, or otherwise, then its
The fulfillment of the design rules as re-
value may be taken from Table 2-3, de-
quired by a structural design code is meant
pending on the type and consequence of
to ensure that a particular prescribed struc-
failure. Note that the numbers given in Table
tural safety level is achieved. The purpose of
2-3 are given for a reference period of one
a code calibration is to determine the set of
year, i.e. they refer to annual probabilities of
partial safety factors to be used with the
failure and corresponding reliability indices.
chosen code format, such that structural de-
Reference is made to NKB (1978).
signs according to the code will meet this

result from a new reliability analysis of the

In the case of a prescribed reliability index, modified component. The geometrical
which is different from the one that results quantities, which can be adjusted to achieve
from an actually executed reliability analysis a specified reliability index, are sometimes
of a structural component, the geometrical denoted design parameters. It is most practi-
quantities of this component must be ad- cable to operate on just one such design pa-
justed. The adjustment is made in such a rameter when adjusting the design in order
way that the required reliability index will to reach the specified reliability index.
Table 2-3 Target annual failure probabilities PFT and corresponding reliability indices βT
Failure consequence
Failure type Less serious Serious Very serious
LOW SAFETY CLASS NORMAL SAFETY CLASS HIGH SAFETY CLASS
(small possibility for (possibilities for per- (large possibilities for
personal injuries and sonal injuries, fatali- personal injuries,
pollution, small eco- ties, pollution, and fatalities, significant
nomic consequences, significant economic pollution, and very
negligible risk to life) consequences) large economic con-
sequences)
Ductile failure
with reserve ca- PF=10−3 PF=10−4 PF=10−5
pacity (redundant βT=3.09 βT=3.72 βT=4.26
structure)
Ductile failure
with no reserve PF=10−4 PF=10−5 PF=10−6
capacity (signifi- βT=3.72 βT=4.26 βT=4.75
cant warning be-
fore occurrence
of failure in non-
redundant struc-
ture)
Brittle failure
(no warning be- PF=10−5 PF=10−6 PF=10−7
fore occurrence βT=4.26 βT=4.75 βT=5.20
of failure in non-
redundant struc-
ture)
A design case is formed as a specific combi- may result for different design cases. A sim-
nation of environmental loading regime, ple example of a calibration of partial safety
type of material, and type and shape of factors is given below for an axially loaded
structure. For a particular design case, which steel truss. A structural design code usually
can be analyzed by a structural reliability has a scope that covers an entire class of
method, a set of partial safety factors can design cases, formed by combinations
thus be determined that will lead to a design among multiple environmental loading re-
which exactly meets the prescribed reliabil- gimes, different structural materials, and
ity. Different sets of partial safety factors several types and shapes of structures. The

design code will usually specify one com- mean value E[σF]=400 MPa and standard
mon set of partial safety factors, which are deviation D[σF]=24 MPa. The cross-
to be applied regardless of which design sectional area of the tower is A. Failure oc-
case is being analysed. This practical simpli- curs when the axial force Q exceeds the ca-
fication implies that the prescribed reliabil- pacity σFA, so a natural format of the design
ity will usually not be met exactly, but only rule is σF,DA≥qD. The subscript D denotes
approximately, when designs are carried out design value. The limit state function is cor-
according to the code. Hence, the goal of a respondingly chosen as
reliability-based code calibration is to de-
termine the particular common set of partial g=σFA−Q
safety factors that reduces the scatter of the
reliabilities, achieved by designs according and the area A is used as the design pa-
to the code, to a minimum over the scope. rameter.
This can be accomplished by means of an
optimization technique, once a closeness Analysis by a first-order reliability method
measure for the achieved reliabilities has leads to determination of A=0.20 m2 in order
been defined, e.g., expressed in terms of a to meet a target reliability index β=4.265,
penalty function that penalizes deviations which corresponds to an annual failure
from the prescribed target reliability. For
probability PF=10−5. The characteristic value
principles and examples of such code opti-
of the axial force is taken as the 98% quan-
mization, reference is made to Hauge et al.
tile in the distribution of the annual maxi-
(1992), Ronold (1999), and Ronold and
mum force, qC=q98% =57.8 MN. The charac-
Christensen (2001).
teristic value of the yield strength is taken as
the 5% quantile in the strength distribution,
2.3.6 Example – Axially Loaded Steel
σF,C=σF,5% =361.8 MPa. One partial safety
Tower
factor, γ1, is introduced as a factor on the
The example given here deals with design of characteristic force, and another one, γ2, is
an axially loaded steel tower against failure introduced as a factor on the characteristic
in ultimate loading. The probabilistic mod- capacity. Substitution of the expressions for
elling required for representation of load and the design force and the design capacity in
capacity is presented. A structural reliability the design equation yields
analysis of the tower is carried out, and a
simple calibration of partial safety factors is γ2σF,CA−γ1qC=γ2⋅361.8⋅0.20−γ1⋅57.8=0
performed.
which gives a requirement to the ratio of the
The design of the axially loaded tower is partial safety factors γ1/γ2=1.251. There is
governed by the maximum axial force Q in a thus an infinite number of pairs (γ1,γ2) that
one-year reference period. The maximum will lead to the required reliability. This
axial force Q follows a Gumbel distribution implies an arbitrariness in selecting the par-
tial safety factor set (γ1,γ2) for the code. The
FQ(q)=exp(−exp(−a(q−b))) reliability analysis gives the design point
values q*=70.6 MN for the force and
in which a=0.428 and b=49.0 correspond to σF*=353.2 MPa for the strength. These are
a mean value E[Q]=50 MN and a standard the most likely values of the governing vari-
deviation D[Q]=3 MN. The yield strength of ables at failure. A robust choice for the par-
steel σF follows a normal distribution with tial safety factors (γ1,γ2) can be achieved by

designing to the design point values from

the reliability analysis. The partial safety x
factors are therefore selected as FX ( x) = 1 − exp(− )
x0
q* σ *
γ1 = = 1.226 and γ 2 = F = 0.979 in which x0=50 kNm is recognised as a
qC σ F ,C Weibull scale parameter. The total number
of bending moment ranges over the design
According to current design practice, a load life TL=20 years is ntot=0.9⋅109. The bending
factor is used as a factor on the characteristic moment ranges X give rise to bending stress
load to give the design load, and a material ranges S=X/W, where W denotes the section
factor is used as a divisor on the characteris- modulus of the blade root. Hence, the
tic resistance to give the design resistance. bending stress range distribution becomes
Hence, these factors become
sW
FS ( s ) = 1 − exp(− )
1 x0
γ f = γ 1 = 1.226 and γ m = = 1.021
γ2
For a given stress range S, the number of
respectively. bending stress cycles N to failure is gener-
ally expressed through an S−N curve, which
Note that the example is purely tutorial to on logarithmic form reads
explain a principle. In reality, the capacity
may be more uncertain than assumed here, ln N = ln K − m ln S + ε
e.g. owing to model uncertainty not ac-
counted for, and larger variabilities in the where the pair (lnK,m)=(114.7,8.0) de-
axial force may also be expected, depending scribes the expected behavior. The zero-
on the source and type of loading and the mean term ε represents the natural variabil-
amount of data available. The resulting par- ity about the expectation and follows a nor-
tial safety factors may then become larger mal distribution with a standard deviation
than the ones found here. σε=0.86. The cumulative damage is calcu-
lated as the Miner’s sum
2.3.7 Example – Fatigue of FRP Blade
Root in Bending ∆n ( s i )
D=å
The example given here deals with design of i N (s i )
an FRP blade root against fatigue failure
dFS
during its design life of 20 years. The prob- ∞ ntot ds
abilistic modelling required for representa- =ò ds
tion of load and resistance is presented. A 0 K exp(ε ) s − m
structural reliability analysis of the blade ntot x
root is carried out, and a simple safety factor = Γ(m + 1)( 0 ) m
K exp(ε ) W
calibration is performed.
The design of the blade root is governed by where Γ denotes the gamma function.
the long-term distribution of the bending
moment range X. In the long term, the According to Miner’s rule, fatigue failure
bending moment ranges are assumed to fol- occurs when the cumulative damage exceeds
low an exponential distribution a threshold of 1.0. A natural format of the

design rule is DD≤1, where DD is the design The design damage DD is then obtained as
damage. The limit state function is chosen as
ntot γ x
DD = Γ(m + 1)( m 0 ) m
g=1−D KC W
and the section modulus W is used as the

in which KC=Kexp(−2σε) reflects the chosen
design parameter.
characteristic S−N curve. Substitution of
Analysis by a first-order reliability method numbers into the design equation DD=1.0
leads to determination of W=0.00209 m3 in leads to the following requirement to the
order to meet a target reliability index material factor
β=3.29, which corresponds to a target failure
γm=1.149
probability PF=0.5⋅10−3 over the design life
of 20 years.
Discussion
Note that this example is purely tutorial to
For design against fatigue failure it has
explain a principle. In reality, the resistance
hardly any meaning to choose the 98%
may be more uncertain than assumed here,
quantile of the annual maximum load, or any
e.g. when there is a limited amount of mate-
other quantile for that matter, as the charac-
rial data available such that the estimated
teristic load value. A characteristic load dis-
values of K and m are uncertain. Moreover,
tribution is needed rather than a characteris-
a model uncertainty may be associated with
tic load value. The long-term stress range
the application of Miner’s rule. With such
distribution, FS(s), is chosen as the charac-
uncertainties properly accounted for, the
teristic stress range distribution, and a load
value of the resulting material factor γm will
factor γf=1.0 on all stress ranges according
become larger than the one found here.
to this distribution is prescribed.
This is based on the assumption that the Note also that with γf=1.0 prescribed, all
long-term stress range distribution is known. uncertainty and variability associated with a
The validity of this assumption in the con- fatigue problem as the present is accounted
text of wind turbines is discussed later. It is for by one single safety factor, γm. This fac-
also assumed that variability in the individ- tor is then applied as a safety factor on re-
ual damage contributions from the individ- sistance, regardless of whether some of the
ual stress ranges averages out over the many uncertainty is associated with load rather
contributing stress ranges in the long-term than with resistance. This is in accordance
distribution. This assumption would not with most standards. Note, however, that in
hold if the cumulative damage were domi- the new Danish standard DS409/DS410, a
nated by damage contributions from only partial safety factor γf on load is introduced
one or a very few large stress ranges, which which, under certain conditions, is to be
could be the case for very large m values, taken as a value greater than 1.0. This ap-
say m>10. plies to situations where the loads causing
fatigue damage are encumbered with uncer-
The characteristic S−N curve is taken as the tainty or ambiguity, such as if they are traf-
expected S−N curve minus two standard fic loads, or if the various quantiles of the
deviations. A partial safety factor γm is ap- long-term stress distribution over the design
plied as a divisor on all stress range values life are statistically uncertain.
of the characteristic S−N curve.

For design of wind turbines against fatigue, in a completely different manner than the
the loads causing fatigue damage are domi- variable natural loads (such as wind loads)
nated by wind-generated loads. Whereas the that the component will experience in reality
distribution of the 10-minute mean wind and which it should be designed for. For
speed on a location may be well known, the selection of loads to be used in full-scale
distribution of the turbulence intensity is tests for verification, it is therefore not rele-
usually not well determined, owing to local vant to apply the partial safety factors for
conditions and influence from the presence load that are prescribed for design in codes
of the turbine, nor is the transfer function to and standards. The loads to be used in tests
stress response in the wind turbine always need to be chosen after thorough considera-
clear. The distribution of wind-generated tion of the variability and uncertainty in the
loads in a wind turbine can therefore be ex- strength, given the degree of knowledge
pected to be known only with some uncer- about the strength or capacity available prior
tainty, and a load factor γf greater than 1.0 to the test.
would then be required. However, in prac-
tice, one would account for such uncertainty Note in this context that the effect of proof
or ambiguity in the load distribution by loading represents an increase in confidence
choosing a load distribution “on the safe about the structural strength or capacity,
side”, a conservative “envelope load spec- resulting from prior successful loading of
trum”, so to speak. In the presented exam- the component.
ple, this would imply choice of a conserva-
tively high value of the Weibull scale pa- 2.3.9 Inspection and inspection intervals
rameter x0, and this could then be used in
The design process is only one element in
conjunction with γf=1.0. assuring safe and reliable structures. In fab-
rication and service, other safety elements
2.3.8 Tests and calculations for verifica- can be introduced such as quality control,
tion alignment control, visual inspection, instru-
It is important to demonstrate that the mented monitoring, and proof loading. Each
structural strengths or capacities of the vari- of these items provide information about the
ous components that constitute the wind structure, additional to the information pres-
turbine structure are sufficient. This can be ent at the design stage, and may hence re-
done by calculations according to some the- duce the overall uncertainty associated with
ory or calculation method, or it can be done the structure. The probabilistic model used
by carrying out full-scale tests of the com- in design can then be updated and calibrated
ponent in question. Note that a full-scale test against reality by including the additional
of a structural component may give a more information.
accurate estimate of the component strength
than theoretical calculations, because model The additional information obtained during
uncertainty and bias owing to simplifica- fabrication and in service may be obtained
tions and limitations associated with the either directly as information about some of
applied calculation method will be reduced the governing variables themselves, e.g.
or removed. strength, or indirectly by observing substi-
tute variables, which are functions of the
When carrying out a full-scale test of a governing variables, e.g. cracks or deforma-
structural component such as a blade, it is tions.
important to acknowledge that the loads
used in the test are generated and controlled

It is of interest to update the failure prob- inspection. The time until the failure prob-
ability from its value in the design stage to a ability again will exceed the critical thresh-
value which reflects the additional informa- old and trigger a new inspection can be pre-
tion gained by inspection. Probability up- dicted and thus forms an inspection interval.
dating by inspection is based on the defini- This can be used to establish an inspection
tion of conditional probability. Let F denote plan.
the event of structural failure. In the design
process the probability of failure PF=P[F] is Note in this context that for some limit states
solved according to the procedures de- such as the fatigue limit state, it may actu-
scribed above. Let I denote an event such as ally be a prerequisite for maintaining the
the observation of a governing variable or required safety level over the design life that
the observation of a function of one or more inspections are carried out at specified inter-
of the governing variables, obtained by in- vals.
spection when the structure is in service.
The updated probability of failure is the
probability of failure conditioned on the 2.4 Mechanical safety
inspection event I,
There are several mechanical systems in a
P[F ∩ I ]
wind turbine:
P[F |I ] =
P[I ] • transmission: hub, shaft, gear, cou-
plings, brakes, bearings, generator
The probability in the numerator can be • mechanical control systems: pitch sys-
solved by a reliability analysis of a parallel tem, teeter mechanism, yaw system,
system, for which solutions are available. hydraulic system, pneumatic system
The probability in the denominator can be
solved by a reliability analysis of a structural The safety of mechanical components will
component as described above, once a suit- usually be determined by their structural
able limit state function has been defined, safety as described in a previous section.
and with due account for measurement un- However, as the components are part of me-
certainty and probability of detection which chanical systems, there are several addi-
are two contributing uncertainty sources of tional aspects to be considered in addition to
importance in this context. the structural safety when the safety of me-
chanical systems is to be evaluated.
For some limit states, e.g. crack growth and
fatigue failure, the failure probability PF The structural strength will in many cases
increases as a function of time. For such become limited by surface damages due to
limit states, prediction of the failure prob- wear such as fretting corrosion in connec-
ability as a function of time can be used to tions due to micro movements, or gray
predict the time when the failure probability staining of gear teeth due to poor lubrication
will exceed some critical threshold, e.g. a conditions. Hence, aspects like friction and
maximum acceptable failure probability. lubrication conditions as well as surface
This predicted point in time is a natural treatment are essential for the mechanical
choice for execution of an inspection. The safety.
failure probability can then be updated as
outlined above, depending on the findings Mechanical components are often made of
from the inspection and including improve- rather brittle high-strength material such as
ments from a possible repair following the case-hardened steel for gears and induction-

hardened roller bearing steel. Furthermore, quirement to structural safety shall be a

the strength of mechanical components is lower bound for the requirement to me-
often relying on extremely fine tolerances, chanical safety.
e.g., correction grinding of gear teeth or Machine components may be vulnerable to
mating surfaces in connections. lightning and may fail as a result. They
should therefore be bonded to local ground.
Nonmetallic materials such as rubber are DEFU (1999) may be consulted for re-
often used in mechanical components to quirements to cross sections of equipotential
achieve damping, or they are used as sealing bonding connections. Rotating and movable
in hydraulic components. Aging properties components such as roller bearings may
and temperature dependence are of impor- burn if struck by lightning, owing to their
tance for such materials. small contact areas. They can be protected
by a protection consisting of two parts; a
Mechanical components are in many cases diversion of the current via an alternative
subjected to internal forces such as pressure part with low impedance, and a reduction of
in hydraulic systems, pretension in bolts and current through the movable components,
shrinkage stress in shrink fit connections. In e.g. by means of electrical insulation.
other cases the strength requirements are
based on the rotational speed as for an inter-
nal gear shaft, which will experience one 2.5 Labour safety
complete bending cycle for each revolution.
2.5.1 General
The mechanical safety will thus depend on The safety of personnel working on a wind
several more or less well-defined parame- turbine or nearby should be considered when
ters, often in a rather complex combination. designing the wind turbine, as well as when
It may be difficult to define all governing issuing instructions and procedures for
phenomena adequately, and assumptions transportation and assembly, operation,
may not always hold in practice. A prob- maintenance and repair.
abilistic approach to assessment of the me-
chanical safety in terms of a failure prob- It should be possible to operate the control
ability may therefore not always be feasible. levers and buttons of the wind turbine with
ease and without danger. These levers and
Due to the complexity of mechanical sys- buttons should be placed and arranged in
tems, the required safety needs to rely on such a manner that unintentional or errone-
experience together with consequences of a ous operation that can lead to dangerous
failure. Typically, codes and standards used situations is prevented. The wind turbine
for mechanical design do not define re- should in general be designed in such a
quirements to safety factors. This is the case manner that dangerous situations do not
for the most commonly used standards, occur. If a wind turbine has more than one
which are the gear standards ISO 6336 and control panel or control unit, it should only
DIN 3990, and the bearing dynamic load be possible to operate it from one panel or
rating standard ISO 281. Hence, minimum unit at the time.
safety requirements need to be determined
based on the manufacturers’ experience and, The Danish “National Working Environ-
if available, on requirements from authori- ment Authority Regulation No. 561” of June
ties, certification bodies and wind turbine 24, 1994 as later amended, cf. the “National
developers. As a general guideline, the re- Working Environment Authority Regulation

No. 669” of August 7, 1995, regarding de- turbine. Operation of the wind turbine and
sign of technical facilities, commonly re- access to its local control system should not
ferred to as “Maskindirektivet”, should be require access to electrical circuits with a
complied with at all times. higher voltage than 50V.
2.5.2 Transportation, installation and Wherever screens and shields are used for
commisioning protection, it should be ensured that, during
normal operation, personnel cannot get in
Requirements concerning personnel safety
contact with any rotating, moving or con-
should be described in the instructions in the
ducting parts.
wind turbine manuals and in the procedures
for assembly, installation and commission-
The light in access routes should have an
ing.
intensity of at least 25 lux. This should also
apply when the main switch of the wind
2.5.3 Normal operation
turbine is turned off.
During normal operation of the wind tur-
bine, the safety of personnel inside and out- Working conditions
side the wind turbine should be considered. The wind turbine should be constructed in
The normal operation of the wind turbine such a way that replacing components sub-
should be possible without accessing the ject to service does not entail working pos-
nacelle. tures or movements which are hazardous to
health or otherwise dangerous. It should be
Operational procedures and operation of the possible to block the rotor and yaw system
wind turbine should be described in the user of the wind turbine in a safe and simple
manual, which is furnished to the turbine manner other than by using the ordinary
owner or to the person responsible for the braking and yaw system of the turbine. For
operation of the wind turbine. It should ap- pitch-controlled turbines, fixation of the
pear from any instructions how personnel pitch setting should be possible. Blocking of
safety has been accounted for. the rotor should be done by mechanical
fixation of the rotor and should be capable
2.5.4 Service, maintenance and repair of keeping the rotor fixed at all wind speeds
below the defined normal stop wind speed.
The manufacturer or supplier of the wind
Blocking of the yaw and pitch systems
turbine should provide instructions and pro- should keep the yaw and pitch systems, re-
cedures, which considers wind speeds and spectively, fixed at all wind speeds below
other external conditions in such a manner
the defined normal stop wind speed.
that service, maintenance and repair work on
the wind turbine can be performed safely.
Operation of the blocking mechanisms and
The wind turbine should be designed with a
the area of application for them should be
view to safe access to and safe replacement described the user manual for the wind tur-
of all components to be serviced. bine in order to avoid incorrect use.
Access It should be possible to illuminate working
It should be made clear by means of locks
areas with a light intensity of at least 50 lux.
and/or signs that ascending the wind turbine In addition, the ligthing must be designed
can be dangerous. It should be prevented such that glare, stroboscopic influences and
that unauthorized persons get access to the
control panel and the machinery of the wind

other disadvantageous lighting conditions nisms, internal electrical systems, mechani-

are avoided. cal systems, support structures, and electri-
cal interconnection equipment. The standard
It should be possible to initiate emergency applies to wind turbines with a swept area
shutdown close to the working areas in the equal to or greater than 40 m2. The standard
wind turbine. As a minimum, it should be should be used together with a number of
possible to initiate emergency shutdown at specified other IEC standards and together
the bottom of the tower, i.e. at the control with ISO2394.
panel, and in the nacelle.
NVN11400-0
2.6 Codes and standards
“Wind turbines – Part 0: Criteria for type
DS472 certification – technical criteria,” 1st edition,
Nederlands Normalisatie-institut, The Neth-
“Last og sikkerhed for vindmøllekonstruk-
erlands, 1999.
tioner” (“Load and Safety for Wind Turbine
Structures”, in Danish), DS472, 1st edition,
This is the Dutch standard for safety-based
Dansk Ingeniørforening, Copenhagen, Den-
design of wind turbine structures. It is valid
mark, 1992.
for wind turbines with a swept rotor area of
at least 40 m2. It is much based on the
This is the Danish standard for design of
IEC61400-1, however, since it is to be used
wind turbine structures. This standard, with
also for type certification of wind turbines in
its two annexes A and B, and with the stan-
the Netherlands, it covers requirements on
dards DS409 and DS410 and relevant
additional aspects such as type testing.
structural codes for materials, forms the
Danish safety basis for structural design of
DIBt RICHTLINIEN
horizontal axis wind turbines. The standard
is valid for the environmental conditions of “Richtlinie. Windkraftanlagen. Einwirkun-
Denmark and for turbines with rotor diame- gen und Standsicherheitsnachweise für
ters in excess of 5 m. Turm und Gründung” (in German). Guide-
lines for loads on wind turbine towers and
IEC61400-1 foundations. Deutsche Institut für Bautech-
nik (DIBt), Berlin, Germany, 1993.
“Wind turbine generator systems – Part 1:
Safety requirements”, 2nd edition, Interna-
GL REGULATIONS
tional Electrotechnical Commission, Ge-
neva, Switzerland, 1999. “Regulation for the Certification of Wind
Energy Conversion Systems,” Vol. IV –
This is an international standard that deals Non-Marine Technology, Part 1 – Wind
with safety philosophy, quality assurance Energy, in “Germanischer Lloyd Rules and
and engineering integrity, and specifies re- Regulations,” Hamburg, Germany, 1993.
quirements for the safety of wind turbine
generator systems. It covers design, installa-
tion, maintenance, and operation under REFERENCES
specified environmental conditions. Its pur-
Danske Elværkers Forenings Undersøgelser
pose is to provide an appropriate level of
(DEFU), Lightning protection of wind tur-
protection against damage from all hazards
bines, Recommendation 25, Edition 1, 1999.
during the design life. The standard is con-
cerned with control and protection mecha-

Hauge, L.H., R. Løseth, and R. Skjong, NKB Report No. 36, Copenhagen, Den-
“Optimal Code Calibration and Probabilistic mark, 1978.
Design”, Proceedings, 11th International
Conference on Offshore Mechanics and Ronold, K.O., “Reliability-Based Optimiza-
Arctic Engineering (OMAE), Calgary, Al- tion of Design Code for Tension Piles,”
berta, Canada, Vol. 2, pp. 191-199, 1992. Journal of Geotechnical and Geoenviron-
mental Engineering, ASCE, Vol. 125, No. 8,
Madsen, H.O., S. Krenk, and N.C. Lind, August 1999.
Methods of Structural Safety, Prentice-Hall
Inc., Englewood Cliffs, N.J., 1986. Ronold, K.O., and C.J. Christensen, “Opti-
mization of a Design Code for Wind-
Nordic Committee on Building Regulations Turbine Rotor Blades in Fatigue,” accepted
(NKB), “Recommendations for Loading and for publication in Engineering Structures,
Safety Regulations for Structural Design,” Elsevier, 2001.


3. External Conditions natural choice for H. The expression for A is

based on a logarithmic wind speed profile
above the ground
3.1 Wind Conditions
The wind climate that governs the loading of u* z
u( z ) = ln
a wind turbine are usually represented by the κ z0
10-minute mean wind speed U10 at the site
in conjunction with the standard deviation
where u* is the frictional velocity, κ=0.4 is
σU of the wind speed. In the short term, i.e. von Karman’s constant, and neutral atmos-
over a 10-minute period, stationary wind pheric conditions are assumed. The fric-
climate conditions are assumed to prevail,
tional velocity is defined as u*=(τ/ρ)1/2, in
i.e., U10 and σU are assumed to remain con- which τ is the surface shear stress and ρ is
stant during this short period of time. Only the air density.
when special conditions are present, such as
tornados and cyclones, a representation of A homogeneous terrain is defined as a ter-
the wind climate in terms of U10 and σU may rain with a constant z0 over the terrain.
be insufficient. Typical values for z0 are given in Table 3-1
for various types of terrain. For offshore
3.1.1 10-minute mean wind speed locations, where the terrain consists of the
The 10-minute mean wind speed will vary sea surface, the roughness parameter is not
from one 10-minute period to the next. This constant, but depends on
variability is a natural variability and can be • wind speed
represented in terms of a probability distri- • upstream distance to land
bution function. In the long run, the distri- • water depth
bution of the 10-minute mean wind speed • wave field
can be taken as a Weibull distribution
Table 3-1 Wind speed parameters for
u various terrain types
FU (u ) = 1 − exp(−( ) k )
10
A Terrain type Roughness Ex-
parameter po-
in which the shape parameter k and the scale z0 (m) nent
parameter A are site- and height-dependent α
coefficients. The scale parameter A at height Plane ice 0.00001
z can be calculated as follows Open sea without 0.0001
waves
z Open sea with waves 0.0001- 0.12
ln
z0 0.003
A = AH Coastal areas with 0.001
H
ln onshore wind
z0
Open country with- 0.01
out significant build-
in which z0 is a terrain roughness parameter, ings and vegetation
AH is the scale parameter at a reference Cultivated land with 0.05 0.16
height H. A common choice for the refer- scattered buildings
ence height is H=10 m. However, in the Forests and suburbs 0.3 0.30
context of wind turbines, the hub height is a City centers 1-10 0.40
3 – External Conditions 31
A widely used expression for the roughness where the exponent α depends on the terrain
parameter of the open, deep sea far from roughness. Note that if the logarithmic and
land is given by Charnock’s formula exponential expressions for A given above
are combined, then a height-dependent ex-
pression for the exponent α results
2
u*
z 0 = Ac
g
æ z ö
ç ln ÷
in which g is the acceleration of gravity, and
lnçç z0 ÷
H ÷
τ0 çç ln ÷÷
u* = è z0 ø
ρ α=
æ z ö
lnç ÷
is the frictional velocity expressed as a èHø
function of the shear stress τ0 at the sea sur-
face and the density ρ of the air. Ac=0.011 is Note also that the limiting value α=1/ln(z/z0)
recommended for open sea. As an approxi- as z approaches the reference height H has
mation, Charnock’s formula can also be an interpretation as a turbulence intensity,
applied to near-coastal locations provided cfr. the definitions given in Sections 3.1.2
Ac=0.034 is used. Expressions for Ac, which and 3.1.3. As an alternative to the quoted
include the dependency on the wave velocity expression for α, values for α tabulated in
and the available water fetch, are available Table 3-1 may be used.
in the literature, see Astrup et al. (1999).
Based on a logarithmic wind speed profile, The basic wind speed, vB, is the 50-year re-
Charnock’s formula leads to the following turn value of the 10-minute mean wind
expression for the roughness parameter for a speed at 10 m height above land with terrain
water surface roughness z0=0.05. The 10-minute mean
wind speed with 50-year return period at
2 other heights and other terrain roughnesses
A æ κU 10 ö
z0 = c ç ÷ can be found as
g ç ln( z z ) ÷
è 0 ø
z
from which z0 can be determined implicitly, v10 min, 50 yr = v B k t ln
z0
and from which the dependency on the wind
speed in terms of U10 is evident. This im-
plies that for offshore locations, determina- in which kt=0.19(z0/0.05)0.078.
tion of z0 and the distribution of U10 involves
3.1.2 Standard deviation of wind speed
an iterative procedure. κ=0.4 is von Kar-
man’s constant. For given value of U10, the standard devia-
tion σU of the wind speed exhibits a natural
For engineering calculations it may some- variability from one 10-minute period to
times prove useful to apply the following another. This variability of the wind speed is
empirical approximation for the scale pa- known as the turbulence, and σU is therefore
rameter A often referred to as the standard deviation of
the turbulence components. Measurements
α
æ z ö from several locations show that σU condi-
A = A10 ç ÷
èHø
32 3 – External Conditions
tioned on U10 can often be well represented These quantities will, in addition to their
by a lognormal distribution dependency on U10, also depend on local
conditions, first of all the terrain roughness
ln σ − b0 z0, which is also known as the roughness
Fσ |
U 10
(σ ) = Φ( ) length. When different terrain roughnesses
U
b1
prevail in different directions, i.e. the terrain
is not homogeneous, E[σU] and D[σU] may
in which Φ() denotes the standard Gaussian vary with the direction. This will be the case
cumulative distribution function. The coeffi- for example if a house is present nearby.
cients b0 and b1 are site-dependent coeffi- Houses and other “disturbing” elements will
cients dependent on U10. See Ronold and in general lead to more turbulence, i.e.,
Larsen (1999) for an example.
larger values of E[σU] and D[σU], than will
be found in smoother terrain. Figures 3-1
The coefficient b0 can be interpreted as the
and 3-2 give examples of the variation of
mean value of lnσU, and b1 as the standard
E[σU] and D[σU] with U10 for an onshore
deviation of lnσU. The following relation- and an offshore location, respectively. The
ships can be used to calculate the mean difference between the two figures mainly
value E[σU] and the standard deviation consists in a different shape of the mean
D[σU] of σU from the values of b0 and b1, curve. This reflects the effect of the in-
creasing roughness length for increasing U10
E [σ U ] = exp(b0 +
1 2 on the offshore location.
b1 )
2
D[σ U ] = E [σ U ] exp(b1 ) − 1
2 2,5
(m/sec)
mean value
2
st. dev.
3 1,5
E [σ U ]
mean value
(m/sec)
2,5 st. dev. 1

D [σ U ]
2
0,5
E [σ U ]
1,5
0
1 0 5 10 15 20 25
D [σ U ]
U 10 (m/sec)
0,5
0
0 5 10 15 20 Figure 3-2 Mean value and standard devia-
U 10 (m/sec) tion of σU as functions of U10 – offshore
location
Figure 3-1 Mean value and standard devia- In some cases, a lognormal distribution for
tion of σU as functions of U10 – onshore lo- σU conditioned on U10 will underestimate
cation the higher values of σU. A Frechet distribu-
tion may form an attractive distribution

E [σ U ] = U 10 Ax κ
1
model for σU in such cases, hence z
ln
z0
σ0 k
Fσ U |U10 (σ ) = exp(−( ) )
σ for homogeneous terrain, in which κ=0.4 is
von Karman’s constant,z is the height above
The distribution parameter k can be solved terrain, z0 is the terrain roughness, which is
implicitly from also known as the roughness length, and Ax
is a constant which depends on z0. Meas-
2 urements from a number of locations with
Γ(1 − )
D[σ U ] 2 k −1 uniform and flat terrain indicate an average
( ) =
E [σ U ] 1 value of Ax equal to 2.4, see Panofsky and
Γ 2 (1 − ) Dutton (1984). Dyrbye and Hansen (1997)
k
suggest Ax=2.5 for z0=0.05 m and Ax=1.8 for
and the distribution parameter σ0 then re- z0=0.3 m. A conservative fixed choice for σU
sults as is desirable for design purposes, i.e., a char-
acteristic value, and DS472 suggests
E [σ U ]
σ0 = 1
1 σ U ,c = U 10
Γ(1 − ) z
k ln
z0
where Γ denotes the gamma function.
Note that this value, although higher than
Caution must be exercised when fitting a the mean value of σU, may not always be
distribution model to data. Normally, the sufficiently conservative for design pur-
lognormal distribution provides a good fit to poses.
data, but use of a normal distribution, a
Weibull distribution or a Frechet distribution The IEC61400-1 standard requires use of a
is also seen. The choice of the distribution characteristic standard deviation for the
model may depend on the application, i.e., wind speed
whether a good fit to data is required to the
U 10,15 + aU 10
entire distribution or only in the body or the σ U ,c = I 15
upper tail of the distribution. It is important a +1
to identify and remove data, which belong to
10-minute series for which the stationarity in which U10,15=15 m/s is a reference wind
assumption for U10 is not fulfilled. If this is speed, IT,15 is the characteristic value of the
not done, such data may confuse the deter- turbulence intensity at 15 m/s, and a is a
mination of an appropriate distribution slope parameter. IT,15=0.18 and a=2 is to be
model for σU conditioned on U10. used in the category for higher turbulence
characteristics, while IT,15=0.16 and a=3 is to
Based on boundary-layer theory, the fol- be used in the category for lower turbulence
lowing expression for the mean value of the characteristics. The expression for the char-
standard deviation σU, conditioned on U10, acteristic value σU,c is based on a definition
can be derived of the characteristic value as the mean value
of σU plus one standard deviation of σU.
3.1.3 Turbulence intensity 3.1.4 Lateral and vertical turbulence

The turbulence intensity IT is defined as the The 10-minute mean wind speed, the stan-
ratio between the standard deviation σU of dard deviation of the wind speed, and the
the wind speed on the one hand, and the 10- turbulence intensity presented above all re-
minute mean wind speed U10 on the other, fer to the wind speed in the constant direc-
i.e., IT=σU/U10. tion of the mean wind during the considered
10-minute period of stationary conditions.
Note that the presence of a wind turbine will During this period, in addition to the turbu-
influence the wind flow locally, and the tur- lence in the direction of the mean wind,
bulence in the wake behind the turbine will there will be turbulence also laterally and
be different from that in front of the turbine. vertically. The mean lateral wind speed will
This phenomenon of a wind turbine influ- be zero, while the lateral standard deviation
enced turbulence is known as a wake effect. of the wind speed can be taken as
Typically, the presence of the wind turbine σUy=0.75σU according to Dyrbye and Han-
will lead to increased turbulence intensity in sen (1997) and σUy=0.80σU according to
the wake. Wake effects need to be consid- Panofsky and Dutton (1984). The mean ver-
ered for wind turbines installed behind other tical wind speed will be zero, while the ver-
turbines. This is of particular interest wher- tical standard deviation of the wind speed
ever wind farms with many turbines in sev- can be taken as σUz=0. 5σU. These values all
eral rows are to be installed. refer to homogeneous terrain. For complex
terrain, the wind speed field will be much
The Frandsen method can be used to take more isotropic, and values for σUy and σUz
wake effects into account. By this method, very near the value of σU can be expected.
the free-flow turbulence intensity IT is modi- Beware that calculations for changes in the
fied by the wake turbulence intensity IT,wake direction of the wind in complex terrain may
to give the following total turbulence inten- come out very wrongly if values for σUy and
sity σUz valid for homogeneous terrain are ap-
plied.
I T , total = I T + I T , wake
2 2
Very often, the wind climate at a location

cannot be documented by site-specific
to be used for design. The wake turbulence
measurements. For such situations, the dis-
intensity can be calculated as
tribution of U10 can still, usually, be repre-
sented well, for example on the basis of
k n CT wind speed measurements from some loca-
I T , wake =
s2 tion nearby. However, the distribution of σU
will usually be harder to obtain, because it
in which CT=4a(1−a) is the upwind thrust will be very dependent on the particular lo-
coefficient with the axial interference factor cal roughness conditions, and it can thus not
a defined in Chapter 4, s is the relative necessarily be inferred from known wind
spacing between successive wind turbines in speed conditions at locations nearby. At a
a row (measured in terms of number of rotor location where wind speed measurements
diameters), and kn is a wind farm configura- are not available, the determination of the
tion parameter. kn=0.4 is suggested for a distribution of the standard deviation σU of
situation with severe wind farm effects. the wind speed is therefore often encum-
bered with ambiguity. It is common practice
to account for this ambiguity by using con- Lu=100Czm

servatively high values for σU for design
purposes, viz. the characteristic values for where z is the height and C and m depend on
σU given in DS472 and IEC61400-1 and the roughness length z0 as given in Figure 3-
referenced above. 3. This spectrum is used in Eurocode 1.
3.1.5 Stochastic turbulence models IEC61400-1 requires that towards the high
frequency end of the inertial subrange the
Wind in one direction is considered, i.e. in power spectral density used for design shall
the direction of the 10-minute mean wind approach the form:
speed. The wind speed process U(t) within a
−2 / 3
10-minute period of constant U10 and σU is æ λ ö
S U ( f ) = 0.05 ⋅ σ U , c ç ÷ −5 / 3
2
considered and can be assumed to be sta- çU ÷ f
è 10 ø
tionary. The spectral density for the wind
speed process expresses how the energy of
the wind turbulence is distributed between in which the turbulence scale parameter λ
various frequencies. Several models for the depends on the height z above the terrain
spectral density exist. A commonly used
model for the spectral density is the Harris ì 0 .7 z for z < 30 m
λ=í
spectrum î 21 m for z > 30 m
L
3.66
U
SU ( f ) = σ U
2 10
3 2πfL 2 5 / 6 10
(1 + ( ) )
2 U 10
1
m
C, m
in which f denotes the frequency, and L is a C
characteristic length, which relates to the 0,1
integral length scale Lu by L=1.09Lu. A cali-

bration to full scale data indicates values for 0,01
0,001 0,01 0,1 1 10
L in the range 66-440 m with L≅200 m used
z 0 (m)
to match the high frequency portion of the
spectrum. Based on experience, the Harris
spectrum is not recommended for use in the Figure 3-3 Coefficients C and m for the
low frequency range, i.e. for f<0.01 Hz. integral length scale of the Kaimal spectrum
Another frequently used model for the For design it is common to relate calcula-
power spectral density is the Kaimal spections to wind conditions at the hub, i.e., U10
trum, and σU are to refer to the wind speed at the
L hub height. Note that the turbulence scale
6.8 u parameter λ relates to the integral length
U 10
SU ( f ) = σ U scale L through L=4.76λ. This gives the
2
fL
(1 + 10.2 u ) 5 / 3 following expression for the Kaimal spec-
U 10 trum, which is well known from the
IEC61400-1,
in which the integral length scale is given by
Lk Spectral moments are useful for representa-

4 tion of the wind speed process U(t). The jth
U 10
SU ( f ) = σ U
2
spectral moment is defined by
fL
(1 + 6 k ) 5 / 3
U 10 ∞
m j = ò ω j SU (ω )dω
with Lk=8.1λ.
0
In the short term, such as within a 10-minute

For calculation of spectral densities for the
period, the wind speed process U(t) can usu-
lateral and vertical wind speeds, the above
ally be represented as a Gaussian process,
formulas can be used with σUy and σUz, re- conditioned on a particular 10-minute mean
spectively, substituted for σU, and with wind speed U10 and a given standard devia-
λy=0.3λ and λz=0.1λ, respectively substi- tion σU. The arbitrary wind speed U at a
tuted for λ. considered point in time will then follow a
normal distribution with mean value U10 and
Note that there is some arbitrariness in the standard deviation σU. This is usually the
models for power spectral density. Each case for the turbulence in homogeneous ter-
model implies idealizations and simplifica- rain. However, for the turbulence in com-
tions and are usually calibrated to provide a plex terrain it is not uncommon to see a
good fit to data within some limited fre-
skewness of −0.1, which implies that the
quency range. In particular at low frequen- Gaussian assumption is not quite fulfilled.
cies, the models show significant differ-
ences. In Figure 3-4 three models for power
Note that although the short-term wind
spectral densities are plotted on dimension-
speed process will be Gaussian for homoge-
less scales and compared. neous terrain, it will usually not be a nar-
row-banded Gaussian process. This comes
about as a result of the spectral density. This
Harris
10 is of importance for prediction of extreme
Kaimal values of wind speed, and such extreme val-
(Eurocode)
ues and their probability distributions can be
IEC limit
1 expressed in terms of the spectral moments.
Reference is made to textbooks on stochastic
process theory for details.
2
U
0,1
fSu/
At any point in time there will be a variabil-

ity in the wind speed from one point to an-
0,01
other. The closer together the two points, the
higher is the correlation between their re-
0,001 spective wind speeds. The wind speed will
0,01 1 100 form a random field in space. A commonly
used model for the autocorrelation function
fLu/U10 for the wind speed field can be derived from
the exponential Davenport coherence spec-
trum,
Figure 3-4 Comparison between the Harris,
r
Kaimal and IEC power spectral densities Coh(r , f ) = exp(−cf )
u
where r is the distance between the two

1 0000
points, u is the average wind speed over the
distance r, f is a frequency, and c is non- S trongly stable
dimensional decay constant, which is re- S table
ferred to as the coherence decrement, and N ear neu tral
N eutral
which reflects the correlation length of the 1000 U nsta ble
wind speed field. The autocorrelation func- S trongly un stable
tion can be found as
height above ground (m)

∞
1
ρ (r ) =
σU
2 ò
0
Coh(r , f ) S U ( f )df 100
in which SU(f) is the power spectral density

of the wind speed. The coherence model can
10
be refined to account for different correla-
tion lengths horizontally and vertically. Note
that it is a shortcoming of the Davenport
model that it is not differentiable for r=0.
Note also that, owing to separation, the lim- 1
iting value ρ(0) will often take on a value 0 5
wind speed (m/s)

10 15
somewhat less than 1.0, whereas the Daven-

port model always leads to ρ(0)=1.0. Figure 3-5 Wind shear profiles for various
stability conditions at a location with rough-
Note that the integral length scale Lu, refer- ness z0=0.02 m and geostrophic wind speed
enced above as a parameter in the models G=12 m/s
for the power spectral density, is defined as
Wind shear profiles can be derived from the
∞ logarithmic model presented in Section
Lu = ò ρ (r )dr 3.1.1, modified by a stability correction. The
0
stability-corrected logarithmic wind shear
profile reads
3.1.6 Wind Shear
Wind shear is the variation with height of u* z
u( z ) = (ln − ψ )
the wind speed. Its effects are not considered κ z0
important for small wind turbines, on the
order of 10 m diameter. Wind shear may be
in which ψ is a stability-dependent function,
important for large and/or flexible rotors. A which is positive for unstable conditions,
number of failures have been attributed to negative for stable conditions, and zero for
wind-shear-induced blade loads. neutral conditions. Unstable conditions typi-
cally prevail when the surface is heated and
The wind profile depends much on the at- the vertical mixing is increasing. Stable
mospheric stability conditions, see Figure 3- conditions prevail when the surface is
5 for an example. Even within the course of cooled, such as during the night, and vertical
24 hours, the wind profile will change be- mixing is suppressed. Figure 3-6 shows ex-
tween day and night, dawn and dusk. amples of stability-corrected logarithmic
wind shear profiles for various conditions at z

a particular location. L MO = in unstable air
R
60 1 − 5R
L MO = z in stable air
neut ral R
50 st able
unst able
40
Table 3-2 Monin-Obukhov length
Atmospheric conditions LMO(m)
height (m)
30 Strongly convective days −10

Windy days with some solar −100
20 heating
Windy days with little sunshine −150
10
No vertical turbulence 0
0 Purely mechanical turbulence ∞
6 7 8 9 10 11 12 13 Nights where temperature strati- >0
fication slightly dampens me-
w ind speed (m /s) chanical turbulence generation
Nights where temperature strati- >>0
Figure 3-6 Wind profiles for neutral, stable fication severely suppresses me-
and unstable conditions chanical turbulence generation
The stability function ψ depends on the non- In lieu of data, the Richardson number can
dimensional stability measure ζ=z/LMO, be computed from averaged conditions as
where z is the height and LMO is the Monin- follows
Obukhov length. The stability function can
be calculated from the expressions g
(γ d − γ )
0.07
ψ=−4.8ζ for ζ≥0 R= T 2 (1 + )
æ ∂u ö æ ∂v ö
2
B
ç ÷ +ç ÷
ψ=2ln(1+x)+ln(1+x2)−2tan−1(x) for ζ<0 è ∂z ø è ∂z ø
in which x=(1−19.3ζ)1/4. in which g is the acceleration of gravity, T is

the temperature, γ=−∂T/∂z is the lapse rate,
The Monin-Obukhov length LMO depends on and γd≈9.8°C/km is the dry adiabatic lapse
the heat flux and on the frictional velocity rate. Further, ∂u / ∂z and ∂v / ∂z are the
u*. Its value reflects the relative influence of vertical gradients of the two horizontal aver-
mechanical and thermal forcing on the tur- age wind speed components u and v ; andz
bulence. Typical values for the Monin- denotes the vertical height. Finally, the Bo-
Obukhov length LMO are given in Table 3-2. wen ratio B of sensible to latent heat flux at
the surface can near the ground be approxi-
If data for the Richardson number R are mated by
available, the following empirical relation-
ships can be used to obtain the Monin- c p (T2 − T1 )
Obukhov length, B≈
L MO (q 2 − q1 )
in which cp is the specific heat, LMO is the The most important transient wind condi-
Monin-Obukhov length, T1 and T2 are the tions to consider are listed below
average temperatures at two levels denoted 1 • extreme of wind speed gradient, i.e.,
and 2, resepctively, and q1 and q 2 are the extreme of rise time of gust
• strong wind shear
average specific humidities at the same two
• simultaneous change in wind direction
levels. The specific humidity q is in this
and wind speed
context calculated as the fraction of moisture
• extreme changes in wind direction
by mass. Reference is made to Panofsky and
Dutton (1984).
Topographic features such as hills, ridges

and escarpments affect the wind speed.
Certain layers of the flow will accelerate
near such features, and the wind profiles will
become altered. Theories exist for calcula-
tion of such changed wind profiles, see Jen- Figure 3-8 Example of extreme rise time for
sen (1999). An example of effects of a ridge gust, from www.winddata.com.
is given in Figure 3-7.
100
downwind
upwind
ridge crest
Figure 3-9 Example of strong wind shear.
height (m)
Notice that the wind speed is inversely pro-

10
portional to height, i.e., so-called negative
wind shear. From www.winddata.com.
1
0 5 10 15 20
wind speed (m/s)
Figure 3-7 Wind profiles observed upwind,

at the crest, and at the foot downwind of a
two-dimensional ridge. Figure 3-10 Simultaneously change in wind
direction and speed.
3.1.7 Transient Wind Conditions From www.winddata.com.
It is important to be aware of transient wind
conditions which occur when the wind speed These are all wind events, which by nature
or the direction of the wind change. These fall outside of what can normally be repre-
are rare events, so there is usually not much sented by stationary wind conditions. Note
data available. that the simultaneous change in wind direc-
tion and wind speed occurs when fronts pass
and is usually not covered by commonly κu

available analysis tools for simulation of u*=
ln( z z 0 )
turbulence. Figures 3-8, 3-9, and 3-10 show
examples of the most import transient wind
events. These examples are all extracted in which κ=0.4 is von Karman’s constant,z
from a comprehensive data base of meas- is the height over the terrain, and z0 is the
ured wind data, maintained by DTU, roughness parameter. More complex for-
http://www.winddata.com/ . mulas for the transformation from u to u*
are available in the literature, by which ef-
3.1.8 Extreme winds – gusts fects of different terrain roughnesses in dif-
ferent directions are taken into account, viz.
Extreme winds and gusts consist of extremes “WASP cleaning”.
of the wind speed or extremes of the short-
term mean of the wind speed, e.g. over a 10- The friction velocities u* derived by the
sec period. Extreme wind speeds can be above transformation of the original data
treated in traditional manner as extremes of refer to the prevailing local roughness z0. It
the wind speed process during stationary 10- is often desirable to transform the friction
minute conditions. An example of a wind velocities to data that relate to a reference
speed record containing extreme gust is roughness z0=0.05 m which is different from
given in Figure 3-11. the true local roughness parameter. This can
be done by geostrophic mapping, utilizing
that the geostrophic wind speed is constant
equal to
u* u*
G= (ln( ) − A) 2 + B 2
κ fz 0
Figure 3-11 Example of extreme gust, from
for all roughnesses z0. Here, the Coriolis
www.winddata.com.
parameter is f=2×(earth’s rate of rotation in
Extreme value analysis radians per second)×sin(latitude)≈1.2⋅10−4
Extreme winds are usually given in terms of rad/s at latitude 55.5°, and the coefficients A
10-minute mean wind speeds, which occur and B take on values A=1.8 and B=4.5. The
with some prescribed recurrence period, procedure is as follows: The geostrophic
e.g., the 50-year wind speed. The 50-year wind speed G is calculated explicitly by the
wind speed is the 10-minute mean wind given formula when the true roughness z0
speed, which is, on average, exceeded once and the corresponding u* are given. For this
every fifty years. Determination of the 50- value of G, the same formula is used to im-
year wind speed requires an extreme value plicitly solve a new value of u* that corre-
analysis of available wind speed data to be sponds to the desired new reference rough-
carried out. It proves useful to carry out such ness z0.
an extreme value analysis on the friction
velocity pressures derived from the wind When the friction velocities u* have been
speed data, rather than on the wind speed determined from the wind speed data as de-
data themselves. For this purpose, the ob- scribed above, the corresponding velocity
served wind speeds u are transformed to pressures q are calculated from
friction velocities u* by
1 where κ=0.4 is von Karman’s constant, and

q= ρu *2 ρ0=1.225 kg/m3 is the density of air.
2
where ρ is the density of air. The above method of extreme value analysis
focuses on the maximum wind speed within
The original wind speed data u are now a specified period of time such as one year
transformed into a set of velocity pressure and is useful for estimation of wind speeds
data q. The q data are grouped into n subre- of specified recurrence periods. Note, how-
cords of specified duration, e.g. one year, ever, that other methods of extreme value
and the maximum value of q in each of the n analysis are also available, which may prove
subrecords is extracted. When the duration useful, the most important of which is the
of a subrecord is one year, then these n peak-over-threshold method.
maximum values of q constitute an empiri-
cal distribution of the annual maximum ve- When the wind speed with 50-year return
locity pressure. The annual maximum of the period is given, i.e. U10,50-yr, then the wind
velocity pressure is expected to follow a speed with return period T years can be
Type 1 extreme value distribution, i.e. a found as
Gumbel distribution,
T
q −b U 10,T = U 10, 50 − yr 0.57 + 0.11 ⋅ ln
F (q) = exp(− exp(− )) ln(1 p)
a
in which p=exp(−nT) is the probability of no
which has two distribution parameters a and exceedance in T years, and n is the number
b. The values of a and b are determined by of exceedances per year. For T=50 years,
fitting to the n observations of the annual n=0.02. Reference is made to DS410.
maximum velocity pressure. The shift pa-
rameter b is the mode of the distribution and
Hurricanes
is interpreted as the value of q which has a Saffir-Simpson’s hurricane scale, see Table
return period of one year. The value of q 3-3, groups and ranks hurricanes according
which has a return period of T years can be to the wind speeds involved and gives con-
found as sequences in terms of implied storm surges
and associated resulting damages.
T
q T = b + a ln
T0 Some parts of the world are characterized by
having other weather phenomena than those
where T0=1 year. This, in particular, can be considered here. One such weather phe-
used to find the 50-year velocity pressure for nomenon is cyclones. As this is a weather
T=50 years. phenomenon totally different from the
storms dealt with above, one cannot just
The corresponding 10-minute mean wind extrapolate results for such storms.
speed with return period T at height z and
terrain roughness z0 can be found as 3.1.9 Site assessment
It appears that the two variables U10 and σU
1 2q T z are essential as parameters in the models
U 10,T = ln
κ ρ0 z0 available for representation of the wind
speed. The most common method for deter-
mination of these two parameters is estima- Table 3-3 Saffir-Simpson Hurricane Scale
tion from wind speed measurements. Both Wind Storm Damage description
U10 and σU refer to a 10-minute reference speed surge
period, U10 being the 10-minute mean wind (m/s) (m)
speed and σU being the standard deviation of *
the wind speed over the 10 minutes. Note 1 33-42 1.0-
Some damage to trees,
that if wind speed measurements are ob- 1.7
shrubbery, and unan-
tained over intervals of other durations than chored mobile homes
10 minutes, the mean wind speed and stan- 2 43-49 1.8- Considerable damage
dard deviation over these other durations 2.6 to shrubbery and tree
need to be transformed to values referring to foliage; some trees
a duration of 10 minutes. The following blown down. Damage
approximate formula applies to transforma- to poorly constructed
tion of the mean wind speed UT in a period signs and buildings’
of duration T to the 10-minute mean wind roofs. Major damage to
speed U10 mobile homes.
3 50-58 2.7- Foliage torn from trees;
UT 3.8 large trees and poorly
U 10 ≈ constructed signs
T
1 − 0.047 ln blown down. Some
10
damage to roofing ma-
in which T is to be given in units of minutes. terials and structures of
Reference is made to Gran (1992) and DNV buildings. Mobile
(1998). homes destroyed.
4 59-69 3.9- Shrubs and trees blown
When the measurement period T is less than 5.6 down; all signs and
10 minutes, the standard deviation of the mobile homes down
wind speed measurements will come out and destructed. Exten-
smaller than the sought-after value for σU. sive damage to roofing
materials, windows and
The following approximate formula applies doors. Complete fail-
to transformation of the standard deviation ures of roofs on many
σU,T obtained from wind speed measure- small residences.
ments in a period of duration T to the 10- 5 70+ 5.7 Shrubs and trees blown
minute standard deviation σU down. Very severe
damage to windows
and doors. Complete
T 2
σ U ≈ σ U ,T + U 10 (0.047 ln
2 2
) failure of roofs on
10 many residences and
industrial buildings.
in which T is to be given in units of minutes. Some complete build-
The formula is valid for T<10 minutes. In ing failures. Small
cases where T>10 minutes, the formula to be buildings overturned or
used reads blown away.
* maximum 1-minute average
T 2
σ U ≈ σ U ,T − U 10 (0.047 ln
2 2
)
10
As described in Section 3.1.1, the long-term Program − is a method for prediction of

distribution of the 10-minute mean wind properties of winds in moderately complex
speed U10 can be represented by a Weibull terrain with relevance for loads on wind
distribution, turbines and other large structures,
http://www.wasp.dk/ . For detailed wind tur-
u bine siting, WASP allows for modelling
FU10 (u ) = 1 − exp(−( ) k )
A complications in the form of terrain inho-
mogeneities, sheltering obstacles, and terrain
For a specific location, the distribution pa- height differences. The method is based on
rameters A and k can be estimated from theory for attached flow and allows for site-
wind speed measurements. specific prediction of
• wind shears and wind profiles
For such an estimation, it is important to • extreme wind speeds
carry out wind speed measurements during a • turbulence
sufficiently long period of time, i.e., several when data for orography and roughness
years. Measurements from a period of time variations are given for the site, and when
as short as one year will not be sufficient to reference wind data are available from a
capture year-to-year variations, and if meas- reference site subject to the same overall
urements are limited to an even shorter pe- weather regime as the prediction site. The
riod of time, they will neither be sufficient mean wind speed can increase considerably
to capture seasonal variations. The meas- on a hill relative to that in flat terrain. The
urement period must be long enough to av- extreme wind speeds on the hill will be cor-
erage out year-to-year variations and sea- respondingly higher. The hill will also affect
sonal variations. Reference is made to Euro- the turbulence structure. Typically, the tur-
pean Wind Atlas. bulence in the direction of the mean wind
will be attenuated, while the vertical turbu-
Some locations have a topography, which lence will be amplified. Separation will usu-
leads to different wind regimes in different ally occur when the hill slope exceeds about
directions, and which thus warrants use of 30°, and the assumption of attached flow
different distribution models for U10 and σU will then no longer be valid.
in different directions. A location on the
coast is a good example of this, as wind In case of very complex, mountainous ter-
from the ocean will adhere to one distribu- rain, the prediction accuracy of WASP can
tion model and wind from land will adhere be assessed by means of the so-called RIX
to another. When such two distributions, number. The RIX number is a site-specific
conditional on direction for example for U10, ruggedness index, defined as the fractional
are combined to one unconditional long- extent of the terrain which is steeper than a
term distribution, this may well come out as critical slope. The critical slope is typically
a bimodal distribution, e.g. represented by chosen as the separation slope of 30°. The
parts of two Weibull distributions put to- RIX number can be estimated from meas-
gether. urements on a topographical map with a
sufficiently fine representation of height
WASP Engineering contours. The RIX number is a coarse meas-
Site-specific loads can be derived from site- ure of the extent of flow separation, and
specific wind conditions, which can be thereby of the extent to which the terrain
modelled by WASP engineering. WASP − violates the requirement of WASP that the
the Wind Atlas Analysis and Application surrounding terrain should be sufficiently
gentle and smooth to ensure mostly attached eral instruction can be found in DS412, by
flows. The operation envelope of WASP which one may set the lowest temperature of
thus corresponds to RIX≈0%. Reference is operation at –10°C for Danish locations,
made to Mortensen and Petersen (1997). because the frequency of lower temperatures
is very limited. For wind turbine structures,
When the prediction site is more rugged DS472 gives a temperature interval (−10°C,
than the reference site, then the wind speeds 30°C) for operation as an example. How-
at the prediction site will be overpredicted ever, the temperature interval for normal
by WASP. Similarly, the wind speeds will operation should always be chosen accord-
be underpredicted when the prediction site is ing to the recommendations specified for the
less rugged. The difference in RIX numbers respective materials, which are being used
between the two sites is a fairly coarse for the construction of the wind turbine. In
measure of the significance of the problem this context, the following issues may be
and provides estimates of magnitude and critical:
sign of the prediction error. As a rule of • Components and connections, which
thumb, when there is a difference ∆RIX in involve two or more materials with dif-
RIX numbers between the prediction site ferent coefficients of expansion. Glass,
and the reference site, then the approximate concrete and steel expand about the
magnitude of the relative prediction error in same, whereas plastics expand more.
the wind speed will be 2∆RIX. For example, • Choice of fluids for lubrication and hy-
if RIXprediction=20% and RIXreference=10%, draulic systems.
then ∆RIX=20%−10%=10%, and the overall • Materials whose mechanical properties
relative error in the predicted wind speeds change when the temperature changes,
will be about 2×10=20%. e.g. rubbers for seals, gaskets and
dampers become brittle at low tem-
Note that WASP engineering can give accu- peratures, and polyesters behave like
rate results outside its operation limits, pro- glass when the temperature falls below
vided the difference in RIX numbers be- the so-called glass transition tempera-
tween the actual site and the reference site is ture.
small and the topographical data are ade- It is important to consider the temperature
quate and reliable. interval for normal operation when materials
are selected for the construction of the wind
turbine and when other conditions of im-
3.2 Other External Conditions portance for the safety of the turbine are to
be evaluated.
3.2.1 Temperatures
Operational temperatures Extreme temperatures
A temperature interval for normal operation For structural parts which can be damaged
of the wind turbine is to be chosen. Under by extreme temperatures, extreme values of
normal functional conditions, the wind tur- temperatures need to be considered. Extreme
bine is considered to be in operation when values of high and low temperatures should
the temperature of the air is within this in- be expressed in terms of most probable
terval.. highest or lowest values with their corre-
sponding recurrence periods.
Different structural design codes have dif-
ferent approaches to how to handle tem- Table 3-4 gives values of the daily minimum
peratures. For steel structures, a rather gen- temperature and the daily maximum tem-
perature for various recurrence periods at occur simultaneously with strong winds. The
two Danish locations. This is based on 123 most severe temperature event to expect and
years of temperature observations, reported consider in the context of wind turbine de-
by Laursen et al. (1999a). Note that the tem- sign is an event with extremely low tem-
perature values tabulated for recurrence pe- perature in conjunction with power failure,
riods of 1000 and 10000 years are achieved in particular power failure of some duration.
by extrapolation and should therefore be For Swedish locations, Statens Planverks
used with caution. Författningssamling (1980) gives acceptable
design values for the daily mean temperature
For many materials, peak temperatures such for structural design against low tempera-
as a daily maximum or a daily minimum tures. Advice is given for how to transform
temperature have too short a duration to these values to acceptable one-hour mean
pose any problems for the materials and temperatures.
their behaviour. The duration is too short for
the materials to be cooled down or heated Table 3-5 Monthly mean temperatures at Danish
up. For such materials it will be more apt to locations
Recur- Kastrup Fanø
consider extreme values of the one-hour rence Monthly Monthly Monthly Monthly
mean temperature, of the daily mean tem- period mean mean mean mean
perature, or of the monthly mean tempera- (years) tempera- tempera- tempera- tempera-
ture (°C) ture (°C) ture (°C) ture (°C)
ture. low high low high
1 −0.4 16.4 0.2 16.0
Table 3-4 Daily minimum and maximum tempera- 10 −5.0 18.4 −4.2 18.3
tures at Danish locations 50 −6.0 19.3 −6.0 19.5
Recur- Landbohøjskolen Fanø 100 −7.7 19.6 −6.7 19.9
rence Daily Daily Daily Daily 1000 −9.7 20.6 −8.4 21.1
period min. max. min. max.
10000 −11.2 21.3 −9.8 22.1
(years) tempera- tempera- tempera- tempera-
ture (°C) ture (°C) ture (°C) ture (°C)
1 −13.6 29.3 −13.3 29.2 3.2.2 Density of air
10 −18.7 31.7 −18.2 32.0
50 −22.3 33.1 −21.5 33.7 The density of air is dependent on tempera-
100 −23.7 33.7 −23.0 34.3 ture and atmospheric pressure. The standard
1000 −28.8 35.3 −27.7 36.3 density of air at the sea surface at 10°C and
10000 −33.9 36.8 −32.4 38.0 1 atm pressure is
Table 3-5 gives low and high values of the ρ=1.225 kg/m3
monthly mean temperature for various re-
currence periods at two Danish locations. The air density averaged over 10 minutes
This is based on 30 years of temperature can be determined from the measured abso-
observations, reported by Laursen et al. lute air temperature T (in units of °K) aver-
(1999b). Note that the temperature values aged over 10 minutes and the measured air
tabulated for recurrence periods of 100, pressure B averaged over 10 minutes by
1000 and 10000 years are achieved by ex-
trapolation and should therefore be used
B
with caution. ρ 10 min =
RT
Note that for a coastal climate like the cli-
mate of Denmark, extreme temperatures in which R=287.05 J/kg/°K is the gas con-
such as 35°C and –20°C will usually not stant. At high temperatures it is recom-
mended also to measure the humidity and to determined. The density of ice can be set
correct for it. equal to ρ=700 kg/m3. For non-rotating
parts, an ice formation of 30 mm thickness
High wind speeds usually occur at low pres- on all exposed surfaces can be considered
sures, for which the density of air takes on for Denmark, The Netherlands, and North-
relatively low values. It is important to be ern Germany. For the wind turbine at stand-
aware that the density of air attains high still, it is relevant to consider also the rotor
values in arctic regions, whereas it attains blades with an ice cover of this thickness.
low values at high altitudes in tropical re- For the rotating wind turbine, it is relevant
gions. to consider a situation with all blades cov-
ered by ice, and a situation with all blades
It appears that − depending on temperature but one covered by ice.
and pressure conditions – the density of air
will actually follow some probability distri- For wind turbines on offshore locations, sea
bution, and this may be important to con- ice may develop and expose the foundation
sider when fatigue assessments are carried of the turbine to ice loads. Loads from later-
out. ally moving ice should be based on relevant
full scale measurements, model experiments
3.2.3 Humidity which can be reliably scaled, or on recog-
nized theoretical methods. When determin-
Humidity is a function of temperature and
ing the the magnitude and direction of ice
atmospheric pressure. If the wind turbine is
loads, considerations should be given to the
to be located in humid areas or in areas with
nature of ice, mechanical properties of the
high humidity part of the day and/or part of
ice, ice structure contact area, shape of
the year, components sensitive to such con-
structure, direction of ice movements etc.
ditions have to be sufficiently protected. A
The oscillating nature of ice loads due to
relative humidity of up to 95% should usu-
build-up and fracture of moving ice should
ally be considered for design.
also be considered.
Increase of corrosion rates due to humidity
Where relevant, ice loads other than those
and especially cyclic exposure to humidity
caused by laterally moving ice, such as loads
should be estimated. Bacterial and fungal
due to masses of ice frozen to the structure
growth should be considered for compo-
and possible impact loads during thaw of the
nents sensitive to such activity.
ice, should be taken into account.
3.2.4 Radiation and Ultraviolet Light
Possible increase in area due to to icing
The intensity of solar radiation should be should be considered when determining
considered for components, which are sen- wind or wave loads on such areas.
sitive to ultraviolet radiation and/or tem-
perature. It is common to consider a solar 3.2.6 Rain, Snow and Hail
radiation intensity of 1000 W/m2 for design.
Rain storms should be considered for com-
Further details about solar radiation can be
ponents, which may be subject to water
found in NASA (1978).
damage. If no special provisions are made,
leakages through covers should be consid-
3.2.5 Ice
ered. If there is a possibility for accumula-
When the turbine is located in an area where tion of snow on the wind turbine, the in-
ice may develop, ice conditions should be crease of weight due to such accumulation
should be considered. The density of heavy Offshore wind turbines will be exposed to
snow is in the range 100-150 kg/m3. corrosion owing to their saline and marine
environment. Their foundations will have
In general, caution should be exhibited structural parts whose locations in the wave
about installation of wind turbines in areas splash zone will be particularly exposed to
where hail storms are known to prevail. corrosion.
When designing wind turbines for areas
where hail storms are common, possible The expected corrosion rate depends on the
damage due to impact from hail stones environment.
should be considered. The nacelle and the
coating on the blades may be specially vul- 3.2.8 Earthquake
nerable to such damage, in particular if the
The effects of earthquakes should be consid-
hail stones fall during strong winds, i.e. they
ered for wind turbines to be located in areas
come in at an angle with the vertical. The
that are considered seismically active based
velocity by which a hail stone falls towards
on previous records of earthquake activity.
the ground is governed by drag and not in-
fluenced by the weight of the hail stone.
For areas where detailed information on
When the speed of the hail stone is known
seismic activity is available, the seismicity
together with the wind speed, then the angle
of the area may be determined from such
by which the hail stone will hit a wind tur-
information.
bine structure can be determined.
For areas where detailed information about
Extreme hail conditions may exist on some
seismic activity is not generally available,
locations. Such conditions can be critical for
the seismicity should preferably be deter-
a wind turbine and may force the turbine to
mined on the basis of detailed investiga-
stop. Extreme hail events may in some cases
tions, including a study of the geological
govern the design of the leading edges of the
history and the seismic events of the region.
rotor blades. The density of hail is usually
higher than that of snow. NASA (1978) re-
If the area is determined to be seismically
ports a density of hail of about 240 kg/m3.
active and the wind turbine will be affected
by a possible earthquake, an evaluation
3.2.7 Atmospheric Corrosion and Abra-
should be made of the regional and local
sion
geology in order to determine the location
Abrasive action by particles transported by relative to the alignment of faults, the epi-
the wind should be considered for exposed central and focal distances, the source
surfaces. mechanism for energy release, and the
source-to-site attenuation characteristics.
The corrosive environment should be repre- Local soil conditions need to be taken into
sented by generally recognized methods. account to the extent they may affect the
The classification of the environment should ground motion. The evaluation should con-
be such that a sufficient corrosion protection sider both the design earthquake and the
system may be established. maximum credible earthquake.
Industrial environments may be rather harsh When a wind turbine is to be designed for
to wind turbine structures. installation on a site which may be subject to
an earthquake, the wind turbine has to be
designed to withstand the earthquake loads.
Response spectra in terms of so-called such that it suffices to establish the accel-
pseudo response spectra can be used for this eration spectrum and then use this to com-
purpose. pute the other two spectra. The three spectra
can be plotted together for various γ as
shown in the example in Figure 3-12. In this
example, the period T=2π/ω is used as the
spectral parameter instead of ω.
It is important to analyze the wind turbine

structure for the earthquake-induced accel-
erations in one vertical and two horizontal
directions. It usually suffices to reduce the
analysis in two horizontal directions to an
analysis in one horizontal direction, due to
the symmetry of the dynamic system. The
vertical acceleration may lead to buckling in
the tower. Since there is not expected to be
much dynamics involved with the vertical
motion, the tower may be analyzed with
respect to buckling for the load induced by
the maximum vertical acceleration caused
by the earthquake. However, normally the
only apparent buckling is that associated
with the ground motion in the two horizontal
directions, and the buckling analysis for the
vertical motion may then not be relevant.
For analysis of the horizontal motions and
Figure 3-12 Example of pseudo response accelerations, the wind turbine can be repre-
spectrum sented by a concentrated mass on top of a
vertical rod, and the response spectra can be
Pseudo response spectra for a structure are used directly to determine the horizontal
defined for displacement, velocity and ac- loads set up by the ground motions. An ex-
celeration. For given damping ratio γ and ample of a pseudo response spectrum is
angular frequency ω, the pseudo response shown in Figure 3-12. For a typical wind
spectrum S gives the maximum value of the turbine, the concentrated mass can be taken
response in question over the duration of the as the mass of the nacelle, including the ro-
response. This can be calculated from the tor mass, plus ¼ of the tower mass.
ground acceleration history by means of
Duhamel’s integral. The following pseudo 3.2.9 Lightning
response spectra are considered
• SD, response spectral displacement Lightning strokes imply discharging of cur-
rent to the earth following a separation of
• SV, response spectral velocity
charge in thunderstorm clouds. Median val-
• SA, response spectral acceleration
ues of lightning current, charge transfer, and
For a lightly damped structure, the following
specific energy produced by a single stroke
approximate relationships apply
are 30 kA, 10 C, and 50 kJ/Ω, respectively.
Maximum recorded values of these quanti-
SA≈ω2SD and SV≈ωSD
ties are 300 kA, 400 C, and 20 MJ/Ω, re- REFERENCES

spectively.
Astrup, P., S.E. Larsen, O. Rathmann, P.H.
s
Madsen, and J. Højstrup, “WA P Engineer-
When lightning strokes hit wind turbines,
ing – Wind Flow Modelling over Land and
lightning current has to be conducted
Sea”, in Wind Engineering into the 21st
through the wind turbine structure to the
Century, eds. A.L.G.L. Larose and F.M.
ground, passing through or near to practi-
Livesey, Balkema, Rotterdam, The Nether-
cally all wind turbine components and ex-
lands, 1999.
posing them to possible damage. Large wind
turbine components such as blades and na-
Dansk Ingeniørforening, Last og sikkerhed
celle cover are often made of composite
for vindmøllekonstruktioner, in Danish,
materials incapable of sustaining direct
DS472, 1st edition, Copenhagen, Denmark,
lightning stroke or conducting lightning cur-
1992.
rent, and are therefore vulnerable to light-
ning damage.
Dansk Standard, Norm for last på konstruk-
tioner, in Danish, DS410, Copenhagen,
In areas with a high probability of damage to
Denmark, 1999.
wind turbines due to lightning, a protection
system should be installed. The purpose of a
Dansk Standard, Norm for stålkonstruk-
lightning protection system is to reduce the
tioner, in Danish, DS412, 3rd Edition, Co-
damage caused by lightning to a tolerable
penhagen, Denmark, 1999.
level.
DEFU, “Lightning protection of wind tur-
The design of any lightning protection sys-
bines,” Recommendation No. 25, Edition 1,
tem should take into account the risk of
DEFU, 1999.
lightning striking and/or damaging the wind
turbine structure. The risk of lightning
DNV, Environmental Conditions and Envi-
striking the wind turbine structure is a func-
ronmental Loads, Classification Notes No.
tion of the height of the structure, the local
30.5, Det Norske Veritas, Høvik, Norway,
topography, and the local level of lightning
1998.
activity. The lightning risk is usually higher
in low mountainous areas than in coastal
Dyrbye, C., and S.O. Hansen, Wind Loads
areas. For structures taller than 60 m, side
on Structures, John Wiley and Sons, Chich-
flashes occur, i.e., some of the lightning
ester, England, 1997.
flashes strike the side of the structure rather
than the tip. Such side flashes are a cause for
Gran, S., A Course in Ocean Engineering,
concern in connection with wind turbines, as
Elsevier, Amsterdam, The Netherlands,
blades struck on the sides may be severely
1992. The Internet version, located at
damaged. Statistically, as many as 40-50%
http://www.dnv.com/ocean/ , provides on-
of all lightning stroke events for wind tur-
line calculation facilities within the fields of
bines cause damage to the control system,
ocean waves, wave loads, fatigue analysis,
whose availability is of vital importance for
and statistics.
the operation of a wind turbine.
IEC, Wind turbine generator systems – Part
For lightning, reference is made to DEFU
1: Safety requirements, International Stan-
(1999).
dard, IEC61400-1, 2nd Edition, 1999.
Jensen, N.O., “Atmospheric boundary layers curacy of wind flow modelling in complex
and turbulence”, in:Wind engineering into terrain,” Proceedings, European Wind En-
the 21st century. Proceedings of Tenth In- ergy Conference, Dublin, Ireland, 1997.
ternational Conference on Wind Engineer-
ing, Copenhagen, Denmark, 21-24 June National Aeronautics and Space Admini-
1999. Larsen, A.; Larose, G.L.; Liversey, stration (NASA), “Engineering Handbook
F.M. (eds.), A.A. Balkema, Rotterdam/ on the Atmospheric Environmental Guide-
Brookfield, Vol. 1, pp. 29-42, 1999. lines for Use in Wind Turbine Generator
Development,” NASA Technical Paper
Laursen, E.V., J. Larsen, K. Rajakumar, J. 1359, 1978.
Cappelen, and T. Schmith, “Observed daily
precipitation and temperature from six Dan- Norwegian Petroleum Directorate, Acts,
ish sites, 1874-1998,” Technical Report No. regulations and provisions for the petroleum
99-20, Danish Meteorological Institute, Co- activity, Stavanger, Norway, 1994.
penhagen, Denmark, 1999a.
Panofsky, H.A., and J.A. Dutton, Atmo-
Laursen, E.V., R.S. Thomsen, and J. Cap- spheric Turbulence, Models and Methods
pelen, “Observed Air Temperature, Humid- for Engineering Applications, John Wiley
ity, Pressure, Cloud Cover and Weather in and Sons, New York, N.Y., 1984.
Denmark – with Climatological Standard
Normals, 1961-90,” Technical Report No. Risø National Laboratory, European Wind
99-5, Danish Meteorological Institute, Co- Atlas, The Handbook of European Wind
penhagen, Denmark, 1999b. Resources, Risø, Denmark, 1989.
Madsen, H.O, S. Krenk, and N.C. Lind, Statens Planverks Författningssamling,

Methods of Structural Safety, Prentice-Hall “Svensk Byggnorm, SBN avd. 2A, Bärande
Inc., Englewood Cliffs, N.J., 1986. konstruktioner, med kommentarer,” in
Swedish, LiberFörlag, Stockholm, Sweden,
Mortensen, N.G., and E.L. Petersen, “Influ- 1980
ence of topographical input data on the ac-
4. Loads • maintenance and repair

• testing
This list of conditions can be used as a tool
4.1 Load cases to identify design situations, which can be
As a part of the design process, a wind tur- expected to be encountered, and to assess
bine must be analysed for the various loads whether listed load cases are relevant, and
it will experience during its design life. A whether relevant load cases are missing.
prime purpose in this respect is to verify that
the turbine will be able to withstand these 4.1.2 Wind events
loads with a sufficient safety margin. This The wind conditions can be divided into
task is systematised by analysing the wind normal and extreme conditions. The
turbine for a number of relevant load cases. IEC61400-1 standard makes a further dis-
tinction and defines the following condi-
Load cases can be constructed by combining tions:
relevant design situations for the wind tur-
• normal wind profile, see Section 3.1.1
bine with various external conditions. The
• normal turbulence, see Section 3.1.2
design situations mainly consist of various
operational conditions of the wind turbine. • extreme coherent gust
The external conditions for the most part • extreme direction change
consist of various wind conditions or wind • extreme operating gust
events. • extreme wind speed, see Section 3.1.8
• extreme wind shear, see Section 3.1.8
As a guidance, lists of design situations,
wind events, and design load cases are pro- 4.1.3 Design load cases
vided in the following subsections. These The design load cases to be analysed during
lists are meant as an aid and are not neces- the design process for a wind turbine are
sarily complete. For design of a wind tur- constructed by combination of relevant de-
bine, it is important to identify all load sign situations and external conditions. The
cases, which are relevant for the wind tur- following combinations constitute a mini-
bine. Failure mode and effects analysis is a mum number of relevant combinations:
useful tool in assessing which load cases are • normal operation and normal external
relevant and which are not, see Chapter 2. conditions
• normal operation and extreme external
4.1.1 Design situations
conditions
The relevant design situations consist of the • fault situations and appropriate external
most significant conditions that a wind tur- conditions, which may include extreme
bine may experience. The design situations external conditions
can be divided into operational conditions • transportation, installation and mainte-
and temporary conditions. The operational nance situations and appropriate exter-
conditions include nal conditions
• normal operation and power production Table 4-1 lists a number of design load cases
• cut-in, cut-out, idling, and stand-still to consider according to DS472. Table 4-2
The temporary conditions include presents a similar list extracted from
• transportation IEC61400-1. These lists are examples, and it
• installation, assembly is the duty of the designer to validate that
• faults, such as control system fault they are adequate.
4 – Loads 53
When calculating loads it is important to • misaligned wind flow relative to the

keep in mind a few factors which may influ- rotor axis, e.g. owing to a yaw error
ence the involved wind conditions and the
magnitude of the loads: Note that some of the load cases that are
• tower shadow and tower stemming, listed in Tables 4-1 and 4-2 are load cases,
which are the disturbances of the wind which refer to site-specific wind turbines
flow owing to the presence of the tower only.
• wake effects wherever the wind turbine
is to be located behind other turbines,
such as in wind farms
Table 4-1 Design load cases, cf. DS472

Limit state
Load Design situation Wind condition Fati- Ulti- Acci-
case type gue mate dent
Normal Power production Normal x x
Start and switch Normal x x
Stop and transition to Normal x x
idling
Stand-still and idling Normal x
Extra- Power production Extreme stationary wind (x) x
ordinary with U10=v10min,50yr
Power production Transient wind condi- (x) x
tions for direction
change 0-90° and wind
speed change 10-25
m/sec in 30 secs
Transport, assembly Largest allowable U10 to x
and erection be specified
Manual operation of To be defined according x x
wind turbine to relevance
Emergency stop 1.3 times cut-out wind x x
speed
Activation of air brakes 1.3 times cut-out wind x x
speed
Idling at yaw error 0.5v10min,50yr x x
Fault conditions Normal x x
Acci- Serious failure Cut-out wind speed (x) x
dental
x to be included
(x) to be included unless it can be established as unnecessary for the current design of the
wind turbine
54 4 – Loads
Table 4-2 Design load cases, cf. IEC61400-1

Design situation Wind condition Other conditions Type of
analysis
Power production Normal turbulence Ultimate
Normal turbulence Fatigue
Extreme coherent gust Ultimate
with direction change
Normal wind profile External electrical Ultimate
fault
Extreme operating gust, Loss of electrical Ultimate
one-year recurrence connection
Extreme operating gust, Ultimate
fifty-year recurrence
Extreme wind shear Ultimate
Extreme direction change, Ultimate
fifty-year recurrence
Extreme coherent gust Ultimate
Power production plus Normal wind profile Control system fault Ultimate
occurrence of fault
Normal wind profile Protection system Ultimate
fault or preceding
internal electrical
fault
Normal turbulence Control or protection Fatigue
system fault
Start-up Normal wind profile Fatigue
one-year recurrence
Extreme direction change, Ultimate
one-year recurrence
Normal shutdown Normal wind profile Fatigue
one-year recurrence
Emergency shutdown Normal wind profile Ultimate
Parked (standing still or Extreme wind speed, Possible loss of elec- Ultimate
idling) fifty-year recurrence trical power network
Normal turbulence Fatigue
Parked and fault condi- Extreme wind speed, one- Ultimate
tions year recurrence
Transport, assembly, To be stated by manu- Ultimate
maintenance and repair facturer
4 – Loads 55
4.2 Load Types load, i.e. the mean flapwise bending mo-
ment, and a reduction in the stiffness.
The external loads acting on a wind turbine
are mainly wind loads. As a wind turbine
The load response in a rotor blade is very
consists of slender elements such as blades
dependent on the damping. The total damp-
and tower, inertia loads will be generated in
ing is a combination of aerodynamic damp-
addition to the gravity loads that act on these
ing and structural damping. The aerody-
elements. Loads due to operation such as
namic damping depends on
centrifugal forces, Coriolis forces and gyro-
• choice of blade aerodynamic profile in
scopic forces must also be considered.
conjunction with chosen blade twist
The loads on a wind turbine can thus in most • operational condition
cases be classified as follows • wind speed
• aerodynamic blade loads • rotor frequency
• gravity loads on the rotor blades • vibration direction of blade cross sec-
• centrifugal forces and Coriolis forces tion
due to rotation • motion of blade section relative to in-
• gyroscopic loads due to yawing coming flow
The structural damping depends much on
• aerodynamic drag forces on tower and
the blade material. The aerodynamic load
nacelle
response is much a result of the lift and drag
• gravity loads on tower and nacelle.
forces in conjunction with the blade profile
properties and damping, and with effects of
Gravity loads on the rotor blades cause
the motion of the rotor structure included.
bending moments in the blades in the edge-
wise direction. For a pitch-controlled tur-
The following subsections give a brief intro-
bine, gravity loads will also cause bending
duction to the most important load types
moments in the flapwise direction. Due to
encountered for a wind turbine with empha-
the rotation of the blades, the gravity load
sis on the physics behind them. More details
effects in the blades will be cyclically vary-
about loads and how to predict them are
ing bending moments. The larger the rotor
given in Sections 4.3 through 4.5.
diameter, the greater are the gravity load
effects. Typically, the bending moment at
4.2.1 Inertia and gravity loads
the blade root will follow a fourth-power
law in the rotor diameter. Considering that Rotor
the rotor area follows a quadratic power law The inertia and gravity loads on the rotor are
in the rotor diameter, this forms one of the mass dependent loads. The cross-sectional
greater challenges in making wind turbines centrifugal force Fc depends on the angular
larger. rotor speed, the radial position, and the mass
of each blade element. At the blade root, this
Centrifugal forces induced by the rotation of force is
the blades can be utilized in conjunction
with rearwards coning of the blades to com- n
pensate for the effects of some of the wind Fc = å m i ri ω 2

i =1
loads and also to provide more stiffness.
Forwards coning of the blades would, in in which mi [kg] is the mass of the ith blade
contrast, lead to an increase in the mean element, ω [rad/s] is the angular rotor speed,
and ri [m] is the radial position of the ith
56 4 – Loads
blade element in a discretisation of the blade respectively.

into n elements.
In many cases it is possible to neglect gyro-
The gravity force is simply given as scopic effects, because the angular velocity
of the yaw system is usually rather small.
n
However, flexible rotor-bearing supports
Fg = å m i g or Fg = mblade g can lead to significant gyroscopic forces
i =1
(rotor whirl), and gyroscopic forces should
in which g [9.82 m/s2] is the acceleration of never be neglected for present MW turbines.
gravity, mi [kg] is the mass of the ith blade
element, and mblade is the total blade mass. Note that the quoted formulas for MK and
MG need to be adjusted for a possible rotor
In general, gyroscopic loads on the rotor will tilt.
occur for any flexible rotor support. In par-
ticular, gyroscopic loads on the rotor will 4.2.2 Aerodynamic loads
occur whenever the turbine is yawing during Rotor
operation. This will happen regardless of the The true wind flow in the vicinity of the
structural flexibility and will lead to a yaw wind turbine rotor is rather complex, be-
moment MK about the vertical axis and a tilt cause the rotor induces velocities. Hence it
moment MG about a horizontal axis in the is common practice to use a simplified
rotor plane. method for calculating rotor loads to be used
for a wind turbine design.
For a three-bladed rotor, the net resulting
yaw moment due to the gyroscopic load The wind speed conditions at a blade cross
effects is zero, MK=0, whereas a non-zero section are illustrated in Figure 4-1.
constant tilt moment is produced,
MG=3M0/2, in which
n
M 0 = 2ω K ω å m i ri
2
i =1
where ω is the angular velocity of the rotor,

ωK is the angular yaw velocity, and mi is the
ith mass located at radius ri in a discretisa-
tion of the rotor blade into n discrete masses.
For a two-bladed rotor, the gyroscopic load

effects lead to a cyclic yaw moment and a
cyclic tilt moment,
Figure 4-1 Diagram of air velocities at a
MK=2M0cosωtsinωt blade cross section
and The wind speed perpendicular to the rotor

plane is V0. This is the inflow wind speed.
MG=2M0cos2 ωt, When the wind passes through the rotor
plane, this wind speed becomes reduced by
4 – Loads 57
an amount aV0 due to axial interference. The yaw system may also be present. The most
rotor rotates with angular rotor speed ω, so a important functional loads can be catego-
blade element at a distance r from the rotor rised as follows
axis will be moving at a speed ωr in the ro- • brake loads from mechanical and aero-
tor plane. When the wind passes through the dynamic brakes
rotor plane and interacts with the moving • transient loads in the transmission sys-
rotor, a tangential slipstream wind speed tem, e.g. caused by engagement of the
a’ωr is introduced. The resulting relative generator
inflow wind speed that the rotor blade will • yawing loads, i.e. loads produced di-
experience comes out as shown in Figure 4- rectly by yawing
1 and is denoted W. This resulting relative • loads caused by pitching the blades or
wind speed gives rise to aerodynamic forces engaging the air brakes, initiated by the
on the blade, viz. a lift force control system
1 4.2.4 Other loads

FL = C L ρcW 2
2 Other loads or load effects to consider are
those induced by tower shadow, vortex
and a drag force shedding by the tower, damping, and insta-
bilities like stall-induced flapwise and
1 edgewise vibrations. Blade vibrations might
FD = C D ρcW 2
2 be both flapwise and edgewise. Either mode
shape (and possibly others) might be excited
in which CL and CD are lift and drag coeffi- due to negative aerodynamic damping.
cients, respectively, ρ is the density of air,
and c is the chord length of the blade. For For wind turbine structures installed in wa-
more details about how to calculate aerody- ter, it is necessary also to consider loads set
namic forces, see Sections 4.3 through 4.5. up by the marine environment such as wave
loads, current loads, and ice loads.
Tower and nacelle
The aerodynamic drag force, Fd on the tower A general reference is made to Andersen et
and the nacelle can be calculated on the ba- al. (1980).
sis of the projected area perpendicular to the
flow 4.3 Aeroelastic load calculations
4.3.1 General
Fd = 0.5 ρ A V02 CD
Load calculations for a wind turbine struc-
in which CD [-] is the aerodynamic drag co- ture are usually performed by means of a
efficient and A [m2] is the projected area computer program based on an aeroelastic
perpendicular to the flow. calculation procedure. The purpose of an
aeroelastic wind turbine analysis is to solve
4.2.3 Functional loads the equations of motion for a given arbitrary
set of forces acting on the structure and for
Functional loads on a wind turbine occur forces generated by the structure itself. Of-
when the turbine is subject to transient op- ten, such a code applies a geometrically non-
erational conditions such as braking and linear finite element approach or a modified
yawing, or when generators are connected to modal analysis approach. Reference is made
the grid. Further functional loads from the to Figure 4-2.
58 4 – Loads
In any case, it is required that the code must structure. In this context, most components
be able to include and simplify the complex are subjected to investigations concerning
mechanical structure of the wind turbine as both extreme loading and fatigue loading.
well as being able to model arbitrary deter-
ministic and stochastic forces acting on the It is inevitable that some engineering judg-
turbine. The general formulation of the dif- ment will always be involved when prepar-
ferential equations of motion is ing for aeroelastic load calculations. It is
therefore important that the modelling as far
Mx + Cx + Kx = F as possible is supported by measured data,
which may be available from various
in which sources. These include type testing of the
• M is a mass matrix rotor blades and prototype testing of an ac-
• C is a damping matrix tual turbine, and they are both important to
• K is a stiffness matrix consider when an aeroelastic model is to be
• F is a force vector acting on the struc- validated.
ture and typically varies with time
• x and its derivatives are unknown vec- 4.3.2 Model elements
tors containing translations and/or rota- Wind field modelling
tions and their derivatives The wind field contains three wind speed
components
• longitudinal wind speed
• transversal wind speed
• vertical wind speed
The wind field is usually divided into
• a mean wind field with shear and slope
• a fluctuating wind field, i.e., turbulence
Wind field simulation is an important part of
a structural wind turbine analysis. For wind
turbines, spatial variations in the turbulence
must be considered, and three-dimensional
wind simulation is required. A prime pur-
pose of wind field simulation is to predict
time series of the wind speed in a number of
points in space, e.g. a number of points
across the rotor disc of a wind turbine. Such
time series of the wind speed form useful
input to structural analysis models for wind
turbines.
Figure 4-2 Example showing representation The parameters used to describe the mean
of a wind turbine structure by finite element wind field are usually those describing the
model (HawC). wind profile and the vertical component of
the wind vector. For flat homogeneous ter-
The loads that are derived from an aeroelas- rain, the vertical component of the wind
tic model are used for design of the various vector can usually be taken as zero.
components that constitute the wind turbine
4 – Loads 59
The mean wind field is superimposed by a the standard deviation of the wind speed,
fluctuating wind field, also referred to as the and c is a coherence decay factor.
turbulence field, with wind speed compo-
nents in three directions, i.e. the longitudi- The original Veers model is extended from a
nal, transversal and vertical turbulence com- one-component longitudinal turbulence
ponents, respectively. The parameters used model to a full-field three-dimensional and
to describe the fluctuating part of the wind three-component model by using the calcu-
field depend on which model is used for its lation method for the longitudinal direction
representation. in the transversal and vertical directions as
well. No cross correlation between the three
In order to predict the wind field in a num- components is modelled. The parameters
ber of points in space, the spatial coherence used in the Veers model are:
of the wind field must be properly accounted • mean wind speed U
for. There are two models available for gen- • standard deviation of wind speed com-
erating a synthetic wind field over a rotor ponents σu, σv, σw
disc, • integral length scales Lu, Lv, Lw of the
• the Veers model by Sandia (Veers, 1988) turbulence components
• the Mann model by Risø (Mann, 1994) • coherence decay factors cu, cv, cw
The Veers model uses a circular grid in the in which the indices u, v and w refer to the
rotor plane, while the Mann model applies a longitudinal, transversal and vertical com-
quadratic grid. Both models generate a syn- ponents, respectively. Reference is made to
thetic set of time series of turbulent wind by Section 3.1.5 for details about the integral
an inverse Fourier transform together with a length scales.
“defactorization” of the coherence.
The Mann model is based on a spectral ten-
The Veers model is based on a method de- sor formulation of the athmospheric surface
veloped by Shinozuka. It is based on a sin- layer turbulence. The model is developed for
gle point spectral representation of the tur- homogeneous terrain and the parameters
bulence and a coherence function. A Kaimal used are:
formulation is chosen as the spectral model, • mean wind speed U
• height above terrain z
6 .8
Li • roughness length z0
Si ( f ) = σ i
2 U The Mann model is capable of representing
fL
(1 + 10.2 i ) 5 / 3 the cross correlation between the three wind
U speed components.
and an exponential Davenport coherence Measurements are used to establish the pa-
model is used, rameters that are used in the models. The
turbulence intensity, defined from the stan-
r dard deviation of the longitudinal wind
Cohi (r , f ) = exp(−c i f ) speed by IT=σu/U, is usually measured by
U
means of a cup anemometer. This measure-
in which the index i identifies the compo- ment corresponds to vectorial summation of
nent, f is the frequency, r is the distance or the longitudinal and transversal wind speed
components. Hence, the turbulence intensity
spatial separation, L is the length scale, σ is
calculated on this basis usually comes out
somewhat higher than the sought-after tur-
60 4 – Loads
bulence intensity for the longitudinal com- • rotational speed

ponent alone, however, it is often used to • rotor radius
represent the longitudinal intensity when no • overall layout of the rotor blade
other estimate is available. • twist
• taper
Data for the relationship between σu, σv, and • thickness distribution
σw are given in Section 3.1. The longitudinal • airfoil properties
length scale Lu is in general dependent on • thickness
the height and the terrain roughness, see
• camber
Section 3.1 for details. The transverse and
• smoothness of surface
vertical length scales can be assumed to re-
• leading edge thickness
late to the longitudinal length scale by
Lv=0.3Lu and Lw=0.1Lu. Typical values of • roughness insensitivity
the longitudinal decay factors ci are in the • blunt/sharp trailing edge
range 2-27.
Both the Veers model and the Mann model

can be used to simulate time series of the
wind speed in a series of points in a plane
perpendicular to the mean direction of the
wind, e.g. a rotor plane. For details about the
models, reference is made to Veers (1988)
and Mann (1994, 1998). Figure 4-3 Distributions of aerodynamic
forces on an airfoil, and their resultants
Aerodynamic modelling
Aerodynamics deals with the motion of air A brief introduction to aerodynamic model-
and the forces acting on bodies in motion ling is given in the following with a special
relative to the air. Aerodynamic theory view to application to rotor blades. The most
makes it possible to perform quantitative important concepts are introduced, and the
predictions of the forces set up by the air so-called blade element momentum theory
flow on the rotor. for calculation of aerodynamic forces on a
rotor blade is presented. By this method,
On stall-regulated wind turbines, large re- momentum balance is used for calculation of
gions of separated flow will occur during blade element forces. As with any other
operation at high wind speeds. The stall method, this method implies a number of
regulation controls the power output from simplifications and idealizations, which
the turbine by exploiting the decrease in the make it practical for use. Other methods do
lift force that takes place when the wind exist, but they are time-consuming and not
turbine blades stall and thereby limit the considered any better.
power output and the loads. The overall
three-dimensional flow on a rotor is a very Consider first a two-dimensional flow past a
complex unsteady flow depending on a profile. Two-dimensional flow is confined to
number of variables, which include a plane. The out-of-plane velocity is thus
• wind speed zero. In order to obtain a two-dimensional
• wind shear flow, it is necessary to extrude the profile
• atmospheric turbulence into a blade of infinite span. For a true
• yaw angle blade, the shape, the twist and the profile
4 – Loads 61
change over the span, and the blade starts at FD

a hub and ends in a tip. However, for long CD =
1
slender blades, the spanwise velocity com- ρW 2 c
2
ponent is usually small relative to the
streamwise component, such that aerody-
namic solutions for two-dimensional flow respectively, in which ρ is the density of air
have a practical interest for wind-turbine and c is the chord length of the airfoil, and
rotor blades. The resultant force F from the in which FL and FD are forces per unit
flow on the blade comes about as the inte- length.
gral of pressure and frictional forces, see
Figure 4-3. The resultant force F is decom- The lift and drag coefficients are functions
posed into one component FL perpendicular of the inflow angle α, as shown in Figure 4-
to the direction of the resulting relative wind 5, of the airfoil shape, and of the Reynolds
speed W and one component FD parallel to number Re=cW/ν, in which ν is the kine-
this direction. FL is the lift and FD is the matic viscosity. CL increases linearly with α
drag, see Figure 4-4. Both the lift FL and the up to the stall angle, αstall, where the profile
drag FD depend on the inflow angle α. stalls. For greater values of α, CL reaches a
maximum value, followed by a decrease in
CL for further increases in α. For small α,
CD is almost constant, but increases rapidly
after stall. The stall phenomenon is closely
related to the separation of the boundary
layer from the upper side of the airfoil. The
manner in which the profile stalls is depend-
ent on the geometry of the profile. Thin pro-
files with a sharp nose, i.e. a high curvature
around the leading edge, tend to stall more
Figure 4-4 Definition of lift and drag on abruptly than thicker airfoils. This is a result
airfoil in 2-D flow of differences in the way the boundary layer
separates. Values of the coefficients CL and
Stall is a nonlinear phenomenon that results CD can be looked up in tables, calibrated to
in a dramatic loss of flow attachment and wind tunnel measurements, see Abbott and
airfoil lift when some limiting inflow angle von Doenhoff (1959).
αstall has been reached. The stall angle αstall
is somewhat arbitrary.
Lift and drag coefficients CL and CD are

defined as
FL
CL =
1
ρW 2 c
2
and
Figure 4-5 Lift and drag curve for an air-
foil (Bak et al., 1999a).
62 4 – Loads
An example of the variation of CL and CD CLs is the lift coefficient at onset of stall
with the inflow angle α is given in Figure 4-
5. It appears that two-dimensional wind tun- To describe the physics in a mathematical
nel measurements are used to obtain the way, the wind turbine rotor can be consid-
coefficient values in the pre-stall region. ered as a disc, which is able to absorb en-
Computational fluid dynamics with correc- ergy from the wind by reduction of the wind
tions for three-dimensional effects is used to speed. The wind speed and the pressure
obtain them in the stall region. In the post- conditions around the disc are illustrated in
stall region, measurements are typically used Figure 4-6.
to determine the coefficient values. Other-
wise, a method by Viterna and Corrigan
(1981) can be used for their prediction. This
method assumes rotors with zero twist an-
gle, and results by the method therefore need
modification when this assumption is not
fulfilled. By this method, the maximum drag
coefficient at inflow angle α=90o is
c D , max = 1.11 + 0.018 ⋅ AR
in which AR denotes the aspect ratio of the

airfoil. The aspect ratio AR is defined for a
blade as the ratio between the length of the
blade and a representative chord. The drag
coefficient in the post-stall region is given
by
C D = B1 sin 2 (α ) + B 2 cos(α ), 15 o ≤ α ≤ 90 o
Figure 4-6 Influence of a wind turbine on
where wind speeds and air pressures
B1 = C D max
B 2 = (1 / cos(α s )) ⋅ (C Ds − C D , max sin 2 (α s ))
αs = inflow angle at stall onset (usually 15o)
CDs = drag coefficient at stall onset
The lift coefficient is given by
C L = A1 sin 2 (α ) + A2 (cos 2 (α ) / sin(α )),

15 o ≤ α ≤ 90 o
where Figure 4-7 Wake illustrated by means of

A1 = B1 / 2 smoke
A2 = (C Ls − C D max sin(α s ) cos(α s )) ⋅
(sin(α s ) / cos 2 (α s )) The three-dimensional flow will produce a
continuous sheet of tangential vorticity be-
4 – Loads 63
hind the trailing edge of the blade, and the dT = 4πrρV 0 a (1 − a)dr
2
wake behind the rotor will be rotating. The

wake is illustrated in Figure 4-7.
and
Calculation procedure for the blade ele-
ment momentum method dQ = 4πr 3 ρV0 ω (1 − a)a' dr
By the blade element momentum method,
the flow area swept by the rotor is divided At this point it is necessary to make an ini-
into a number of concentric ring elements. tial choice for a and a’. It is suggested to
These elements are considered separately assume a=a’=0 as an initial guess.
under the assumption that there is no radial
dependency between them, i.e., what hap- The flow angle φ is the angle between the
pens at one element cannot be felt by any rotor plane and the direction of the relative
other elements. Usually, each ring element is wind velocity Vrel on the rotating blade. The
divided into a number of tubes, which are flow angle can be calculated from
also assumed to be independent. This ap-
proach allows for nonsymmetric induction. (1 − a)V 0
The wind speed is assumed uniformly dis- tan φ =
(1 + a' )ωr
tributed over each ring element. The forces
from the blades on the flow through each
ring element are assumed constant. This The local inflow angle is α=φ−θ, where the
assumption corresponds to assuming that the pitch angle θ is the local pitch of the blade
rotor has an infinite number of blades. This relative to the rotor plane. See Figure 4-8.
assumption is corrected for later.
Determine the lift and drag coefficients CL
Consider now the ring element of radius r and CD for the blade from adequate tables.
and thickness dr. The thrust on this element They are functions of α and of the blade
from the disc defined by the rotor is thickness relative to the chord length. Trans-
form these coefficients to normal and tan-
dT = 2πrρu (V 0 − u1 )dr gential coefficients CN and CT by
CN=CLcosφ+CDsinφ
in which V0 is the wind speed before the
rotor and u1 is the wind speed in the wake
behind the rotor, and u=½(V0+u1) is the CT=CLsinφ−CDcosφ
wind speed through the rotor plane. The
torque on the ring element is The solidity σ is defined as the fraction of
the cross-sectional area of the annular ele-
ment which is covered by the blades. The
dQ = 2πr 2 ρuCθ dr
solidity depends on the radius r of the an-
nular element and can be found as
when the tangential wind speed at radius r is
zero upstream of the rotor and uw in the c(r ) B
wake. Introducing the axial induction factor σ (r ) =
2πr
a=1−u/V0 and the tangential induction factor
a’=½uw/(ωr), where ω denotes the angular
in which B denotes the number of blades.
velocity of the rotor, then the expressions for
Calculate the influence factors a and a’ by
the thrust and the torque can be rewritten as
64 4 – Loads
1 profile close to the blade tip. Recalling the

a= assumption of an infinite number of blades
æ 4 F sin φ ö
2
ç + 1÷÷ sweeping the rotor area, Prandtl’s tip loss

ç σC
è N ø factor – by its expression – also implies a
correction for the actual finite number of
and blades of the rotor.
1 Examine these values of a and a’. If they

a' = deviate significantly, i.e. by more than some
æ 4 F sin φ cos φ ö
çç − 1÷÷ acceptable tolerance, from the values of a
è σC T ø and a’ assumed for the calculations, go back
to where the values for a and a’ were ini-
tially assumed and use the new values of a
and a’ to recalculate the flow angle φ. Re-
peat the calculation procedure from there on
to obtain a new updated set of a and a’. This
iterative procedure needs to be repeated until
a convergent set of values for a and a’ re-
sults. Note that simple momentum theory
breaks down when a becomes greater than
about 0.3. Formulas exist for correction of a
when this situation occurs. Whenever a> ac,
where ac≈0.2, Glauert’s correction can be
applied, and implies that a is replaced by
1
a= (2 + K (1 − 2a c )
2
− ( K (1 − 2a c ) + 2) 2 + 4( Ka c − 1) )
2
in which
4 F sin 2 φ
K=
Figure 4-8 Velocity components
σC N
in which Note that it is of particular importance to

apply Glauert’s correction in order to com-
2 B R−r pute the induced velocities correctly for
F= arccos(exp(− )) small wind speeds.
π 2 r sin φ
Once a convergent set of a and a’ has been
where R is the rotor radius. Note that F is determined, it can be used to calculate the
known as Prandtl’s tip loss factor. This is a local forces on a rotor blade at distance r
reduction factor, which corrects for the finite from the axis of rotation, i.e., the force nor-
length of the blades that implies that it is mal to the rotor plane and the force tangen-
difficult to fully keep the pressure difference tial to the rotor plane. The normal force per
between the top and bottom of the blade length unit of the blade is
4 – Loads 65
1 V (1 − a)
2 2 Moments of stiffness inertia about the vari-
FN = ρ 0 2 cC N ous axes defined in Figure 4-9 can be cal-
2 sin φ
culated by means of standard formulas,
which can be found in structural engineering
and the tangential force per length unit of textbooks. The angle α between the refer-
the blade is ence axis X’ and the principal axis X can be
calculated as
1 V0 (1 − a)ωr (1 + a' )
FT = ρ cC T
2 sin φ cos φ 1 2[ED X 'Y ' ]
α= arctan
2 [EI Y ' ] − [EI X ' ]
The procedure is repeated for all ring ele-
ments modelled, i.e., for all radius values r,
and the result thus consists of distributions
along the rotor blade of the normal and tan-
gential forces per unit length. These distri-
butions form the basis for calculating
stresses, forces and moments in any cross
section along the blade. In particular, rFT
can be integrated along the blade to give the
contribution from one blade to the total shaft
torque.
Structural modelling
Rotor blades are slender such that they from
a structural point of view will act like
beams, and beam theory can be applied.
Reference is made to Section 5.1. For analy- Figure 4-9 Section of blade showing
sis of a rotor blade by means of beam the- principal axes, chord line, tip chord line,
ory, a number of definitions relating to the and angles to chord line
blade profile are useful. Consider a section
of the blade. The elastic axis is perpendicu- where
lar to the section and intersects the section in
a point where a normal force (out of the [ED ] = ò EX ' Y ' dA
X 'Y '
is the deviation mo-
plane of the section) will not give rise to A
bending. The shear center is the point where ment of inertia, and [EI X ' ] = ò EX ' 2 dA and
an in-plane force will not rotate the profile A
in the plane of section. The two in-plane [EI ] = ò EY '

Y'
2
dA are the bending stiff-
principal axes are mutually perpendicular A
and both cross the elastic axis. The principal nesses about the X’ and Y’ references axes,
axes are defined by the phenomenon that respectively. The bending stiffnesses about
whenever a bending moment is applied the principal axes can now be computed as
about one of them, the beam will only bend
about this axis. Applying a bending moment [EI ] = [EI ] − [ED ] tan α
X X' X 'Y '
about any other axis will induce bending
also about another axis than the one corre- and
sponding to the applied moment.
66 4 – Loads
[EI ] = [EI ] − [ED ] tan α

Y Y X 'Y '
angle β and of the tip speed ratio λ=ωRR/u.
ωR is the angular frequency of the rotor, and
respectively. R is the rotor radius.
The most important structural stiffness data The pitch angle β and the rotor speed ωR are
for a rotor blade have been dealt with here. the two parameters which can be used for
Note that since present wind turbine blades the control of the turbine. The control of the
are usually relatively stiff in torsion, the turbine is usually based on one of the fol-
torsional stiffness has usually not been con- lowing two approaches:
sidered. Note also that this may change for • Optimization of the power below the
future wind turbine designs. nominal power, i.e., the rotor is kept as
close as possible to the maximum value
In structural modelling and analysis, it is of CP, which occurs for a particular op-
important to be aware of the flutter phe- timal set (βopt, λopt) of β and λ. This is
nomenon, which may result from coupled achieved by choosing β equal to its op-
torsional and flapping motion. Low ratios of timal value βopt, and keeping λ constant
torsional and flapwise frequencies and high at its optimal value λopt. However, λ
tip speeds indicate rise of flutter, which may cannot be controlled directly, since u is
be destructive for the rotor blade. Reference hard to measure with sufficient accu-
is made to Section 4.3.6. racy. Therefore, in stead, information
about which power production is opti-
Control system modelling mal for a given value of ωR is utilized,
The control system is to keep the operating and ωR is used as the parameter to con-
parameters of the wind turbine within speci- trol the turbine and optimize the power
fied ”normal” limits. This, in turn, is to keep rather than λ. This information can be
the loads on the wind turbine within certain established from the aerodynamic data
limits. Because the operating parameters are for the rotor, first of all profile data, and
usually controlled by monitoring their cur- computer software is available for this
rent values and/or their first or second de- purpose.
rivatives, a regulation algorithm can always
• Limitation of the power by keeping it as
be set up and coded for use together with an
close as possible to the nominal power.
aeroelastic code for load prediction.
The aerodynamic driving torque varies con-
For a pitch-regulated turbine with variable
tinuously due to the turbulence of the wind.
speed, a method exists for representation of
This moment is transferred to electric power
the control system. The description in the
by the transmission system and is used as a
following pertains to the control system of a
primary indicator for the loading on the
wind turbine with variable speed.
transmission system. By the pitch- and
speed-regulation of the turbine, changes in
The mechanical effect of the wind turbine is
the aerodynamic power are absorbed as
changes in the angular velocity of the rotor
1 instead of inducing changes in the torque
P= ρAu 3 C P ( β , λ )
2 which is transferred to the gearbox. This
implies that relative to a conventional fixed-
in which ρ is the density of air, A is the rotor speed turbine, a lower level of gearbox
area, u is the wind speed, and CP is the rotor forces can be kept.
efficiency, which is a function of the pitch
4 – Loads 67
4.3.3 Aeroelastic models for load pre- 4.3.4 Aerodynamic data assessment
diction
Choice of aerodynamic coefficients for use
Aeroelasticity is a discipline where mutual as input to any aerodynamic analysis method
interaction between aerodynamic and elastic requires careful consideration. In this con-
deflections is investigated. Aeroelastic mod- text, it is important to give special attention
els for load prediction are usually based on to which regulation strategy is adopted for
the blade element momentum method for the wind turbine, which is subject to design.
transformation of the wind flow field to
loads on the wind turbine structure. When the turbine is pitch-controlled, the
major task is to program as correctly as pos-
Two important and commonly applied sible the regulation routine to be used with
methods used for discretisation in connec- the aeroelastic computer code. The aerody-
tion with structural modelling of wind tur- namic coefficients are fairly easy to obtain
bines are the finite element method (FEM) when the turbine is pitch-controlled, since
and the modal analysis method (or methods the turbine will only operate in the linear
strongly related to modal analysis). Both part of the lift curve and three-dimensional
methods are implemeted in computer codes effects are of little importance. However, the
for load prediction. The computer code initial part of the lift curve in the stall region
HawC is an example of an aeroelastic code should still be correctly modelled, because
based on the finite element method. The no pitch regulation mechanism can be con-
computer code FLEX4 is an example of an sidered so perfect that the blades will never
aeroelastic code based mainly on methods experience a stalled condition.
related to modal analysis.
When the turbine is stall-regulated or active-
A number of computer programs for aero- stall-regulated, an appropriate choice of the
elastic analysis with load and deformation aerodynamic coefficients is more difficult to
prediction for wind turbines exist. Some of achieve. A first step in deriving representa-
these programs are commercially available tive values for the aerodynamic coefficients
and can be purchased, while others are de- consists of identifying the profile series used
veloped by wind turbine manufacturers and for the rotor blade. When the profiles are
are not available to the public. Most of the well known, it is easy to find two-
programs provide solutions in the time do- dimensional wind tunnel data to support the
main. However, a few programs exist which choice of coefficient values. Data for so-
offer solutions in the frequency domain. called NACA profiles can readily be found
in Abbott and von Doenhoff (1959). When
Regardless of which aeroelastic code is ap- the profiles are not well known, it is recom-
plied to prediction of wind turbine loads, it mended to inspect a visualization of the pro-
is essential that it be validated. The IEC re- file shape and find a representative and well
quires that the load model used to predict known profile series for which measured
loads for design verification of wind tur- aerodynamic two-dimensional data exist.
bines be validated for each design load case.
This validation of the load prediction is to be The next step is to assign values to the aero-
based on a representative comparison be- dynamic coefficients with due consideration
tween measured and predicted loads on a of possible three-dimensional effects. It is
similar wind turbine. The comparison and recommended to evaluate or calibrate the
the subsequent calibration shall include both aerodynamic coefficients from power curve
the peak loads and the fatigue loads. data or thrust curve data. A representative
68 4 – Loads
thrust curve can easily be derived from the these high angles. A method for such ex-
tower bottom bending. When reliable meas- trapolation can be found in Eggleston and
ured data are available, this will provide a Stoddard (1987). Alternatively, the method
good basis for deriving ”correct” data. by Viterna and Corigan (1981), described in
Section 4.3.2, can be used.
In most cases, measurements are not avail-
able. The assignment of values to the aero- Note that proper selection of values for the
dynamic coefficients must be based on a aerodynamic coefficients is a very important
general impression and on experience. Al- step in the design analyses for a wind tur-
ternatively, it can be considered to apply bine, since a reliable prediction of the dy-
guidelines for three-dimensional corrections namic response of the wind turbine is very
of two-dimensional data according to Bak et dependent on correct choices of aerody-
al. (1999b). These guidelines can be summa- namic coefficients.
rized as follows
• the lift coefficients are lower at the 4.3.5 Special considerations
blade tip in the stall region when 3-D
effects are included Structural damping
• the lift coefficients are unchanged at a In order to achieve a realistic response from
distance approximately two thirds of the an analysis by an aeroelastic code, specifi-
rotor radius from the rotor axis cation of the structural damping must be
made with caution. The structural damping
• the lift coefficients are higher on the
model is included in the equations of motion
inner part of the blade, i.e. closer to the
in order to assure dissipation of energy from
rotor axis, when 3-D effects are in-
the structural system. Quite often, measure-
cluded
ments of the structural damping is only ob-
• the drag coefficients are unchanged on
tained for a limited number of mode shapes,
the outer part of the blade
if available at all.
• the drag coefficients are slightly lower
on the inner part of the blade for inflow Table 4-3: Modal damping in logarithmic
angles up to approximately 20° when 3- decrement for a 19m blade.
D effects are included. For higher in-
flow angles, the drag coefficients are Mode Freq. Damp. [% log.]
higher than indicated by 2-D data. shape [Hz] Mean Std.dev.
1st flap 1.636 1.782 0.080
Beware that these guidelines are based on 2nd flap 4.914 2.021 0.011
calibration of computational results for one 3rd flap 9.734 2.468 0.026
particular blade, viz. the LM19.1 blade, and 4th flap 16.22 3.227 0.033
cannot necessarily be projected to apply to 1st edge 2.943 3.603 0.011
other blades without validation. 2nd edge 10.62 5.571 0.041
1st torsion 23.16 5.807 0.062
It is usually not possible to find aerody-
namic data for inflow angle outside the Experience shows that the structural damp-
range –20°-20°. As high inflow angles out- ing in terms of the logarithmic decrement is
side this range do occur on wind turbines, in usually of the order of 3% for the blades and
particular during extreme conditions formed of 5% for the shaft and the tower. An exam-
by extreme yaw errors or extreme wind ple of logarithmic decrements found for a
speeds, it is often necessary to extrapolate 19m rotor blade is shown in Table 4-3.
the available aerodynamic data to values for (Baumgart et al, 2000.)
4 – Loads 69
In order to assure dissipation of energy, it is where C, M, K is the damping, mass and

required that the part of the damping matrix stiffness matrices, respectively, p is the
C in the equations of motion that originates number of damping ratios, and ak are mod-
from structural damping is positive definite elling constants
or at least semi-positive definite
The modelling constants ak are determined
x Cx ≥ 0
T from the p simultaneous equations
Positive definiteness of the structural dam- 1 é a0 3 2 p −3 ù

ing matrices ensures dissipation of energy at
ξi = ê + a1ω i + a 2ω i + L + a p −1ω i ú
2 ëω i û
all velocities.
A commonly used model for representation i=1,p, in which ξi is the ith damping ratio
of damping is the Rayleigh damping model, and ωi is the angular frequency for the vi-
whose major advantage is a decoupling of bration mode that ξi was obtained for.
the equations of motion whenever a modal
formulation is used. The damping model has The disadvantage of using this generalized
the form damping model is that the damping matrix C
– in the general case – is a full matrix. This
C =α M + β K may cause the simulations in an aeroelastic
analysis to become rather timeconsuming.
For p=2 this model reduces to a Rayleigh
where C, M, K is the damping, mass and
damping model.
stiffness matrices, respectively, and α and β
are model constants
For appreciation of the results presented
herein, recall that the ith damping ratio is
Reference is made to Bathe (1982). The
defined as
Rayleigh damping model enables an accu-
rate fit of two measured damping ratios
c
only. Here, α and β are determined from ξi =
two damping ratios that correspond to une- 2 mi k i
qual frequencies of vibration.
in which mi is the ith mass, ci is the ith
A disadvantage of the use of the Rayleigh damping, and ki is the ith stiffness. The cor-
model is that it is known to overpredict the responding logarithmic decrement is
damping at high frequencies of vibration.
2πξ i
In a more general damping model it is as- δi = ≈ 2πξ i
1−ξi
2
sumed that p damping ratios have been de-
termined. Based on this, the damping matrix
is represented by a Caughey series: The logarithmic decrement is the natural
logarithm of the amplitude decay ratio, i.e.
k
the natural logarithm of the ratio of the am-
å [ ]
p −1
C=M a k M −1 K plitudes in two consecutive displacement
k =0 cycles.
70 4 – Loads
Verification of the structural damping

When some initial guesses of the damping
coefficients have been chosen for the aero- When the damping model is a Rayleigh
elastic model, it must be verified that the damping model, then it is only possible to fit
model leads to a correct representation of the correct damping properties at two fre-
the damping. This is done by fixing all de- quencies. If the damping properties are
grees of freedom of the turbine, except the known at more frequencies, such as in the
one for which the damping model is to be example in Table 4-5, then one is restricted
checked. An external periodic force is ap- to select some average representation of the
plied to the structural component in question damping. In this context it is important to
(e.g. a blade) using the natural frequency of aware of the overprediction of the damping
that component as the excitation frequency. by the Rayleigh model at high frequencies,
The structure is only excited for a few sec- e.g., when analysing high-frequency impulse
onds. The subsequent decay of the induced forces on fibre-glass materials.
vibration is measured, and the damping co-
efficient is easily determined from a curve Stall-induced vibrations of blades
fit of the equation Examples exist of wind turbines, which have
developed significant edgewise vibrations of
the blades, even during relatively moderate
wind situations such as those characterized
f ( x) = C ⋅ exp(−δ ⋅ f n ⋅ t ) by 10-minute mean wind speeds of about 15
m/sec. Such edgewise vibrations are unde-
in which t denotes time, fn is the excitation sirable from a structural point of view and
frequency, δ is the logarithmic decrement, need to be avoided. Edgewise vibrations
and C is a calibration constant from the fit. occur whenever the negative aerodynamic
An example of such a fit is shown in Figure damping exceeds the structural damping.
4-10. The dynamic coefficients in the model The aerodynamic forces supply more energy
must be adjusted until the correct logarith- to the vibrations than the structural damping
mic decrement δ results from this exercise. can absorb. An example of such vibrations,
simulated for a 500 kW turbine, are shown
in Figure 4-11 and Figure 4-12. Figure 4-11
shows the vibrations for the basic configu-
ration of the turbine, whereas Figure 4-12
shows the vibrations for the new configura-
tion, with increased shaft stiffness. Both
figures are extracted from Petersen et al.
(1998). Note that flapwise vibrations are
equally important to consider.
Recommendations concerning the design of
new blades and modifying old blades, when
edgewise vibrations pose a problem, are
Figure 4-10 Damping test of rotor blade in given by Petersen et al. (1998) and repro-
edgewise vibration duced in the following.
4 – Loads 71
Figure 4-11 Edgewise blade root moment for the Figure 4-12 Edgewise blade root moment for the
500 kW turbine in basic configuration. 500 kW turbine with an increased shaft stiffness.
5. When the single blade analysis results

For a new turbine design, where a great deal in appropriate aerodynamic damping,
of freedom is at hand, the following proce- continue the analysis by use of the full
dure is recommended: aeroelastic model, including verified
models for stall hysteresis on both air-
1. Base the airfoil on well documented foil lift and drag. The structural models
airfoil data. must have a sufficiently detailed repre-
2. Choose airfoils, blade planform (chord, sentation of the blade bending modes
aerodynamic twist) and blade structural and the rotor whirling modes.
properties (stiffness, mass, mode shapes 6. Carry out full aeroelastic calculations
– i.e. amplitude and direction of defor- for the wind speeds of interest. These
mation, structural damping, modal wind speeds can to a wide extent be
mass, modal stiffness and modal natural identified through a single blade analy-
frequency) considering not only the ro- sis.
tor performance, but also the aerody- 7. Investigate the damping characteristics
namic damping characteristics. Espe- by looking at the responses at the blade
cially, the blade twist should be consid- natural flapwise and edgewise frequen-
ered, not only in relation to power per- cies, for instance based on the power
formance, but also in relation to mode spectral densities of the blade root
shape vibration direction. bending moments.
3. Use a quasi-steady approach (Petersen 8. Repeat from Step 2 if the results are
et al., 1998) to calculate the basic aero- unsatisfactory.
dynamic damping characteristics for a
single blade, considering both the fun- When problems with flapwise and edgewise
damental flapwise and edgewise mode vibrations have been recorded on a particu-
shapes. If not satisfactory, repeat from lar turbine, then the turbine must be re-
Step 2. designed. In this case, only a limited number
4. Extend the quasi-steady analysis and of possibilities remain open:
include verified models for dynamic
stall on both the airfoil lift and drag, 1. In general, the scheme above should be
still considering a single blade. If not applied to the existing design.
satisfactory repeat from Step 2. 2. The aerodynamic properties of the
blades may be modified by using differ-
72 4 – Loads
ent aerodynamic devices, e.g. stall strips and possibly lead to its failure. Flutter vi-
and/or vortex generators. brations consist of a coupling of flapwise
3. The structural properties of the turbine and torsional blade vibrations, which are
may be changed by adding local masses sustained by the airflow around the blade.
or increasing local stiffnesses. These There are several undesirable conditions that
changes may change the relevant mode must exist for flutter to occur. The two most
shapes and natural frequencies in a fa- critical conditions occur when the frequen-
vourable manner. cies of a flapwise bending mode and the first
4. Mechanical damping devices may be torsional mode are not sufficiently sepa-
used either on the blades or on the sup- rated, and when the centre of mass for the
porting structure. It should be men- blade cross-sections is positioned aft of the
tioned that these devices do not remove aerodynamic center, which is approximately
the cause of the vibrations, which is the at the quarter chord point, which is marked
supply of energy by the aerodynamic in Figure 4-4 at a distance “c/4” from the
forces. leading edge. In the design of blades, it is
therefore important to ensure a high tor-
Figure 4-13 Airfoil lift data (unmodified). Figure 4-14 Airfoil drag data (unmodified).
Figure 4-16 Airfoil data with vortex gen-

Figure 4-15 Airfoil data with stall strips. erators.
Flutter sional stiffness (e.g. first torsional frequency

Flutter is an aeroelastic instability that may larger than ten times the first flapwise fre-
lead to large-amplitude vibration of a blade, quency), and – if applicable – to ensure that
4 – Loads 73
the centre of mass for the blade cross-

sections is located between the leading edge α (t ) = α 0 + β ⋅ sin(ω ⋅ t )
and the quarter chord point.
in which α0 is the mean inflow angle, β is
Use of stall strips and vortex generators the amplitude, t is the time, and ω is the an-
Stall strips can be used on the outer section
gular frequency.
(typically the far one-third) of a rotor blade
as an ‘emergency’ action when problems
The phenomenon of delayed onset of stall is
with heavy vibrations of the blades occur. denoted dynamic stall, and it influences the
Stall strips in general increase the damping lift and drag coefficients. It is important to
on local sections of the blade by increasing
account for dynamic stall when it occurs,
the drag after the onset of stall.
since lift and drag coefficients are usually
derived from wind tunnel tests with constant
Vortex generators are often used to increase
inflow angle. Various models for dynamic
the lift and thus to improve the aerodynamic stall can be found in Petersen et al. (1998).
properties of a blade. The vortex generators
are used on the inner section of the blade
An example of dynamic stall measured in a
and improve the maximum lift of the sec-
wind tunnel for β=2° is given in Figure 4-
tion. An example of an airfoil with stall strip
17.
and vortex generator modifications is shown
in Figures 4-13 through 4-16, (Petersen et
al., 1998).
Dynamic stall
The onset of stall becomes delayed when the
inflow angle is varying periodically, e.g.
according to
Figure 4-17 CL and CD loops, dynamic stall
74 4 – Loads
4.4 Load analysis and synthesis nth cycle. Consider now n cycles at constant
stress range S. Under certain assumptions
4.4.1 Fatigue loads
the damage indicator is not an explicit func-
Introduction tion of the actual stress range S, but only of
Fatigue loads are cyclic loads, which cause the ratio n/N, where N is the number of cy-
cumulative damage in the materials of the cles to failure, and where N is a function of
structural components, and which eventually S. It can then be shown that the cumulative
lead to structural failure. Failure takes place damage indicator can be expressed as the so-
by the initiation and propagation of a crack called Miner’s sum
until the crack becomes unstable and propa-
gates fast, if not suddenly, to failure. Fatigue ni
loads are usually loads well below the load D=å
i N (S i )
level that will cause static failure, and many
load cycles are required before a fatigue
in which ni is the number of stress cycles
failure will take place. This is commonly
with stress range Si, and N(Si) is the number
referred to as high-cycle fatigue. However,
of cycles to failure at stress level Si. The
for some materials with particularly high
sum is over all stress ranges Si in a suitable
S−N curve slopes, such as some epoxy mate-
discretization of the stress range axis. It thus
rials, the loads of importance for fatigue are
appears that prediction of the accumulated
close to those that will cause static failure.
damage D requires knowledge of the distri-
For such materials fatigue becomes an ex-
bution of the stress ranges. This is dealt with
treme value problem as far as the loads are
in the following subsections. It also appears
concerned with only a few load cycles re-
that the dependency of N on S is required.
quired to cause fatigue failure. This is com-
The most commonly used relation between
monly referred to as low-cycle fatigue.
N and S is the S−N curve model N=KS−m, in
which K is a constant and m is a slope pa-
It is common for many damage accumula-
rameter. This model can be calibrated from
tion laws to deal with a simplified measure
laboratory test data of the material. The fail-
of damage in terms of a scalar damage indi-
ure criterion by the Miner’s sum is D=1.
cator D. The value of D is taken as zero in
the initial state and as one at failure. The
damage accumulation is determined entirely Note that the Miner’s sum formulation for
the cumulative damage used with the quoted
by the stress variation, and stresses are as-
sumed to vary in cycles with stress ranges Si. S−N curve model disregards a possible de-
The damage indicator is a non-decreasing pendency on the average stress level. This
function of the number of stress cycles. The can be remedied by a so-called Goodman
formulation, e.g., by making the coefficient
damage increment ∆Dn in the nth stress cy-
cle in general depends on the damages ac- K dependent on the mean stress. However,
cumulated at the end of each of the preced- there is usually not enough data available to
calibrate such a refined, mean-stress-
ing n−1 stress cycles, and on the stress range
dependent model.
of the nth stress cycle. Under the assumption
that the damage accumulated after n cycles
Rain-flow counting
is independent of the order in which the
Cycle counting methods are used to estab-
stress cycles occur, ∆Dn will only depend on lish distributions of stress ranges from a
the accumulated damage at the beginning of stress history. Several methods of cycle
the nth cycle (i.e., at the end of the n−1 pre- counting exist, for example
ceding cycles) and on the stress range of the
4 – Loads 75
Figure 4-18 Illustration of the rain-flow counting method: (a), (b) application to a stress history, from
Wirsching and Shehata (1977); (c) application to a low-cycle strain history, from Madsen et al. (1986)
• peak counting For simplicity, the mean level is taken as

• range counting zero in the description.
• rain-flow counting
These three counting methods give the same In the rain-flow counting method, the stress
result for a pure sinusoidal stress history and history is first converted into a series of
for an ideal narrow-band stress history. peaks and troughs as shown in Figure 4-18
Rain-flow counting will be considered here with the peaks evenly numbered. The time
and is described for stationary processes. axis is oriented vertically with the positive
direction downward. The time series is then
76 4 – Loads
viewed as a sequence of roofs with rain fal- stant-amplitude data in order to calculate the
ling on them. The rain-flow paths are de- accumulated damage.
fined according to the following rules
(Wirsching and Shehata, 1977): In the rain-flow method small reversals are
treated as interruptions of the larger ranges
1. A rain flow is started at each peak and and the rain-flow method also identifies a
trough. mean stress for each stress cycle. A com-
2. When a rain-flow path started at a parison between test results and results pre-
trough comes to the tip of the roof, the dicted by the three mentioned cycle counting
flow stops if the opposite trough is more methods shows that the rain-flow counting
negative than that at the start of the path method generally gives the best results. This
under consideration (e.g., path [1-8], method is the only one of the three that
path [9-10], etc.). A path started at a identifies both slowly varying stress cycles
peak is stopped by a peak which is more and more rapid stress reversals on top of
positive than that at the start of the rain these. The peak counting method will in
path (e.g., path [2-3], path [4-5], path general assign larger probabilities to larger
[6-7], etc.). stress ranges, while the range counting
3. If the rain flowing down a roof inter- method will assign larger probabilities to
cepts flow from a previous path, the smaller stress ranges. Compared to the rain-
present path is stopped (e.g., path [3- flow counting method, the peak counting
3a], path [5-5a], etc.). will therefore result in larger estimates for
4. A new path is not started until the path the accumulated damage, while the range
under consideration is stopped. counting method will predict smaller values
of the damage. Analytical results for the
Half-cycles of trough-originated stress range stress range distribution obtained through
magnitudes Si are projected distances on the rain-flow counting are very difficult to ob-
stress axis (e.g., [1-8], [3-3a], [5-5a], etc.). It tain and the method is generally used only
should be noted that for time series suffi- with measured or simulated stress histories.
ciently long, any trough-originated half-
cycle will be followed by another peak- The rain-flow method identifies a mean
originated half-cycle of the same range. This stress for each stress cycle. An attractive
is also the case for short stress histories if representation of the resulting stress range
the stress history starts and ends at the same distribution from cycle counting by the rain-
stress value. flow method is therefore to form a matrix
with one row for each mean stress level and
The rain-flow method is not restricted to one column for each stress range. Each ele-
high-cycle fatigue but can also be used for ment of this matrix will then contain the
low-cycle fatigue where strain range is the number of stress cycles associated with a
important parameter. Figure 4-18 shows a particular stress range and a particular mean
simple example. In this sequence four events stress. Each row of the matrix will contain a
that resemble constant-amplitude cycling are discretized stress range distribution condi-
recognized, 1-6-9, 2-3-3a, 4-5-6, and 7-8-8a. tioned on a particular mean stress. When
These events are closed hysteresis loops and S−N curves are available for various ratios R
each event is associated with a strain range between the compressive stress amplitude
and a mean strain. Each closed hysteresis and the tensile stress amplitude, this repre-
loop can therefore be compared with con- sentation will allow for prediction of the
partial damage in each element of the matrix
4 – Loads 77
by applying the appropriate S−N curve for • stop at cut-in and stop at cut-out
that element. The total fatigue damage D can • idling and standstill
subsequently be determined by summing up • yaw misalignment
the partial damages over all elements in the Start and stop are transient conditions, for
matrix, which the stress distributions in the consid-
ered structural component are not easily
n ij ( S j ) determined. During production, stationary
D = åå conditions can be assumed to prevail in the
i j N ij ( S j )
short term, e.g. during 10-minute periods.
During a 10-minute period, the wind climate
in which nij denotes the number of stress parameters such as the 10-minute mean
cycles in the matrix element corresponding wind speed U10 and the turbulence intensity
to the jth stress range, Sj, and the ith mean IT at the hub height can be assumed to be
stress, and Nij is the number of stress cycles constant. During such short periods of sta-
to failure in this element. Reference is made tionary conditions, the load response proc-
to Figure 4-19. esses that give rise to the stress ranges in the
considered structural component can be
Stress range taken as stationary processes. Under station-
ary conditions, stress ranges are often seen
(col. j)
to have distributions which are equal to or

Mean stress
close to a Weibull distribution,
(row i) nij FS(s)=1−exp(−(s/SA)B)
Note that when the exponent B equals 1, this

distribution turns into an exponential distri-
Figure 4-19 Matrix representation of rain- bution, and when it equals 2 it becomes a
flow counted stress range distribution Rayleigh distribution. For representation of
stress range distributions, which are not
This approach is useful for damage predic- quite Weibull distributions, it will often suf-
tions for structural materials, such as blade fice to apply a somewhat distorted Weibull
materials, for which a non-zero mean stress distribution. The simplest distribution model
is of importance. For such materials, the among this family of distributions is a three-
dependency on the non-zero mean stress will parameter Weibull distribution,
not be adequately accounted for by a
Miner’s sum model, which is based on only FS(s)=1−exp(−((s−a)/SA)B)
one single S−N curve associated with full
stress reversals about a zero mean stress. Other models for moderate distortions of a
parent Weibull distribution to form a para-
Stress range distributions metric distribution model, which fits data
To establish the stress range distributions for well, exist (Ronold et al., 1999).
a structural component in a wind turbine, it
is important to consider the various condi- The coefficients a, B and SA are distribution
tions that the turbine can be in. There are parameters, and they can often be expressed
various operational conditions, including as functions of the wind climate parameters
• production U10 and IT. When the short-term stress range
• start at cut-in and start at cut-out distributions, conditional on U10 and IT have
78 4 – Loads
been established, when the long-term distri- cycle is a repetitive period of operation,
butions of U10 and IT are known as outlined which is characterised by a typical succes-
in Section 3.1, and when the total number of sion and duration of different modes of op-
stress cycles during the production life of eration. The duty cycle can be specified by a
the turbine is assessed, then the compound sequence of parameters, which define the
distribution of all stress ranges during this mode of operation for consecutive 10-
production life can be established. This minute periods for a representative time
compound distribution can itself often be span. Information about the duty cycle
represented by a Weibull distribution, and it should as a minimum contain the number of
can be represented as shown in Figure 4-20, starts and stops during the time span, which
where n denotes the number of stress cycles the duty cycle covers, together with the wind
which exceed a stress range S during the climate parameters in this time span.
production life. This distribution forms a
significant contribution to the loads that Starting the wind turbine is considered to be
cause fatigue damage. Note that if the curve potentially critical with respect to fatigue.
in the semi-logarithmic diagram in Figure 4- One reason for this is that connection of an
19 had been a straight line, the distribution electric generator to the grid can cause high
would have been an exponential distribution, transient loads in the drive train and the ro-
which is a special case of the Weibull distri- tor blades. It is recommended to distinguish
bution. between start at the cut-in wind speed and
start slightly below the cut-out wind speed.
For stopping, it is recommended to distin-

250
guish between stopping slightly below the
cut-in wind speed and stopping at the cut-
200 out wind speed, as the latter is associated
150 with rather large aerodynamic loads.
S
100
For standstill and idling below the cut-in
50 wind speed, loads are considered insignifi-
cant with respect to fatigue.
0
0 2 4 6 8 10
Yaw misalignment can be critical with re-
log10n
spect to fatigue.
Figure 4-20 Example of compound stress
distribution, a socalled load spectrum Note that stresses in a particular location of
the wind turbine structure, considered criti-
To obtain the distribution of all fatigue loads cal with respect to fatigue, may be due to a
in the design life, this compound distribution combination of stresses arising from differ-
has to be supplemented by the stress ranges ent sources or load processes. When the
owing to start, stop, standstill, idling, and stress levels are sufficiently low, elastic
yaw misalignment, to the extent that these material behavior can be assumed for the
are considered to contribute to the cumula- combination of the stresses before fatigue
tive fatigue damage. damage calculations are carried out. It can
further – somewhat conservatively – be as-
For this purpose, information about the duty sumed that load peaks of two load processes
cycle of the turbine is essential. The duty occur simultaneously, and that high-
4 – Loads 79
amplitude cycles can be combined with purpose, and one estimate of the equivalent
high-amplitude cycles. Mean values of two load range can be obtained from each simu-
load processes should be combined in the lated time series. For a given wind climate
most unfavourable manner to produce the (U10,IT), the uncertainty in the estimated
largest mean stress value. For more details equivalent load range can be expressed in
on how to combine load spectra, see IEA terms of the coefficient of variance (COV)
(1990). of the estimate. COV can be expected to be
proportional to 1 / NT , where N is the
Equivalent loads
number of simulated time series for the
Once the load spectrum has been established
given wind climate, and T is the duration of
with contributions from all operational
the simulated time series, e.g. 10 minutes.
modes over the design life of the wind tur-
bine, it is often convenient to define a so-
Consider now lifetime load spectra and cor-
called damage-equivalent load range S0 to be
responding equivalent life time load ranges.
used with an equivalent number of cycles
Time series of load response are simulated
neq. This is a constant load range S0, which
for various wind climates. Compound life-
in neq cycles will lead to the same accumu-
time load spectra are established on this ba-
lated damage as the true load spectrum that
sis by appropriate weighting according to
consists of many different load ranges Si and
the long-term distribution of the wind cli-
their corresponding cycle numbers ni. When
mate parameters. A total of n simulated life-
the equivalent number of cycles neq is cho-
time load spectra are established this way,
sen or specified, then the equivalent load
and one equivalent load range S0i is inter-
range S0 can be found as
preted from each such simulated lifetime
1m
load spectrum. The central estimate of the
æ å n i S im ö equivalent lifetime load range is taken as the
ç ÷
S0 = ç i ÷ arithmetic mean
ç n eq ÷
è ø
1 n
S0 = å S 0i
n i =1
in which m denotes the S−N curve slope of
the material in question. This definition of a
damage-equivalent load range is well known The standard deviation of the S0i’s is esti-
in fatigue analysis and is often used in wind mated by
turbine load analysis. An example of appli-
cation can be found in Stiesdal (1991). Note 1 n
that the equivalent load concept only applies s= å ( S 0i − S 0 ) 2
n − 1 i =1
to materials whose S−N curves are described
by one slope m, i.e. it can not be used for A Gaussian assumption can usually be made
materials with bilinear S−N curves. for the simulated equivalent load ranges S0i.
On this background, a two-sided confidence
Uncertainties interval for the estimated equivalent load
Load spectra are usually not known with range with confidence 1−α can be estab-
certainty, but are predicted, e.g. from a lim- lished as
ited number of time series of load response,
obtained by simulation according to some
s
aeroelastic analysis scheme. Time series of S0 ± t α
1− , n −1
10-minute duration are usually used for this n 2
80 4 – Loads
in which t1−α/2,n−1 is the 1−α/2 quantile in the cut-out. The loading during a duty cycle can
Student’s t distribution with n−1 degrees of also be random in nature. The state of dam-
freedom. Table 4-4 tabulates these quantiles age is considered only at the end of the duty
for selected degree-of-freedom values for a cycle. The probability distribution of dam-
couple of common choices for the confi- age after a duty cycle is assumed to depend,
dence 1−α. Quantiles of the Student’s t dis- in a probabilistic manner, only on the duty
tribution can in general be found in most cycle itself and on the damage accumulated
published tables of statistical distributions. at the start of the duty cycle. It is, however,
independent of how the damage was accu-
Table 4-4 Quantiles of mulated up to the start of the duty cycle.
the Student’s t distribution, These assumptions are the Markov assump-
t α tions and the damage process is viewed as a
1−
2
, n −1 discrete-time, discrete-state Markov process.
Degrees of Confidence 1−α The time t is measured in units of duty cy-
freedom, n−1 0.90 0.95 cles, hence t=1,2,.... Equal increments in t
1 6.31 12.71 need not correspond to equal increments in
2 2.92 4.30 chronological time, as different duty cycles
can be present in the lifetime, such as in the
3 2.35 3.18
wind turbine example already mentioned.
5 2.02 2.57
10 1.81 2.23
A discretization of damage into the set of
20 1.72 2.09 states i=1,2 ,..., b is made. Here state b de-
50 1.68 2.01 notes a state of failure. The damage accu-
∞ 1.64 1.96 mulation process is then a discrete-time,
discrete-state Markov process and as such
Markovian cumulative damage model can be viewed as a Markov chain. The prob-
Rain-flow counting, S−N curves and Miner’s ability distribution of damage is completely
rule for prediction of cumulative damage determined by the transition matrix for each
can be used to form a probabilistic model for duty cycle and by the initial damage present
damage accumulation when random vari- at t=0. The initial state of damage is speci-
ables are introduced in place of the model fied by the vector p0={πi}, where πi is the
parameters such as the parameters of the probability of damage being in state i at t=0.
S−N curves.
p 0 = [π 1 , π 2 ,...π b ] ; π i ≥ 0 ;
b
Another way to introduce a probabilistic åπ

i =1
i
=1
structure in a cumulative damage model is to
assume an evolutionary probabilistic struc-
The type of initial damage is not specified
ture from the start. Such an approach can be
and may arise from material defects, manu-
based on the model by Bogdanoff et al.
facturing defects, or other types of initial
(1978,1980). A basic element in this model
defects.
is a duty cycle which is a repetitive period of
operation in the life of a structural compo-
The transition matrix for a duty cycle is
nent in which damage can accumulate. Un-
der constant-amplitude loading, a duty cycle P={Pij}, where Pij is the probability of the
can correspond to a certain number of load damage being in state j after the duty cycle
given that the damage is in state i at the be-
cycles. Another example is the operation of
ginning of the duty cycle. It is first assumed
a wind turbine from cut-in to the subsequent
that damage accumulation during a duty
4 – Loads 81
cycle can only be of one unit. The transition duty cycles are of the same severity, the
matrix is then of the form expression for pt reduces to
é p1 q1 0 0 . 0 0 ù pt = p0Pt
ê0p q 0 . 0 0 úú
ê 2 2
ê 0 0 p3 q3 . 0 0 ú where P is the common transition matrix.

P=ê ú Since matrix multiplication is generally not
ê . . . . . . . ú commutative, it follows that not only the
ê0 0 0 0 . p b −1 q b −1 ú severity but also the order of the duty cycles
ê ú influences the damage accumulation.
ëê 0 0 0 0 . 0 1 ûú
A generalization of the model consists in
where pi+qi=1, pi≥0, and qi≥0 for all i. The filling out the upper portion of the transition
corresponding Markov chain is pictured in matrix. Then damage increases by more than
Figure 4-21. The Markov chain has b−1 one unit during a duty cycle are possible.
transient states and one absorbing state i=b. The expression for pt still holds, so the cal-
culation of the probability distribution is just
as simple. The number of parameters in the
model is, however, increased.
The probability distribution of various ran-

dom variables associated with the damage
accumulation process can now be deter-
mined. The time Wb to failure, i.e. the time
to absorption at state b, has the cumulative
distribution function FW(t,b) given by
Figure 4-21 Markov chain with transition
probabilities, from Madsen et al. (1986) FW(t,b)=P(Wb≤t)=pt(b), t=1,2,....
The state of damage at time t is given by the The mean value and the variance of the life-
vector pt={pt(i)}, where pt(i) is the probabil- time Wb are
ity that damage is in state i at time t.
E [Wb ] = å (1 − FW (t , b))
b
p t = [ p t (1), p(2) t ,... p t (b)] ; p t (i ) ≥ 0 ; t =1
å p (i ) = 1
i =1
t
Var [Wb ] = 2å t (1 − FW (t , b)) + E [Wb ] − E [Wb ]
b
2
t =1
It follows from Markov chain theory that

The state of damage at time t is denoted
p t = p 0 P1 P2 ⋅ ⋅ ⋅ Pt −1 Pt D(t). It has the probability density function
P(D(t)=i)=pt(i), i=1,2,....
where Pj is the transition matrix for the jth
duty cycle. By this expression for pt, the
from which the cumulative distribution
probability distribution of damage is com-
function, the mean value and variance, and
pletely specified at any time and is calcu-
other statistics are computed directly.
lated by simple matrix operations. If the
82 4 – Loads
The results above are all functions of the σm. For design purposes, the characteristic
πi’s and the Pij’s. For use of these results, load response is usually taken as some
the initial state of damage in terms of the quantile of the distribution of the maximum
πi’s needs to be established together with the load response in 10 minutes.
transition probabilities Pij associated with
transititions between the different damage There are two fundamentally different ap-
states. For further details, reference is made proaches to predicting the maximum load
to Madsen et al. (1986). response and particular quantiles of its dis-
tribution
4.4.2 Ultimate loads • statistical model, which utilizes the in-
formation about the maximum load re-
Extreme value distributions are of interest
sponse obtained from the n simulated
when extreme load responses are needed
time series in terms of n simulated
such as for design against failure in ultimate
maximum values xm
loading. Reference is usually made to a par-
• semi-analytical model, which − based
ticular load case, e.g.
on stochastic process theory − utilizes
• normal operation at a 10-minute mean
the information about the underlying
wind speed near the cut-out wind speed
load response process in terms of the
• stand-still at a rare 10-minute mean
four statistical moments µ, σ, α3 and α4,
wind speed such as the one with a 50-
year recurrence period and the crossing rate νµ.
The two approaches are presented in the
• faulty operation at a high wind speed
following sections, and their accuracies are
due to error in the safety system
discussed.
For the considered load case, it is assumed
that a total of n 10-minute time series of the Statistical model
The maximum load response Xm in 10 min-
load response X has been generated by aero-
utes can be assumed asymptotically to fol-
elastic simulations. The following quantities
low a Gumbel distribution
associated with the load response X can be
interpreted from each of the n time series
F X ( x m ) = exp(− exp(−α ( x m − β )))
• mean value µ m
• standard deviation σ
• skewness α3 in which α is a scale parameter and β is a
• kurtosis α4 location parameter. From the n simulated
• rate νµ of upcrossings of level µ time series there are n observations of the
• maximum xm in 10 minutes maximum load response Xm. For estimation
of α and β, the n values of Xm are ranked in
The maximum value Xm of the load response increasing order, xm,1,...xm,n. Two coeffi-
X in the 10 minutes is of interest. The cients b0 and b1 are calculated from the data
maximum value will not be a fixed value,
but will have a natural variability, which can 1 n 1 n r −1
be represented by a probability distribution.
b0 = å
n r =1
x r and b1 = å
n r =1 n − 1
xr
The natural variability is reflected in terms
of different values for xm in the n simulated and α and β are estimated by
time series. The mean value of the maxi-
mum load response in 10 minutes is denoted ln 2 γ
µm, and the standard deviation is denoted αˆ = and βˆ = b0 − E
2b1 − b0 αˆ
4 – Loads 83
in which γE=0.57722 is Euler’s constant. When a characteristic value with a specified

confidence is aimed for, then it is usually
The mean value and standard deviation of taken as the upper confidence limit, i.e. a
Xm are estimated by one-sided confidence interval is considered.
The characteristic value with confidence
γ π 1−α then becomes
µˆ m = βˆ + E and σˆ m =
α ˆ αˆ 6
xˆ m ,θ + t 1−α , n −1 ⋅ se( xˆ m ,θ )
respectively.
Note that a high number of simulation runs n
The θ-quantile in the distribution of Xm can may be necessary to achieve a sufficiently
be estimated by accurate estimate of xm,θ or µm.
xˆ m ,θ = µˆ m + k θ σˆ m Example
Consider a load response process X whose
extreme value Xmax in 10 minutes has been
in which estimated on the basis of n=5 simulated 10-
minute time series. The following estimates
6 1 pertaining to the extreme value distribution
kθ = (− ln(ln( )) − γ E )
π θ have resulted:
The standard error in the estimate of the θ- α̂ =3.69, βˆ =3.87, µ̂ m =4.02, σˆ m =0.35.
quantile is estimated by
An estimate of the θ=95% quantile of Xmax
σˆ m with 1−α=95% confidence is sought. This
se( xˆ m ,θ ) = 1 + 1.14k θ + 1.1k θ2
n gives kθ=1.866. The central estimate of the
95% quantile of Xmax is
This reduces to
xˆ m , 95% = 4.02 + 1.866 ⋅ 0.35 = 4.673 .
σˆ m
se( µˆ m ) =
n The standard error in this estimate is
for the special case that the θ-quantile is 0.35

replaced by the mean value µm. se( xˆ m , 95% ) = 2.64 = 0.413 .
5
Assuming a normal distribution for the esti-
The pertinent quantile in the Student’s t dis-
mate of the θ-quantile, the two-sided confi-
tribution is t1−α/2,n−1=2.78, and the two-sided
dence interval for the θ-quantile with confi-
confidence interval for xm,95% becomes
dence 1−α becomes
4.673±2.78⋅0.413=4.673±1.148. This indi-
cates a rather wide interval about the central
xˆ m ,θ ± t ⋅ se( xˆ m ,θ )
1−
α
2
, n −1 estimate. If n is changed from 5 to 100, then
the interval is narrowed considerably to
in which t1−α/2,n−1 is the 1−α/2 quantile in the 4.673±0.183.
Student’s t distribution with n−1 degrees of
freedom.
84 4 – Loads
Semi-analytical model estimates of the Gumbel distribution pa-

The semi-analytical model owing to Daven- rameters α and β can be found as
port (1961) utilizes more information about
the n 10-minute time series of the load re- π
sponse than just the n maximum response αˆ =
values xm. The load response X can be σˆ m 6
viewed as a stochastic process during the γ
βˆ = µˆ m − E
10-minute series. The process X can be con- αˆ
sidered as a quadratic transformation of a
parent standard Gaussian process U, The standard error in the mean value esti-
mate µ̂ m is
X = ζ + η (U + εU 2 ), ε << 1
σˆ m
A first-order approximation gives the fol- se( µˆ m ) =
lowing expressions for the coefficients n
α3 where n is the sample size, i.e., the number

ε= of 10-minute series available for the estima-
6
η =σ tion of µm by the above formula. When σm
can be considered well determined by the
ζ = µ − εσ
above formula for σˆ m , then a the two-sided
confidence interval for µm with confidence
from which it appears that use is made of the
1−α becomes
mean value µ, the standard deviation σ and
the skewness α3 of the load response process
X. σˆ m
µˆ m ± u α ⋅
1−
2 n
The mean value and standard deviation of
Xm are estimated by in which u1−α/2 is the 1−α/2 quantile of the
standard normal distribution function.
µˆ m = η + ζ ( 2 ln(ν µ T )
Comparison and recommendations
+ ε 2 ln(ν µ T )) The semi-analytical results will usually pro-
γ E ζ (1 + 2ε 2 ln(ν µ T ) ) vide a higher accuracy of the extreme value
+ estimates than the statistical results. This
2 ln(ν µ T ) comes about because the semi-analytical
results utilize much more of the available
and information than the statistical results do. In
other words, a smaller sample size n is re-
quired to achieve the same accuracy when
πη 1 + 2ε 2 ln(ν µ T )
σˆ m = semi-analytical results are used for the ex-
6 2 ln(ν µ T ) treme value estimates than when only statis-
tical results for the extreme values are used.
respectively, where T denotes duration and It is recommended always to predict extreme
is usually the length of a simulated time loads by means of the semi-analytical
series, i.e. T=10 minutes. Corresponding method, based on the statistics of the simu-
4 – Loads 85
lated load response process, rather than us- not in operation. For the operating load
ing only the observed maximum values. cases, the periodic nature of the response
mean and standard deviation for some loads
Two or more arbitrarily selected simulated must be accounted for. A method, based on
10-minute time series may give considerably azimuthal binning, can be used for this pur-
different extreme values. This implies that pose. By such a method, the rotor disc is
the practice of performing a few simulations divided into a number of sectors, each iden-
and selecting the average extreme load or tified by its azimuth angle. When the rotor
the largest extreme load as the ultimate load disc is discretized into M sectors of equal
without proper consideration of the stochas- angle of aperture, sector-specific mean val-
tic nature of the extremes will not give re- ues and standard deviations of the load re-
producible results. Further, the results can- sponse process can be established as
not be extrapolated to a characteristic value
defined by a quantile or to a different dura- 2π 1
tion of the load case than 10 minutes. The µi = µ( (i − ))
M 2
semi-analytical approach takes the stochastic
nature of the extremes into account and pro- and
vides a rationale for analysis of extreme
loads from simulated time series of load 2π 1
responses. σi =σ( (i − ))
M 2
As an alternative to the presented quadratic in which µ and σ are the mean value and
transformation of the parent Gaussian proc- standard deviation, respectively, of the load
ess U to the physical load response process
response X. Let α and β denote the distribu-
X, a cubic transformation can be applied. tion parameters in the Gumbel distribution
This can be made according to a fourth- of the maximum value during the time T of
moment Hermite polynomial expansion as
the normalized process (X−µ)/σ. They can
described by Winterstein (1988). This will
be determined by the analytical model as
allow for representation of the load response
described above. A lower bound for the
process in terms of also the kurtosis α4.
mean value µm of the largest extreme Xmax of
X in all sectors during the time T is
It is not recommended to consider transfor-
mations, which involve higher-order statisti-
ì ln M − γ E ü
cal moments of the load response X. This is µˆ m ,lower = max íµ i + σ i ( β − )ý
because the lengths of available time series î α þ
of the load response are usually much too
short to allow for a sufficiently accurate An upper bound for the mean value µm of
estimation of such higher-order statistical the largest extreme Xmax of X in all sectors
moments. In other words, higher-order mo- during the time T is
ment estimates based on available simulated
time series of load response can usually not ì γ ü
be trusted. µˆ m ,upper = max íµ i + σ i ( β + E )ý
î α þ
Correction for periodic loads
More details about this method can be found
The presented semi-analytical model pro-
in Madsen et al. (1999). A recommended
vides good accuracy for prediction of ex-
value for discretization of the rotor disc into
treme values for wind turbines, which are
sectors is M=36.
86 4 – Loads
4.5 Simplified load calculations tion that the airflow load F0 is applied with
an eccentricity e=R/6.
4.5.1 Parametrized empirical models
Various simplified models for load calcula- For calculation of blade loads, it is assumed
tion exist and are presented in the following that the airflow load is distributed evenly
subsections. Today, where available com- between the NV blades, such that the flap-
puter resources usually do not prohibit exe- wise airflow load on one blade becomes
cution of computer-intensive calculations,
most load calculations for wind turbines are Table 4-5 Design rotor loads by simplified method
carried out by means of aeroelastic codes. Load component Sym Static Dynamic load
-bol load amplitude
The simplified models are therefore in- Horizontal force FX 0 0
cluded mainly for historical reasons and to in rotor plane
provide tools for preliminary calculations Moment about MX eF0 0.25eF0
and quick checks of results. horizontal axis
in rotor plane
Horizontal force FY F0 0.25F0
4.5.2 The simple load basis along rotor axis
Moment about MY 1.3Me,nom 0.25⋅1.3Me,nom
Based on a systematization of measurements rotor axis
and experiences from typical stall-regulated Vertical force FZ −mg 0
wind turbines with active yaw and approxi- Moment about MZ eF0 0.25eF0
mately constant rotor speed, a simplified vertical axis
method for calculation of wind turbine loads
has been established. The method is useful PV=F0/NV
for preliminary design of wind turbines with
rotor diameters between 5 and 25 m and This load is assumed to be the resultant of a
rotor speeds between 35 and 50 m/s. triangular-shaped flapwise line load along
the blade with the maximum value occurring
The method is based on three load quantities at the blade tip.
as follows
• a static horizontal airflow load For calculation of blade loads in the rotor
F0=300A, where F0 is in units of N, plane, i.e. edgewise blade loads for the indi-
A=πR2 is the swept area of the rotor, vidual blades, the gravity loads need to be
and R is the rotor radius in units of m. considered. For calculation of effects on
• a driving torque Me,nom=Pnom/(2πnrη), machinery and tower, the dynamic load am-
where Pnom is the nominal power of the plitudes given in Table 4-5 may be reduced
wind turbine, nr is the rotor frequency, by 33%. For design of the main shaft, a
and η is the nominal efficiency, usually torque from the mechanical brake of two
times the value of MY is to be assumed.
η≤0.9.
• the weight of the rotor mg, where m is
The design loads given here are compatible
the rotor mass and g is the acceleration
with the Danish codes. When applied for
of gravity
design purposes, they therefore need to be
checked against design capacities calculated
The rotor loads are expressed in terms of
according to the Danish codes.
these three quantities as summarized in Ta-
ble 4-5. Moments are based on the assump-
4 – Loads 87
4.5.3 Quasi-static method ìæ 2

z ö
ï ç ln( ) + 3.1 ÷
The quasi-static method presented in the ï ç z0 ÷ for n 0 L > n *
following gives simple expressions for four ïç z ÷ U 10
different design loads as derived for a still- ï çç ln( ) ÷÷
standing wind turbine in rather severe wind ψ =íè z0 ø
ï 2 kb + kr
conditions: The blade load is calculated for a ï1 + 3.9 n L
vertical blade above the hub. The root mo- for 0 ≤ n *
ï z U 10
ment is used as the design tilt moment. The ï ln( )
î z
axial force is determined by summing the 0
wind load over the blades. The yaw moment

is determined as the root moment from a where z is the hub height, z0 is the roughness
horizontal blade. length, L is the rotor radius, n0 is the eigen
frequency of the vibration mode associated
By the quasi-static method, the load per unit with the particular design load which is to be
length of the blade is calculated by calculated, and the threshold n* is given in
Table 4-6.
1
p(r ) = ψρU 10 D (r )C
2
Moreover, the background turbulence effect
2
kb in the expression for ψ is approximated
by
in which C is the maximum value of the lift
coefficient CL or the drag coefficient CD. ì L
Typical maximum values of CL and CD are ï0.9 − 2.5 l for blade load
kb = í
in the range 1.3-1.5. U10 is the 10-minute L
ï 0.75 − 3 for axial force
mean wind speed with a recurrence period î l
of fifty years at a height h. The height h and
the coefficient C are given in Table 4-6, and the resonance effect kr in the same ex-
depending on which design load is to be pression is given by
calculated. ρ is the density of air, and D(r) is
the chord length of the blade at distance r n0 l
from the hub.
U 10 π2
kr = F (n 0 )
The quasi-static gust factor ψ in the load nl 2δ
(1 + 1.5 0 ) 5 3
expression is to be calculated as U 10
Table 4-6 Quasi-static load specification parameters

Load Eigen frequency n0 cor- Threshold C Height h
responding to n*
Blade load Blade flapwise bending 1.7 CL,max 2
z+ L
3
Axial force Tower bending 0.45 CD,max z
Tilt moment Blade flapwise bending 1.7 CL,max 2
z+ L
3
Yaw moment N/A 0 CD,max z
88 4 – Loads
Here, ℓ=6.8Lu, in which Lu is the integral Periodical maximum method

length scale of the Kaimal spectrum, see The distribution of the maximum load in a
Section 3.1.5. δ=2π(ζ0+ζa) is the logarithmic period of specified duration such as one year
increment of damping expressed in terms of can be assumed to be a Gumbel distribution
the structural damping ratio ζ0 and the aero-
dynamic damping ratio ζa. F(n) is the aero- F(q)=exp(−exp(−α(q−β)))
dynamic admittance function whose value
depends on which design load is being con- When the distribution parameters α and β
sidered refer to the distribution of the maximum
load in one year, then the load with recur-
ì rence period T years can be estimated by
ï
ï 1 T 1
ï 1 q T= β − ln(ln ) ≈ β + ln T
ï for blade load α T −1 α
nL
ï 1+ 3
ï U 10 4.5.5 Parametrized load spectra
ïï 1
F ( n) = í for axial force Parametrized load spectra are useful for cal-
ï 1 + 12 nL culations where not only the extreme load is
ï U 10 of interest, but the entire load distribution,
ï nL
ï 2.7 such as for design against fatigue failure.
ï U 10 ì for rotor Parametrized load spectra are simplified and
ï í idealized load distributions expressed in
nL nL 2 îmoments
ï1 + 4.4 + 21.8( ) terms of a number of characteristic parame-
îï U 10 U 10
ters.
A useful parametrized load spectrum is the

4.5.4 Peak factor approach for extreme one which is presented in DS472. This load
loads spectrum is meant for calculation of blade
Extreme loads are often specified in terms of loads and is based on the characteristic
the load which has a certain recurrence pe- aerodynamic line load p0. The value of p0
riod, such as fifty years, i.e. the load that on can be calculated as
average will be exceeded once every fifty
years. Two approaches are presented here. 1
p0 = ρW 2 cC L
2
Peak-over-threshold method
By the so-called peak-over-threshold in which ρ is the density of air, conserva-
method, the load with a recurrence period T tively set equal to 1.28 kg/m3, c is the char-
can be estimated by acteristic chord length of the blade at a dis-
tance r=2/3R from the hub, CL is the lift
qT=q0+αln(λT) coefficient at the same distance r=2/3R, and
W is the resulting wind speed to be calcu-
in which q0 is some chosen threshold for the lated from
load q, λ is the mean rate of exceedances of
the threshold q0 by the load q, and α is the 4π
W2 =( n r R) 2 + V0
2
mean value of the exceedances ∆q=q−q0. 3
4 – Loads 89
where nr is the rotor frequency, R is the rotor load range distribution consists of load
radius, and V0 is the nominal stall wind ranges of constant magnitude and their
speed at the height of the hub, defined as the number is equal to the total number of rota-
minor of the following two wind speeds tions NR of the rotor during the design life.
• the nominal 10-minute mean wind The constant magnitude load range of this
speed Vnom at which the turbine reaches deterministic load will typically be a func-
its nominal power Pnom tion of the mass m per unit length of the
• the 10-minute mean wind speed at rotor blade and of the acceleration g of
which stall just extends to the entire gravity.
blade for airflow parallel to the rotor
shaft The cyclically varying stochastic load ap-
The load distribution along the blade is rep- pears for example as a cyclically varying
resented as a triangular line load whose flapwise bending moment at the blade root
value is 0 at the hub and p0 at the blade tip a and results from the airflow forces set up on
distance R away from the hub, such that the the blades by the wind. The corresponding
value of the line load at a distance r from the load range distribution is represented by a
hub can be calculated as generic standardised distribution, whose
unitless range values F∆* are to be multi-
p0 r plied by a design constant, which depends
p= on load type and direction, to give the
R
sought-after load range F∆.
and such that the resulting bending moment
at the blade root becomes According to DS472, the generic standard-
ised load range distribution to be used for
the stochastic load ranges is given by the
R2
M root = p 0 following expression
3
F∆*(N)=β⋅(log10(NF)−log10(N))+0.18
For fatigue calculations, loads are in general
represented by some mean load, which is with the additional condition F∆*(N)≤2kβ
superimposed by some cyclically varying
load. The cyclically varying load is consid- in which
ered in the following. The load ranges F∆ of
the cyclically varying load are represented β=0.11kβ(IT+0.1)(A+4.4)
by a probability distribution. This probabil-
ity distribution is expressed such that F∆(N) IT is the characteristic turbulence intensity at
is the load range which is exceeded N times the hub height according to formula from
during the design life of the wind turbine. DS472, see Section 3.1.2 and 3.1.3. Note
that this generic standardised load range
The cyclically varying load consists of a distribution is valid for rotor diameters less
deterministic part owing to gravity loads and than 25 m. Application to larger rotor di-
a stochastic part owing to aerodynamic ameters may lead to overconservative re-
loads. The cyclically varying deterministic sults.
load appears for example as a cyclically
varying edgewise bending moment at the In general, kβ=1 such as for loads on indi-
blade root and results from the rotation of vidual blades, however kβ=2.5 for calcula-
the blades about the hub. The corresponding tion of rotor pressure from all three blades.
90 4 – Loads
NF=nCTL(exp(−(Vmin/A)k)−exp(−(Vmax/A)k) is load range value, the second-largest deter-

the number of load ranges in the design life ministic load range value is added to the
TL of the turbine corresponding to a charac- second largest stochastic load range value,
teristic load frequency nC. and so on, until all load range values of the
two distributions have been combined to
A and k are scale and shape parameter, re- form the combined distribution. An example
spectively, in the long-term Weibull distri- of such a combination of a deterministic and
bution of the 10-minute mean wind speed stochastic load range distribution is given in
U10, see Section 3.1.1. Figure 4-22 for the edgewise line load on a
rotor blade.
Vmin and Vmax are cut-in and cut-out wind
speed, respectively, for the operation of the F∆*(N)
wind turbine, see Section 2.2. (sum)
For combination of deterministic loads, a

harmonic variation with time can be as- (deterministic)
sumed, and the following formula can be
used, based on the assumption that the blade (stochastic)
is vertical at time t=0:
1 log10(N)
p(t ) = p + p ∆C cos(2πn C t )
2
1 0 log10(NR) log10(3NR)
+ p ∆S sin( 2πn C t )
2 Figure 4-22 Standardised distribution
where p may denote, for example, line load Table 4-7 gives expressions for determinis-
for blades. p is then the mean line load, tic and stochastic load components for line
and p∆C and p∆S are load ranges for cosinu- loads for rotor blades. The characteristic
soidal and sinusoidal load components, re- frequency nC to be used is also indicated and
spectively. expressed in terms of the rotor frequency nR.
For combination of deterministic load Rotor loads

ranges and stochastic load ranges it is rec- Rotor loads are expressed as the sum of a
ommended, conservatively, to simply add deterministic mean load and a harmonically
the load range values from the two distribu- varying load with stochastic range
tions. The largest deterministic load range
value is then added to the largest stochastic
Table 4-7 Line load on rotor blade
Mean Load range distributions
Direction Deterministic Stochastic
p p ∆C p ∆S p ∆ , stochastic
Edgewise 2Mnom/(3R2) 0 +2mg 0.3F∆*(N)p0
Flapwise 1.5p0r/R 0 0 F∆*(N)p0r/R
Along blade (2πnR)2mr −2mg 0 0
Frequency nC nR nR 3nR
4 – Loads 91
Table 4-8 Rotor loads

Load compo- Sym- Determi- Stochastic load Fre- Lowest resonance
nent bol nistic range F∆ quen- frequency n0
mean F cy nC (oscillation form)
Horizontal force FX 0 0
in rotor plane
Moment about MX 0 0.33F∆*(N)kRp0R2 3n R nROTOR (rotor, tilt,
horizontal axis asymm.)
in rotor plane
Horizontal force FY 1.5p0R 0. 5F∆*(N)p0R 3n R nTOWER (tower,
along rotor axis bending)
Moment about MY 0.5Mnom 0. 45F∆*(N)p0R2 3n R
rotor axis
Vertical force FZ −Mg 0
Moment about MZ 0 0. 33F∆*(N)kRp0R2 3n R nROTOR (rotor, yaw,
vertical axis asymm.)
loads. The expressions are based on the

1 blade line load p0, the nominal torque Mnom,
F = F + F∆ cos(2πn C t ) the rotor mass M, and the standardised load
2
range distribution given in terms of F∆*(N)
where t denotes time, and nC is the fre- as outlined above. Some of the expressions
quency. include a correction factor kR that accounts
for amplification effects, which are a func-
tion of the ratio between the resonance fre-
1,8 quency n0 and the rotor frequency nR. The
1,6 correction factor kR is given in Figure 4-23.
1,4
The resonance frequency n0 is the lowest
resonance frequency for the associated os-
1,2
cillation form, i.e. nTOWER for towers in
1 bending and nROTOR for the collective
kR
0,8 asymmetric rotor oscillation at stand-still,
0,6
where one blade oscillates out of phase with
illicit
the two other blades.
0,4
frequency
range
0,2
4.6 Site-specific design loads
0
0 1 2 3 4 5 6 7 8 As part of a scheme to reduce the cost of
n 0/n R produced energy, site-specific characteristics
can be included directly in the design proc-
ess of wind turbines. The idea is to optimise
Figure 4-23 Amplification factor kR vs. fre- the wind turbine design by minimising the
quency ratio n0/nR cost of produced energy, given the charac-
teristics at a particular site. Extreme loads
Table 4-8 gives expressions for determinis- and fatigue loads on a wind turbine are site-
tic and stochastic load components for rotor dependent. Site-specific design loads will
92 4 – Loads
therefore form part of such a cost-optimal 4.7 Loads from other sources than
design process. wind
Installation of wind turbines in shallow wa-
Two approaches to site-specific designs are
ters of up to 15 m water depth can be fore-
envisaged
seen. A support structure or foundation
• adapt existing wind turbine designs to
structure is used for transfer of loads from
specific site by minor adjustments
the wind turbine and its tower to the sup-
• design from scratch porting soils at the seabed. The foundation
structure will be exposed to wave loads,
Site-specific design has a potential when- current loads and ice loads. The prediction
ever wind turbines are to be installed in of wave loads are dealt with in the follow-
complex terrain such as in mountainous ar- ing. A very brief introduction to current
eas, where special conditions may prevail. It loads and ice loads is also given.
can be used to verify that all relevant load
cases have been considered. 4.7.1 Wave loads
Note that using site-specific design loads Wave climate
and carrying out site-specific wind turbine The wave climate at a location can usually
designs is in contrast to the current trend be considered stationary within periods of
within the wind turbine industry. In order to typically three hours duration. The wave
keep down manufacturing costs, the current climate is represented by the significant
trend is not to site-optimise wind turbines, wave height HS and the peak period TP. The
but rather to produce a selection of standard significant wave height HS is a measure of
wind turbines. The task is then to choose a the intensity of the wave climate and is de-
standard wind turbine from this selection fined as four times the standard deviation of
and verify that it is suitable for a given loca- the sea elevation process η. Some sources
tion. The tower and the foundation may still define the significant wave height as the
be site-optimised if desirable, and site- average of the highest one third of the wave
specific loads will be required for this pur- heights. For a narrow-banded Gaussian sea
pose. The foundation design will always elevation process, the two definitions con-
have to be site-specific in that it needs to be verge. The peak period TP is related to the
designed for the prevailing local soil condi- mean zero-crossing period of the sea eleva-
tions. tion process. The significant wave height
and the peak period can be taken as constant
WASP Engineering within each three-hour period.
Site-specific loads can be derived from site-
specific wind conditions, which can be Long-term distributions of HS and TP are
modelled by WASP engineering. WASP − site-dependent. The long-term distribution
the Wind Atlas Analysis and Application of HS can often be represented well by a
Program − is a method for prediction of Weibull distribution, whereas the distribu-
properties of winds in moderately complex tion of TP conditioned on HS is usually well
terrain with relevance for loads on wind represented by a lognormal distribution
turbines and other large structures, whose distribution parameters are functions
http://www.wasp.dk/. For details, reference of HS. Examples can be found in Bitner-
is made to Section 3.1.9. Gregersen and Hagen (2000).
4 – Loads 93
Wave spectrum in which η is a spectral width parameter.

The frequency content of the sea elevation The maximum wave height Hmax during
process can be represented by the power some time span TL is often of interest for
spectral density. The spectral density of the design. Let N denote the number of zero-
sea elevation process can be represented by upcrossings of the sea elevation process in
the JONSWAP spectrum this period of time, i.e., N=TL/TZ. The distri-
bution of Hmax can then be approximated by
−5
5 æ ωT ö
S η (ω ) =
2
H S T P çç P ÷÷ 2h 2
32π è 2π ø FH max (h) ≈ exp(− N exp(− ))
(1 − ν 2 ) H S
2
−4
5 æ ωT ö
⋅ exp(− çç P ÷÷ )
4 è 2π ø and the expected value of the maximum
æ
exp çç −
1 æ ωT P ö ö÷
çç
2
−1 ÷÷ ÷ wave height can be approximated by
ç 2σ 2 è 2π ø ÷ø
⋅ C (γ )γ è
(1 − ν 2 ) ln N
E [H max ] ≈ H S
in which ω denotes the angular frequency, 2
ω>0, σ=0.07 for ω<2π/TP, σ=0.09 for
ω>2π/TP, and C(γ)=1−0.287lnγ. The peak- A first-order approximation yields the fol-
enhancement factor γ can be taken as lowing value for the spectral width parame-
ter, ν=0.43. With this value of ν, the fol-
TP TP lowing relationship between the maximum
γ = exp(5.75 − 1.15 ) ; 3.6 ≤ ≤5
HS HS wave height Hmax and the maximum wave
crest Zmax holds,
in which HS is in meters and TP in seconds. Hmax≈1.8Zmax
The following approximate relationship ex-
ists between the peak period TP and the zero- In shallow waters, shoaling effects imply
crossing period TZ that wave crests become more peaked while
wave troughs become flatter and not quite as
5+γ deep. The sea elevation process has become
TZ = TP
11 + γ somewhat ”skewed” and will not quite be
Gaussian, and the individual wave heights
will not quite be Rayleigh distributed. Tech-
Reference is made to Gran (1992) and DNV
niques are available to account for the
(2000).
skewness introduced by the shoaling. Refer-
ence is made to Winterstein et al. (1991) and
Wave heights
U.S. Army Coastal Engineering Research
In deep waters, the sea elevation process η Center (1973).
is a Gaussian process and the individual
wave heights H will follow a Rayleigh dis-
Note that in shallow waters, the wave
tribution when HS is given, heights will be limited by the water depth, d.
The maximum possible wave height at a
2h 2 water depth d is approximately equal to the
FH (h) = 1 − exp(− )
(1 − η 2 ) H S
2
water depth,
94 4 – Loads
H max,lim ≈ d 4 2π 2kd
h0 = H max (1 + ) tanh[kd ]
HS sinh [2kd ]
and the Rayleigh distribution of the wave
heights will become distorted in the upper in which k is the wave number, which re-
tail to approach this limit asymptotically. sults implicitly as the solution to
ω 2 = gk tanh [kd ]
Wave periods
Once the significant wave height HS is
given, the zero-upcrossing period TZ is usu-
ally well represented by a shifted lognormal where g is the acceleration of gravity. The
distribution wave length L can be calculated from the
wave number as L=2π/k. The angular veloc-
ln(t − δ ) − a1 ity ω is found from the wave period T as
FTZ (t ) = Φ ( ) ; t≥δ ω=2π/T.
a2
Wave forces by Morison’s equation
in which Φ denotes the standardized normal Wave forces on slender structural members
distribution function, the distribution pa- such as a cylinder submerged in water can
rameters a1 and a2 are functions of HS, and be predicted by Morison’s equation. By this
the shift parameter δ can be approximated equation, the horizontal force on a vertical
by element dz of the structure at level z is ex-
pressed as
δ ≈ 2.2 H S
dF = dFM + dFD
when HS is given in meters and δ is given in D2 D
= C M ρπ xdz + C D ρ x xdz
seconds. This is based on braking consid- 4 2
erations, see Haver (1990).
where the first term is an inertia force and
The mean zero-upcrossing period TZ is an the second term is a drag force. Here, CM
average wave period associated with a and CD are inertia and drag coefficients,
seastate of a given significant wave height respectively, D is the diameter of the cylin-
HS. The wave period T associated with the der, ρ is the density of water, x is the hori-
maximum wave height Hmax in this sea state zontal wave-induced velocity of water, and
can be represented by a Longuet-Higgins x is the horizontal wave-induced accelera-
distribution
tion of water. The level z is measured from
stillwater level, and the z axis points up-
TZ 2 2 h
Φ ((1 − ( ) (ν + 1)) 0 ) wards. Thus, at seabed z=−d, when the water
FT (t ) = t 2ν depth is d. The movement of the structure is
h0
Φ( ) considered to be very small.
2ν
According to first-order linear wave theory,
in which Φ denotes the standardized normal the horizontal wave-induced velocity is
distribution function, ν is the spectral width
parameter as referenced above, and h0 is a cosh[k ( z + d )]
x = AW ω sin(ωt )
normalized maximum wave height in deep sinh [kd ]
water
4 – Loads 95
and the acceleration is Note that the integration from –d to 0 ig-

nores contributions to the force from the
cosh[k ( z + d ]) wave crest above the stillwater level at z=0.
x = AW ω 2 cos(ωt )
sinh [kd ] However, this is a minor problem for the
inertia force FM, since this has its maximum
when a nodal line at the stillwater level
in which AW is the wave amplitude.
passes the structure. The drag force FD, on
the other hand, has its maximum when the
The resulting horizontal force F on the cyl-
crest passes the structure, and if this force is
inder can be found by integration of Mori-
dominating, a significant error can be intro-
son’s equation for values of z from –d to 0,
duced by ignoring the contribution from the
wave crest.
F = FM + FD
0
D2 H 2 Note also that Morison’s equation is only
= ò (C M ρπ ω cos(ωt ) valid when the dimension of the structure is
−d 4 2
small relative to the wave length, i.e. when
cosh[k ( z + d ])
⋅ )dz D<0.2L, and that it is only valid for non-
sinh [kd ] breaking waves. In deep water, waves break
0
D H2 2 when H/L exceeds about 0.14.
+ ò (C D ρ ω sin(ωt ) sin(ωt )
−d 2 4
The inertia coefficient depends on the cross-
cosh 2 [k ( z + d )] sectional shape of the structure and of the
⋅ )dz
sinh 2 [kd ] orientation of the body. Typically, CM is in
the range 1.6-2.5. For a vertical cylinder,
Figure 4-24 Relative magnitude of inertia and drag forces
96 4 – Loads
CM=2.0. For a cylinder with in-service ma- FM = AM cos(ωt )

rine roughness, e.g. owing to marine growth,
CM should normally not be less than 1.8. and the drag force as
The drag coefficient CD is never less than
0.6, and for a smooth cylinder CD=1.0.
FD = AD sin(ωt ) sin(ωt )
An example of application of Morison’s
equation is given in Figure 4-25 for a cylin- in which AM and AD denote the respective
der with 4 m diameter in 10 m water depth. amplitudes of the two forces. Let A denote
The upper part of this figure shows the sea the ratio between the amplitudes, A=AM/AD.
elevation process, the middle part shows the The following relationship can be estab-
horizontal water particle velocity and accel- lished
eration at the stillwater level, and the lower
part of the figure shows the resulting hori- H C sinh 2 [kd ]
=π M
zontal wave force and overturning moment D C D ((sinh[2kd ] / 4 + kd / 2) A
at the seabed. The example is based on
CM=2 and CD=1.2. and can be represented as shown in Figure
4-24. For a given structure and location,
Figure 4-24 can be used to quickly establish
whether the inertia force or the drag force is
the dominating force, once the ratios H/D
and d/L have been calculated.
With reference to Figure 4-24, structures

located above the curve for A=1 experience
loads which are dominated by the drag term
in Morison’s equation. Structures located
below the curve for A=1 are dominated by
inertia loads. The two dashed asymptotes in
Figure 4-24 are valid for shallow water
waves (d/L<1/20) and deep water waves
(d/L>0.5), respectively. The asymptotes are
derived from asymptotic results for small-
amplitude wave kinematics.
Example
A cylinder with diameter D=5 m is consid-
ered subjected to the wave climates at the
Middelgrunden and Rødsand sites in Danish
Figure 4-25 Example of sea elevation proc- waters. Water depths and wave data for a
ess, associated water particle velocities and number of cases at these two locations are
accelerations, and resulting Morison forces tabulated in Table 4-9 with the resulting
at the seabed for a cylindrical example ratios
structure H/D, d/L and A. It appears that for both sites
A values occur, which indicate that the iner-
The inertia force can be expressed as tia force dominates the loading.
4 – Loads 97
Table 4-9 Examples of ratio between inertia the wavesplash zone to absorb or reduce ice
force and drag force loads.
Site Wave Depth Wave H/D d/L A
height d length
H (m) L Table 4-10 Coefficients ξ and α
(m) (m)
MG 3.8 5.5 40 0.76 .138 7.1
MG 2.6 5.5 28 0.46 .196 16.2
RS 3.5 8.0 45 0.70 .178 9.8
RS 3.5 9.5 47 0.70 .202 10.9
RS 3.5 11.0 50 0.70 .220 11.7
RS 6.2 8.0 64 1.24 .125 4.0
RS 6.7 9.5 76 1.34 .125 3.7
RS 6.7 11.0 88 1.34 .125 3.7
MG=Middelgrunden
RS=Rødsand
Wave forces by diffraction theory

When the dimension of the structure in
question is large compared with the wave
length, typically when D>0.2L, Morison’s
equation is no longer valid. The inertia force
will then be dominating and can be pre-
dicted by means of diffraction theory. For a
cylinder of radius R=D/2 installed in water
of depth d and subjected to a wave of am-
plitude A, this theory gives the following
maximum horizontal wave force
4 ρgA sinh [k (d + A sin α )]

F X , max = ξ
k2 tanh[kd ]
whose vertical arm measured from the sea

Note also that in shallow waters waves may
floor is
break locally over a sloping seabed, if they
are large enough, and thereby violate the
kd sinh[kd ] − cosh[kd ] + 1 kinematic assumptions behind the presented
hF = d
kd sinh[kd ] formulas. However, for a horizontal seabed,
such local breaking is not expected, as too
The coefficients ξ and α are given in Table large waves will have broken prior to arriv-
4-10, extracted from Gran (1992). ing at the location. For these waves, the as-
sumed wave kinematics will regenerate be-
This can be used for prediction of wave fore the waves arrive at the location, while
forces on foundation structures shaped like the waves heights will conform to the limit
cylinders, such as monopiles and some given by the water depth.
gravity-based structures. Note, however, that
the formulas may lead to erroneous results if Wave loads are strongly dependent on the
the structural geometry deviates much from water depth. For this reason it is important
the assumed cylindrical shape, such as when to consider effects of local variations in the
a conical structural component is present in
98 4 – Loads
water depths, including astronomical tides Consider the combination of wind and wave
and storm surges. loads. The short-term wind climate is usu-
ally represented by the 10-minute mean
4.7.2 Current loads wind speed U10, and the short-term wave
climate is usually represented by the signifi-
Morison’s equation can be used to predict
cant wave height HS. U10 and HS may be
current loads. Note in this context that the
interpreted as intensities of the correspond-
velocity x and the acceleration x in Mori-
ing wind speed and sea elevation processes,
son’s equation need to be taken as the re- respectively. The wind and the waves at a
sulting combined current and wave velocity particular location often have a common
and acceleration, respectively. As in the case cause such as a low pressure. The waves are
of pure wave load prediction, Morison’s driven by the wind and often generated lo-
equation is only applicable as long as the cally. At the same time, the roughness im-
wave length is longer than five times the plied by the wave-affected sea surface influ-
diameter of the cylindrical structure. ences the wind. A high wave intensity will
imply a high wind intensity, and vice versa.
4.7.3 Ice loads
It is current practice to distinguish between It is important to consider this usually strong
static ice loads and dynamic ice loads. For dependency – the simultaneous occurrence
conical structures, Ralston’s formula for ice of wind and wave climates of high intensi-
loads can be applied, see Ralston (1977). ties – in design. In a probabilistic analysis,
For ice loads in general, reference is made to this can practically be done by modelling
API (1995). one of the climate variables as a so-called
independent variable by means of its mar-
4.7.4 Earthquake loads ginal cumulative distribution function, and
then model the other variable as a dependent
For prediction of earthquake loads, reference variable by means of a distribution condi-
is made to Section 3.2.8. tioned on the independent variable. For the
considered wind and wave example, one
could represent the significant wave height
4.8 Load combination HS by its marginal long-term distribution,
Principles for how to combine loads arising typically a Weibull distribution, and then
from different concurrent load processes are model the 10-minute mean wind speed U10
outlined in this section. conditional on HS. The distribution of U10
conditioned on HS will typically be a log-
When several load processes are acting con- normal distribution
currently, their combined load response in
the structure needs to be considered for de- ln u − b1
FU (u ) = Φ( )
sign. For example, the foundation structure 10 | H S
b2
of an offshore wind turbine will be subject
to the combined action from wind and wave
in which Φ denotes the standard normal
loads, and the resulting structural response
distribution function, and the distribution
from this action governs the design. Another
parameters b1 and b2 are functions of the
load combination, which is possible, is the
significant wave height HS, i.e. b1=b1(HS)
combination of wind and ice loads, and cur-
and b2=b2(HS). In some cases, other generic
rent may combine with any of the other load
distribution types than the lognormal distri-
types mentioned.
bution may provide the best representation
4 – Loads 99
of U10 conditional on HS, e.g. a Weibull dis- ìïmax Q1 (t ) + Q2 (t )

tribution. Qmax = max í 0≤t ≤T
ïîQ1 (t ) + max
0 ≤ t ≤T
Q2 (t )
Once the wind and wave climates are given
in terms of a specific set of concurrent U10 Turkstra’s rule indicates that a natural code
and HS values, the wind speed process con- format for a combination of two loads for
ditioned on U10 and the wave process condi- use in deterministic design is
tioned on HS can be considered independent.
It is therefore, for example, not reasonable ìγ q + γ 2ψ 2 q 2 k
to expect that the maximum wind speed will q design = maxí 1 1k
occur at the same time as the maximum îγ 1ψ 1 q1k + γ 2 q 2 k
wave height.
in which q1k and q2k are characteristic values
For design, it is reasonable to consider some of Q1 and Q2, γ1 and γ2 are associated partial
relatively rare combination of wave and safety factors, and the ψ-factors are load
wind climate as the characteristic climate combination factors.
and then find the maximum load response
that occurs for this climate over its duration. For a wind turbine structure, loads do not
Practically, one may for example consider always combine linearly to give the sought-
the wave climate for the significant wave after maximum load response. Aeroelastic
height with a 50-year return period in com- wind load calculations may be non-linear,
bination with a wind climate conditioned on and the combined wave and wind load re-
this wave climate, e.g. the expected value of sponse may not necessarily come out as the
U10 conditional on the 50-year significant linear combination of the separately calcu-
wave height, or some higher quantile of U10. lated wave load response and the separately
With such a rare characteristic wave and calculated wind load response. For such
wind climate in mind, the following para- nonlinear cases, the combined response is to
graphs outline how the combined load re- be calculated from some appropriate struc-
sponse, e.g. the horizontal force in some tural analysis for the concurrent characteris-
section of the foundation structure, during tic wave and wind load processes without
this climate can be found. applying any partial load factors. The re-
sulting maximum load response from this
For linear load combinations, Turkstra’s rule analysis can be interpreted as a characteristic
plays a central role. The rule states that the load response, which reflects the combined
maximum value of the sum of two inde- wave and wind loading. A common partial
pendent random processes occurs when one safety factor is then to be applied to this
of the processes has its maximum value. characteristic load response to give the de-
Application of Turkstra’s rule to the combi- sign value qdesign, i.e. it will no longer be
nation of two load processes, e.g., wave load possible to distinguish between different
and wind load, implies that the combined partial safety factors for wave loads and
load will have its maximum either when the wind loads.
wave load has its maximum or when the
wind load has its maximum. Let Q1 and Q2
denote the two load processes. Mathemati- REFERENCES
cally, the maximum combined load Qmax
over a time span T will be Andersen, P.S., U. Krabbe, P. Lundsager,
and H. Petersen, “Basismateriale for
beregning af propelvindmøller,” Report No.
100 4 – Loads
Risø-M-2153 (in Danish), Risø National Bogdanoff, J.L., and F. Kozin, “A New Cu-
Laboratory, revised version, 1980. mulative Damage Model, Part 4,” Journal of
Applied Mechanics, Vol. 47, pp. 40-44,
Abbott, I.H., and A.E. von Doenhoff, The- 1980.
ory of Wing Sections, Dover Publications
Inc., New York, N.Y., 1959. Bogdanoff, J.L., and W. Krieger, “A New
Cumulative Damage Model, Part 2,” Jour-
API, Recommended practice for planning, nal of Applied Mechanics, Vol. 45, pp. 251-
designing and constructing structures and 257, 1978.
pipelines for arctic conditions, RP2N, 2nd
edition, American Petroleum Institute, 1995. Davenport, A.G., “The Application of Sta-
tistical Concepts to the Wind Loading of
Bak, C., P. Fuglsang, N.N. Sørensen, H.A. Structures,” Proc. Inst. of Civil Engineers,
Madsen, Wen Zhong Shen, J.N. Sørensen, Vol. 19, 1961.
“Airfoil Characteristics for Wind Turbines,”
Risø R-1065(EN), Risø National Labora- DNV, Environmental Conditions and Envi-
tory, 1999a. ronmental Loads, Classification Notes No.
30.5, Det Norske Veritas, Høvik, Norway,
Bak, C., H.A. Madsen, and N.N. Sørensen, 2000.
“Profilkoefficienter til LM19.1 vingen
bestemt ud fra 3D CFD,” 1999b. Eggleston, D.M, and F.S. Stoddard, Wind
Turbine Engineering Design, Van Nostrand
Bathe, K.J., Finite Element Procedures in Reinhold Co. Inc., New York, N.Y., 1987.
Engineering Analysis, Prentice-Hall, 1982
Gran, S., A Course in Ocean Engineering,
Baumgart, A., I. Carlén, M. Hansen, G. Lar- Elsevier, Amsterdam, The Netherlands,
sen, S.M. Petersen, “Experimental Modal 1992. The Internet version, located at
Analysis of a LM 19.1 m blade,” unpub- http://www.dnv.com/ocean/, provides on-
lished work. line calculation facilities within the fields of
ocean waves, wave loads, fatigue analysis,
Bitner-Gregersen, E., and Ø. Hagen, “As- and statistics.
pects of Joint Distribution for Metocean
Phenomena at the Norwegian Continental Hallam, M.G., N.J. Heaf, and L.R. Whoot-
Shelf,” ASME Paper No. OMAE-2000- ton, “Dynamics of Marine Structures: Meth-
6021, Proceedings, International Conference ods of Calculating the Dynamic Response of
on Offshore Mechanics and Arctic Engi- Fixed Structures Subject to Waves and Cur-
neering, 2000. rent Action,” CIRIA Underwater Engineer-
ing Group, 6 Storey’s Gate, London SW1P
Bogdanoff, J.L., “A New Cumulative Dam- 3AU, Report UR8, 1978.
age Model, Part 1,” Journal of Applied Me-
chanics, Vol. 45, pp. 246-250, 1978. Haver, S., “On a Possible Lower Limit for
the Spectral Peak Period”, Statoil Report
Bogdanoff, J.L., “A New Cumulative Dam- No. F&U-MT 90009, Stavanger, Norway,
age Model, Part 3,” Journal of Applied Me- 1990.
chanics, Vol. 45, pp. 733-739, 1978.
IEA, “Expert Group Study on Recom-
mended Practices for Wind Turbine Testing
4 – Loads 101
and Evaluation, 3. Fatigue Loads,” 2nd edi- Ronold, K.O, J. Wedel-Heinen, and C.J.
tion, 1990. Christensen, “Reliability-based fatigue de-
sign of wind-turbine rotor blades,” Engi-
Larsen, G., and P. Sørensen, “Design Basis neering Structures, Elsevier Science Ltd.,
2,” Proceedings, IEA Symposium “State-of- Vol. 21, No. 12, pp. 1101-1114, 1999.
the-Art of Aeroelastic Codes for Wind Tur-
bine Calculations,” pp. 137-145, Lyngby, Stiesdal, H., “Rotor Loadings on the BO-
Denmark, 1996. NUS 450 kW Turbine,” Proceedings,
EWEC’91, Amsterdam, The Netherlands,
Madsen, H.O., S. Krenk, and N.C. Lind, 1991.
Methods of Structural Safety, Prentice-Hall
Inc., Englewood Cliffs, N.J., 1986. U.S. Army Coastal Engineering Research
Center, Shore Protection Manual, Vols. I-
Madsen, P.H., K. Pierce, and M. Buhl, “The III, Washington, D.C., 1973.
use of aeroelastic wind turbine response
simulations for prediction of ultimate design Veers, P.S., “Three-Dimensional Wind
loads,” Proceedings, 3rd ASME/JSME Joint Simulation,” Report No. SAND88-0152,
Fluids Engineering Conference, Paper No. Sandia National Laboratories, Albuquerque,
FEDSM99-S295-10, San Francisco, Cal., N.M., 1988.
1999.
Viterna, L.A. and R.D. Corrigan, “Fixed
Mann, J., “Wind field simulation,” Prob- pitch rotor performance of large horizontal
abilistic Engineering Mechanics, Elsevier axis wind turbines,” DOE/NASA Workshop
Science Ltd., Vol. 13, No. 4, pp. 269-282, on Large Horizontal Axis Wind Turbines,
1998. Cleveland, Ohio, July 28-30, 1984.
Mann, J., “The spatial structure of neutral Winterstein, S.R., “Nonlinear Vibration
atmospheric surface-layer turbulence,” Modes for Extremes and Fatigue,” Journal
Journal of Fluid Mechanics, No. 273, pp. of Engineering Mechanics, ASCE, Vol. 114,
141-168, 1994. No. 10, pp. 1772-1790, 1988.
Petersen, J.T., “Geometric non-linear finite Winterstein, S.R., E. Bitner-Gregersen, and

element model for a horisontal axis wind K.O. Ronold, “Statistical and Physical Mod-
turbine,” Risø National Laboratory, 1990. els of Nonlinear Random Waves,” Pro-
ceedings, International Conference on Off-
Petersen, J.T., H.A. Madsen, A. Björk, P. shore Mechanics and Arctic Engineering
Enevoldsen, S. Øye, H. Ganander, D. (OMAE), Vol. II , pp. 23-31, Stavanger,
Winkelaar, “Prediction of Dynamic Loads Norway, 1991.
and Induced Vibration in Stall,” Risø-R-
1045, Risø National Laboratory, 1998 Wirsching, P.H., and A.M. Shehata, “Fa-
tigue under Wide Band Random Stresses
Ralston, T.D., “Ice Force Design Consid- Using the Rain-Flow Method,” Journal of
erations for Conical Offshore Structures,” Engineering Materials and Technology,
Proceedings, Fourth POAC Conference, ASME, July 1977, pp. 205-211.
Vol. 2, pp. 741-752, St. John’s, Nfld., Can-
ada, 1977.
102 4 – Loads
5. Rotor It is important that the blade sections near

the hub are able to resist forces and stresses
from the rest of the blade. Therefore, the
5.1 Blades blade profile near the root is both thick and
Rotor blades are usually made up by a ma- wide. Further out along the blade, the blade
trix of fibre glass mats, which are impreg- profile becomes thinner to get acceptable
nated with a material such as polyester, aerodynamic properties. As the blade speed
hence the term glass-fibre reinforced poly- increases toward the tip, an approximately
ester, GRP. The polyester is hardened after constant lift force along the length of the
it has impregnated the fibre glass. Epoxy is blade is achieved by decreasing the chord
sometimes used instead of polyester. Like- width toward the tip. In other words, the
wise, the basic matrix is sometimes made up blade tapers from a point somewhere near
wholly or partly by carbon fibres, which the root towards the tip, see Figure 5-2. In
form a lighter, but more expensive material general, the blade profile is a compromise
with a high strength. Wood-epoxy laminates between the desire for strength and the de-
are sometimes used in large rotor blades. sire for good aerodynamic properties. At the
root, the blade profile is usually narrower
5.1.1 Blade geometry and tubular to fit to the hub.
The outer contour of a wind-turbine rotor

blade is designed based on aerodynamic
considerations. The cross section of the
blade has a streamlined asymmetrical shape,
Figure 5-2 Side view of a blade
with the flattest side facing the wind. Once
the aerodynamic outer contour is given, the
blade is to be designed to be sufficiently
strong and stiff from a structural point of
view. The blade profile is a hollow profile
usually formed by two shell structures glued
together, one upper shell on the suction side,
and one lower shell on the pressure side. To
make the blade sufficiently strong and stiff,
so-called webs are glued onto the shells in
the interior of the blade and form a boxlike
structure and cross-section, see Figure 5-1.
From a structural point of view, this will act Figure 5-3 View of a blade from the tip,
like a beam, and simple beam theory can be illustrating the twist of the blade: The wind
applied to model the blade for structural comes in from the left, and the pitch is 0° at
analyses in order to determine the overall the tip and about 25° at the hub.
strength of the blade.
The blade is twisted along its axis so that it
may follow the change in the direction of the
resulting wind along the blade that the blade
will experience when it rotates. Hence, the
pitch will vary along the blade. The pitch is
Figure 5-1 Section of a blade showing the angle between the chord of the blade
upper and lower shell and two webs. profile and the rotor plane, see Chapter 4.
5 – Rotor 103
Figure 5-3 illustrates the twist of a blade. act in the direction of rotation at the root
Note in this context that the pitch angle, than at the tip. The force at the root has a
which id referred to throughout this docu- smaller torque arm relative to the rotor axis
ment, usually refers to the collective rotation than the force at the tip and will therefore
of the entire blade relative to the rotor plane. give about the same contribution to the
starting torque as the force at the tip.
For a blade, four noncoincident trajectories
can be defined as follows During operation, the wind approaching the
• mass axis, the spanwise locus of section blade profile is the vectorial sum of the
mass centers farfield wind speed perpendicular to the
• elastic axis, the spanwise locus of points rotor plane and the head wind due to the
about which no section is incurred with rotational movement of the blade through
bending deflection the air. Smaller aerodynamic forces are pro-
• control axis, the axis of mechanical duced near the root than at the tip. However,
feathering, completely determined by the forces produced near the root are more
the blade retention and pitching mecha- aligned with the direction of rotation than
nism the forces near the tip. The change in mag-
• aerodynamic axis, is at the blade section nitude and direction of the forces along the
quarterchord for a conventional airfoil blade contributes to determine the shape and
shape within linear performance limits design of the blade, including the width,
These four trajectories are not axes in the thickness and twist of the blade.
true sense, since they – owing to the ge-
ometry of the blade – are not straight lines. The design loads for a blade can be deter-
However, for calculations for a particular mined by means of blade momentum theory
section of the blade, four axes, perpendicular and aeroelasticity. For details, reference is
to the section, can be defined as the tangents made to Chapter 4.
to the trajectories at their respective inter-
sections with the section. 5.1.3 Materials and manufacturing
Introduction
5.1.2 Design Loads Wind turbine blades have to be large to ex-
In principle, the same airflow conditions tract useful power from a low-density me-
would apply at all sections along a blade as dium such as air. Moreover, the blades have
long as the profile stays the same, while the to be aerodynamically shaped, smooth, light,
magnitude and direction of the forces would dimensionally accurate and capable of sus-
change depending on the distance to the tip. taining a very large number of load cycles
However, in practice, the profile and the over the design life. A rotor blade typically
blade thickness vary along the blade and consists of three parts – two outer shells and
thereby make the airflow conditions more an internal loadbearing web or spar. In addi-
complex. tion comes various auxiliaries such as blade
extenders and special details at the blade
At stand-still, the wind pressure force will – root.
depending on the load case – be somewhat
larger at the root than at the tip, as the blade Fibre-reinforced plastics are the favoured
is wider at the root. The force is acting material for rotor blades. Fibre-reinforced
roughly at a right angle to the flat side of the plastics are formed by a viscous resin such
blade profile. As the blade is more twisted at as a polyester, which is reinforced by the
the root, a larger component of the force will embedment of glass fibres or carbon fibres
104 5 – Rotor
and which is subsequently cured by means • fibre reinforcements, e.g., glass fibres,
of a catalyst to become a hard solid. The aramid fibres, and carbon fibres
main alternative material is wood, which is a • polyester resins and vinylester resins
natural composite and exhibits similar char- • epoxy
acteristics and properties. The layout of a • gelcoats and topcoats
blade may involve outer shells laid up as a • sandwich core materials
sandwich structure, which consists of two • sandwich adhesives and cements
layers of fiber-reinforced plastics, separated
• adhesives (glues)
by a core material. Adhesives are used to
join separate blade parts, and a coating may Fibre reinforcements
be applied on the outer surfaces of the blade The chemical composition of the glass used
for protection. Aluminium may be used in in glass fibres should be of the E quality for
some details of a blade. Steel is used in con- which the sum of the Na2O and K2O con-
nection components, e.g., at the blade root. tents is to be less than 2% by weight. A cer-
tificate showing the chemical composition
General requirements should be presented, or a chemical analysis
Materials are to be adequately marked for should be carried out documenting that the
identification. The marking should include requirements to E glass have been met. The
name and trademark of manufacturer, mate- requirements to the chemical composition of
rial grade, and batch number. Only materials E glass are given in Table 5-1.
with guaranteed properties may be used.
Properties to be considered in this context
Reference is made to ASTM D578 and
include, but are not limited to
DNV SfC 2.9 TAP 1-501.8 (Jan. 1999). Fi-
• strength bres made of other glass qualities may be
• toughness (at low temperatures if ap- used provided their mechanical properties
propriate) and hydrolytic resistance are equally good or
• cold deformability better.
• aging characteristics
• suitability for welding Table 5-1 Chemical composition by
• resistance to rot weight, E quality glass
Note that for some materials, some of these
Element Requirement
properties may not be relevant. Guaranteed
property values may be given in terms of SiO2 52-62 %
manufacturer’s nominal value, manufac- CaO 16-25 %
turer’s specified value, or manufacturer’s Al2O3 12-16 %
specified minimum value.
B2O3 0-10 %
A type approval of a material, e.g. according Na2O + K2O 0-2 %
to DNV standards, can be used as a part of MgO 0-5 %
the quality control and design documenta- TiO2 0-1.5 %
tion, but is in itself not sufficient for ap- Fe2O3 0-0.8 %
proval of the material for its intended use; F2 0-1.0 %
full scale tests of a sample blade are required
to verify the strength of the blade, statically
as well as in fatigue. When used for blade Glass fibres are to be produced as continu-
ous fibres. They can be tested in the par-
manufacturing, the following materials
ticular mould in which they are to be used
should be considered for type approval:
5 – Rotor 105
during the subsequent lay-up and manufac- Properties, which are suggested to be veri-
turing of the laminate or rotor blade. fied, and associated relevant test methods
for the cured resin are summarised in Table
Coupling agents, sometimes referred to as 5-4. A number of tensile tests are listed for
“finish”, is a coating used on filament bun- determination of the tensile strengths. The
dles to aid the coupling to the resin and to objective of this testing is to verify the
preserve the integrity of the fibre bundles. For strength properties of the laminate. Tensile
glass fibres, coupling agents of silane com- tests should be performed in the main fibre
pound or complex chromium compound are directions of a fabric/weave, and in two di-
usually used. For carbon fibres, coupling rections for chopped strand mats. The lami-
agents of epoxy are often applied. nate to be tested should fulfil the following
conditions and requirements:
For roving, which will be applied by spray- • The test results are only valid for the
ing, a demonstration should be made, pref- type of resin used but can be used also
erably in the presence of a certifying for resins with higher tensile strength
authority, in order to show that the roving is and higher elongation at failure.
suitable for the application in question. • It is recommended that the laminate is
cured at room temperature and atmos-
Properties to be verified and associated rele- pheric pressure. Another curing cycle
vant test methods for glass-fibre reinforce- may, however, be chosen by the manu-
ments are summarised in Table 5-2. The facturer. It is recommended that a cur-
tests include tensile and flexural tests on ing cycle which can be carried out in a
specimens of the finished fibre-reinforced workshop is selected.
laminate. • The laminate should have a fibre frac-
tion by weight, measured according to
Resins for structural laminates ISO 1172 (1996), when moulded in
The resin is the polymeric matrix into which polyester:
the fibre reinforcement is embedded. The - for weaves and fabrics: 50%±5%.
most common type of resin is formed by the It is recommended that a fibre fraction
so-called thermosetting resins such as poly- obtainable in the workshop is selected.
ester, vinylester and epoxy resins. They are • In general, all fabrics/weaves should be
usually supplied as viscous liquids which set oriented in the same direction in the
to a hard solid when activated by a catalyst, laminate. Exceptions can be made if
and remain solid on further heating. symmetric laminates are needed for
testing.
Consider a polyester in the following: The • Test specimens should be wide enough
polyester should be suitable for lamination to cover at least four repeats of the
by hand lay-up, spraying or vacuum injec- structure of the weave/fabric/mat.
tion. It should have good wetting properties Unless otherwise specified by the manufac-
and should cure satisfactorily at normal turer, the following curing procedure should
room temperature. Polyester intended for be used:
other production methods may need special
• 1% catalyst MEKP in a 50% solution
consideration.
• Curing: 24 hours at 20 °C
Properties to be verified and associated rele- • After-curing: 24 hours °C
at 50
vant test methods for the resin in liquid con- Use of a standard MEKP with 9.0-9.2% ac-
dition are summarised in Table 5-3. tive oxygen is preferred.
106 5 – Rotor
When the polyester contains waxes or other prior treatment of the primary laminate
substances that might lower the secondary surface. Curing minimum one week at
bonding ability, a delamination test should 20°C.
be carried out in addition to the tests re- For both the primary and the secondary
quired as specified in the above tables. The laminate, the fibre fraction by weight should
sample for the additional delamination test be 50%±5%. A tension test is then to be
should be prepared as follows: carried out on the prepared sample in the
• A primary laminate consisting of 3 lay- flatwise plane and according to ASTM
ers of 450 g/m2 emulsion mat and resin C297. The fracture stress shall be minimum
rich surface is cured at 20°C for 48 9 MPa. The fracture is not to be evidencing
hours. brittle fracture with even surfaces.
• A secondary laminate also consisting of
3 layers of 450 g/m2 mat is built on top
of the primary laminate without any
Table 5-2 Requirements to Glass Fibre Reinforcements

Property Test method Number and selection Acceptance criterion
of parallels (required value for approval
testing)
Moisture content ISO 3344 (1997) Min. 3 parallels, Maximum 0.2 %.
sampling according to
the test standard
Loss on ignition ISO 1887 (1995) min. 5 parallels, Manufacturer’s nominal value
sampling according to Tolerance limits for the
(ISO 1987??) the test standard. various materials are subject
to approval in each separate
case.
Weight per unit Rovings (yarn): min. 5 parallels, The arithmetic mean ± 2 x
length or area ISO 1889 (1997) sampling according to standard deviation shall be
the test standard. within the manufacturer's
value ± 10 %.
Mats:
ISO 3374 (1990)
Woven Fabrics:
ISO 4605 (1978)
Tensile min. 5 parallels, Continuous Fibres/Weaves
ISO 527-4,5 (1997) from at least 2 batches and fabrics 1)
- strength Manufacturer’s specified
minimum value
- modulus Manufacturer’s specified
value
- elongation Manufacturer’s specified
minimum value
1)
To be tested in main fibre directions
5 – Rotor 107
Table 5-3 Testing of Polyester and Vinylester in liquid condition

Property Test standard Acceptance Criterion (required value
for approval testing)
Viscosity1 ISO 2555 (1989) MSV ± 250 mPas
(Brookfield) or
ISO 2884 (1974)
(Cone/plate)
Acid value ISO 2114 (1975) MNV ±10%
Monomer content ISO 3251(1993) MSV ± 2 % or MNV±5%
Mineral content DIN 1645 (1976),
B1.1, 4.10
Gel time2 ISO 2535 (1997) for curing time at room temperature:
<60 min.: MSV± 5 min
120-60 min.: MSV± 10 min
>120 min: MSV± 15 min
Linear curing shrinkage ASTM D 2566-69 MNV
1
For polyester and vinylester the following parameters should be used; Viscometer type A,
rotational frequency 10, temperature 23 °C. Viscosity can be accepted with Manufacturer’s
specified value ± 20 %.
2
Specify activator and initiator and % of each.
MNV = Manufacturer’s nominal value
MSV = Manufacturer’s specifed value
Table 5-4 Testing of mechanical properties of the cured end product (laminate of
fibre-reinforced plastics)
Property Test standard Number of tests
Density ISO 1675 (1985 ) min. 3

Volumetric curing shrinkage ISO 3521 (1997) min. 3 *
1
Ultimate tensile strength ISO 527-1(1993) min. 5
Tensile modulus ISO 527-1,2(1993) min. 5
Fracture elongation ISO 527-1,2(1993) min. 5 *
Ultimate flexural strength ISO 178 (1993) min. 5
Flexural modulus ISO 178 (1993) min. 5
2
Barcol hardness EN 59 (1990) Measured on all test speci-
ASTM D 2583 (1995) mens
Heat deflection temperature (HDT) ISO 75-2 (1993) min. 5 *
1
Test samples for tensile testing ISO 527-2/1A/50; test specimen 1A and test speed 50
mm/min.
2
Resin may deviate from these values, provided a minimum value of 30 is met and the
manufacturer can demonstrate adequate cure.
* Barcol hardness should be measured on each specimen and comply with manufacturer's
specified value.
108 5 – Rotor
Gelcoat and Topcoat patible with resins based on polyester, vi-

Gelcoat and topcoat should be produced of nylester and epoxy and core materials such
polyester satisfying the requirements out- as PVC and balsa. Limited compatibility
lined above. The resin must not contain may require special considerations.
more than 15% pigments and other filling
compounds and must cover completely at a Properties to be verified for sandwich adhe-
thickness of max. 500 µm. sives and mortar materials are:
Sandwich Core Materials For uncured material:

Core materials must have stable long term • Viscosity, ASTM 1084 method B for
properties. Continuous chemical processes, free-flowing adhesive, or equivalent.
diffusion, etc. must not affect the physical • Linear curing shrinkage, ASTM 2566 or
properties of the core or the core-laminate equivalent.
interface. The surface of the material should For cured material in the joint:
normally be such that no further machining • Tensile strength, ASTM C 297
or grinding is required to obtain proper • Shear strength, ISO 1922
bonding of the material. Typical sandwich • Embrittlement, to be assessed by meas-
core materials are closed-cell polymeric uring fracture elongation at low and
foams and end-grain balsa cores. Core mate- high temperatures
rials should normally be compatible with
resins based on polyester, vinylester and Fatigue properties should be investigated
epoxy. Core materials with a limited com- when relevant.
patibility will need special considerations.
Connections
Properties to be verified for sandwich core It is crucial that embedded bolts and other
materials include, but are not necessarily metallic parts embedded in the rotor blade
limited to, the following: material are absolutely clean and free from
any dirt. Sandblasting may be necessary to
• Compressive strength at 60°C not less meet this requirement.
than 50% of the value at 23°C.
• Density: nominal value± 10%. Note that a too weak hub may cause exces-
sive bending of the bolts that connect the
For core materials subject to significant cy- blades to the hub.
clic shear loading, an extensive test program
will normally be required including a fatigue Corrosion protection
test. Choice of materials and corrosion protection
measures to counteract the effects of possi-
Sandwich Adhesives and Mortars bly aggressive environmental conditions
Sandwich adhesives must have stable long needs thorough consideration with a view to
term properties. Continuous chemical proc- the fact that rotor blades are not easily ac-
esses, diffusion, etc. must not affect the cessible and renewed application of a pro-
physical properties of the adhesives or the tective coating may not be feasible.
core-laminate interface. The adhesive should
normally be such that no further surface Note that carbon fibres used as reinforce-
treatment is required to obtain proper bond- ment of the rotor blades will cause galvanic
ing to the core and laminate surfaces. Sand- corrosion of any steel parts they get in touch
wich adhesives should normally be com- with, unless stainless steel is used for the
5 – Rotor 109
making of these parts. Stainless steel should the general loads on the blades and may
therefore be used for any steel components need extra attention.
in blade parts which are not easily accessi-
ble. As blades are getting larger, large unsup-
ported panels will be present between the
5.1.4 Strength analyses webs and the leading and trailing edges.
This may have an impact on the stability of
General
the blades. Therefore, the buckling capacity
Structural analyses of the rotor blades must
of a blade must be verified by a separate
be carried out for all relevant load cases in
calculation, in addition to the full-scale test.
order to verify that the strength of the blades
For this calculation, a FEM analysis will
is sufficient to withstand the loads that these
normally be required. Furthermore, buckling
load cases exert on the blades. By the
of the webs may also have to be considered.
strength calculations in these analyses, it
must be verified that both the ultimate
As rotor blades become longer, evaluation
strength and the fatigue strength, for a given
of the stability against buckling becomes
design life, are sufficient. For structural
still more important because of the large
parts in compression, also stability against
unstiffened panel segments, which are then
buckling must be considered.
usually involved.
For each load case, a set of design loads is
Standards
established by multiplying the relevant char-
The following standards are normally used
acteristic loads by partial safety factors for
for verification of wind turbine blades.
load. The standards applied for this purpose
• DS472, “Last og sikkerhed for vind-
should be quoted when the design loads are
møllekonstruktioner” (“Load and Safety
documented. In principle, each load case can
for Wind Turbine Structures”, in Dan-
be defined in terms of six load components
ish), DS472, 1st edition, Dansk In-
and their variation over the blade span. The
geniørforening, Copenhagen, Den-
resolution used to specify this variation must
mark, 1992.
be fine enough to allow for sufficiently ac-
curate calculations in all points of interest • DS456 “Konstruktioner af glasfiber-
along the blade, especially in all critical areas, armeret umættet polyester” (“Structural
e.g., wherever changes in geometry or mate- Use of Glass Fibre Reinforced Unsatu-
rial occur. rated Poyester”, in Danish), DS456, 1st
edition, Dansk Ingeniørforening, Co-
Environmental conditions, which affect the penhagen, Denmark, 1985.
material behaviour, should be considered
and taken into account. Such conditions in Ultimate strength
particular include humidity and temperature, When the direction of the load is time-
which may both lead to degradation of dependent, information about phase and fre-
strength and stiffness, and their design ef- quency should be given. For each section of
fects calculated by multiplying characteristic interest along the blade, the design loads have
effects by appropriate partial safety factors to be calculated. In principle, all six load
should be applied in the strength analysis. components need to be calculated. Normally,
the bending moments and the shear forces are
Loads on critical components such as tip most critical, but also the torsional moment
brakes are often different in character from and the axial force can in some cases be im-
portant for the design.
110 5 – Rotor
Once the necessary information regarding teristic value for σF,T. Note that very errone-
material strength and stiffness, geometry of ous numbers may result for the characteristic
blade and lay-up of laminate is established, values of σF,T if they are interpreted on the
the capacity of the blade can be calculated basis of a Weibull distribution assumption
section by section. This can be done more or for σF,T.
less by hand, or by some calibrated finite ele-
ment program. In principle, all six capacity The characteristic value for a strength prop-
components need to be calculated, e.g. flap- erty is usually defined as a particular quan-
wise and edgewise bending moment capaci- tile in the probability distribution of the
ties. When characteristic values for material property. This is often a quantile in the
strength are used as input, characteristic ca- lower tail of the distribution, e.g., the 2% or
pacity values result. The design capacities are 5% quantile. Note that different standards
then found by dividing the characteristic ca- may define the characteristic value differ-
pacities by a materials factor, i.e. a partial ently, i.e., it is not always defined as the
safety factor for materials, cf. Section 2.3. same percentile in the different standards.
Accordingly, different standards may pre-
It is, in general, to be checked for each sec- scribe different partial safety factors to be
tion of interest along the blade, and for all used with their respective characteristic val-
six load components in that section, that the ues for design. As an example, DS472 de-
calculated design load does not exceed the fines the characteristic value of an FRP ma-
corresponding calculated design capacity. terial as the 5% quantile and requires a ma-
terials factor of 1.7, while DS456 defines the
In this context, the tensile strength σF,T in the characteristic value as the 10% quantile and
direction of the fibres is one of the strengths, requires a materials factor of 1.8. It is essen-
which is important to consider. This strength tial in design that the characteristic value at
is dominated by the strength σF,B of the fibre stage is combined with the correct partial
bundles. The strength σF,B of a fiber bundle is safety factor. It is not licit to combine the
proportional to the mean failure stress σ of characteristic value of one standard with the
the individual fibres, partial safety factor of another, as this can
lead to erroneous results and unsafe designs.
exp(1 / m)
σ F ,B = σ Stability
1
m1 m Γ (1 + ) The longer the blades, the more likely is
m
stability against buckling to govern the de-
sign of the blades instead of ultimate
where m is a material constant and Γ denotes
strength. The stability against buckling can
the gamma function. Reference is made to
most easily be verified by calculations by
Beaumont and Schultz (1990). The individ-
means of a properly calibrated finite element
ual fibre strengths σ very often follow a program. Performing such a calibration by
Weibull distribution, however, their mean hand is more difficult owing to the complex
will under the central limit theorem follow a geometry of a rotor blade, and well-proven
normal distribution, and σ and hence also tools for this purpose, such as closed-form
σF,T can then be deduced to be normally solutions for the buckling capacity, are not
distributed. The normal distribution prereq- available.
uisite for σF,T is important when a particular
lower-tail quantile of the distribution of σF,T When designing for a sufficient stability
is to be interpreted and used as the charac- against buckling, it is a standard approach to
5 – Rotor 111
apply an extra design margin to take care of There is usually different S−N curves for
effects of geometrical imperfection, fibre different values of the ratio R between
misalignment, workmanship etc. minimum and maximum stress. For a fully
reversed loading about a zero mean stress,
Fatigue Strength R=−1. For a pure tensile loading about a
Sufficient fatigue strength must be docu- non-zero mean stress, R≥0. The effect of R
mented. This applies to all sections along the on the S−N curve is often disregarded and
blades and to all directions in each location. one common S−N curve is applied regard-
For this purpose, all load components should less of R. However, data for interpretation of
be given in all points of relevance along the such a common S−N curve are usually ob-
blade, including phase and frequency infor- tained from tests carried out for one fixed R
mation. This is automatically fulfilled when value, whose value and sign depend on
the six load components are given as time whether the test specimens are subject to
series. From the time series of the six load bending or pure tension. For the case that
components, long-term stress distributions
one common S−N curve is applied, the cu-
can be established in all points of relevance.
mulative fatigue damage over the design life
In principle, this includes distributions of the
can be predicted by the Miner’s sum,
mean stress as well as distributions of the
stress range that represents the variation
∆n ( S i )
about the mean stress. Rain-flow counting is D=å
a commonly used method for this purpose. i N (S i )
The total number of stress cycles in the de-
sign life can also be extracted from the time in which the sum is over the stress range
series and can be used to transform the stress values Si on the discretised stress range axis,
range distribution into a design lifetime his- ∆n(Si) is the number of stress cycles in the
togram of stress ranges. A sufficiently fine ith stress range bin along this axis as deter-
discretisation of the stress range axis needs mined from the design histogram of stress
to be chosen for this purpose. ranges, and N(Si) is the number of cycles to
failure at the stress range Si of this bin as
Fatigue strength data are usually reported as determined from the design S−N curve.
S−N curves, which are sometimes referred to
as Wöhler curves. S−N curves give the num- In each section of interest along the blade it
ber of cycles N to failure for specified stress must be verified that the fatigue strength is
range S. S−N curves which are representa- not exceeded. This is in practice done by
tive for the blade material must be estab- checking that the Miner’s sum, calculated
lished. S−N curves are usually linear on a for the design stress range histogram in
log-log scale, lnN=lnK−mlnS. Test data usu- conjunction with the design S−N curve, does
ally give a mean S−N curve and a natural not exceed a critical value which is usually
variability of lnN about this mean, often equal to 1.0.
expressed in terms of a standard deviation σ.
A characteristic S−N curve is often ex- When the effect of R on the S−N curve is to
pressed as the mean curve minus two stan- be accounted for, the assessment of the cu-
dard deviations, hence lnN=lnK−mlnS−2σ. mulative damage can be carried out by
The design S−N curve is established by di- means of Goodman diagrams and becomes
viding all S values of the characteristic S−N somewhat more cumbersome.
curve by a materials factor.
112 5 – Rotor
Fatigue in edgewise bending is dominated insufficient to withstand the shear loads that
by gravity loads, and depends to a large ex- occur in the blades. It is in this context im-
tent on the weight of the blade and the actual portant to consider also a sufficient fibre
number of rotations of the rotor during the strength perpendicular to the blade axis.
design life. Fatigue in flapwise bending is
dominated by the blade response to aerody- 5.1.5 Tip deflections
namic loads exerted by the wind. Note that
A deflection analysis of a blade must be
transient loads during start/stop and loads
carried out. As a part of the deflection
due to yaw errors may give significant con-
analyses it must be proven, for all load
tributions to the cumulative fatigue damage
cases, that the ultimate tip deflection (caused
and need to be given thorough consideration
by a static load or by an interaction of dy-
in addition to the loads that occur in normal
namic loading and structural response) is
operation. For an example of fatigue calcu-
acceptable.
lations and fatigue design, see Section 2.3.7.
The tip deflection can normally be calcu-
Frequency
lated by means of an aeroelastic computer
As a minimum, the two lowest eigenfre-
code, but it can also be calculated by hand or
quencies of the rotor blade, both in flapwise
by some finite element program. Knowing
and edgewise oscillation, should be calcu-
the initial distance from the blade tip to the
lated. These eigenfrequencies should be
tower in the no-load condition, this allows
compared to the rotational frequencies of the
for determination of the clearance between
wind turbine. A sufficient margin to these
the rotor blade and the tower. The clearance
frequencies must be available to avoid reso-
should be determined by using the most un-
nance of the blade. It is recommended to
favourable combination of geometrical tol-
keep the eigenfrequencies outside a range
erances and characteristic stiffnesses of the
defined as the rotational frequency ±12%.
rotor blade and its supports. The effect of
damping may be important and should be
Calibration of Design Tools considered. Further, if the rotor or its sup-
It is essential only to use well-proven and
ports is subject to creep, shrinkage, tem-
properly calibrated design tools. Especially
perature deformations or degradation with
in cases of advanced computer programs
time, this must be allowed for in the clear-
such as finite element programs, it is im-
ance measure.
portant to tune or calibrate the results from
use of the computer program models against
When comparing the resulting available
results achieved from full-scale tests on ro-
clearance between the rotor blade tip and the
tor blades, so as to obtain models which best
tower, a specified minimum clearance must
possible reflect reality. Note that caution
be met. Danish rules require the clearance to
must be exercised when it comes to gener-
be calculated for the characteristic extreme
alization of model calibration results, as
load on the blade times a safety factor of
calibration results obtained for a model of
1.3, and requires this clearance to be at least
one blade cannot automatically be trans-
0, i.e. the blade shall not reach to the tower
ported and rendered valid for a model of
when subjected to the extreme design load.
another blade.
Dutch rules have a similar requirement, but
this is stricter in the sense that the safety
Delamination factor on the load is to be taken as 1.5 rather
Delamination may occur if the shear
than 1.3.
strength of the fibre-reinforced laminate is
5 – Rotor 113
In order to increase the distance between the

blade tip and the tower during operation, the The latter approach is the most common
rotor can be coned, the rotor plane can be approach. By this approach, a lightning con-
tilted, and the blade can be produced with a ductor is mounted in the blade tip, and
predeflection. through a commutator lug in the interior of
the blade, the current is directed to the blade
Normally, the stiffness and mass properties root. The effectiveness of this method is
of the blade as used in the tip deflection cal- expected to be much dependent on the size
culations have to be verified through the test of the blade and on the amount of metals or
of the blade. carbon fibres used in the blade. Note that a
lightning protection like the ones presented
5.1.6 Lightning protection here cannot be expected to work in abso-
lutely all cases. A blade may be hit by many
Lightning striking a rotor blade may cause
lightning strokes over its design life, and
damage to the blade such as – in the extreme
there may be cases for which the protection
case – peeling at the back edge. Lightning
will fail, e.g. if the conductor cable fuses.
protection of rotor blades can in principle be
For large blades, i.e. blades longer than
designed in two different ways:
about 20 m, it may be necessary to secure
• The lightning can be prevented from the blades against damages from lightning
penetrating into the blade by diverting that strikes the blade in other locations than
the current along a prepared current the blade tip. Note also that it is necessary to
path, such as a conductive tape, on or in account for the limited conductivity of the
the surface of the blade. carbon fiber materials that are used in tip
• The lightning can be directed from the shafts. From the aircraft industry, methods
point of stroke on the blade through the exist for lightning protection of glass fiber
blade by way of a conductor cable. and carbon fiber materials by which these
materials are made electrically conductive
by means of metallic sheets, nets, or threads,
rather than by mounting metallic conductors
on the material surfaces. Reference is made
to DEFU (1999). Examples are shown in
Figure 5-4.
5.1.7 Blade testing

The purpose of rotor blade testing is to ver-
ify that the laminates in the blade are struc-
turally safe, i.e. that the plies of the rotor
blade do not separate by delamination in
static or cyclic loading. The purpose is also
to verify that the fibres do not fail under
repeated loading. Full-scale blades are used
for the rotor blade tests, and tests are carried
out for verification of fatigue strength as
well as static strength.
Figure 5-4 Examples of how lightning pro-
tection of blades can be arranged, from For every new type of blade to be manufac-
DEFU (1999) tured, one blade is to be tested statically and
114 5 – Rotor
dynamically to verify the strength of the trum including a test factor of 1.3. Nor-
blade design. mally, the test is carried out at a fixed load
level and for a number of load cycles of
The production of a test blade is to be in- between 2⋅106 and 5⋅106 cycles.
spected in order to verify that the blade is
representative for the blade design. Alterna- Infrared cameras can be used to reveal local
tively, the blade may be sampled at random build-up of heat in the blade. This may ei-
from the blade production. However, this ther indicate an area with structural damp-
alternative approach is usually not feasible, ing, i.e. an area where the blade designer has
because it is often desirable to carry out the deliberately laid out fibres which convert the
blade test early during a production, and bending energy into heat in order to stabilise
therefore the test blade is often selected on a the blade, or it may indicate a zone of de-
deterministic basis as one of the first blades lamination or a zone where the fibres are
in the production series. close to failure.
The test blade is to be equipped with strain Furthermore, the lowest natural frequencies
gauges and displacement transducers. The and corresponding damping ratios of the
measurement results from the strain gauges blade are measured as these are important
are continuously monitored on computers. input parameters for the load calculations.
Nonlinear variations in the pattern of bend- This is achieved by excitation of the blade at
ing may reveal a damage in the rotor blade different frequencies and in different direc-
structure. tions. It is essential that the natural frequen-
cies of the blade do not coincide with the
The blade is to be tested statically by a static rotational frequencies of the wind turbine,
load in two opposite directions flapwise and and the purpose of this part of the test is to
in two opposite directions edgewise. The make sure that the natural frequencies of the
static load must at least equal the extreme blade differ from the rotational frequency
design load. Tests in two opposite directions with sufficient margin. Note that also the
instead of just in one direction are necessary aerodynamic damping is measured, in par-
due to lack of symmetry in the blade. Meas- ticular in the flapwise direction.
urements by means of the strain gauges and
displacement transducers are used to moni- In case the natural damping of the blade is
tor and verify that the strains and displace- insufficient to avoid vibrations, e.g. edge-
ments stay within the design limits and de- wise vibrations, a damper may be built into
sign calculations during the entire test. Non- the blade. The damper may consist of mate-
linearity in the measured strains may indi- rials with high internal damping, which is
cate buckling or damage to the blade. built into the blade, or it may be a mechani-
cal damper. It is essential that this damper is
Dynamic testing is also carried out for also included in the test blade, since the de-
loading in both flapwise and edgewise di- sign of the connection between the damper
rections. Normally the blade is tested for and the blade may introduce weak areas or
loading in one direction at a time, but a load large stiffness changes.
with simultaneous components in both di-
rections may also be applied. The dynamic A rotor blade can also be tested for its resid-
testing is carried out as an accelerated test, ual static strength (and thus its ability to
i.e. at a load level, which causes the same withstand extreme loads in the long term) by
damage to the blade as the true load spec- being bent once by a very large force. The
5 – Rotor 115
magnitude of this force is usually deter- 5.2 Hub

mined by the load level of the static test de-
The hub is the fixture for attaching the
scribed above. The residual strength test is
blades to the rotor shaft. It usually consists
usually performed after the blade has been
of nodular cast iron components for distri-
subject to fatigue testing, and the purpose is
bution of the blade loads to the wind support
to verify that the static strength of a blade,
structure, i.e., ultimately to the tower. A
which has been in operation for a substantial
major reason for using cast iron is the com-
amount of time, is sufficient and that the
plex shape of the hub, which makes it hard
stiffness – although it may have changed – is
to produce in any other way. In addition
still acceptable.
comes that it must be highly resistant to
metal fatigue, for which reason any welded
A test specification for the blade test should
hub structure has to be left out as less feasi-
be worked out. The specification should
ble.
specify the loads and the measurement
setup, including means of avoiding overload
of the blade. This is an important aspect, as
the dynamic test is normally carried out at a
frequency close to the lowest flap-
wise/edgewise natural frequency, at which
even small variations in the energy input
may significantly alter the load amplitude.
The test loads must be determined in such a
way that major parts of the blade are assured
to become tested to the required load level.
Note that in case the test is running outdoor,
weather conditions such as sun and rain may
influence the resulting blade deflections.
During testing the stiffness of the test rig

must be considered. Ideally, the stiffness of
the test rig should be comparable to that of
the blade-hub connection when the blade is Figure 5-5 View of wind turbine hub
installed on the turbine. As a minimum, the
stiffness of the test rig must be measured to The moments and forces transmitted to the
allow for adjustment of the measured de- hub and tower depend on the type of hub.
flections. Three types of hub are common
• hingeless rigid hub, has cantilevered
A guidance on blade testing is given in IEC blades and transmits all moments to the
61400-23 TS Ed. 1: “Wind turbine generator tower.
systems - Part 23. Full-scale structural test- • teetering rotor, has two rigidly con-
ing of rotor blades for WTGS's”. nected blades supported by a teeter-pin
joint, which can only transmit in-plane
5.1.8 Maintenance moments to the hub. Flapwise moments
The blade should be kept clean to assure that are not transmitted.
the aerodynamic properties of the blade re- • articulated hub, has free hinges in flap-
main unchanged. ping and lead-lag, so there is no me-
116 5 – Rotor
chanical restraint moment on the blades • full flapping moment

in either flapping or lead-lag. • flapping shear, resulting thrust on one
The hingeless hub is the most common con- blade
figuration for wind turbine hubs. • lead-lag moment, power torque of one
blade, and gravity loads
Figure 5-5 shows an example of such a hub, • lead-lag shear, in-plane force that pro-
and Figures 5-6 and 5-7 show the hub in the duces power torque
context of the transmission system, in which • centrifugal forces
it forms a part of the link between the rotor • pitching moments of one blade
blades and the generator. Figures 5-6 and 5- Design loads can be calculated from the
7 provide examples of two different bearing blade loads in accordance with this list.
arrangements with one and two main bear-
ings, respectively. For details, see Section It is obviously structurally beneficial to re-
6.2. duce the loads at the blade-hub interfaces, in
particular the large blade-flapping moment.
The blade-root flapping moment can be
traded off against tip deflection. Flexible
blades have the advantage of relieving some
of the load into centrifugal terms. Other con-
siderations apply in this context, since yaw
stability and fatigue life are both signifi-
cantly affected by blade flexibility.
Figure 5-6 Transmission system consisting The hub should be constructed in such a
of hub, one main bearing, main shaft, gear way that it will be possible to adjust the tip
and coupling. Courtesy Bonus Energy A/S. angle of the blade and tighten up the bolt
connections.
5.2.2 Strength Analyses

Cast components should be designed with
smooth transitions and fillets in order to
limit geometrical stress concentrations.
The characteristic resistance for cast iron

components can be taken as the 5th or 95th
percentile of test results, whichever is the
most unfavourable.
Figure 5-7 Transmission system consisting The structural resistance in the ultimate limit
of hub, two main bearings, main shaft and state is to be determined by elastic theory.
gear, and with the generator marked out When design is carried out according to the
allowable stress method, the maximum al-
5.2.1 Determination of Design Loads lowable stress should be taken as the char-
The loads at the blade-hub interfaces to be acteristic resistance divided by a safety fac-
considered at the blade root for design of the tor.
hub consist of the following
5 – Rotor 117
Fatigue design can be carried out by meth- • Treaded steel bushes are mounted di-
ods based on fatigue tests and cumulative rectly into the blade root and fixed to
damage analyses. When design is carried out the blade by glue.
according to the allowable stress method, In both cases, the bolts from the blade pass
the allowable fatigue stress range is to be through a flange on the cast hub. The bolt
taken as the characteristic resistance divided holes in this flange can be made somewhat
by a safety factor. The characteristic resis- elongated to enable adjustment of the tip
tance for a given number of cycles to failure angle.
is defined as the stress range that corre-
sponds to 95% survival probability. The For these bolt connections, a bolt tension
characteristic resistance should be modified procedure is usually required. Such a proce-
to account for size effects, surface condi- dure should usually specify
tions and mean stress. When limited test • bolt, nut, washer type, dimension and
data are available for estimation of the char- quality
acteristic resistance, the characteristic resis- • flatness tolerances for surfaces
tance should be given with 95% confidence. • roughness of surface
• surface treatment and protection
The layout of a wind turbine hub makes it • bolt tensioning sequence and method
often hard to determine which section is the
• treatment of treads, e.g., waxing
structurally most critical section of the hub.
• torque to be applied
In this context, the Finite Element Method
The relation between bolt torque and bolt
(FEM) forms a suitable tool for strength
tension may be subject to test.
analysis of the hub and can be used in con-
junction with state-of-the-art fatigue analysis
5.2.4 Hub Enclosure
techniques to determine the fatigue life and
to optimize the design with respect to The hub enclosure, which is sometimes re-
strength and cost. It is advisable to qualify ferred to as the nose cone, is usually made
the loading assumptions for such FEM up from glassfiber-reinforced polyester. In
analyses by measurements. cases where the hub enclosure is large, it is
recommended to consider the wind load it
The FEM analyses can in particular be used will be exposed to.
to document that a satisfactory strength is
available in critical sections, such as at stress 5.2.5 Materials
concentrations and stiffness transitions, as
Spheroidal graphite cast iron, also known as
well as at the shaft-hub connection and the
nodular cast iron, is the preferred material
interfaces to the blades.
for the hub. Cast iron is classified according
to its mechanical properties, such as strength
5.2.3 Analyses of Bolt Connections
and hardness, in EN1563. Cast hubs are
The blades are usually bolted to the hub. usually tested by nondestructive testing
There are two techniques for mounting the (NDT) for verification of the mechanical
bolts in the blades: properties and for detection of possible de-
• A flange is established at the blade root fects and internal discontinuities. The fol-
by moulding the glass-fiber reinforced lowing NDT methods are available:
plastic to form a ring, in which steel • ultrasonic inspection
bushes for the bolts are embedded. • magnetic particle inspection
• visual inspection
• hardness measurements
118 5 – Rotor
Ultrasonic inspection can be carried out as EN1369 Founding – Magnetic particle in-
point testing or, more thoroughly, as com- spection. European Standard. CEN, 1996.
plete scanning. For ultrasonic inspection, it
is common to assign different acceptance
REFERENCES
criteria, e.g. in terms of different allowable
defect sizes, to different areas of the cast Beaumont, P.W.R., and J.M. Schultz, “Sta-
hub. These areas and the assigned accep- tistical Aspects of Fracture,” Chapter 4.6 in
tance criteria should be indicated on the Failure Analysis of Composite Materials,
drawing of the casting. Usually a strict ac- Volume 4, ed. by P.W.R. Beaumont, J.M.
ceptance criterion is assigned to an area with Schultz, and K. Friedrich, Technomic Pub-
high stresses, whereas a more lax acceptance lishing Co., Inc., Lancaster, Penn., 1990.
criterion is assigned to an area with low
stresses. The stricter the acceptance criterion Danske Elværkers Forenings Undersøgelser
in a particular area, the more thorough is the (DEFU), ”Lynbeskyttelse af vindmøller. Del
ultrasonic inspection in that area, and the 7: Vinger,” DEFU Report No. TR394-7 (in
stricter are the requirements to the allowable Danish), 1999.
size of detected discontinuities.
DNV, Rules for Classification of High
It is important to consider whether the cho- Speed and Light Craft, Materials and
sen structural material possesses the neces- Welding, Part 2, Chapter 4, Composite Ma-
sary ductility. Low temperatures can be terials, Det Norske Veritas, Høvik, Norway,
critical for cast hubs in this respect, and the 1999.
choice of hub material should therefore be
made with due consideration of the tem- Hück, M., “Berechnung von Wöhlerlinien
peratures of the surroundings. für Bauteile aus Stahl, Stahlguss und
Grauguss,” Verein Deutscher Eisenhütten-
Note that repair of cast hubs by means of leute Bericht Nr. ABF 11 (in German), July
welding is not permitted. 1983.
5.2.6 Standards Hück, M., W. Schütz, and H. Walter, “Mod-

ern Fatigue Data for Dimensioning Vehicle
EN1563 Founding – Spheroidal graphite
Components of Ductile and Malleable Cast
cast irons. CEN, 1997.
Iron,” first published in Automobiltechnis-
che Zeitschrift 86, Nos. 7, 8 and 9, 1984.
5 – Rotor 119
120 5 – Rotor
Guidelines for Design of Wind Turbines – DNV/Risø
6. Nacelle Figure 6-1 illustrates the loads and reactions

that the main shaft is subjected to. Under
normal circumstances reactions in the main
6.1 Main Shaft shaft can be calculated from equilibrium
The main shaft transmits the rotational en- considerations.
ergy from the rotor hub to the gearbox or
directly to the generator. Moreover the pur- mgear My
pose of the main shaft is to transfer loads to
the fixed system of the nacelle. In addition
to the aerodynamic loads from the rotor, the
main shaft is exposed to gravitational loads Mz
and reactions from bearings and gear. R2v
mhub
R1h
The main shaft is also subjected to torsional Fy
vibrations in the drive train. Such vibrations Fx R1v
My
will usually be of importance for possible Mx
frictional couplings like shrink fit couplings Figure 6-1. Loads and reactions on the main
between shaft and gear. shaft.
A wind turbine can be exposed to large tran- For each load case, a set of design loads are
sient loads. Therefore, it has to be consid- established by multiplying the relevant char-
ered whether the chosen structural material acteristic loads by a partial safety factor for
possesses the necessary ductility. This is load. The loads can be combined by simple
particularly important if the turbine is to be superposition.
operated at low temperatures. It should be
ensured that the shaft is protected against Fatigue loads consist of histories of stress
corrosion, because corrosion may imply a amplitudes. There is one history per load
considerable reduction of the assumed fa- component. Hence, combination of fatigue
tigue capacity. Suitable quality assurance loads implies combination of stress histories.
should be implemented to make sure that the Unless the phase differences between the
geometrical and mechanical assumptions for individual load components are known, the
the design are fulfilled, e.g., surface rough- largest stress amplitude of one load compo-
ness, that the specified values of material nent is to be added to the largest stress am-
parameters are met, and that the imperfec- plitude of each of the other load compo-
tions of the material do not exceed any criti- nents, the second largest amplitude is to be
cal level. added to the second largest amplitude of
each of the other load components, and so
6.1.1 Determination of Design Loads forth.
In the following, it is assumed that all rele-
vant load cases are taken into consideration. 6.1.2 Strength Analysis
Selection of the relevant load cases and de- Structural analysis of the main shaft shall be
termination of the characteristic values of carried out for all relevant load cases in or-
the individual load components on the rotor der to verify that the strength of the shaft is
can be made according to procedures given sufficient to withstand the loads it is sub-
in Sections 4.4-4.8 for both extreme loads jected to.
and fatigue loads.
6 – Nacelle 121
It shall be verified that both the ultimate Geometrical size effect

strength and the fatigue strength are suffi- The reduction due to the geometrical size
cient for the actual design life. effect is based on the fact that specimens of
different dimensions exhibit differences in
6.1.3 Fatigue Strength the fatigue properties, despite they are made
from materials of identical initial dimen-
The fatigue strength can be expressed in
sions. This can be accounted for by means
terms of the particular stress amplitude that
of the influence factor K2, applied to the
leads to failure after a specified number of
fatigue strength. K2 can be determined from
constant-amplitude stress cycles. This can be
Figure 6-3. If fatigue strength in pure ten-
expressed in terms of an S−N curve, also sion-compression is considered as a basis for
known as a Wöhler curve, which gives the the calculations, K2 can be set equal to 1.
number of cycles N to failure at stress am-
plitude S.
Effects of surface roughness
Fatigue cracks are often initiated from un-
Various factors influence the strength rela- evennesses in the surface and from small
tive to reference values determined from surface cracks. An ideal “reference surface”
small specimens in laboratory tests, and may for the test specimens corresponds to a pol-
lead to a reduction of this strength. In this ished surface with a surface roughness of
respect, the following elements of impor-
Ra=0.05-0.1µm. The surface roughness that
tance should, as a minimum, be considered:
can be expected from a careful machining is
• Technological size effect
Ra=0.4-1.6µm. A reduction factor Kr to ac-
• Geometrical size effect count for the effect of surface conditions
• Surface roughness deviating from the reference surface can be
• Stress concentrations determined from Figure 6-4.
• Stress ratio R, defined as the ratio be-
tween the minimum stress and the Technological size effect
maximum stress with tensile stresses

defined as positive and compressive 1
stresses as negative.
0,95
structural
Technological size effect steel
0,9
Reduction due to the technological size ef-
fect is based on the fact that specimens of Alloy steels
K1 0,85
identical dimensions, made from materials
of the same kind but of different dimensions, 0,8
have different fatigue properties. Test
specimens are made with relatively small 0,75
dimensions (typical diameter d=5-10 mm)
and have had their mechanical properties 0,7
improved as a result of the reduction of the 1 10 100 1000
cross section by forging or rolling. The d [mm]
technological size effect can be accounted

for by means of the influence factor K1, ap- Figure 6-2 Tecnological size effect (d is
plied to the fatigue strength. K1 can be de- diameter of considered part of shaft), from
termined from Fig. 6-2. Roloff and Matek (1994).
122 6 – Nacelle
couplings and tight bearing fits. A consid-

Geometrical size effect
erably amount of stress concentration factors
for various common shapes are given in
1
Peterson (1974). The sensitivity of the mate-
0,95 Tension-
rial to such stress concentrations primarily
compression depends on the ratio between the yield
0,9 strength and ultimate strength of the material
Bending and and, on the stress gradient in the considered
K2 0,85 torsion part of the structure. This so-called notch
sensitivity can be accounted for by means of
0,8 the notch sensitivity factor q
0,75
1
q=
0,7 1+ A
1 10 100 1000 ρ
d [mm]
in which ρ is the notch radius and A is the

Figure 6-3 Geometrical size effect (d is di-
Neuber factor, which can be determined
ameter of considered part of shaft), from
from Figure 6-5. Note that the notch radius
Roloff and Matek (1994).
appears as “rc” in the drawing in Figure 6-6.
Note that Ra denotes an average surface
roughness. In some literature, the peak sur-
face roughness Rz is given rather than Ra.
There is no unambiguous relation between
Ra and Rz, but for preliminary calculations
one may consider the peak surface rough-
ness Rz equal to approximately 6 times Ra.
Figure 6-5 Notch sensitivity factor q, from

Sundström (1998).
Figure 6-4 Reduction factor KR vs. surface

roughness Ra, from Sundström (1998). Figure 6-6 Shaft lay-out in vicinity of
bearings, from SKF (1989).
Stress concentrations
Stress concentrations will occur at local Whenever a notch is encountered, the notch
changes in geometry, e.g., where cross sec- sensitivity factor q can be used to modify the
tions alter. This is also the case at shrink fit
6 – Nacelle 123
stress concentration factor α into a notch

Stress concentration factor, torsion
factor β to be used instead:
2,1
β=1+q(α-1) h/r=4
2
A commonly encountered notch is found by h/r=3
the recess, which is to transfer the axial load 1,9
from a main bearing. A common lay-out of a
α 1,8
shaft in the vicinity of bearings is shown in h/r=2
Figure 6-6, and data for the involved geo-
1,7
metrical quantities are given in Table 6-1.
1,6
rs ba ha rc
mm mm mm mm
1 2 0.2 1.3 1,5
1.1 2.4 0.3 1.5 0,005 0,01 0,015 0,02
1.5 3.2 0.4 2 r/D
2 4 0.5 2.5
2.1 4 0.5 2.5 Figure 6-8 Stress concentration factor, from
3 4.7 0.5 3 Roloff and Matek (1994).
4 5.9 0.5 4
5 7.4 0.6 5
6 8.6 0.6 6 The following symbols are used in Figures
7.5 10 0.6 7 6-7 and 6-8:
Table 6-1 D=larger diameter of shaft
d=smaller diameter of shaft
The geometrical stress concentration factor h=(D-d)/2
is denoted α and can be determined from r=radius of notch
Figures 6-7 and 6-8.
Stress concentrations due to tight bearing
Stress concentration factor, bending
fits, corresponding to the ISO 286 shaft tol-
erances from m to r can be accounted for by
3
h/r=5 a stress concentration factor α=1.1-1.2.
2,9
2,8 h/r=4 For shrink fit couplings a stress concentra-
2,7 tion factor α=1.7-2.0 can be used.
2,6 h/r=3
α 2,5 Influence of mean stress
2,4 Most material data are obtained from tests
2,3
h/r=2 carried out as either fully reversed bending
2,2
tests or as pulsating tension-compression
2,1
tests, i.e., the stress ratio R=σmin/σmax is −1.
For tests in torsion, R is usually equal to −1.
2
0,005 0,01 0,015 0,02
When other stress ratios prevail than those
r/D represented in available tests it may be nec-
essary to reduce the maximum allowable
Figure 6-7 Stress concentration factor, from stress range. However, for main shafts the
Roloff and Matek (1994). stress ratio often assumes values near to –1
such that it may not be necessary to consider
124 6 – Nacelle
any such reduction. In more accurate calcu- and torsion (and possibly tension-
lations, one may take into account the influ- compression). Such data are always avail-
ence of the mean stress by means of meth- able for standard reference materials. When
ods available for this purpose, e.g. a Haig σD is not available for the actual material,
diagram. Reference is made to Gudehus and one may for tension-compression use the
Zenner (1999), Bergmann and Thumser following estimate, which refers to the mean
(1999) and VDI. S−N curve with 50% failure probability:
Fatigue resistance and characteristic S−N σD50% = (0.436×Re+77)

curve
In lack of endurance tests of the actual mate- where Re denotes the yield strength in MPa.
rial, a number of different methods to estab- This number in units of MPa applies to pol-
lish a synthetic characteristic S−N curve are ished test specimens of small dimension, 7-
available. Each method has its own advan- 10 mm, made from mild steel or low-alloy
tages and disadvantages. Some of the meth- steel.
ods can be found in the Gudehus and Zenner
(1999), Bergmann and Thumser (1999) and To achieve results corresponding to 2.3%
Sundström (1998). General aspects of fa- failure probability rather than 50% failure
tigue calculation can be found in Appendix probability, σD50% has to be reduced as fol-
C. In general, it is not recommended to lows
combine two different methods because this
may cause the assessment of the overall σD2.3% = σD50% −2·s
safety level to become nontransparent.
in which s represents a standard deviation.
Fatigue strength data are usually presented For mild steel and low-alloy steel, s can be
on a logS-logN plot, in which the S−N curve taken as 6% of σD50%, see Sundström (1998).
appears as a straight line. Usually the mean
S−N curve through the data is given, i.e. the Determination of design S−N curve
curve corresponding to 50% failure prob- On a logS−logN scale, the S−N curve can be
ability and 50% survival probability. For considered linearly decreasing from (10, Rm)
design of wind turbine components, it is to (106, σD), where Rm denotes the ultimate
common to use an S−N curve with only 2- strength. For N>106, the curve may be taken
3% failure probability. as a horizontal line at S=σD.
In the following, only one method for estab- Stresses exceeding the yield stress Re are
lishment of a characteristic S−N curve will usually not allowed in a structure subjected
be referenced, which is simple and suitable to fatigue loading, so the S−N curve is cut
for preliminary calculations. off at Re.
It may be difficult to establish S−N curves Due to the variation of the influence factors
for alloy steels. If test data are available, the from point to point within the shaft, the S−N
test assumptions are often not documented. curve to be used for the actual shaft is
If reliable S−N curves for the applied steel unique for each individual location consid-
quality are not available, a synthetic S−N ered within the shaft.
curve may be used, based on static strength
data for the material in conjunction with One may, conservatively, use the following
fatigue strength σD under rotational bending design rules:
6 – Nacelle 125
σ⋅γf < σD2.3%⋅K1⋅K2 /( Kr⋅β⋅γm) in order to ensure a sufficiently low failure

probability in the design life of the shaft,
τ ⋅γf < τD2.3%⋅ K1⋅K2/(Kr⋅β⋅γm)
åni/Ni< 1
in which σ and τ are the actual characteristic
normal stress and torsional stress, respec- in which ni is the number of stress cycles at
tively, in the observed cross section. σD2.3% the stress amplitude σi and Ni is the number
is the characteristic value for the endurance of cycles to failure at this stress amplitude.
limit, calculated from σD50% as specified The sum is over all stress amplitudes in the
above. assumed discretisation of the stress ampli-
tude axis. For main shafts, both bending and
For the notch factor β, a somewhat better torsion occur at the same time. One can take
approximation can be achieved by account- this into account by using the reference
ing for the smaller sensitivity of the material stress
to notches at lower numbers of cycles. At
N=10 one can set β=1 and then let β in-
σ i ,ref = σ i + 3τ i
2 2
crease linearly with logN to β=1+q(α−1) at

N=106 as shown in Figure 6-9.
in which σi is the normal stress and τi is the
Notch factor
torsional stress. As the bending stress is os-
cillating with a stress ratio near –1 while the
β µαξ = 1 + q ( α - 1 ) shear stress is oscillating with a positive
stress ratio between 0 and approximately
0.8, use of the reference stress σi,ref implies
an approximation. A more accurate ap-
β
proach can be found in Gudehus and Zenner
(1999).
6.1.4 Ultimate strength

The stress calculation is to be carried out
1
according to standard mechanical engineer-
0 1 2 3 4 5 6 7 8 ing methods and is usually straight-forward.
Log10(N) For materials with a high ultimate strength,
it may be necessary to account for stress
Figure 6-9 Notch factor concentrations in extreme load cases. The
influence of the ultimate strength can be
Fatigue loads are usually specified in terms accounted for by the following factor
of a distribution of stress amplitudes, i.e. a
so-called load spectrum, which on discre- βextreme = 1+(α−1)⋅(Rm/1000)2
tised form gives the number of stress cycles
in each interval of a suitably discretised
in which the ultimate strength Rm is to be
stress amplitudes axis. The stress cycles in
given in units of MPa, and where the stress
each stress amplitude interval contribute to
concentration factor α is to be taken from
the total fatigue damage. The total fatigue
Figures 6-7 and 6-8. The following criterion
damage can be calculated by Palmgren-
is to be fulfilled in design:
Miner’s method and has to be less than 1.0
126 6 – Nacelle
σ⋅γf < Re/(γm⋅βextreme) somehow be connected to the transmission

input shaft. Formerly, hollow-axis gears
where σ is the actual stress in the observed were used by which the torsional moment
section, and Re is the yield strength. was transferred from the main shaft to the
shaft bushing of the gear by means of a
For stress ratios > 0 and dissimilar stress keyway connection. This principle for tor-
distribution, a plastic strain up to 0.2% may sional moment transfer has now been aban-
be allowed for single extreme load cycles. doned, except for small simple wind tur-
Reference is made to Gudehus and Zenner bines, as the keyway connections are only
(1999). The amount of plastic strain may be poorly suited to transfer varying and re-
estimated from Neuber’s correction, see versing loads. Assembly and dismantling
Gudehus and Zenner (1999) and Sundström can also be difficult. Usually some kind of
(1998). For repetitive extreme load cycles shrink fit coupling is used, e.g. in conjunc-
with stress ratios < 0, a low-cycle fatigue tion with a shrink disk.
calculation should be conducted, see Gude-
hus and Zenner (1999) and Sundström Shrink disks
(1998). A correct functioning of clamp couplings
requires a careful dimensioning and assem-
Partial safety factors bly. If the coupling, in addition to torque,
The requested safety level is normally also is to transfer bending moments, it is
achieved by multiplying the characteristic important to include this in the calculations.
loads with a load safety factor γf and reduc- When applying catalogue values for trans-
ing the characteristic material properties ferred torque, it is recommended to apply a
with a material safety factor γm. safety factor of 1.5 to the peak torque in
order to be on the conservative side.
The load safety factor accounts for the ran-
dom nature of the load components (e.g. To achieve the prescribed safety, it is im-
fatigue loads, extreme loads, gravitational portant to make sure that the roughness con-
loads etc.) and the uncertainty in the calcu- ditions of the surface and the tolerances for
lation method. The material safety factor both parts are in accordance with the rec-
accounts for the scatter in the material prop- ommendations of the manufacturer. During
erties and the level of quality control for assembly, correct bolt pretensions should be
material. applied, and the required cleanness of the
frictional surfaces should be accounted for.
The values of these safety factors are nor-
mally given in the relevant regulations for Couplings
wind turbines, e.g. DS 472, and the associ- In certain cases where the gear is rigidly
ated material standards. For machine com- mounted on the machine foundation, a cou-
ponents, the certifying bodies or the owner pling is applied which can only transfer
sometimes define their own values for γm. torque and which is flexible with respect to
bending moments. In these cases, the hub
6.1.5 Main Shaft-Gear Connection parts of the couplings are usually shrunk
both on main shaft and gear shaft. For the
In cases where the main shaft does not con- design of the coupling itself, reference is
stitute an integral part of the gear and in made to special literature.
cases where it forms the generator shaft in
multipolar generators, the main shaft should
6 – Nacelle 127
6.1.6 Materials tions by some suitable non-destructive test-

ing procedure such as ultrasonic testing, and
In most cases, non- or low-alloyed machin-
that the shaft is delivered with a materials
ery steel is used, i.e. steel with a carbon
certificate.
content of 0.3-0.7% and with less than 5%
alloy of metals such as Mn, Cr, Mo, Ni and
6.1.7 Standards
V.
Materials:
This classification covers steels with rather Structural steels EN 10 025
different mechanical properties, extending Tempered steels DIN 17200
from non-alloyed steels with an ultimate Nodular irons DIN 1693 Parts 1 and 2.
strength of about 500 MPa and failure elon-
gation of 15% to low-alloyed steels with Technical terms of delivery and testing: DIN
ultimate strength up to 1500 MPa and failure 1690 Parts 1 and 2, DIN 54111 Part 2. Mag-
elongation less than 10%. netic particle inspection SIS 114401; Capil-
lary Liquid Testing: Nondestructive Testing,
In addition to an adequate ultimate strength, PI-4-2 Liquid Penetrant testing. General
the material should exhibit a high fracture Dynamics.
toughness and a low brittle transition tem-
perature. Steels within this category are Certificates EN 10 204.
standardised in DIN 17200.
Such steels are suitable also as a basis for 6.2 Main Bearing
forged materials for main shafts. This is a
6.2.1 General
very common design approach, which im-
plies an improvement of the structure and a The main bearing of a wind turbine supports
good transition to the flange for the rotor the main shaft and transmits the reactions
hub. from the rotor loads to the machine frame.
On account of the relatively large deforma-
In some cases, cast main shafts are used. tions in the main shaft and its supports, the
The casting provides a great degree of free- spherical roller bearing type is often used,
dom as far as the shaping of the shafts goes, see Figure 6-10 for an example.
whereas it poses some limits in terms of
relatively low ultimate strength and failure
elongations.
Usually, nodular iron in the qualities

GGG.40 or GGG.50, according to DIN
1693, is used.
The main shaft is one of the most critical

components in a wind turbine structure. It is
important to make sure, by appropriate
quality assurance, that the assumed material
quality is met, and that the manufacturing Figure 6-10 Spherical roller bearing, from
has not caused development of surface Bonus (1999).
cracks or other imperfections. This implies
that the material is checked for imperfec-
128 6 – Nacelle
Figure 6-11 Bearing frame with main shaft and bearing arrangement for wind turbine with two main bear-
ings; courtesy Vestas
6 – Nacelle 129
Spherical roller bearings have two rows of rangement for a turbine with two main
rollers with a common sphered raceway in bearings in the same housing. Two other
the outer ring. The two inner ring raceways examples are given in Figures 6-12 and 6-
are inclined at an angel to the bearing axis. 13. A couple of additional examples appear
The bearings are self-aligning and conse- from illustrations of the hub and transmis-
quently insensitive to errors of alignment of sion system in Section 5.2.
the shaft relative to the housing and relative
to shaft bending. In addition to high radial
load capacity, the bearings can also accom-
modate axial loads in both directions.
The allowable angular misalignment is nor-

mally 1-2.5 degrees depending on the bear-
ing series. This is sufficient to compensate
for deformations in shafts, housing and ma-
chine frame caused by the rotor loads and
subsequently to prevent excessive edge
loads, which would result in possible dam-
age to the bearing.
Figure 6-13 Nacelle with transmission sys-
tem with two main bearings. The second
bearing on the main shaft forms the fore-
most bearing in the gearbox. This arrange-
ment makes it easy to replace the gearbox.
Courtesy NEG Micon.
yg
ys
yr
Figure 6-12 Nacelle with transmission sys- MzR

tem, for which the main bearing is integrated
Fs
in the gearbox; courtesy NEG Micon. FzR Fa
FyR Frh
The main bearings are mounted in bearing FxR Frv
housings bolted to the main frame. The MyR MxR
quantity of bearings vary among the differ-
ent types of wind turbines: Many wind tur- Figure 6-14 Main shaft.
bines have two bearings, each with its own
flanged bearing housing. Some turbines with FxR - Side force on rotor and nacelle
two bearings use the hub as a housing. Some FyR - Thrust on rotor
turbines have only one main bearing, as the FzR - Weight of rotor
gearbox functions as a second main bearing. MxR - Tilting moment at rotor
Each bearing arrangement has advantages MyR - Driving torque at rotor
and disadvantages. Figure 6-11 shows an MzR - Yaw moment at rotor
example of the main shaft and bearing ar- FS - Shaft mass
130 6 – Nacelle
yr - Distance rotor to bearing centre the support. These requirements are fulfilled
yg - Distance bearing centre to gear stay by spherical roller bearings, see Figure 6-10.
ys - Distance shaft c.o.g. to gear stay
6.2.4 Operational and Environmental
6.2.2 Determination of Design Loads Conditions
Figure 6-14 shows the conventional design The main shaft speed range of a conven-
of a rotor shaft. The shaft is supported by a tional wind turbine of rated power in the
bearing placed adjacent to the rotor. Another range 500-2500 kW is about 10 to 30 rpm at
bearing at the opposite end of the rotor shaft nominal load. Depending on the operational
is integrated in the main gearbox. strategy, the wind turbine will experience all
speeds from stand-still to nominal speed for
The loads in Figure 6-14 are drawn in direc- varying amounts of time. These conditions
tions according to a conventional coordinate have to be included in the load spectrum for
system and not in the direction they will the bearing. The temperature range can vary
normally have. considerably and has to be evaluated under
consideration of the actual site. In the IEC
From the example in Figure 6-14, the front 61400-1, a normal ambient temperature
main bearing loads can be calculated from range from −10º C to +40º C is specified.
simple beam theory as: Environmental conditions such as salinity,
chemically active substances and abrasive
Fa = −FyR particles have to be considered.
1
M1 + M 2
2 2
Fr = Lightening currents passing through the
yg
bearings may have to be considered for
some sites.
with
6.2.5 Seals, Lubrication and Tempera-
M 1 = M xR − FS y S − FzR ( y r + y g ) tures
M 2 = FxR ( y r + y g ) + M zR Seals
Bearing seals are needed partly to hold back
The loads are usually specified in terms of a bearing lubrication and partly to keep out
load spectrum or distribution of loads, which contaminants. Especially since the main
on discretised form gives the number of bearing is often placed relatively unpro-
hours of operation within each defined load tected, close to the outside, dirt and rain-
interval in the discretisation. For each inter- water can easily come in contact with the
val, the associated operational and environ- bearing.
mental conditions have to be taken into ac-
count. All relevant load cases are to be in- Non-rubbing seals (labyrinth seals) are ap-
cluded in this load spectrum. propriate since they exhibit practically no
friction and no wear. A labyrinth seal forms
6.2.3 Selection of Bearing Types a good supplement to other seals.
The main bearing must be able to accomo-
When using rubbing seals it should be aimed
date axial as well as radial forces from the
to mount the seal on the shaft and let the
rotor. Further, the bearing must allow mis-
bearing housing form the sealing surface.
alignment from deflection of the shaft and
This is to reduce the risk of scratches in the
6 – Nacelle 131
shaft. The compatibility of the grease with from contamination and expensive circula-
the seal material has to be checked. tion system is avoided. Rolling bearing
greases are standardised in DIN 51825.
Lubrication
The main purpose of lubrication is to create Grease consists of a base oil with thickeners
a lubricant film between the rolling elements and possibly additives added. The following
to prevent metal-to-metal contact. This is to grease types are distinguished:
avoid wear and premature rolling bearing • Mineral oil with metal soaps as thick-
fatigue. In addition, lubrication reduces the ener.
development of noise and friction, thus im- • Mineral oil with non-soap thickener
proving the operating characteristics of a • Synthetic oils with non-soap thickener
bearing. Additional functions may include
protection against corrosion and enhancing A possible choice of grease lubrication
the sealing effect of the bearing seals. would be a lithium soap base grease of
penetration class 2-3 with EP additives and
When selecting lubrication, some of the maybe corrosion and oxidation inhibitors.
matters that needs to be considered are vis-
cosity, consistency, operating temperature Lithium soap base grease has the quality of
range, the ability of protection against corro- being water resistance and usable at a wide
sion and the load carrying ability. of range temperature –35 °C to +130 °C.
Water in the lubrication leads to corrosion, Other possibilities are sodium grease or cal-
degrading of the lubrication, forms aggres- cium soap base grease.
sive substances together with the oil addi-
tives, and affects the formation of a load Sodium grease absorbs large amounts of
carrying lubricating film. water and is useful in environment with
condense. It may however soften to such an
Since the rotational speed varies between extent that it flows out of the bearing.
zero and the nominal speed, a boundary lu-
brication condition will consist a consider- Calcium soap base greases of penetration
able part of the operational time. This will class 3 do not absorb any water which is
normally call for a lubricant with EP (ex- advantageous where the bearing seals are
treme pressure) additives and as high a vis- exposed to splash water.
cosity as practically possible. In this respect,
it is essential to consider possible load cases The stiffness of the grease is determined by
where the radial load is low or even zero the consistency class. A stiff grease, which
combined with low temperature (high vis- belongs to consistency class 3 or higher, can
cosity) because it may cause sliding in the contribute to the sealing of the bearing and
bearing. keep out contaminants by lying in a laby-
rinth seal or in the contact area of a rubbing
Grease lubrication seal.
The most commonly used lubrication in the
main bearings is grease. However, for high P/C load ratios, greases
of consistency class 1-2 should be selected.
Grease has the advantage of being easily In a dusty environment a stiff grease of
retained in the bearing arrangement, it con- penetration class 3 should be used.
tributes to sealing the bearing arrangement
132 6 – Nacelle
The relubrication interval corresponds to the duce the harmful effects of metal-to-metal
minimum grease life F10 of standard greases contact, which occurs in some places. The
in accordance with DIN 51 825. The grease suitability of EP additives varies and usually
service life is dependent on the type and depends largely on the temperature. Their
amount of grease, bearing type and size, effectiveness can only be evaluated by
loading, speed, temperature and mounting means of tests in rolling bearings. (FAG
conditions. For larger bearings (> 300 mm) WL-81 115/4.)
the relubrication is recommended to be more
frequently than F10. In some cases continu- The intervals between oil changes depend on
ous lubrication is established. lubrication system and circulation, contami-
nation and ageing of the oil.
Great care should be taken if the grease type
is to be changed. If incompatible greases are Before going into operation, oil must be
mixed, their structure can change drastically, supplied to the bearing. In case of circulat-
and the greases may even soften considera- ing oil lubrication, the oil pump should be
bly. started before the turbine goes into opera-
tion. In other cases the bearing must be
Concerning the amount of grease, a rule of manually lubricated before first start-up and
thumb is to fill the bearing completely with total drain of the bearing during service
lubrication while the housing is half filled. must be avoided.
SKF, 1989).
6.2.6 Rating Life Calculations
Oil lubrication
The fatigue load carrying capacity of the
Since temperatures normally are relatively
bearing is characterised by the basic dy-
low, the lubricant does not need to function
namic load rating C. This quantity can be
as heat dissipator. Then there is no need for
calculated according to ISO 281 (1990).
circulation of the lubricant, which simplifies
the design. On the other hand, when the lu-
The static load carrying capacity of the
brication is not circulated there is no possi-
bearing is characterised by the static load
bility/opportunity for filtering, and relubri-
rating C0. This quantity can be calculated
cation is therefore necessary.
according to ISO 76 (1990).
Oil lubrication makes it necessary to moni-
The general methodology for selecting and
tor the lubrication system because of the risk
calculating rolling bearings is given in Sec-
of leakage. A disadvantage compared to
tion 6.3 for both dynamic and static loads. In
grease lubrication is that it demands a better
general, for a design life of 20 years, the
sealing and a circulation system.
required basic rating life L10h should equal or
exceed 300,000 hours. The modified rating
Straight oils and preferably corrosion- and
life L10mh should at the same time reach
deterioration-inhibited oils can be used. If
175,000 hours. In the context of grease lu-
the recommended viscosity values are not
brication and oil lubrication without filter,
maintained, oils with suitable EP additives
the contamination factor ηC should not be
and anti-wear additives should be selected.If
chosen higher than 0.2 if special precautions
the bearings are heavily loaded (P/C > 0.1)
are not taken to obtain and verify higher
or if the operating viscosity ν is smaller than
values. For oil lubrication with off-line fil-
the rated viscosity ν1, oils with anti-wear tering, a contamination factor ηC of 0.5-0.7
additives should be used. EP additives re- may be obtained.
6 – Nacelle 133
The C/P ratio for the largest load interval should be a possible difference in tempera-
should not exceed 2.5. ture and expansion of inner ring respectively
compression of outer ring during mounting.
The above-mentioned figures apply to me-
dium size wind turbines (600-1000 kW). It The bearing clearance should always be
should be borne in mind that large bearings checked after mounting on the shaft.
are relatively less sensitive to contamination
particles of a certain size than smaller bear- 6.2.8 Bearing Housing
ings.
The bearing must be firmly supported by the
whole circumference to achieve a proper
The static safety factor C0/P0 based on the
load transmission.
extreme load cases should not be lower than
4.
For bearings with normal tolerances, the
dimensional accuracy of the cylindrical
6.2.7 Connection to main shaft
seating in the housing should be at least IT
The thrust from the rotor should be consid- grade 7, and on the shaft at least IT grade 6.
ered to be taken up by shoulder or by fric-
tion or a combination of both. One should The tolerance for cylindrical form should be
then be aware of the stress concentration in at least one IT grade better than the dimen-
the main shaft at the shoulder. sional tolerance.
In cases where there is a possibility of the Since cast bearing houses often will have a
wind turbine being in a situation where the complex geometry, an obvious method to be
wind is coming from the back one should used for verification would be a finite ele-
consider a stop ring as shown in Figure 6- ment analysis, see Appendix D.
15.
6.2.9 Connection to Machine Frame
back wind The connection between the bearing housing
secure ring
and the machine frame will most likely be a
bolt connection. The connection should be
capable of transferring the combination of
axial and radial forces from the bearing to
the main frame by friction or by shear in
bolts by tight fit, depending on the geometry
Figure 6-15 Back wind stop ring of the connection (see Appendix A).
A tight fit, i.e. complete support of the 6.2.10 Standards

bearing ring over its entire circumference,
makes a full utilisation of the bearing's load ISO 76 Roller bearings – Static load ratings.
carrying capacity possible.
ISO 281 Roller bearings – Dynamic load
The bearing clearance should be as small as ratings and rating life + Amendment 1 and
possible to ensure an accurate guidance 2.
however large enough in order not to get
stuck in any situation. When considering the IEC 61400-1 Wind generator systems Part 1.
bearing clearance things to take into account Safety requirements.
134 6 – Nacelle
6.3 Main Gear curved spiral teeth. Spiral bevel gears

result when the teeth are helical or
The purpose of the main gear is to act as a
curved. See Figure 6-18.
speed increaser and to transmit energy be-
• Worm gears consist of a worm-thread
tween the rotor and the generator.
wheel, or “endless screw”, working in
conjunction with a cylindrical toothed
6.3.1 Gear types wheel. The axes are nonparallel and
nonintersecting, see Figure 6-19. Worm
The most common gear types used in main
gears are not used for main gears in
gears for wind turbines can be identified and
wind turbines, but are sometimes used
classified as follows, based on their geomet-
as yaw gears in yaw systems.
rical design:
• Hypoid gears are a cross between worm
• Spur and helical gears consist of a pair
gears and bevel gears in that they are
of gear wheels with parallel axes. Spur
similar to bevel gears but have
gears have cylindrical gear wheels with
nonintersecting axes. Hypoid gears are
radial teeth parallel to the axes. In heli-
turned into spiral bevel gears in the
cal gears, the teeth are helical, i.e. they
limiting case when the offset between
are aligned at an angle with the shaft
the axes approaches zero. Reference is
axes. Double-helical gears have two
made to Figure 6-20.
sets of helical teeth on each wheel.
Helical gears are sometimes referred to
as spiral gears or oblique gears. See
Figure 6-16.
• Epicyclic or planetary gears consist of
epicyclic trains of gear wheels, i.e., they
are gears where one or more parts – so-
called planets – travel around the cir-
cumference of another fixed or revolv-
ing part. See Figure 6-17. Figure 6-16 Examples of spur and helical
gears, from Niemann and Winter (1985)
Planetary gears in combination with one or
more parallel axis gears form the most
commonly applied gear type for the main
gear in wind turbines. Gears in which the
power is transferred from one wheel to two
or more meshing wheels are referred to as
gears with a split power path.
Other types of gears, rarely or never used in

Figure 6-17 Planetary gear principle, outer
wind turbine main gears, include the fol-
fixed annulus with three revolving planets
lowing, which are listed here merely for
and one rotating planet carrier in the middle,
completeness:
from Niemann and Winter (1985)
• Bevel gears consist of a pair of toothed
conical wheels whose working surfaces
Tooth form
are inclined to nonparallel intersecting
The tooth forms used practically universally
axes. Bevel gear wheels can be de-
in spur, helical, bevel and worm gears are
signed with spur teeth, helical teeth, and
so-called involute teeth. This tooth form
6 – Nacelle 135
implies that rotation of the base circle at a ings. Different bearing types applied in
uniform rate is associated with uniform dis- gears include
placement. The path of contact is a straight • ball bearings
line, which coincides with the line of action. • cylindrical roller bearings
• spherical roller bearings
• tapered roller bearings
Examples of bearings are shown in Figure 6-

21. Two bearings should be used to support
each gear shaft, one for support of both ra-
dial and thrust forces, the other for support
of only radial forces and free to allow for
axial growth under thermal changes. Bearing
fits should be tight to prevent damage to the
bearing or the housing and to prevent spin-
ning of inner and outer bearing races.
Figure 6-18 Examples of bevel gears, from

Niemann and Winter (1985).
Figure 6-21 Examples of roller bearings:

Spherical bearing (left) and cylindrical
bearing (right). From SKF (1997).
Figure 6-19 Worm gear
The bearing fits should be carefully selected
in order to avoid spinning bearing rings as
well as squeezing of the rollers.
For roller bearings in gears, it is normally

sufficient to have a predicted L10h lifetime of
at least 100,000 hrs referring to the cubic
mean load, or 30,000 hrs referring to the
nominal power.
The loads should be specified in terms of a

Figure 6-20 Hypoid gear, from Niemann load duration spectrum, which on discretised
and Winter (1985) form gives the number of hours of operation
within each of a number of suitably chosen
Bearing types load intervals. The equivalent power, which
Bearings for wind turbine gears should all is defined as the cubic mean power, can be
be rolling element, anti-friction type bear- calculated as follows:
136 6 – Nacelle
æ
1 3
Table 6-2 Reliability factor a1
3 h ö
Peq = çç å ( Pi i ÷÷ Failure Reliability Reliability
è i ht ø
probability n 1−n factor a1
(%) (%)
where Pi denotes the power in the ith defined 10 90 1.00
load interval, hi denotes the number of hours 5 95 0.62
of operation in this interval, and ht denotes 4 96 0.53
the total number of hours of operation. 3 97 0.44
2 98 0.33
Rolling bearings
1 99 0.21
The design of roller bearings for a gear is
made using a load duration spectrum. The
Bearing manufacturers recommend values
load duration spectrum may be transformed
for the factor product a23=a2a3. These values
to an equivalent cubic mean load, Peq. Ac-
may vary from manufacturer to manufac-
cording to ISO281/1, the basic rating life is
turer, and a23 will usually depend on the
defined as
viscosity ratio
p
æ C dyn ö ν
L10 = ç ÷ ⋅10 6 revolutions κ=
ç P ÷ ν1
è eq ø
in which Cdyn is the basic dynamic capacity in which ν is the operational viscosity of the
of the bearing, and p is an exponent, which lubrication, and ν1 is the viscosity required
takes on the value 3 for ball bearings and for an adequate lubrication, i.e. a lubrication
3.33 for roller bearings. The bearing manu- which is sufficient to avoid metallic contact
facturers usually quote values for Cdyn. The between the rolling elements and the race
index 10 refers to a 10% failure probability ways.
associated with the bearing life L10. ISO
281/1 suggests the following formula for New methods for bearing life prediction are
calculation of adjusted rating life being developed, e.g. by SKF and INA. SKF
defines a modified rating life
p
æ C dyn ö
L na = a1 a 2 a 3 ç ÷ ⋅10 6 revolutions p
ç P ÷ æ C dyn ö
è eq ø L naa = a1 a SKF çç ÷ ⋅10 6 revolutions
÷
è P ø
in which the suffix n denotes the associated
probability of failure. The corresponding in which the adjustment factor aSKF depends
survival probability is 1−n, which is also on the cleanliness of the lubricant and the
referred to as the reliability. The coefficient load ratio Pu/P, where Pu denotes the endur-
a1 is a reliability factor, which depends on ance load limit and P is the actual load.
the actual reliability, see Table 6-2. The co- Since the aSKF factor will vary from one load
efficients a2 and a3 are factors for the bear- level to another, use of this approach to life-
ing material and service conditions, respec- time predictions requires the entire load
tively. For a commonly applied reliability of spectrum of loads Pi to be applied, not only
90%, a conventional bearing material and an equivalent load Peq. The total rating life
normal operating conditions, a1=a2=a3=1.0. can accordingly be obtained as follows
6 – Nacelle 137
The major function of the material selection

ht and subsequent heat treatment of steel mate-
L10 mh = rials to be used for gears is to achieve the
hi
å L10 mh , i
desired microstructure at the critical loca-
i
tions so that, in particular, the teeth will
have the desired contact and tooth root
in which the numerator ht denotes the total strength capacity. Common heat treatments
number of hours of operation, and the defor steel include:
nominator denotes the sum of all relative life • Preheat treatments (anneal, normalise,
consumptions over a suitable discretisation temper)
of the load spectrum. • Heat treatments
• through-hardening (anneal, nor-
Other methods exist which require advanced malise, normalise and temper,
calculation programs developed by the ma- quench and temper),
jor bearing manufacturers. • surface hardening (flame and in-
duction harden, carburise, carboni-
Lubrication conditions are essential for the tride, and nitride)
bearing design. The oil viscosity should • Postheat treatment (stress relief)
preferably result in a viscosity ratio κ>1.0.
In the region from κ=0.4 to κ=1.0, the Case hardening implies that carburisation is
bearing manufacturers recommend lubrica- carried out after a prior heat treatment and is
tion oils with approved EP (“extra pres- followed by hardening and tempering. The
sure”) additives. case hardening implies that the surface layer
becomes harder than the interior.
The oil supplied to bearings should have a
cleanliness and a temperature corresponding Requirements to materials and hardening
to the design assumptions. The cleanliness treatments for gear transmissions can be
of the lubricant as defined in ISO 4406 con- found in DNV Rules for Classification of
sists of three numbers, for example Ships, Part 4, Chapter 2.
17/14/12, which corresponds to certain
numbers of particles per 100 ml sample 6.3.2 Loads and Capacity
greater than 2µm, 5µm and 15µm, respec-
tively. For gears, the load levels are in focus rather
than the load ranges, which is in contrast to
The static capacity of roller bearings is de- what is the case for many structural details.
fined in ISO 76 and is the load for which the Procedures for calculation of loads and for
total permanent deflection of the bearing is prediction of load capacities for gears are
1/10000 times the rolling element diameter. outlined in the following and include pre-
The margin to this capacity should not be diction of surface durability, tooth strength,
less than 4.0 for design extreme loads. and scuffing load capacity. Other damages
such as wear, gray staining (micropitting)
Materials and fractures starting from flanks may also
Gears should preferably be made from sepa- limit the gear capacity, although limited or
rate steel forgings. Heat treatment and case no calculation procedures are given for the
hardening can be used to harden the steels. associated capacities.
The nominal tangential load is given as
138 6 – Nacelle
istribution of load in multiple-path

2T transmissions.
FT =
d • KV is an internal dynamic factor to ac-
count for internally generated dynamic
in which T is the applied torque and d is a loads in the gear.
reference diameter. • KHβ is a face load factor for contact
stress and scuffing.
Surface durability • KHα is a transverse load distribution
Contact stress. The design contact stress σH factor for contact stress and scuffing.
is derived from the nominal tangential load •
as follows The factor Z is defined as the product
FT (u + 1) Z=ZBDZHZEZεZβ
σH = Z K
d 1bu
in which
in which u is the gear ratio per stage, b is the • ZBD is a zone factor for inner point of
face width, and d1 is the reference diameter single pair contact for pinion or wheel
of the pinion. The factors K and Z are com- • ZH is a zone factor for pitch point
pound influence factors to account for vari- • ZE is an elasticity factor that accounts
ous effects. The factor K is defined as the for influence of modulus of elasticity
product and Young’s modulus. This is a factor
whose squared value is in units of
K=KAKγKVKHβKHα stresses.
in which • Zε is a contact ratio factor that accounts
• KA is an application factor defined as for influence of transverse contact ratio
the ratio between maximum repetitive and overlap ratio.
torque and nominal torque and accounts • Zβ is a helix angle factor.
for dynamic overloads external to the
gearing. Note that all factors except ZE are dimen-
• Kγ is a load-sharing factor defined as the sionless. Formulas and details for calcula-
ratio between the maximum load tion of the various factors can be found in
through the actual path and the evenly ISO6336, DIN3990, and DNV CN41.2
shared load and accounts for the mald-
Table 6-3 Endurance limits
Steel grade σHlim (N/mm2)
Alloyed case-hardened steel of special approved high grade 1650
Alloyed case-hardened steel of normal grade 1500
Nitrided steel of approved grade 1250
Alloyed quenched and tempered steel, bath or gas nitrided 1000
Alloyed, flame or induction hardened steel (HV=500-650 N/mm2) 0.75HV+750
Alloyed quenched and tempered steel 1.4HV+350
Carbon steel 1.5HV+250
Note that these values refer to forged or hot-rolled steel. For cast steel, values need to be re-
duced by 15%. HV denotes surface hardness.
6 – Nacelle 139
Capacity. The characteristic endurance limit Tooth root strength

σHlim for contact stresses is the stress that Tooth stress. The following section deals
can be sustained for 5⋅107 cycles without the with the tooth root strength as limited by
occurrence of progressive pitting. This limit tooth root cracking. The local tooth stress σF
usually defines the beginning of the endur- for pinion and wheel can be predicted as
ance stress range, i.e., it defines the lower
knee of the σ−N curve. Values of σHlim are FT
given in Table 6-3, but should be used only σF = YF YS Yβ K A K V K Fβ K Fα
bmn
for materials, which are subjected to rele-
vant quality control. Results of suitable fa-
in which
tigue tests may also be used to establish val-
ues of σHlim.
• YF is a tooth form factor
• YS is a stress concentration factor
The design endurance strength, also known
as the permissible contact stress, is • Yβ is a helix angle factor
• b is the face width at the tooth root,
either for pinion or for wheel
σ H lim
σ HP = Z N Z L ZV Z R ZW Z X • mn is the normal module
SH
Other coefficients and variables are defined
in which elsewhere, see Section 6.3.1. Expressions
for the tooth form factor, the stress concen-
• SH is the required safety factor. tration factor and the helix angle factor can
• ZN is a life factor for endurance be found in ISO6336, DIN3990 and DNV
strength, reduces the endurance strength CN41.2.
when the design life of the gear is
greater than 5⋅107 cycles, and increases Capacity. The permissible local tooth root
it when the design life is less than 5⋅107 stress σFP for pinion and wheel can be pre-
cycles. dicted as
• ZL, ZV, and ZR are lubricant factor,
speed factor and roughness factor, re- σ FE Yd YN
spectively, all referring to various as-
σ FP = Yδ T YR T YX YC
SF rel rel
pects of oil film influence.

in which
• ZW is a work hardening factor
• σFE is the local tooth root bending en-
• ZX is a size factor durance limit of the reference test gear.
The stress concentration factor for the
Formulas and details for calculation of the test gear is normally 2.0.
various factors can be found in ISO6336,
• Yd is a design factor, which accounts
DIN3990 and DNV CN41.2.
for other loads than constant load direc-
tion, e.g., idler gears, temporary change
Design rule. The design rule to be fulfilled
of load direction, prestress due to
is
shrinkage etc.
• YN is a life factor for tooth root stresses
σ H ≤ σ HP related to the reference test gear dimen-
sions. It is used to take into account the
higher load bearing capacity for a lim-
140 6 – Nacelle
ited number of load cycles, and a re- along the path of contact. Usually the flash
duced load bearing capacity when the temperature criterion will govern the design
number of load cycles is large. against scuffing failure.
• SF is a required safety factor.
• YδrelT is a relative sensitivity factor of Two inequalities should be fulfilled to meet
the gear, related to the reference test the flash temperature criterion:
gear
• YRrelT is a relative surface condition θ S − θ oil
θB ≤ + θ oil and θ B ≤ θ S − 50°
factor of the gear, related to the refer- SS
ence test gear
• YX is a size factor in which
• YC is a case depth factor considering
subsurface fatigue. • θS is the scuffing temperature as deter-
mined from FZG tests
Formulas and details for calculation of the
• θoil is the oil temperature before it
various factors can be found in ISO6336, reaches the mesh, i.e., the normal alarm
DIN 3990 and DNV CN41.2. temperature
• θB is the maximum contact temperature
Design rule. The design rule to be fulfilled
along the path of contact, calculated as
is
the sum of the bulk temperature θMB
and the maximum flash temperature
σ F ≤ σ FP
θflamax along the path of contact
• SS is the required safety factor, usually
Scuffing load capacity taken as 1.50.
High surface temperatures due to high loads
and sliding velocities can cause lubricant The integral temperature criterion reads
films to break down in the gear. This will
lead to seizure or welding-together of areas
θS
of tooth surfaces between the wheel and the θ int ≤
pinion. This phenomenon is known as SS
scuffing and may lead to failure. In contrast
to pitting and fatigue failure, which both in which
exhibit a distinct incubation period, a single
short overloading can lead to a scuffing fail- • θint is the integral temperature, calcu-
ure. lated as θint=θMC+1.5⋅θflaint, where θMC is
a bulk temperature and θflaint is the mean
Two criteria are to be fulfilled to ensure a flash temperature along the path of
sufficient safety against scuffing failure. contact.
Both criteria are formulated in terms of cri- Formulas for calculation of θS, θMB, θMB,
teria on temperature, i.e., the local contact θflamax, and θflaint can be found in DNV
temperature may not exceed some permissi- CN41.2.
ble temperature. The one criterion is a so-
called flash temperature criterion, based on Note that gray staining may occur under the
contact temperatures, which vary along the same conditions that may lead to scuffing,
path of contact. The other criterion is an and this may even happen without or before
integral temperature criterion, based on the the occurrence of a literal scuffing failure.
weighted average of contact temperatures
6 – Nacelle 141
Capacity of shaft/wheel connections vention of corrosion. However, the principal

Several types of couplings between shaft function of the oil is the protection of the
and wheel appear in the context of gears. rubbing surfaces of the gear teeth. There are
The major types are listed in Section 6.4 many requirements for the lubrication sys-
with issues of importance for their design tem. The following items are recommended
dealt with. as a minimum.
Other capacities The choice of a lubricant and lubrication

Other capacities to be considered in the system should be the joint responsibility of
context of gears include the gearbox manufacturer and the gear box
• Bearing calculations purchaser.
• Shaft capacities for torsional loads and
tooth loads Wind turbine gears have relatively low
• Gearbox housing and suspension. The pitchline velocity and high gear tooth loads.
capacity of the gearbox is much de- These conditions require either synthetic or
pendent on the type of gearbox, e.g., mineral gear oil with antiscuff additives and
whether it is cast or welded, whether it the highest viscosity that is practical. The
is small or large, and whether it is one- choice of lubricant depends on many factors
or two-pieced. The gearbox housing including viscosity, viscosity index, pour
shall have sufficient strength and stiff- point, additives and costs.
ness to avoid mesh misalignment due to
deflection of the gearbox housing. It is recommended to evaluate the ability of
a lubricant to resist micropitting.
6.3.3 Codes and Standards
Viscosity classes for lubricants according to
DIN 3990, “Tragfähigkeitsberechnung von ISO are given in Table 6-4, and recom-
Stirnrädern”, Part 1-5, Deutches Institut für mended criteria for acceptance of gear lubri-
Normung e.V., 1987. cants are given in Table 6-5.
ISO 6336-1 to 6336-5 Monitoring
Oil level indication by means of a sight glass
AGMA, “Recommended Practices for De- or dipstick or equivalent is to be provided.
sign and Specification of Gearboxes for The lubrication oil temperature is to be
Wind Turbine Generator Systems,” Ameri- monitored. A temperature in excess of the
can Gear Manufacturers Association, Alex- approved maximum should result in an
andria, Virginia, 1997. automatic shutdown. The monitoring system
should be arranged so as to imply shutdown
6.3.4 Lubrication in case of malfunction.
The functioning of a gear involves relative
motion of surfaces in contact under load. For gears with a forced lubrication system,
Separation of the surfaces by a thin film of the pressure is to be monitored. Oil pressure
oil is a key factor in achieving smooth op- below minimum, with a running gear,
eration and good service life. Lubrication is should result in an automatic shutdown. The
used for this purpose. Oils for the lubrication requirement does not apply if the forced
of gearing have to meet diverse performance system mainly is arranged for oil cooling
requirements including stability, resistance and the gear can work satisfactorily with
to foaming, separation from water, and pre- splash lubrication. However, an indication
142 6 – Nacelle
of the insufficient oil pressure is to be pro- expected working positions in the bearings.
vided. The mesh contact should be consistent with
that which would result in the required load
Table 6-4 Viscosity classes according to distribution at full load.
ISO
Viscosity Average Limits of kine- The gear transmission is to be spin tested in
class viscosity matic viscosity at the workshop, and checked with regard to oil
at 40°C 40°C (mm2/s) tightness. The spin test can also be used for
(mm2/s) gear mesh verification. For the prototype
min. max. gear as well as for selected gears from serial
VG 2 2.2 1.98 2.42 production, special testing will be necessary,
VG 3 3.2 2.88 3.52 in particular with respect to face load distri-
VG 5 4.6 4.14 5.06 bution, lubrication/temperatures and vibra-
VG 7 6.8 6.12 7.48 tions.
VG 10 10 9.0 11.0
VG 15 15 13.5 16.5 Assembly and testing in the nacelle:
The function of the lubrication oil system and
VG 22 22 19.8 24.2
monitoring is to be tested.
VG 32 32 28.8 35.2
VG 46 46 41.4 50.6
If the arrangement is made so that external
VG 68 68 61.2 74.8 bending moments (e.g. due to the rotor or
VG 100 100 90 110 torque reaction) can influence the gear mesh
VG 150 150 135 165 alignment the contact pattern has to be veri-
VG 220 220 198 242 fied under real or simulated conditions for
VG 320 320 288 352 some gears in a series. This can be achieved
VG 460 460 414 506 by applying a thin suitable lacquer to the
VG 680 680 612 748 teeth before the test. The tooth contact pat-
VG 1000 1000 900 1100 tern at the actual test load is to be analyzed
VG 1500 1500 1350 1650 with respect to load distribution at the rated
VG 2200 2200 1980 2420 load. This requirement also applies to gears
VG 3200 3200 2880 3520 where no part load or full load testing has
been made in the workshop.
Installation of gearing
The gear is to be installed so that appropriate 6.3.5 Materials and testing
alignment and running conditions for the
The quality requirements for materials and
gear are maintained under all operating con-
heat treatment for gears are divided in three
ditions. In case of flexible mounting, harm-
levels according to DIN3990 T5 and
ful vibrations are to be avoided. Excessive
ISO6336-5.The three levels are denoted ME,
movements of flexibly mounted gears are to
MQ and ML, respectively, and ME is the
be limited by stopper arrangements. Design
highest quality level.
of the oil systems and the maintenance
methods of changing the oil should be de-
Specific requirements for quality control and
veloped to minimise oil leaks and spills.
material requirements and testing are given
for each material type and quality level. The
Assembly and testing in the workshop
strength values for both pitting and bending
The accuracy of meshing is to be verified for
fatigue are dependent on the quality level.
all meshes. The journals should be in their
Values are given in DIN3990 T5 and
6 – Nacelle 143
ISO6336-5. The DNV Classification Note sification Note 41.2 is based on are given
N41.2 is based partly on ISO6336. The ma- below.
terial and testing requirements that the Clas-
Table 6-5 Recommended criteria for acceptance of gear lubricants

Parameter Methodology Recommended criteria for acceptance
Viscosity (mm2/s) ISO 3104 ±10%
Viscosity indices ISO 2909 Min. 90
Oxidation stability ASTM D 2893 Increase in viscosity of a test sample
oxidised at 121°C should not exceed
6% of reference value
Corrosion properties, iron ISO 7120 No rust after 24 hours with synthetic
sea water
Corrosion properties, copper ISO 2160 #1b strip after 3 hours at 100°C
Foaming properties ASTM 892 Sequence 1: max. 75/10 10:00
Sequence 2: max. 75/10 10:00
Sequence 3: max. 75/10 10:00
Load carrying property DIN 51 354 Load stage min. 12
Micropitting resistance test FVA. No. 54 Stage 10
Filterability ISO/DIS 13357-1,2 As stated in standard
• ultrasonic test
Testing and inspection of gearing
Requirements to testing and inspection of For steel that will be case hardened, the me-
gearing depend on which type of heat treat- chanical properties need not be documented
ment is applied. The following types of heat before the heat treatment process, as they are
treatment are dealt with in separate subsec- generally documented after the final heat
tions: treatment.
• case hardened
• alloyed through hardened (quenched The impact energy (KV) at ambient tem-
and tempered) perature in tangential direction to pin-
• nitrided ion/wheel is not to be less than specified for
the approved material type and in no case
Case Hardened Gears less than 30 J.
For wind turbine gears, the test requirements
for case hardened steels normally corre- For case hardening the following is required:
spond to ordinary and intermediate grades of
quality as defined in DNV Classification • Core hardenability (Jominy) is ran-
Note 41.2. A higher level of material and domly checked.
heat treatment quality control may result in • Suitable heat treatment is made prior to
acceptance of increased endurance values machining in order to avoid excessive
(see Classification Note 41.2). distortions during quenching.
• Carburisation is made by gas in a con-
Certificates are required for trolled atmosphere furnace. The furnace
• chemical composition shall be equipped with carbon potential
• mechanical properties (including controls and continuously recorded.
Charpy-V)
144 6 – Nacelle
The entire case hardening process is checked It is required that the depths to 550 HV, 400
at regular intervals with regard to: HV and 300 HV and the core hardness are
documented to be within the approved speci-
• The case microstructure: to be marten- fication. Further, the hardness at any point
site with allowance for up to 15% re- below the surface is not to exceed the sur-
tained austenite and fine dispersed indi- face hardness (before grinding) with more
vidual carbides. (Higher percentage of than 30 HV. The core impact energy is not
retained austenite may be accepted pro- to be less than approved specification but in
vided increased safety factor against no case less than 30 J.
scuffing.)
• Decarburisation: not to be visible at a As mentioned above, the case hardening pro-
magnification of 500. cess (the coupons) is normally to be docu-
• The core microstructure: to be marten- mented for each hardening batch and each
sitic/bainitic with no free ferrite in criti- material charge. The use of coupons made of
cal tooth root area. same material type but not same charge may
be accepted provided that the manufacturer
The case hardening process is normally to has a quality system, which ensures sufficient
be documented for each hardening batch and reproducibility. In particular, the limits of
each material charge with a certificate, elements in the chemical composition com-
which includes the following: bined with the respective heat treatment proc-
• Hardness profile. esses must ensure that the required core prop-
• Core impact energy (KV). erties are obtained. This part of the quality
If no alternative procedure is approved, the system has to be specially evaluated. A re-
certificate is to be based on a coupon test. duced extent of impact testing may also be
The coupon is normally to be made of ma- considered.
terial from the same charge as the actual
gear and heat-treated along with this charge. Case hardened gears are to have a minimum
The coupon is to be sampled in the tangen- tooth root space hardness of 58 HRC over
tial direction and is not to be separately the entire face width. Otherwise a reduction
forged. If it is not possible to sample cou- of permissible tooth root stresses applies,
pons tangentially, longitudinal or radial see Classification Note 41.2 Part 3, Sec. 7.
samples may be accepted. Depending on material type this may be
difficult to obtain for large gears and control
The coupon diameter is not to be less than 2 testing may be required. Therefore, manu-
times the normal modulus, minimum 20 facturers may carry out special procedure
mm. Further, the size is to be sufficient for tests in order to document the permissible
making 2 test pieces for impact energy (KV) sizes for their various material types. Com-
of the core. ponents of smaller sizes than those tested
need no documentation of tooth root hard-
The hardness profile (hardness as a function ness.
of depth) is to be determined by hardness
measurements with a load of 10-50 N. The If a component exceeds the tested size, or if
measurements are to be made from the sur- the manufacturer has not carried out a pro-
face to the core. The expected amount of cedure test, the tooth root space hardness is
grinding is to be subtracted. to be checked at mid face. If this hardness is
less than the specified minimum (58 HRC if
nothing else is specified), this measurement
6 – Nacelle 145
is to be carried out over the entire face The tooth accuracy of pinions and wheels is
width. to be documented with reference to ISO
1328-1975 or to a corresponding national
Nitrided Gears standard.
For nitrided gears the process is to be docu-
mented for each nitriding batch and material Visual inspection is to be carried out with
charge with a certificate containing: respect to:
• Hardness profile. • Surface roughness of flanks.
• White layer thickness. • Surface roughness of root fillets.
If no alternative procedure is specified, the • Root fillet radius.
certificate is to be based on a coupon test. • Possible grinding notches of root fillet.
The coupon is to be made of material from Any grinding in the root fillet area, and
the same charge as the actual gear and heat- in particular when leaving notches will
treated along with this charge. result in reduction of permissible
stresses. See Classification Note 41.2
The coupon diameter is not to be less than 2 Pt. 3 Sec. 7.
times the normal modulus. The hardness pro-
file (hardness as a function of depth) is to be 6.4 Couplings
determined by hardness measurements with a
load of 10-50 N. The measurements are to be The major types of couplings are listed in
made from the surface to the core. If further the following with issues of importance for
grinding is intended (and is approved), the their design dealt with:
expected amount is to be subtracted.
6.4.1 Flange Couplings
It is required that the depth to 400 HV and the The flange thickness just outside the flange
core hardness are documented to be within fillet is normally to be at least 20% of the
the approved specification. required shaft diameter.
The white layer thickness is not to exceed Coupling bolts are to be prestressed so that a
10µm. suitable amount of pre-stress remains even
under the most severe running conditions, in
As mentioned above, the nitriding process particular with regard to bending moments.
(the coupons) is normally to be documented The safety is to be demonstrated in both the
for each hardening batch and each material Ultimate Limit State and the Fatigue Limit
charge. The use of coupons made of the State. Minimum Safety Factors as for shaft
same material type but not from the same design apply, see Section 6.1.
charge may be accepted provided that the
manufacturer has a quality system, which If the torque transmission is based only on
ensures sufficient reproducibility. friction between the mating surfaces of
flange couplings, the friction torque (in-
Inspection cluding the influence of axial forces and
100% surface crack detection by means of bending moments) is not to be less than 1.5
wet fluorescent magnetic particle method is times the characteristic peak torque.
required for the toothed area including the
ends of the teeth. Upon request, liquid The torque transmission may also be based
penetrant may be considered. on a combination of shear bolts and friction
between the mating flange surfaces. A basic
146 6 – Nacelle
principle is that both the friction alone (in- 6.4.4 Torsionally Elastic Couplings.
cluding the influence of axial forces and
Rubber couplings are to be designed such
bending moments) and the shear bolts alone
that a failure of a rubber element does not
should be able to transmit the characteristic
cause loss of the connection between the
peak torque.
rotor and the brake.
6.4.2 Shrink Fit Couplings
6.4.5 Tooth Couplings
The friction connection is to be able to
Tooth couplings are to have a reasonable
transmit at least 1.5 times the characteristic
safety with respect to surface durability and
peak torque without slipping. Bending mo-
tooth strength. This is subject to special con-
ment influence is to be considered.
sideration.
For tapered mating surfaces where a slip-
page due to torque and/or axial force may 6.5 Mechanical Brake
cause a relative axial movement between the 6.5.1 General
tapered members, the axial movement is to
be prevented by a nut or similar. When a nut Mechanical brakes are usually used as a
is required, the pre-stress is to be of the backup system for the aerodynamic braking
same magnitude as the axial force compo- system of the wind turbine and/or as a
nent from the tape. parking brake, once the turbine is stopped,
e.g. for service. Mechanical brakes are
The permissible material stress depends on sometimes also used as part of the yaw sys-
relative wall thickness, material type, and tem. In a mechanical brake, brake calipers,
whether the coupling is demountable or not, brake discs and brake pads form crucial
and the usual range of permissible equiva- parts. A hydraulic system is usually used for
lent stress (von Mises) is 70% to 110% of the actuation and release of the brake.
the yield strength of the hub.
6.5.2 Types of brakes
6.4.3 Key Connections Mechanical brakes can be active or passive,
The connection is to be able to transmit the depending on how the hydraulic system of
characteristic peak torque. the brake is applied:
The shear stress in the key is not to exceed • active brake: the pressure of the hy-
50% of the yield strength in shear. The pres- draulic system actively pushes the brake
sure on the side of the keyway is not to ex- pads against the brake disc.
ceed 85% of the yield strength of the key. • passive brake: the pressure of the hy-
The yield strength to be applied in checks draulic system keeps a spring tight.
according to these two criteria is not to ex- Once the pressure is released, the spring
ceed 2/3 of the tensile strength of the key, and is also released and will then push the
it is not to exceed twice the yield strength of brake pads against the brake disc.
the shaft or the hub, whichever is involved. In either case, the hydraulic pressure of the
hydraulic system is crucial in order to be
In principle, there is to be no clearance be- able to operate the brake as intended. The
tween the hub and the shaft, however, a cer- hydraulic pressure is usually provided by
tain amount of minimum interference fit is means of an accumulator. For active sys-
required, e.g. approximately 0.02% of the tems, it is particularly important to make
shaft diameter. sure that the pressure in the accumulator is
always available, and it is important to have
6 – Nacelle 147
1 Cover for cardan 19 Washer

2 Suspension for screw pump 20 Set screw
3 Foundation plate for brake arrange- 21 Washer
ment 22 Set screw
4 Brake disc for hub 23 Washer
5 Middle plate for distance pipe 24 Hexagonal socket head cap screw
6 Fitting for cover 25 Set screw
7 Lock screw for brake disc 26 Set screw
8 Bushing 27 Set screw
9 Bushing 28 Washer
10 Cover for brake 29 Hexagonal head bolt
11 Side cover for brake, top 30 Slotted raised countersunk head
12 Side cover for brake, bottom screw
13 Sensor pick-up plate at cardan 31 Brake caliper
14 Fitting for sensor 32 Blade caliper air bleed
15 Angle iron for cover 33 Screw pump
16 Washer 34 Brake lining
17 Hexagonal head bolt 35 Fork sensor
18 Hexagonal head bolt 36 Shims
Figure 6-22 Example of brake arrangement; courtesy Vestas.
148 6 – Nacelle
redundancy in this respect, i.e. an extra pres- 6.5.4 Brake torque sequence
sure source is necessary for backup.
The rise time from zero to maximum brake
torque will influence the dynamic response
The type of spring used in a mechanical
of the turbine heavily. Because the turbine is
brake to keep up a pressure is often a coil
rotating when the brake is actuated, full
spring of the disc spring type. This type of
brake force will be mobilised immediately
spring is nonlinear and has the advantage
upon the actuation of the brake.
that it is capable of maintaining an approxi-
mately constant spring pressure over a con-
siderable range of deflection.
An example of a brake arrangement is given

in Figure 6-22.
6.5.3 Brake discs and brake pads

Brake discs and brake pads must be able to
withstand temperature loading, since the
friction during braking leads to dissipation
of energy in terms of heat and causes high
temperatures to develop locally.
Brake pads may be made from different Figure 6-23 Temporal evolution of torque at
kinds of materials. Ceramic brake pads do and after grid loss with subsequent actuation
not withstand high temperatures (tempera- of the brake
tures in excess of 300-400°C) very well, in
the sense that they lose their frictional re- For design of the brake system, it is impor-
sistance. For high temperatures, brake pads tant to consider the maximum torque during
made from sinter bronze can be used. the course of the braking. Depending on the
dynamics, the maximum torque may well
Brake discs must be subject to temperature occur towards the end of this course, and
calculations or temperature measurements. transient vibrations may sometimes follow.
They must meet requirements to planarity in An example of the temporal evolution of the
order to work properly, i.e., they must not torque during the course of braking is given
warp when subjected to temperature loading. in Figure 6-23. The turbine is rotating when
In general, the thicker the disc, the better is a grid loss occurs, and the torque drops to
its ability to absorb temperature loading. zero. When the connection to the grid is lost,
the brake is actuated, and the torque raises
The possible variation in the frictional coef- and develops as shown in the figure.
ficient poses a problem, which must be
given due consideration when the brake 6.6 Hydraulic systems
system is being dimensioned: If the fric-
6.6.1 General
tional coefficient is too big, the brake force
will become too large. If the frictional coef- In a hydraulic system, power is transmitted
ficient is too small, the brake system will be and controlled through a liquid under pres-
unable to brake. sure within an enclosed circuit. Hydraulic
systems are used in wind turbines, for ex-
ample in terms of a hydraulic accumulator
6 – Nacelle 149
for pitching of the rotor blades. A hydraulic rated if their mixture would be dangerous or
system must be protected against exceeding would result in the contamination of the
the maximum admissible pressure. A pres- hydraulic fluid and/or loss of gas through
sure release valve can be used for this pur- absorption.
pose and will prevent explosion in the event
of fire. All components of the hydraulic Each accumulator is to be protected on both
system must be easily accessible for assem- its gas side and its hydraulic fluid side by a
bly, adjustment and maintenance. safety device such as a relief valve, a fuse
plug or a rupture disc to prevent excessive
Pressure shocks should be kept to a mini- pressure if overheated. When the accumu-
mum. Pressure shocks or a large pressure lator is an integral part of a system with such
drop must not lead to a dangerous condition. a safety device, the accumulator itself need
A safe condition must be guaranteed in the not be supplied with such a safety device.
event of power supply failure and in the
subsequent event of restoration of the power 6.6.4 Valves
supply.
Valves are used in hydraulic systems to
control the hydraulic effect between a pump
The following external factors must not af-
and an engine, cylinder, or actuator. The
fect the operation of a hydraulic system:
purpose of valves is to govern the direction
• salt and other corrosive substances
and amount of the volume flow rate or to
• sand and dust block the volume flow rate, and it is also to
• moisture limit or control the pressure of the fluid. A
• external magnetic, electromagnetic and distinction can be made between four major
electric fields types of valves, viz.
• sunlight • shut-off valves, which block flow in one
• vibrations direction and allow for partly or full
flow in the opposite direction
When a hydraulic system forms a part of the • directional control valves, which control
safety system, grid failures and extreme the direction of the volume flow rate,
temperatures must not compromise the op- and which can block the volume flow
eration of the system. rate or adjust the amount of volume
flow rate
6.6.2 Arrangement • pressure valves, which limit or control
Hydraulic systems should have no connec- the hydraulic pressure
tions with other piping systems. The hy- • flow control valves, which adjust the
draulic fluid is not to have a flash point volume flow rate to the hydraulic ac-
lower than 150°C and is to be suitable for tuator such that the desired speed of the
operation at all temperatures that the system hydraulic actuator is achieved.
may normally be exposed to. Means for fil-
tration and cooling of the fluid are to be in- 6.6.5 Application in safety systems
corporated in the system wherever neces- If a hydraulic system forms part of a safety
sary. system, the design and construction of the
hydraulic system must comply with the re-
6.6.3 Accumulators quirements to the safety system, and it must
For gas and hydraulic fluid type accumula- be designed, constructed and used as a fail-
tors, the two media are to be suitably sepa- safe or redundant system. Hydraulic sys-
150 6 – Nacelle
tems, which form parts of safety systems, blockage of the pipe as the result of
can be divided into three categories as fol- maloperation.
lows: 2. For actively released brakes, which are
being actuated hydraulically or pneu-
1. systems in which the brake is actively matically, the actuation must be accom-
released by a hydraulic or pneumatic plished by means of a pressure accu-
pressure medium. mulator, and the following additional
2. systems in which the brake is actively requirements apply:
released (mechanically, hydraulically, • the connecting pipe between the
pneumatically or electrically), but actu- accumulator and the actuated com-
ated hydraulically or pneumatically. ponent, e.g. the blade adjustment
The active release thus takes place mechanism, must be as short as
against a ”passive” oil or air pressure. possible.
3. systems in which the brake is released • no other components, such as
in the neutral state (passively released) valves, couplings and rotating ele-
and is actuated hydraulically or pneu- ments, are allowed in this connec-
matically. tion.
• the pressure in the accumulator
It is recommended to comply with the fol- must be monitored at a level which
lowing requirements to these three types of is sufficiently high to guarantee in-
hydraulic systems: dependent braking action.
1. When brakes (mechanically or aerody- • the other safety system of the wind
namically) are actively released, the turbine must be actuated mechani-
pressure medium shall be able to flow cally, and the actuating element
away in a reliable manner during a shall be designed and constructed
braking action. In the hydraulic system, ”safe-life”, e.g. a mechanical
the following features shall be present spring.
as a minimum: 3. The use of a passively released braking
• two valves must be placed parallel system, by which the braking action
to one another as a switching ele- takes place by means of build-up of
ment. An incorrect switch position pressure, is only permissible when the
of a valve must lead to a safe situa- following conditions are met:
tion. It must be possible to test each • the monitoring, control and actua-
valve individually. tion systems are designed and con-
• the return pipes must be made suf- structed with redundancy, i.e. they
ficiently strong by their design and are designed and constructed at
construction, or by protection, such least in duplicate, and function in-
that they cannot be closed off by dependently of the electrical grid.
external damage. • the pressure build-up is supported
• filters must be installed on the pres- by a pressure accumulator to which
sure side of the pump and should the same requirements apply as
be avoided in the return pipe. If those stated above.
filters are installed in the return • the other safety system of the wind
pipe, a bypass is required. turbine must be equipped with a
• no components must be installed in ”safe-life” actuation element.
the return pipe which can lead to
6 – Nacelle 151
In addition, for Categories 2 and 3, an ac- Generators are electromechanical compo-

tively operated hydraulic or pneumatic in- nents. A generator consists of a rotor and a
stallation shall not be used to keep a wind stator. The rotor may consist of a magnet or
turbine in a safe state for a long period after a cage of conducting bars. The rotor is lo-
a safety system has been actuated. cated in the centre of the stator, which con-
sists of a static arrangement of electromag-
6.6.6 Additional provisions nets. The electromagnets are hooked up to a
grid. When a torque is applied to the rotor
It is recommended to take the necessary
and makes it rotate, such as when it is cou-
steps to ensure that failure of a redundant
pled to the main shaft of a wind turbine, the
system can be detected. Long-term stand-by
electromagnets will produce electric power
redundancy should be avoided. In the case
by induction, and transmit the generated
of oil leaks in hydraulic systems, other wind
alternating current to the grid. An example
turbine components or other systems must
of a generator is shown in Figure 6-24.
not be affected.
6.6.7 Codes and Standards

The following guidelines and standards deal
with design of hydraulic systems
Teknisk Forlag, Hydraulik Ståbi, in Danish,

ed. T. Rump, Teknisk Forlag a.s., Copenha-
gen, Denmark, 1996.
ISO, ”Hydraulic fluid power – General rules

for the application of equipment to transmis-
sion and control systems,” International
Standard, ISO4413, 1st edition, 1979.
Figure 6-24 Six-pole generator © Danish
ISO, ”Hydraulic fluid power – Gas-loaded Wind Turbine Manufacturers Association
accumulators with separators – Range of
pressures and volumes, characteristic quan- The produced alternating current, which is
tities and identification,” International Stan- transmitted to the electrical grid, must match
dard, ISO5596, 1st edition, 1982. the frequency of the grid. For a two-polar
generator, this implies that the rotational
speed of its rotor needs to be 50 rotations
6.7 Generator per second, equal to 3000 rotations per min-
ute, when it is connected to a 50 Hz grid.
6.7.1 Types of generators Generators usually have two, four, or six
The generator is the unit of the wind turbine poles. The higher the number of poles in the
that transforms mechanical energy into generator, the slower is the rotational speed
electrical energy. While the blades transfer needed to keep the frequency of the grid.
the kinetic energy of the wind to rotational For a six-pole generator, the rotational speed
energy in the transmission system, the gen- can thus be reduced to 1000 rotations per
erator provides the next step in the supply of minute. The required rotational speed of the
energy from the wind turbine to the electri- generator’s rotor is achieved by means of
cal grid. the gearbox of the wind turbine, since the
wind turbine rotor itself is not allowed to
152 6 – Nacelle
rotate at this high speed for physical reasons.

Rotation of the wind turbine rotor at the high The advantage of the variable slip comes
rotational speed of the generator rotor would about when the wind turbine is operated at
cause aerodynamic problems and supersonic its rated power. At the rated power, power
speeds. Noise would also be a problem, and fluctuations when the wind speed changes
excessive centrifugal forces would be gener- are undesirable. When a wind gust hits the
ated. A multi-pole generator has the advan- wind turbine rotor, the slip enables the gen-
tage of reducing the mechanical complexity erator speed to increase a little in response to
of the generator and makes it possible to the gust without causing a corresponding
reduce the gearbox and sometimes even increase in the generated power output. The
omit it. slip thus ensures a smooth power output and
at the same time contributes to keep the
There are two major types of generators loads on blades, main shaft and gearbox
• synchronous generators down.
• asynchronus generators
The variation of the operating speed with the
A synchronous generator is a generator, applied torque for an asynchronous genera-
which operates at a constant speed, dictated tor is beneficial because it implies a smaller
by the frequency of the connected grid, re- peak torque and less wear and tear on the
gardless of the magnitude of the applied gearbox than for a synchronous generator.
torque. The speed dictated by the frequency This is one of the most important reasons for
of the grid is also known as the synchronous using an asynchronous generator rather than
speed. a synchronous generator in a wind turbine,
which is connected directly to the electrical
An asynchronous generator is a generator, grid. Another reason for using asynchronous
which allows slip, i.e. deviations from the generators is that the involved slip is benefi-
rotational speed dictated by the frequency of cial when there is flexibility in the structural
the connected grid. In other words, the rota- system.
tional speed is allowed to vary somewhat
with the applied torque. This is the most Traditionally, the active materials in a gen-
common generator type used in wind tur- erator consist of magnetically conducting
bines. A variant with coiled rotors is preva- iron and electrically conducting thread ar-
lent. The slip is defined as the difference ranged in a coil. Permanent magnets are
between the rotational speed of the generator becoming still more used, and electrical
and the rotational speed dictated by the fre- components such as temperature sensors are
quency of the grid. The slip is sometimes becoming integral parts of the generators.
expressed in percent of the latter. When a
slip of up to about 1% is possible, the op- With a view to switches and power failures,
eration mode for the asynchronous generator the generator must be able to withstand
is referred to as constant speed, as this slip is switching on in phase opposition at 100%
rather insignificant. When a larger slip is residual voltage, or it must be secured
allowed, say within 10%, and can be varied against the occurrence of this situation by
electronically, e.g. by a rotor current con- special arrangement in the control system.
troller, it is referred to as variable slip. The
pitch or stall control of the wind turbine is Generators are to be constructed in such a
meant to ensure that the allowable slip of the way that when running at any working
generator is not exceeded. speed, all revolving parts are well balanced.
6 – Nacelle 153
Suitable fixed terminal connectors are to be not running, and by heating of closets, when
provided in an accessible position with suf- the self-heating is insufficient to avoid dam-
ficient space for convenient connection of aging condense of water.
the external cables.
Resistance toward saline atmospheres
6.7.2 Climate aspects When the wind turbine is to be located near
a coastline or offshore, the electrical equip-
Temperature
ment shall be constructed in such a way that
The generator shall be designed to be fully
it will not be damaged by the impact from
functional at the temperatures that can occur
the saline and moist environment. The en-
locally, when the external ambient tem-
capsulation, the cooling and the insulation of
perature is within the range [−10,30]°C for the generator are all to be designed in such a
normal operation. These limits refer to manner that it can withstand these impacts.
Danish conditions. For other countries, this For equipment in the turbine, this can be
range is typically expanded to [−20,30]°C. achieved by applying climate control in
The ambient temperature is the instantane- terms of desalination and dehumidification
ous value of the temperature of the air out- systems or by heating to avoid condensation
side the wind turbine. The temperature range and saline deposit.
for the location of the generator should be
documented, and the self-heating of the gen- Electrical immission and emission
erator should also be documented. When the Electric and electronic equipment, whose
generator is assembled at its intended loca- functions can be affected by electrical im-
tion, it needs to be assured that it is not mission, shall meet the requirements given
placed near heat dissipating components, or in the EMC directive as described in
it has to be designed to withstand the associ- DS/EN50082-2, Generic Immunity Stan-
ated temperatures. dard, Industrial Environment. Electric and
electronic equipment, from which electrical
Relative humidity of air emission can occur, shall meet the require-
To avoid leak currents, i.e. low insulation ments given in the EMC directive as de-
resistances, corrosion and other damaging scribed in DS/EN50081-2, Generic Emis-
influences on the components, these compo- sion Standard, Industrial Environment.
nents should − either by their design or by
climate control in the wind turbine − be se- 6.7.3 Safety aspects
cured so that damaging condense cannot
occur. The generator forms one of several links in
the transmission between the rotating system
The electrical components of the turbine and the electrical system of a wind turbine,
shall be fully functional at a relative humid- i.e. between the blades and the grid. Failure
ity in the external ambient air of 95% within of any link in this chain implies a risk. This
the entire temperature range as specified risk is absorbed by the protection system,
above. which brings the wind turbine to a safe con-
dition in which it remains until normal op-
The influence of the humidity of air on the eration can be resumed. The number of fail-
electrical components are always dependent ures that demand activation of the protection
on other climatic parameters, in particular system should be minimised in order to re-
on the temperature and on changes in the duce the burden on the protection system.
temperature. Creation of condense can be The probability of breakdown results from
remedied by heating when the generator is the probability of failure of the protection
154 6 – Nacelle
system combined with the probability of a forms part of both the protection system and
critical error that requires the intervention of the control system. Errors which have com-
the protection system. The probability of mon causes needs to be given special atten-
breakdown shall be less than 0.0002 per tion. Neither blade pitch nor mechanical
machine year. The target for the reliability brake will have the required effect if the
of the protection system is a number, which blades are locked at a position of, for exam-
is large enough to keep the probability of ple, +15°.
breakdown below this level.
For such concepts, the target reliability
The grid-connected asynchronous generator quoted above still applies. Components
with short-circuited cage winding in the which form part of both control and protec-
rotor has been the basis for the safety con- tion functions are to be fail-safe designed, or
siderations behind the Danish approval their probability of failure is to be mini-
scheme and for the associated recommenda- mised.
tions.
The generator is to be designed such that it
For a stall-controlled wind turbine with can produce a sufficiently large torque to
asynchronous generator, it is the generator keep the turbine within its defined range of
and its relatively simple control that most of operation, see Section 2.2.1. Disconnection
the time constitute the system that keeps the (switching off) of the generator should be
turbine in a safe condition. If the generator based on reverse power, on zero-power, or
is disconnected due to an error or due to on a signal from the relay protection against
intervention by the protection system, at electrical errors. The purpose is to utilise the
least one of the two fail-safe brake systems, braking power of the generator in all situa-
which are part of the protection system, shall tions.
begin working. These are usually pitchable
blade tips and mechanical brakes. “Fail- When a frequency converter is used, it needs
safe” is defined as a design philosophy by to be considered together with the generator
which the safety of the turbine is maintained with respect to safety and frequency of er-
even during component failure or grid fail- rors.
ure.
The generator is to be designed for the me-
For a turbine with variable speed and chanical impacts it will be exposed to. At
pitchable blades, and thus a relatively com- start-up of a turbine with fixed unpitchable
plex control, it is the interaction between blades in high wind speeds, the cut-in of the
these features and the generator, which en- generator should be performed at an under-
sures that the turbine is kept in a safe condi- synchronous number of revolutions to limit
tion. If the generator is disconnected due to the acceleration until the full number of
an error or due to intervention from the pro- revolutions is achieved. At pitch control of
tection system, or if an error occurs in the the blades, a control strategy is to be applied
blade pitch system or the speed control, at which ensures limitation of the acceleration.
least one of the two fail-safe brake systems,
which are part of the protection system, shall 6.7.4 Cooling and degree of sealing
begin working. These are usually pitchable
blades and mechanical brakes. Note that for The cooling system of the generator should
turbines with pitchable blades, one of the as a minimum correspond to IC41 for jacket
two brake systems – the pitchable blades – cooling according to DS/EN60034-6.
6 – Nacelle 155
The generator, including its possible exter- Generators are to be capable of withstanding
nal encapsulation and its external cooling the following overspeed during two minutes:
system with cooling agents such as air or 1.25 times the rated maximum speed.
water, should as a minimum be protected
against external impacts corresponding to 6.7.7 Overloading
degree of sealing IP54 in accordance with
The wind turbine is to be automatically
DS/EN60034-5.
controlled in such a way that it will be
brought to either standstill or idling at a low
The machine cabin of the wind turbine is not
rotational speed, if the average produced
considered as a sufficient enclosure of the
power in 10 minutes exceeds 115% of the
generator for protection against unintended
nominal (rated) power Pnom. For a stall-
intrusion of objects such as tools, dust from
controlled wind turbine with a conventional
brakes, and hydraulic liquids. The internal
asynchronous generator, it is in addition
shielding in the cabin is to provide safety for
required that it shall be brought to a stand-
personnel and against unintended objects.
still or idling if the produced one-second
The temperature can be higher in the cabin
mean power exceeds 140% of the nominal
than outside the turbine, in which case the
(rated) power. The corresponding limiting
generator is to be designed for the tempera-
value for a pitch-controlled turbine with a
ture within the cabin, unless some other ex-
conventional asynchronous generator is
ternal cooling system is in place.
200% of the nominal (rated) power Pnom.
6.7.5 Vibrations
For a passively controlled generator, the
The generator shall be balanced such that it instantaneous value of the torque shall as a
as a minimum fulfils the requirements to minimum be 1.35 times the one-second
Class N according to DS/IEC60034-14. The value. For an actively controlled generator,
generator shall be capable of withstanding which forms part of the protection system,
vibrations from other parts of the wind tur- the overload capacity of the generator shall
bine. be documented for each individual concept.
At short-circuiting of the grid and at short-
6.7.6 Overspeed term grid failures, the generator shall be
capable of absorbing the thermal and dy-
The generator is to be constructed such that
namic forces.
it fulfils the requirements to overspeed ac-
cording to DS/EN60034-1 and DS/EN
6.7.8 Materials
60034-3. For stall-controlled turbines with
pitchable blade tips, a typical upper limit for Permanent magnets
overspeed in the context of grid failure and Permanent magnets shall be designed in
activation of the blade tips will be about 1.4 such a manner that the minimum induction
times the operational number of rotations. locally in the magnet during an electrical
fault will not fall short of the break point
Synchronous generators may produce over- value, at which irreversible demagnetisation
voltage, which may be damaging to the fre- sets in at some extreme temperature. It is
quency converter. To prevent overvoltage, recommended to keep a margin to the break
the voltage should be adjusted downward. point value of at least 0.1 Tesla for NeBFe
When permanent magnets are used, the gen- and 0.05 Tesla for ferrite materials. Me-
erator terminals may be short-circuited or chanical stability of NeBFe magnets can be
loaded by brake resistances.
156 6 – Nacelle
protected against corrosion by means of with at-rest heating (heating system for use
coating by tin, zinc or similar. at standstill). In case of grid failure and
longer periods of standstill, it must be en-
Coils sured that the coil has dried out before the
Coils are to be constructed with an insula- turbine is restarted. Note that for a generator
tion, which as a minimum meets the re- in protection class IP54, at rest heating is
quirements to Class F according to normally not considered necessary.
IEC60085. The temperature rise at maxi-
mum load must not exceed the limits set Bearings
forth for the chosen class of insulation in When frequency converters are used, ca-
IEC60034-1. This refers to coils in air- pacitive couplings might produce flow paths
cooled generators. The allowable tempera- through the bearings. This might imply a
ture rise values are reproduced in Table 6-6 reduced lifetime for the bearings. The im-
together with acceptable values for the total pact shall be reduced to a level that the
temperature. bearings can withstand, e.g. by insulation of
the bearings, or by dU/dt filtering.
When a frequency converter is used, the coil
insulation becomes exposed to a larger im- Bearings are to be efficiently and automati-
pact due to large voltages U and large time cally lubricated at all running speeds and
derivatives dU/dt of the voltage. The impact within the service intervals specified by the
can be reduced by application of filters for manufacturers. Provisions is to be made to
smoothing of the voltage. It shall always be prevent the lubricant from gaining access to
ensured that the insulation of the generator windings or other insulated or exposed con-
can resist the impacts that it is exposed to. A ducting parts.
frequently applied maximum value for dU/dt
is 1 kV/µsec. However, in some cases up to 6.7.9 Generator braking
5 kV/µsec is used. As a minimum, the insu- By controlling a frequency converter, the
lation shall be capable of resisting impulse generator can be used to reduce the rota-
voltages of 1300 V as measured on the gen- tional speed to the level that corresponds to
erator clamps. Alternatively, a different wire the minimum frequency of the converter.
can be used to obtain extra insulation. This is referred to as generator braking. By
means of external brake resistances, braking
Table 6-6 Allowable temperatures and tem- by means of the generator can form part of
perature rise values the protection system brakes as well as of
Insu- Temperature Total tempera- the operational brake system, provided that
lation rise (°C) ture (°C) the generator remains magnetised during the
class braking and that braking to bring the turbine
A 50 105 to a safe condition is fail-safe. Generator
E 65 120 braking is suitable for synchronous genera-
B 70 130 tor types, including permanently magnetised
F 90 155 generators and multiple-poled generators
H 115 180 without gearbox, for which mechanical
braking is made difficult by large dimen-
If there is a risk that the internal heating is sions.
insufficient to avoid damaging condensation
of water in the generator, in particular at
standstill, the generator shall be furnished
6 – Nacelle 157
6.7.10 Lifetime Generators should have sufficient momen-

tary and steady short-circuit current in rela-
The generator shall be designed for the same
tion to the release characteristics of switch
lifetime as the rest of the wind turbine. The
and fuse gear on the installation, thereby to
design lifetime shall be at least 20 years.
ensure a reliable release by short-circuits
This requirement does not apply to compo-
anywhere in the grid. This can be verified by
nents subject to wear, for which replacement
random tests.
intervals are specified in the user’s manual
for the actual type of wind turbine.
It is recommended that the steady short-
circuit current should normally not be less
6.7.11 Testing of generators
than 3 times the full-load current.
For new generators, it is recommended that
the manufacturer carries out tests as speci- 6.8 Machine Support Frame
fied in the following. The tests may be ar-
The machine support frame is located on top
ranged as type tests, as production sample
of the tower and supports the machinery
tests, or as routine tests.
including the gear box. Usually it also pro-
vides support for the nacelle cover. In con-
The manufacturer’s test reports should pro-
trast to the tower, the machine support frame
vide information about make, type, serial
is usually a very turbine-specific structure,
number, insulation class, all technical data
which can be constructed in many different
necessary for the application of the genera-
ways and according to many different lay-
tor, and the results of the tests.
outs. It can be a welded, bolted or cast steel
structure. Sometimes it is formed as an inte-
Recommended tests:
gral part of the gear box, i.e. the gear box
• temperature test at full load (minimum itself acts as the machine frame. Sometimes
1.15 times rated load) it is integrated with the yaw system. Some-
• overload test times it is as simple as a big plate. It is in
• overspeed test general a much less standardised product
• high-voltage test than a tower.
• measurement of insulation resistance
• measurement of resistance of windings For design of the machine support frame, the
• measurement of air gap following issues are important to consider
• open-circuit voltage characteristics • a sufficient stiffness of the frame must
• short-circuit current resistance be ensured in order to meet the stiffness
• measurement of excitation current at requirements for the machinery: The
rated current, voltage and power factor gear box should not be able to move
(if possible also at cosφ=0 lag, where φ relative to the bearings, and the yaw
denotes phase angle) system will not get sufficient mesh with
• short-circuit test the gear rim if the frame is not suffi-
ciently stiff.
Generators should withstand terminal short- • sufficiently fine tolerances must be met
circuits at least during one second without during the manufacturing of the frame
damage to the generator itself or to its exci- in order to facilitate the proper posi-
tation equipment. This can verified by type tioning and assembly of the yaw system
test or random test. and the transmissions.
158 6 – Nacelle
• the frame must be designed against fa- Artificial light has to be installed to ensure
tigue owing to its exposure to the rotor safe access and working conditions.
forces.
• the access to the nacelle is through the 6.10 Yaw System
tower and the machine support frame,
which implies that the machine support Yaw is the rotation of the nacelle and the
frame must include a hole of convenient rotor about the vertical tower axis. By yaw-
size for personnel to pass. ing the wind turbine, the rotor can be posi-
tioned such that the wind hits the rotor plane
The machine support frame is exposed to at a right angle. The yaw system provides a
rotor loads consisting of thrust, yaw moment mechanism to yaw the turbine and keep the
and tilt moment. These load components are rotor axis aligned with the direction of the
not necessarily in phase. In addition comes wind. The situation when this alignment is
weight. Also local forces wherever forces not achieved produces a yaw error. The yaw
are being transmitted should be considered, error is defined as the angle between the
i.e. forces at bearing housings, gear shafts, horizontal projections of the wind direction
and yaw system. It is recommended to use and the rotor axis.
finite element methods for structural analy-
sis of the machine support frame. The yaw system can be either passive or
active. A passive yaw system implies that
6.9 Nacelle Enclosure the rotor plane is kept perpendicular to the
direction of the wind by utilisation of the
The purpose of the nacelle enclosure is to surface pressure, which is set up by the wind
protect the machinery and the control system and which produces a restoring moment
of the wind turbine against rain and moisture about the yaw axis. For upwind turbines,
and against salt and solid particles such as this usually requires a tale vane in order to
sand grains in the air. The nacelle enclosure work properly. Also coning of the rotor can
is also meant to protect against noise and is help keeping the nacelle in place. Note that a
therefore often covered with some noise- passive yaw system may pose a problem in
reducing material. The enclosure needs to be terms of cable twist if the turbine keeps
tight to fulfil its purpose. However, it also yawing in the same direction for a long time.
needs to have ventilation to allow for ade- An active yaw system employs a mechanism
quate cooling of the gear system. of hydraulic or electrically driven motors
The nacelle enclosure is to be designed for and gearboxes to yaw the turbine and keep it
wind load. In this context it is important to turned against the wind. Such active posi-
choose correct lift and drag coefficients. tioning of the turbine relative to the wind is
also referred to as forced yaw. Most larger
The nacelle enclosure is usually used also as horizontal axis wind turbines use forced yaw
a walkway and needs to be designed with to align the rotor axis with the wind. An
sufficient dimensions for this purpose. In example of an active yaw system is given in
this context it is particularly important to Figure 6-25.
consider the design of the fixtures of the
enclosure. The mechanism used for an active yaw sys-
tem usually consists of a number of electri-
It is important that the nacelle enclosure can cally operated motors in conjunction with a
be opened to allow for removal of damaged gear that actuates a large toothed yaw ring in
components for replacement, for example by the tower circumference. These components
means of a helicopter.
6 – Nacelle 159
together with yaw brakes and yaw bearings vanes and activates the yaw mechanism ac-
are most often delivered as standard compo- cordingly. In addition to this automated yaw
nents from a supplier, who also provides the of the wind turbine, it should be possible to
pertaining design documentation. Note in yaw the nacelle manually. Manual yaw is
this context that these components when needed during start-up, service of the yaw
used in a yaw system may be exposed to system, and tests of the turbine.
conditions which have not been taken into
account by the supplier. The following sec- 6.10.1 Determination of Design Loads
tions deal with the different components of
Yaw is characterised by the maximum an-
an active yaw system, including the yaw
ring, the yaw drive with the yaw motors, the gular yaw velocity ωk and the fraction of the
yaw bearing, and the yaw brakes. design life during which yaw takes place.
When the duration of yaw is not known for
The yaw error is usually measured by means design, DS472 specifies that yaw can be
of direction sensors such as one or more assumed to take place during 10% of the
wind vanes. The wind vanes are usually time for all wind speeds that occur. Assum-
placed on top of the nacelle. Whenever the ing yaw in 10% of the time is, however,
wind turbine is operating, an electronic con- quite conservative for most sites.
troller checks the orientation of the wind
Yaw drive
YAW DRI-
VE
MACHINE
M FRAME
ashin frame (nacelle)
(NACELLE)
Slewing
bearing Yaw gear wheel
SB
Yaw brake
YAWBRAKE
SB=SLEWING BEARING
Tower top
Figure 6-25 Typical active yaw system involving a slewing bearing
160 6 – Nacelle
M zR
FN
yN
F zR
M yR F yR yR
F xR zN
zR Yaw bearing
M xR Rotor centre
Figure 6-26 Loads acting on the rotor and the yaw bearing
The static load consists of the weight of the M 2 = M xR − FzR y R − F yR z R + FN y N
nacelle and rotor, acting as an axial force on
the bearing. This load is superimposed by
the wind load on the rotor as illustrated in Fr = FyR 2 + FxR 2
Figure 6-26. Fa = FzR + FN
In Figure 6-26, the rotor loads are drawn in The yaw moment depends on the magni-
directions according to a conventional coor- tudes of the yaw error and the wind speed.
dinate system and not in the direction they The direction of the yaw moment depends
will normally have. on the direction of rotation of the rotor in
conjunction with the direction of the yaw
FxR - Side force on rotor and nacelle error.
FyR - Thrust on rotor
FzR - Weight of rotor The extreme design yaw moment, which is
MxR - Tilting moment at rotor the design basis for yaw drives and yaw
MyR - Driving torque at rotor brakes, is likely to appear during operation
MzR - Yaw moment at rotor at a maximum wind speed with a maximum
FN - Weight of nacelle yaw error. Dynamically, the yaw moment
yN - Horizontal distance to nacelle c.o.g. will have a tendency to oscillate with a fre-
yR - Horizontal distance to rotor c.o.g. quency of x times the rotor frequency, where
zN - Vertical distance to nacelle c.o.g. x denotes the number of blades.
zR - Vertical distance to rotor c.o.g.
Figure 6-27 shows a computer simulation of
From these quantities the static loading on the yaw moment as a function of yaw error
the yaw bearing, tilt moment Mtilt, yaw mo- and wind speed for a three-bladed wind tur-
ment Myaw, radial force Fr and axial force Fa, bine. It appears that the yaw error oscillates
can be calculated as: with frequencies equal to the rotor frequency
and three times the rotor frequency. It also
Myaw = MzR + FxR ⋅ yR + Mbrake + Mfriction appears that the sign of the yaw moment
changes when the yaw direction changes,
Mtilt= M 1 + M 2
2 2
and – not surprisingly – the yaw moment
increases when the yaw error and the wind
with speed increase.
M 1 = M yR + FxR z R
6 – Nacelle 161
6.10.2 Yaw Drive speed, the yaw motors needs to be connected

through a gearbox.
The yaw drive is the system of components
used to cause the yaw motion. A large yaw
Such drives can be delivered as standard
speed will produce gyro effects that will
equipment from manufacturers of electri-
result in large loads on the wind turbine. The
cally operated motors.
yaw speed must therefore be small enough
that gyro effects become negligible. Refer-
The yaw drives must have sufficient power
ence is made to Section 4.2.1 for details.
to overcome the largest mean yaw moment
occurring plus friction in the yaw bearing.
A passive yaw system will normally require
Installation, lubrication and service are to be
some sort of damping arrangement to reduce
in accordance with the specification from
the yaw speed.
the manufacturer.
For larger turbines with active yaw systems,
Pinion shaft, pinion and mounting must be
the yaw speed is usually lower than 1°/sec, designed to withstand the maximum yaw
which is small enough that gyro effects can moment including partial safety factors.
be ignored. To achieve such a low yaw
Figure 6-27. Relation between wind speed, yaw error and yaw moment.
162 6 – Nacelle
6.10.3 Yaw Ring ing. Braking can then be performed by

means of hydraulic activated callipers. Pas-
The yaw drive is normally mounted in the
sive hydraulic callipers, i.e. spring-applied
nacelle in such a way that it is in gear with a
brake callipers, are preferable because they
toothed yaw ring mounted on the top of the
can ensure braking also in the case of a leak-
tower. Yaw rings are normally toothed on
age in the hydraulic system.
the inner surface as this provides a better
protection from the surroundings and im-
The brake calliper manufacturer will provide
plies a slightly better mesh.
surface tolerances and geometrical tolerances
for the brake disc. These tolerances must be
The gears connected to the drive and the
complied with.
yaw ring must be designed in such a manner
that tooth failure for the maximum peak yaw
The yaw brake callipers, discs and mounting
moment is prevented.
bolts shall be designed to withstand the
maximum occurring yaw moments. For tur-
Bending stress in a tooth can be calculated
bines with yaw brake systems designed with
according to DIN 3990, Part 3 from:
limited safety (S = 1.15) it shall be proven
that the turbine will withstand free yaw op-
≤ σ Fl
F
σF = ⋅ ⋅ eration, i.e. inertia forces due to acceleration
b ⋅m Y f Yε SF
of the nacelle.
Yf – Tooth shape factor The brake is to be protected against dust,

Yε – Load reduction factor corrosion, oil and any other influence that
F – Tooth force might alter the friction. The brake and con-
b – Tooth width trol system shall be designed such that a
m – Module situation where pads are worn out is avoided.
σFl – Tooth yield stress
SF – Safety factor In addition to using the ordinary yaw brake,
it must be possible to block the yaw mecha-
Scuffing might occur in gears with hardened nism to enable service and maintenance ad-
pinions whose tip edges act as scrapers. jacent to the yaw mechanism without any
Damage due to wear and fatigue should be personal risk.
considered.
Blocking of the yaw system can be done by
It should be noted that the yaw ring is not mechanical fixation of the yaw ring or the
only loaded during yawing. Even though yaw motor shaft. Blocking of the yaw sys-
yaw brakes might be applied a part of the tem must not be dependent on external
yaw moment might be transmitted through power supply. Only in the case of negative
the yaw drive. callipers, the motor is allowed to be electri-
cally braked. A fixation of the ring is prefer-
6.10.4 Yaw Brake able because a possible backlash in the yaw
During power production, turbulence, wind gear can be large enough to cause the na-
shear and smaller inevitable yaw errors will celle to jerk during service.
give rise to a torque moment about the tower
axis. To keep the nacelle in position and to The locking mechanism must be designed
spare the gears, it is common to mount a for wind speeds up to a predefined wind
brake disc in connection with the yaw bear- speed, above which maintenance, which
6 – Nacelle 163
demands locking of the yaw mechanism, is Slewing bearings used in yaw systems are
not permissible. different from normal bearings in that they
are exposed to oscillatory motion. This os-
6.10.5 Yaw Bearing cillatory motion and the low rotational speed
imply that slewing bearings have a tendency
The yaw bearing is the bearing that supports
to exhibit a relatively low ratio between the
the nacelle in a horizontal axis wind turbine.
lubrication film thickness and the surface
It is located between the rotating nacelle and
roughness.
the stationary tower and transmits wind
loads from the nacelle to the tower. For the
Calculation of load rating for bearings in
yaw bearing, it has been common in the past
oscillatory motion is described in NWTC. It
to choose between two different solutions –
is quite complex and involves the variation
slide plates or rolling bearings.
of load and displacement/position as a func-
tion of time. In practice, it is common that
Rolling bearings will often be designed as
the bearing manufacturer provides the de-
slewing bearings, which are capable of ac-
sign calculation or means of how to verify
commodating combinations of axial, radial
the design.
and moment loads. Yaw motion is generated
by gears mounted in the nacelle and being in
Unlike regular rolling bearings, whose
gear with the toothed bearing. Slewing
strengths are represented by the load rating
bearings, as seen in Figure 6-28, are
C, the strength of a slewing bearing will
mounted by bolting to the seating surfaces.
normally be represented in terms of a curve
Usually bolts of quality 10.9 are recom-
that gives the relation between the allowable
mended, see Appendix A.
equivalent tilt moment and the equivalent
axial load. Reference is made to Figure 6-30
Tolerances for the mounting surfaces should
for an example.
be in accordance with specifications from
the bearing manufacturer. To prevent the
The equivalent axial load Feq is calculated
bearing from becoming distorted, the contact
from the radial load Fr and the axial load Fa
surfaces must be carefully machined and
in the same way as for ordinary rolling
attention must be drawn to the stiffness of
bearings,
the surrounding structure. Application of a
plastic grouting may compensate for irregu-
Feq = (X⋅Fa + Y⋅Fr)⋅ KA ⋅ KS
larities in the contact surfaces.
in which X and Y are combination factors
which depend on bearing type and the ratio
between Fa and Fr. Values of X and Y will be
provided by the manufacturer. KA is an ap-
plication factor and is recommended to be in
the range 1.7-2.0 for yaw bearings. The
safety factor KS equals the partial safety
factor for looad for the relevant load case.
The tilt moment as calculated in Section

6.10.1 is denoted Mt, and the equivalent tilt
moment Meq is calculated as
Figure 6-28. Four Point Ball Bearing
164 6 – Nacelle
Meq = Mt ⋅ KA ⋅ KS lent load for a specified load spectrum or a

set of representative load cases can be cal-
in which KA and KS are to be taken as for the culated according to the formula
axial load.
Feq
åP ⋅O
i
p
i
P= p
i
,
åO i
L in which P denotes either the equivalent

axial force Feq or the equivalent tilt moment
P Meq, depending on whether axial force or tilt
moment is considered. Oi denotes the dura-
O Meq tion of the load Pi corresponding to the ith
load block in a discretisation of the load
Figure 6-29. Bearing life curve spectrum or the ith load case when a load
The service life curve, which is exemplified case representation is adopted. The exponent
in Figure 6-29, is usually provided by the p is referred to as the bearing exponent, and
manufacturer of the bearing. The service life is to be taken as 3 for ball bearings and 10/3
curve represents the combinations of for roller bearings.
equivalent tilt moment and equivalent axial
load for which the failure probability at a A curve similar to the one shown in Figure
specified number of revolutions is 10%. 6-29 is used to calculate the stress reserve
Note that different manufacturers specify factor during extreme loading of the bearing.
different numbers of revolutions for their Calculation of equivalent loads Feq and Meq
respective service life curves. The number is also analogues. The different application
of revolutions is a measure of the bearing factors, X, Y etc. will however differ.
life. When Meq and Feq have been calculated,
the corresponding point P can be plotted as When slewing bearings are provided with
shown in Figure 6-29. The extension of the spur gears, the manufacturer will normally
line OP intersects with the service life curve specify the allowable tangential force with
in the point L. The bearing life, specified by reference to the bending stress at the root of
the manufacturer in terms of a number of the tooth in order to prevent tooth failure.
revolutions, can now be increased by multi-
plication by a factor which is to be calcu- The friction torque moment in a slewing
lated as the line length ratio |OL|/|OP|. bearing can be calculated as:
As for any other component in the wind Myaw = µ⋅(a⋅Mt + b⋅Fa⋅Dr + c⋅Fr⋅Dr),
turbine, the dynamic loading of the bearing
can be represented by one or more load in which
spectra, or – on a simpler form – by a num- Myaw – friction moment,
ber of load cases that are assumed to ade- µ – friction coefficient,
quately represent the loading that the bear- Mt – tilting moment,
ing will experience over the design life of Fa – axial force (gravity),
the turbine. In the latter case, at least four Fr – radial force (thrust + side),
load cases should be modelled. The equiva- Dr – raceway diameter.
6 – Nacelle 165
Table 6-7 gives properties for a few types of

For a four-point ball bearing the coefficients relevant materials. The mechanical proper-
can be set to: ties listed apply to the unreinforced material.
The data given are only to be considered as
µ = 0.006, a = 2.2, b = 0.5 and c = 1.9 a guidance, since data from different manu-
facturers vary considerably.
Proper lubrication of the yaw bearing is of
great importance. It is advised to follow the Polyurethane (PUR) has a high ultimate
bearing manufacturer’s instructions regard- strength combined with a large elongation at
ing type and amount of grease, and relubri- failure, i.e. it exhibits a ductile behaviour. It
cation intervals. has good wearing properties and exhibits
constancy towards oil and grease. Different
One way of improving the lubrication in a types of PUR are categorised according to
bearing is periodically to let the bearing ro- their “Shore hardness”. The Young’s
tate through an angle equal to at least one modulus of elasticity of polyurethanes
bearing segment. grows exponentially with the Shore hard-
ness.
Bearing seals should be checked every six
months. Polyamide (PA) is characterised by a com-
bination of mechanical strength and chemi-
DNV (1992) recommends that the safety cal resistance, good sliding properties and
against failure for the slew ring toothing is to high fatigue strength. PA is categorised ac-
be 1.6. cording to the number of C-atoms in its
molecules.
An alternative to a slewing bearing is a slide
bearing, which is often designed as a num- Acetal (POM) offers excellent inherent lu-
ber of plates and claws that engage the na- bricity, fatigue resistance, and chemical re-
celle to the tower, see Figure 6-30. sistance. Acetals suffer from outgassing
problems at elevated temperatures, and are
brittle at low temperatures.
Polyethylene Terephtalate (PET) exhibits

good creep constancy, which means it can
absorb large static loads. It also exhibits
good dimensional stability, wear properties
and low friction. On the negative side, it is
sensitive to notches.
Plates used for slide bearings will normally

need to be greased to obtain a sufficiently
Figure 6-30. Schematic diagram of slide low friction torque and for corrosion protec-
bearing tion of the steel parts.
The materials used for slide bearings are
cast polyamide plates or similar materials A major difference between rolling bearings
which are relatively strong, with good slid- and slide bearings is that slide bearings in-
ing and hard-wearing properties. volve larger frictional resistance. Slide
166 6 – Nacelle
Table 6-7 Material properties for materials used in slide bearings

PA 66 PUR PET POM
Tensile strength 52 MPa 83 MPa 46 MPa 61 MPa
Compressive strength 60 MPa NA 97 MPa 31 MPa
Flexural module 1379 MPa 3447 MPa 2758 MPa 2620 MPa
Hardness Rockwell R 100 119 120 107
Temperature
Maximum 121 °C 110 °C 100 °C N/A
Minimum −79 °C −40 °C −15 °C N/A
Source: www.plasticsusa.com
bearings therefore require larger yaw motors 4 revolutions. The sensor can be designed
but less brake capacity than rolling bearings. quite simply as a switch on the yaw drive
which is activated once per revolution.
6.10.6 Yaw Error and Control
For redundancy in the cable twist monitor-
Yaw errors have a significant impact on the
ing system, the wind turbine can also be
blade loads, the rotor loads and the power
equipped with a pull switch, which becomes
production, hence it is important that the
activated when the cables become too
yaw mechanism is efficient and reliable.
twisted.
Even small yaw errors give rise to increased
fatigue loads in the blades due to the varia-
To emphasise the importance of preventing
tion in the inflow angle during the rotation
cable twist, it suffices to note that the tur-
of the rotor that they cause.
bine can easily experience 50-100 rotations
in the same direction during one year, if it is
If the yaw error becomes too big, i.e. it falls
allowed to rotate unrestricted.
outside the specified or assumed range of
operation, the wind turbine must be stopped.
6.10.8 Special design considerations
Special arrangements to prevent long-term Due to yaw moments being cyclic as men-
yaw errors can be made. One such arrange- tioned in Section 6.10.1, loose fits should be
ment is formed by using a set of independent avoided.
wind vanes. In case one of the vanes gets
stuck, the vanes will misalign. An error REFERENCES
message will then be produced and the wind
AGMA, Recommended Practices for Design
turbine will be stopped.
and Specification of Gearboxes for Wind
Turbine Generator Systems, American Gear
6.10.7 Cable twist
Manufacturers Association, Alexandria, Vir-
In larger power producing turbines, cables ginia, 1997.
are needed to conduct the current from the
wind turbine generator down through the Bergmann, J., and R. Thumser, Forschung
tower. The cables will become twisted if the für die Praxis P 249, Synthetische Wöhler-
turbine keeps yawing in the same direction linien für Eisenwerkstoffe, Studiengesell-
for a long time. The wind turbine must schaft Stahlanwendung e.V., Verlag und
therefore be equipped with a cable twist Vertriebsgesellschaft mbH, Düsseldorf,
sensor, which monitors the number of revo- Germany, 1999.
lution and informs the controller when it is
time to untwist the cables, usually after 2 to
6 – Nacelle 167
Bonus, Bonus Info, Special Issue, The Wind ISO, “Rolling bearings – Dynamic load rat-
Turbine Components and Operation, ings and rating life,” International Standard,
Autumn, Brande, Denmark, 1999. ISO281, 1st edition, 1990.
Danish Energy Agency, “Technical Criteria ISO, “Rolling bearings - Static load ratings,”
for the Danish Approval Scheme for Wind International Standard, ISO 76, 2nd edition,
Turbines,” Copenhagen, Denmark, April 1987.
2000.
Niemann, G., and H. Winter, Maschinenele-
Dansk Ingeniørforening, Last og sikkerhed mente, Band II, Getriebe allgemein, Zahn-
for vindmøllekonstruktioner, in Danish, radgetriebe – Grundlagen, Stirnradgetriebe,
DS472, 1st edition, Copenhagen, Denmark, Springer-Verlag, Berlin, Germany, 1985.
1992.
NWTC, “Guideline DG03, Wind Turbine
Det Norske Veritas, Calculation of Gear Design, Yaw & Pitch Rolling Bearing life,”
Rating for Marine Transmissions, DNV National Renewable Energy Laboratory,
Classification Note No. 41.2, Høvik, Nor- NWTC – Certification Team.
way, 1993.
Peterson, R.E., Stress Concentration Fac-
Det Norske Veritas, “Guidelines for Certifi- tors, John Wiley and Sons, New York, N.Y.,
cation of Wind Turbine Power Plants,” Co- 1974.
penhagen, Denmark, 1992.
DIN 743, Tragfähigkeitsberechnung von Pilkey, W.D., Peterson’s Stress Concentra-
Wellen und Achsen, 1998. tion Factors, 2nd Edition, John Wiley and
Sons, New York, N.Y., 1997.
DIN 743 part 1-3 Tragfähigkeitsberechnung
von Welle und Achsen. Roloff, H., and W. Matek, Maschinenele-
mente. Formelsammlung, Vieweg Verlag,
FAG Technical Information, FAG Rolling Braunschweig/Wiesbaden, Germany, 1994.
Bearings, TI No. WL 43-1190 EA.
SKF, “General catalogue,” 1989.
FAG OEM und Handel AG, Rolling Bearing
Lubrication, Publ. No. WL 81 115/4 EA. SKF, Roller Bearings in Industrial Gear-
boxes, Handbook for the gearbox designer,
Gudehus, H., and H. Zenner, Leitfaden für SKF, Denmark, 1997.
eine Betriebsfestigkeitsrechnung, 4. Auflage,
Verein Deutscher Eisenhüttenleute, Düssel- Sundström, B., Handbok och formelsamling
dorf, Germany, 1999. i Hållfasthetslära, Institutionen for hållfas-
thetsläre, KTH, Stockholm, Sweden, 1998.
IEC, Wind turbine generator systems – Part
1: Safety requirements, International Stan- VDI Berichte 1442, Festigkeitsberechung
dard, IEC61400-1, 2nd Edition, 1999. Metallischer Bauteile.
168 6 – Nacelle
7. Tower mon tower structures. The main features of

the different tower types are briefly dealt
with below.
7.1 Introduction
The tower of a wind turbine supports the Tubular Towers
nacelle and the rotor and provides the neces- Most large wind turbines are delivered with
sary elevation of the rotor to keep it clear of tubular steel towers, which are manufactured
the ground and bring it up to the level where in sections of 20-30 m length with flanges at
the wind resources are. The towers for large either end. The sections are bolted together
wind turbines are typically made of steel, on the site. The towers are conical, i.e. their
but concrete towers are sometimes used. As diameters increase towards the base, in order
of today, most towers are tubular towers, to increase their strength towards the tower
however, also lattice towers are in use. base, where it is needed the most, because
Guyed towers are used only for relatively this is where the load response owing to the
small wind turbines. The tower is usually wind loading is the largest. The conical
connected to its supporting foundation by shape allows for saving on the material con-
means of a bolted flange connection or a sumption, since the necessary shell thickness
weld. becomes reduced when the diameter in-
creases.
In the context of a wind turbine, the tower is
a low-technology component, whose design The maximum length of the tower sections
is easy to optimize, and which therefore – is usually governed by requirements to allow
during the design process – lends itself eas- for transportation. Also the upper limit for
ily as an object for possible cost reduction at the outer diameter of the tower is usually
very little effort. This may come in useful as governed by such requirements, at least for
the cost of a tower usually forms a signifi- land-based turbines – the maximum clear-
cant part of the total cost of a wind turbine. ance under highway bridges in Denmark is
4.2 m. An advantage of tubular towers com-
Figure 7-1 shows an array of the most compared to other towers is that they are safer
Tubular steel tower Tubular concrete Lattice tower Three-legged tower Guy-wired pole
tower tower
Figure 7-1 Various tower structures, from www.windpower.org, © Danish Wind Turbine
Manufacturers Association
7 – Tower 169
and more comfortable for service personnel 7.2 Load Cases

and others that have to enter and climb the
The load cases to be investigated for tower
towers.
design are described in Chapter 4.1. How-
ever, some special load cases may apply
Lattice Towers
particularly to the tower, for example those
Lattice towers are manufactured using
that need to be considered when a long
welded steel profiles or L-section steel pro-
tower is to be transported. In addition, the
files. The basic advantage of lattice towers is
erection of the tower on site may involve
reduced cost, since a lattice tower requires
critical loads. Such special cases should be
only about half as much material as a freely
investigated as appropriate.
standing tubular tower with a similar stiff-
ness. It also gives less wind shade than a
7.3 Design Loads
massive tower. The major disadvantage of
lattice towers is their visual appearance, The design loads are most often determined
although this is a debatable issue. Never- from an aeroelastic analyses as described in
theless, for aesthetic reasons lattice towers Section 4.3 or – less frequently − from a
have almost disappeared from use for large, simplified calculation, see Section 4.5. Re-
modern wind turbines. gardless of which analysis approach is used
for these calculations, the loads should be
Guyed Pole Towers calculated from a model, in which the tower
Many small wind turbines are built with properties (geometry, materials, stiffnesses)
narrow pole towers supported by guy wires. are in agreement with the ones used in the
The advantage is weight savings, and final design. Thus, the design of the tower
thereby reduced costs. The disadvantages may demand an iterative procedure to get
include difficult access around the towers, from an initial design to the final design
which make them less suitable in farm areas. with the correct stiffness.
Finally, this type of tower is more prone to
vandalism, thus compromising the overall The design loads for fatigue are to be deter-
safety. mined by calculations, which are to be sup-
plemented and verified by actual measure-
Other Types ments from a prototype turbine. Since load
Some towers are designed as hybrids of the measurements cannot be made until the tur-
bine has been designed and constructed, it is
No of blades 3
recommended to apply an additional partial
Hub height 71.5 m safety factor of 1.2 on fatigue loads until
Rotor diameter 66 m measurements are carried out and become
Rotational speed 15.9 rpm available. This precaution with the use of an
Rated power 1800 kW additional safety factor is meant to avoid a
Tower Tubular major redesign in the event of increased
Table 7-1: Main Characteristics of design loads.
Example Turbine
The extreme loads for design can only be
above types, for example the three-legged determined by calculations, because these
tower for a 95 kW turbine in Figure 7-1. loads cannot be measured due to the very
long recurrence period between events.
170 7 – Tower
When designing a turbine with fixed speed, sponses are to be determined from the prob-
the frequency of the rotor revolution is of ability distribution of the extreme response,
outmost importance. This frequency, often see Section 4.4. This fluctuation is due to the
referred to as ‘1P’, may induce increased dynamics of the turbine in conjunction with
dynamic loads, e.g. due to rotor unbalances, the stochastic turbulence field.
wind shear and tower shadow. In addition,
the higher ‘P’s’ are of importance, e.g. the 7.3.2 Extreme loads occurring during
‘2P’ and the ‘3P’, which are the frequenciesnormal power production
of blades passing the tower on a 2- and 3-
The results of a response analysis based on
bladed turbine, respectively. When design-
49 realisations (calculations with different
ing a turbine with variable speed, one must
seeds) of the turbulence field is shown in
verify that the rotor speed of the turbine
Table 7-3. The load case is analysed for
does not operate in or near the first natural
U10=24 m/s and IT=11%, and the simulation
frequency of the tower, see Section 7.4.1.
time for each realisation is 10 minutes.
7.3.1 Loads and responses
Table 7-3 shows results for four important
The behaviour of the tower during extreme processes as interpreted from the 49 calcu-
loading is illustrated through an aeroelastic lations. These processes are the wind speed
calculation for an example turbine with a process, the electrical power and the hori-
rated power of 1800 kW. The main charac- zontal force and moment at the foundation
teristics of this turbine are given in Table 7- level. Considering the variation of the proc-
1, and its dynamics are given in Table 7-2. esses and the time of operation, it is obvious
that a proper selection of the characteristic
An aeroelastic response calculation for an value must be rooted in probabilistic meth-
operating turbine – showing the bending ods as described in Section 4.3.2.
moments – is shown as a function of time in
Figure 7-1. The response is calculated for a Table 7-3 provides simulated estimates of
10-minute mean wind speed U10=24 m/s, the first four central moments of the moment
which is near the cut-out wind speed, and a and horizontal force responses, viz. the
turbulence intensity IT=11%. From Figure mean, the standard deviation, the skewness
7-1 it appears that the response is highly and the kurtosis of these two responses. The
fluctuating with time, thus the design re- bottom lines in the table give estimates of
Configuration of turbine
Blades: Blades:
In normal position 90o pitched
Mode shape Freq. [Hz] Damp. (%) Freq. [Hz]
1st tower transversal 0.418 6.0 0.417
1st tower longitudinal 0.419 6.0 0.420
1st rotor torsion 0.805 5.0 0.704
1st rotor torsion 0.979 1.002
1st asymmetric rotor (yaw) 1.000 1.064
1st symmetric rotor (flap) 1.067 3.1 1.769
1st edgewise mode 1.857 3.1 1.032
2nd edgewise mode 1.045
Table 7-2: Dynamic Properties of Example Turbine
7 – Tower 171
Figure 7-1: Tower Response from analysis of a 1.8 MW tur-

bine (left = alongwind response, right = transverse wind re-
sponse). Operational loads for 24 m/s wind speed
the maximum moment and the simultaneous 2. Calculate the statistics (as exemplified
value of the force, and of the maximum in Table 7-3) for each executed simula-
force and the simultaneous value of the tion (at least 10 minutes).
moment. The central estimates are denoted Run Wind Power Moment Horis.
by µ in the table. Statistical uncertainty in statistics M force, F
the estimates is also included in terms of the Mean µ 24.09 1810 14323 210
standard deviations of the estimates, denoted σ 0.00 5.5 28.8 0.413
by σ in the table. It appears that the maxi- Std.dev. µ 2.63 119.6 2046 30.9
mum of the moment response process and
σ 0.01 4.2 152 2.3
the maximum of the force response process
Skewness µ -0.052 -0.187 -0.028 -0.055
do not occur at the same time.
σ 0.218 0.085 0.127 0.239
The characteristic extreme response and the Kurtosis µ 2.953 6.650 2.950 2.953
design extreme response for an operating σ 0.377 0.550 0.256 0.239
turbine should be selected according to the Upcross. µ 6.703 1.714 2.520 1.678
following scheme: period σ 0.881 0.052 0.748 0.896
max[M] µ 28.21986.7 21018 312
1. Simulate/calculate the extreme re- and simul. σ 2.9 90.2 777 12
sponses for the codified load cases and max[H] µ 27.41978.0 20907 314
if necessary include other load cases and simul. σ 1.9 95.9 812 12
that might be critical for the considered
concept (e.g. variable speed etc). For Table 7-3 : Extreme loads during power
each load case at least 5 different seeds production (simulation time 10 minutes).
should be used.
3. Use the procedures in Section 4.4.2 to
project the responses calculated from a
172 7 – Tower
10-minute simulation to a longer period The extreme loads calculated from 60

of operation (e.g. more days – calcu- simulations of the response processes are
lated from the mean wind speed distri- shown in Table 7-4. The mean wind speed
bution – see Chapter 3). is 34 m/s and the turbulence intensity is
4. Choose the characteristic value of the 10%. The central estimates are denoted by µ
response as described in the code. (Not in the table. Statistical uncertainty in the
many codes deal with this issue, and estimates is also included in terms of the
those that do do not necessarily define standard deviations of the estimates, denoted
the characteristic value in the same by σ in the table. From the data it appears
manner.) that the statistical variations for the various
5. Apply a safety factor as described in the considered responses are significant also in
codes (the revised edition of the Danish the parked condition.
code of practice DS472 specifies γf=1.5)
to calculate the design value of the re- The characteristic extreme response should
sponse. be chosen equal to the expected maximum
response in a 10-minute period. This ap-
7.3.3 Extreme loads – parked turbine proach is justified only if the coefficient of
variation on the maximum value is suffi-
An extreme load analysis is also performed
ciently low, see Mørk. An approach by Dav-
with the example turbine in a parked con-
enport can be used for this purpose, see
figuration. The wind is acting perpendicular
Section 4.4.2. The corresponding design
to the nacelle, i.e. with a full drag on the
value is found by applying the appropriate
nacelle, and the blades are pitched to give a
partial safety factor (the current edition of
full drag also on the blades.
DS410 and the revised edition of DS472
Run Wind Moment Horis. specify γf=1.5).
statistics M force, F
7.3.4 Fatigue Loading
Mean µ 33.51 12293.7 234.2
σ 0.001 23.3 0.45 The fatigue loads are also fluctuating loads,
Std.dev. µ 3.333 2555.2 45.79 thus one must assure that a sufficient num-
σ 0.005 79.3 1.09 ber of simulations are performed before the
Skewness µ -0.027 0.208 0.206 design of the tower can be carried out. A 1.5
σ 0.205 0.203 0.210 MW stall-controlled turbine with a tubular
Kurtosis µ 2.939 3.091 3.055 tower as reported in Thomsen (1998) is con-
sidered as an example, and the fatigue loads
σ 0.248 0.266 0.233
are as shown in Table 7-5. The table is cal-
max[U] µ 43.74 17515.3 331.5
culated with the Wöhler curve exponent
and simul. σ 1.03 2362.6 26.4 m=3, which is the value traditionally se-
max[M] µ 40.30 21490.8 395.9 lected for steel, and the equivalent number
and simul. σ 1.72 1052.0 17.4 of cycles is 600. The turbulence intensity is
max[H] µ 40.38 21449.9 396.5 set to 15%.
and simul. σ 1.67 1056.4 17.0
Table 7-5 shows the variation in the
Table 7-4: Extreme loads for parked equivalent load as a function of the mean
turbine (simulation time 10 minutes) wind speed. The table is based on a large
number of simulations for each wind speed,
and both the mean and the coefficient of
variation of the load estimates are included,
7 – Tower 173
denoted by µ and COV, respectively. The

coefficient of variation appears to be rather
large (greater than 10%) in this example,
despite the large number of simulations used
for the estimations. This indicates the im-
portance of carrying out a sufficient number
of simulations. It is recommended not to use
less than five simulations, see Danish En-
ergy Agency (2000), however, five simula-
tions may in many cases be very insuffi-
cient.
Wind speed Statis- Equiva-

[m/s] tical lent load
para- [kNm]
meter
7 µ 828.7
COV 17%
10 µ 1511.1
COV 10%
15 µ 3151.1
COV 13%
20 µ 6059.4 Figure 7-2: Sensitivity of 10-minute
COV 15% equivalent load to wind speed (above) and
24 µ 7703.1 lifetime fatigue load to turbulence intensity
COV 15% (below). Error bars correspond to ±one
Table 7-5 : Equivalent Load for standard deviation in estimates.
Tower Bending in a 10-minute
Simulation. The dependency of the damage-equivalent
load on the wind speed and on the turbu-
The equivalent load as shown in Table 7-5 lence intensity is shown in Figure 7-2. The
must be corrected with respect to the proper upper half of the figure shows the equivalent
operation time at different wind speeds (ad- load in a 10-minute simulation period as a
justing for the distribution of the mean wind function of the wind speed. Also three levels
speed – resulting in the lifetime equivalent of turbulence intensity are analysed. From
load). The mean wind velocity is assumed to the figure, it is clearly seen that the fatigue
follow a Weibull distribution with scale pa- damage increases with the wind speed and
rameter A=10 m/s and slope parameter
k=2.0. The lifetime fatigue equivalent load Quantity Stat. Eq. load
for a 20-year lifetime is given in Table 7-6, par. [kNm]
calculated for 107 equivalent load cycles. Req µ 10292.42
Note that the uncertainty in the lifetime fa- COV 6%
tigue loading is still significant with a coef- neq, L 107
ficient of variance of 6%. This may call for a
Table 7-6: Equivalent Load for Tower
safety factor for fatigue loading different
Bending in a 10-minute Simulation
from 1.0.
174 7 – Tower
Figure 7-3: Time Series: Response at tower bottom.
Figure 7-4: Power Spectrum: Response at tower bottom.

turbulence intensity. Also the uncertainty in found., from which it appears that the major
the damage increases with increasing wind fatigue contribution arises from the ‘normal
speed and turbulence. The error bars in production’ load combinations. Load com-
Figure 7-2 illustrate the mean of the fatigue binations involving failures or errors do
load plus/minus one standard deviation. hardly influence the fatigue damage at all.
The load combinations are identical to those
The lower half of Figure 7-2 gives the life- specified in DS472.
time damage and the lifetime equivalent
load as functions of the site-specific design 7.4 General Verifications for Towers
turbulence intensity. The figure shows that
the lifetime equivalent load increases from 7.4.1 Dynamic Response and Reso-
about 6000 kNm to about 14000 kNm when nance
the turbulence intensity increases from 10% The first natural frequency of the tower
to 20%. The fatigue damage is correspond- should always be measured in connection
ingly – due to the nonlinearity in the fatigue with the erection of a prototype turbine.
Wöhler curve – increased by a factor of 8.
It should be verified that the first natural
An analysis of a 500 kW stall-controlled frequency of the tower does not coincide
turbine gives the relative contribution to the with the rotor and blade-passing frequencies,
fatigue damage from different load cases i.e. the 1P and 3P frequencies, respectively.
and is reported in Thomsen et al. (1997). If it is confirmed that the tower frequency is
The results of this investigation are repro- kept outside ranges defined as the rotor fre-
duced in Error! Reference source not
7 – Tower 175
quency ±10% and the blade passing fre- generator speeds, this investigation should
quency ±10%, respectively, then there will be performed for both corresponding blade-
normally not be any problems due to load passing frequencies. Special attention should
amplification arising from vibrations at or be given to variable-speed turbines, in which
near the natural frequency. cases the turbine should not be allowed to
For turbines with two generators or two operate in a frequency interval defined as
LC U 10 Yaw I T Cont. Comment

no. [m/s] [deg] [%] [%]
1 6 -10 18 0,9
2 8 -10 18 1,7 7,0
3 10 -10 18 3,1
4 12 -10 18 4,1 6,0
LC's: normal power production
5 14 -10 18 5,0
6 16 -10 18 6,5 5,0
7 18 -10 18 6,4
8 20 -10 18 6,1
9 22 -10 18 5,9 4,0
10 24 -10 18 5,6
11 6 +10 18 0,9 3,0
12 8 +10 18 2,3
13 10 +10 18 4,1 2,0
14 12 +10 18 5,0
15 14 +10 18 5,3 1,0
16 16 +10 18 6,6
17 18 +10 18 6,6
0,0
18 20 +10 18 6,1
1 4 7 10 13 16 19 22 25 28 31 34
19 22 +10 18 5,9 Load combination number
20 24 +10 18 5,3
LC's: With errors LC's: starts/stops
21 Low Start 0,0

22 Nom Start 0,1
23 High Start 0,1
24 Low Stop 0,0
25 Nom Stop 0,1
26 High Stop 0,1
27 5 Idle 0,1
28 0.5 Umax -40 18 0,1 Tip brakes activated
29 0.5 Umax 0 18 0,0 Tip brakes activated
30 0.5 Umax +40 18 0,2 Tip brakes activated
31 25 -40 18 0,4 Large yaw error
32 25 40 18 1,0 Large yaw error
33 25 -10 18 2,1 Only one tip brake
34 25 10 18 2,6 Only one tip brake
35 25 -10 18 0,0 Only two tip brakes
Table 7-7: Relative Contribution to the Total Fatigue Damage from Different
Load Combinations (load cases according to DS472, 500 kW turbine, Tower
Subjected to Bending).
176 7 – Tower
the eigen frequency of the tower ±10%. d 0 − γ ⋅ u max > F

When turbines are operating in an over-
synchronous mode, i.e. where the 1P fre- where
quency is larger than the first tower natural d0 is the distance between the tower
frequency, one must include a proper analy- and the blade tip in the un-
sis of the start-up and stopping sequences. loaded/undeflected condition.
This should be done in order to account for umax is the maximum deflection of the
the increased dynamic loading that occurs blade considering all relevant load
when the rotor frequency is passing through cases and based on characteristic
the first natural frequency of the tower. load values and characteristic mate-
rial properties
An example for the 1.8 MW turbine is γ is the partial safety factor on the
shown in Figure 7-4 and Figure 7-4. From maximum deflection of the blade,
the power spectrum it appears that the major to be chosen according to the rele-
part of the energy is found at the natural vant load case, see Table 7-8.
frequency of the tower at 0.418 Hz. F Requirement to the residual clear-
ance between the tower and blade
The 1P to 4P passing frequencies are not
tip, usually 0.0.
close to the first natural frequency and fur-
thermore not much energy is found on the
passing frequencies in the response spec- The definition of the safety factor γ depends
trum. on which set of standards is used for the
critical blade deflection analysis.
It should be noted that the natural frequency
of the tower is very dependent on the ‘effi- According to the IEC61400-1, the safety
ciency’ of the presumably fixed support at factor on the maximum deflection is taken
the foundation level. If the calculation as equal to the corresponding partial safety
model assumes that the tower is completely factor for load γf times the material factor
fixed, then the error in the natural frequency γm. The value of γm to be used for this pur-
of the tower may be up to 20%. Guidelines pose depends on the coefficient of variation
for selection of elastic springs as supports of the strength and on the definition of the
for a tower whose foundation is not rigid can characteristic strength. When the coefficient
be found in Chapter 8. Note that it is be- of variation is 10% and the characteristic
coming practice to use tower dampers to strength is defined as the 5% quantile with
compensate for low damping perpendicular 95% confidence, then γm =1.1. The value of
to the direction of the wind. γf to be used depends on which load case
governs the design and needs to be extracted
When large variations in the foundation from the IEC61400-1.
stiffness are encountered within a particular
project, such as the development of a wind According to the Danish practice, an overall
farm, the natural frequency in bending safety factor γ is used. Note that the Danish
should be measured at all installations. practice with use of γ=1.3 is currently sub-
ject to discussion and should be used with
7.4.2 Critical Blade Deflection Analysis care.
The distance between tower and the blade
tip should comply with the following condi- When considering the blade-to-tower dis-
tion: tance, one should be aware that not only the
blades but also the tower will deflect when
7 – Tower 177
exposed to wind loading, and the deflection which is located at a height H above the
of the tower may be out of phase with the tower base. Note that this height may devi-
deflection of the blades. ate somewhat from the hub height.
Note that the clearance between blade and Section loads in the tower at height h can be
tower is not only governed by the structural calculated from the loads applied at the top
deflections, but also by a possible slip at the of the tower:
yaw bearing, by the perpendicularity of the
tower flange, and by the tolerances on the H
tilt and on the rotor plane. Fz (h) = FzT + ρ t ò A( z) dz

H −h
Large blade deflections are typically en-

countered in extreme wind situations, but M z (h) = M zT
also in the operating condition when the yaw
angle is large or when the terrain is sloping. F y ( h) = F yT
+ F w (h)
Atypical wind profiles may also give rise to
large blade deflections. Special attention
should therefore be given to wind turbine M x (h) = M xT + F yT ⋅ ( H − h) + M w (h)
designs that are sensitive to special load + FzT ⋅ (δ ( H )-δ (h))
cases such as negative wind shear in com-
plex terrain. MzT
The current Danish practice is to use vali-

FzT
dated stiffness data in the aeroelastic calcu-
lations for analysis of deflections. Validated FyT MxT
stiffness data are blade stiffness data, which wind
comply with the experimental data from
static tests of the blades. This implies that
one may have to produce two sets of aero-
elastic calculations. The first is a load cal-
culation, in which the model is tuned to the
H
correct natural frequencies and damping
properties of the structural system. The sec-
ond is a special deflection model where the
deflections are tuned to be in agreement
with the static experiments. Ideally, the two
models would be the same, but in practice,
this is not always the case.
h
7.5 Tubular Towers
7.5.1 Loads and responses
For the purpose of calculating section loads
in the tower, the tower can be viewed as a Figure 7-6 Cantilever beam model of tubu-
cantilever beam as shown in Figure 7-6. lar tower subject to loading at the level of
External loads, denoted by index T in this the hub
figure, are applied at the tower top flange,
178 7 – Tower
in which the symbols are explained as fol-

lows 7.5.2 Extreme loads
For identification of the loads that govern
Fy thrust from wind load
the design, the combination of load compo-
Mx bending moment from wind load
nents that produces the highest stress must
Fz gravity force
be found. This can be quite a task when an
Mz torsional moment
aeroelastic analysis program is used that
ρt density of tower including appurte-
simulates a large number of load cases in
nances
10-minute time series. Further, which load
A(z) cross-sectional area as a function of
case actually governs the design will most
height z
likely vary between different sections of the
δ deflection of tower due to thrust tower.
from wind
Alternatively, loads must be combined by
External loads, here denoted by index T, are taking the maximum of each load compo-
assumed to include the dynamic effect or nent from the particular load case where the
gust factor (referring to a quasi-static ap- most dominant load has its maximum, or
proach). more conservatively they can be combined
by combining the maxima of the various
In particular, in MxT, it is important to in- load components regardless of which load
clude the contribution from the possible case they actually appear in.
eccentricity of the nacelle.
7.5.3 Fatigue loads
The section force Fw(h) and the moment
Mw(h) from the wind load on the tower can Combining fatigue loads is an even more
be calculated as complicated task. When using the rain-flow
method as described in Section 4.4.1, the
H load spectra for the different load compo-
Fw ( h ) = ½ ò ρV ( z ) ϕD ( z )C ( z )dz
2
nents are normally not directly combinable.
H −h
Therefore it might be a good idea, if possi-

H
ble, to combine the time series of the various
M w (h) = ½ ò ( H − h − z ) ρV ( z ) 2 ϕD ( z )C ( z )dz load components resulting from the aero-
H −h
elastic simulation. For example, the result-
in which the symbols are explained as fol- ing bending moment in the direction θ rela-
lows tive to the y-axis can be calculated as
ρ air density
V(z) wind speed M res = M x sin θ + M y cos θ
D(z) outer tower diameter
C(z) form factor in which Mx and My are the bending mo-
ϕ gust factor ments associated with the load components
in the x and y directions, respectively.
C(z) depends on the Reynold’s number,
Re=VD/ν, in which ν denotes the kinematic This could be taken even further to calculate
viscosity of air. At 20°C, ν=15.09⋅10−6 the stress at relevant sections in the tower
m2/sec. For painted steel towers, C(z) can be for every time step during the simulation
set to 0.6.
7 – Tower 179
and then rain-flow count the resulting time Vortex-induced vibrations are normally not
series for the stress. a problem after installation of the tower and
the wind turbine. Once the nacelle is in
7.5.4 Vortex Induced Vibrations place, its weight will lower the critical wind
speed for vortex-induced vibrations to a low
The turbine must be checked for vortex in-
level − typically below 10 m/s − which is
duced vibration. The vortex excitation may
within the interval of power production.
occur during mounting of the turbine, i.e. in
a situation where the rotor and nacelle have
When the blades rotate and pass the tower,
not yet been mounted on the tower. A sug-
they will reduce the wind speed and create
gested procedure to be followed for this
turbulence in the wind that passes the tower
purpose can be found in Eurocode 1 – Sec-
behind the blades, thereby obstructing the
tion 2.4 – Annex C or alternatively in the
generation of vortices.
Danish DS410, according to which the criti-
cal wind speed vr can be calculated as
Another aspect, which contributes to reduce
the effect of vortex-induced vibrations, is
n⋅D
vr = the aerodynamic damping of the blades and
St the nacelle.
in which the symbols are explained as fol- 7.5.5 Welded joints

lows
Welds are in general treated in the same
n tower natural frequency manner as the rest of the structure when a
D tower diameter proper reduction factor for the weld quality
St Strouhal number and base material is included.
For conical towers, D should be set equal to Table 7-8 shows the recommended detail
the top diameter. categories for bolts with rolled threads after
St
heat treatment and common welds in tubular
0.21 towers according to the standards Eurocode
0.20
3 and DS412. The given detail categories
0.19
0.18
assume 100% controlled full penetration
0.17 butt welds of quality level B according to
0.16 DS/ISO 25817.
0.15 H/b
1 10 100
Figure 7-7. The Strouhal number vs. ratio Weld Categories

between tower height H and tower diameter Plate to plate 80
b. Plate to flange 71
Plate to doorframe 80
The analysis might prove that certain wind Axially loaded bolts 71
velocities should be avoided when erecting
the tower. However, the sensitivity to vortex Table 7-8. Detail categories for bolts and
vibrations may be changed by temporary common welds in tubular towers.
guy wiring of the tower or mounting a tem- Figure 7-8 shows typical welds in a tubular
porary mass near the top of the tower. tower.
180 7 – Tower
According to DNV (1987), the stress con-

centration factor for the single sided plate
tapering can be calculated as:
e
SCFtaper = 1 + 6 ,
æ æt ö
2.5 ö
t 2 ç1 + çç 1 ÷÷ ÷
a. b. c. ç t ÷
Typical weld Typical weld Weld be- è è 2 ø ø
at door frame at flange tween two
shells of dif- in which t1 and t2 are the plate thicknesses of
ferent thick-
the lower and upper part of the tower shell,
ness
respectively, and the eccentricity e is given
Figure 7-8 Typical weld details in tubular by
tower
e = ½(t 1 − t 2 )
Note that for the weld between the tower
shell and the flange in Figure 7-8b, the given When using quality levels poorer than B
detail category assumes a small shell thick- according to DS/ISO 25817, it is recom-
ness relative to the flange thickness. mended to apply detail categories a corre-
sponding number of levels lower.
Note also that a weld between two shells of
different thickness as shown in Figure 7-8c The fatigue damage can be calculated using
is symmetrically tapered to avoid stress con- the Palmgren-Miner’s rule as described in
centrations. The slope of the tapering should Appendix C.
not be greater 1:4.
Whenever the weld is perpendicular to the
In case of single-sided tapering, as shown in direction of loading and the material thick-
Figure 7-9, a stress concentration is intro- ness t is greater than 25 mm, the fatigue
duced. strength σfatd should be reduced according to
the following formula
σred,fatd = σfatd⋅(25/t)0.25
t2
In cases where it is not possible to achieve
the required design lifetime for a detail in an
analysis that includes partial safety factors,
e inspection of the detail is required. How-
ever, it is still necessary to meet the re-
quirements to the nominal lifetime when
t1 calculations are made without partial safety
factors on the material properties included.
Note that this approach is only allowed in
the Danish standards.
Figure 7-9. Single sided plate tapering in
tubular tower The time until the first inspection should at
most be set equal to the calculated design
7 – Tower 181
lifetime for the case that partial safety fac-

tors are included. Once one or more inspec- As of today, the fatigue loading often gov-
tions have been carried out, subsequent in- erns the design near openings. In this case, a
spection intervals should be chosen de- suitable fatigue strength curve is to be se-
pending on the results of the previous in- lected. This should be done in agreement
spections. with Section 9.6.3 of Eurocode 3, which
specifies the use of a “Category 71 curve”
7.5.6 Stress Concentrations near for a full penetration butt with permitted
Hatches and Doors welds defects acceptance criteria satisfied.
The fatigue damage is calculated using a
The doors and hatches induce stress con-
Palmgren-Miner’s approach.
centrations near the openings. This stress
concentration is traditionally represented by
7.5.7 Stability analysis
a stress concentration factor (SCF), which
expresses the stress ratio between a sample The buckling strength of the tower usually
with and a sample without the opening. The governs the tower design as far as the shell
SCF could be determined from parametric thickness is concerned. The buckling
equations, from a finite element analysis or strength of the tower can be analysed using
by model experiments. A study by Jørgen- the approach described in Annex D of
sen (1990) found a SCF of about 1.8 for a DS449 combined with DS412, DIN 18800
door opening. However, it is recommended or other recognized standards.
that an individual SCF analysis should be
performed for each detail in question. In the following, the method suggested in
the Danish standard is presented
It is important to consider stress concentra-
tions at the tower door. They depend much Stresses owing to the axial force, σad, and
on how the door flange is carried out, i.e. owing to the bending moment, σbd, are given
whether it is straight, or whether it is curved by
to follow the curvature of the tower wall at
the top and bottom of the door. They also Nd
depend on how the flange is aligned with the σ ad =
2πRt
tower wall, and whether it is placed exter-
nally or internally with respect to the tower
Md
wall. σ bd =
πR 2 t
Further, it must be considered to what extent
the doorframe replaces the missing tower A reduction factor ε is calculated as
shell regarding cross-sectional area, moment
of inertia and center of gravity. Finally, it 0.83
εa =
might be relevant to consider local stability R
of the door region. 1 + 0.01
t
For unstrengthened circular holes, like in-
spection holes and port holes, the stress con- ε b = 0.1887 + 0.8113ε a
centration factor will normally be about 3.
ε a σ ad + ε b σ bd
ε=
For small holes like bolt holes, detail catego- σ ad + σ bd
ries can be found in structural steel codes.
182 7 – Tower
The relative slenderness ratio for global sta-

According to theory of elasticity, the critical bility is
compressive stress is
f yd
Ed λr =
σ el = æ N el ö
R çç ÷÷
3(1 − ν 2 ) è 2πRt ø
t
The relative slenderness ratio for local The core radius k of a tube is given by
buckling is
f yd πR 2 t
k=
λa = 2πRt
εσ el
For cold-formed welded towers, the equiva-
If λa ≤ 0.3, the critical compressive stress σcr lent geometrical imperfection can now be
is given by calculated as
σ cr = f yd e = 0.49(λ r − 0.2)k
If 0.3 < λa ≤ 1, the critical compressive For welded towers, it can be calculated as
stress σcr is given by
e = 0.34(λ r − 0.2)k
σ cr = (1.5 − 0.913 λ a ) f yd
However, if λr≤0.2 then e = 0.
However, if the tower height H does not
If
exceed 1.42 R R / t , then
2
σ cr = f yd e> H
1000
f yd 2 = σ cr then an additional increment
Otherwise 2
∆e = ( e − H)
1000
f yd 2 = f yd
is to be added to e.
From theory of elasticity, the Euler force for
a cantilever beam is given by Finally, the following inequality must be
fulfilled
1 2
π E d πR 3 t Nd N el M +N e
N el = 4 + ⋅ d 2 d < f yd 2
H2 2πRt N el − N d πR t
7 – Tower 183
The symbols used in the above formulas are a+b

explained below: F =Z+R=Z
a
Nd design axial force respectively.
Md design bending moment
R tower radius
The bolt stress, disregarding pretension, in
t tower shell thickness
the L flange connection is given by
H tower height
Ed design modulus of elasticity
F
ν Poisson's ratio σL =
fyd design yield stress Ast
7.5.8 Flange connections where

The tower sections and the connection to the Md design bending moment
foundation are often adjoined with L or T D diameter of tower shell
flange connections, see Figure 7-10. n number of bolts
Ast bolt stress area
Z Z Since tower flanges often have a consider-

D D able thickness, the risk of brittle fracture
b a should be considered. This can be done ac-
b a
cording to Eurocode 3, Annex C or DS412,
Annex A.
7.5.9 Corrosion protection

F Corrosion protection of the tower should be
in accordance with DS/R 454 and for off-
shore turbines according to DS/R 464 or
similar recognized standards. The most
common type of corrosion protection used
R for tubular towers is paint.
7.5.10 Tolerances and specifications

Figure 7-10. L flange connections
Whatever standard is used for the stability
Figure 7-10 shows an L flange connection analysis, the calculations assume certain
along with the deformed shape of the upper maximum imperfections. These tolerances
side of the connection. Using the model in are concerned with maximum deviations
Figure 7-10, the tension force Z in the tower from nominal longitudinal and circumferen-
shell and the bolt force F are calculated from tial dimensions and must be respected. Oth-
erwise, when a tolerance is exceeded, the
4⋅M excess must be included in the factor e in the
Z= stability calculations.
D⋅n
and Elements of large thickness, such as tower

flanges, should be checked for stratification.
184 7 – Tower
Further, the flanges must satisfy certain tol-

erances regarding straightness and circular- Dansk Ingeniørforening, The Danish Code
ity to enable correct tightening of the con- of Practice for Loads for the structural use of
nection. The stresses in the flange should be steel .DS412, 1998.
based on the smallest flange thickness, cf.
the tolerance specifications. Det Norske Veritas, Column Stabilized Units
(Semisubmersible Platforms), Classification
7.6 Access and Working Environment Notes No. 31.4, Høvik, Norway, 1987.
When designing the tower it must be veri- Eurocode 1, Basis of design and actions on
fied that proper access to the nacelle is pos- structures, Section 2-4: Wind actions on
sible. It must be assured that the access does structures.
comply with the requirements for personnel
safety and maintenance. In addition, the Eurocode 3. Design of steel structures-Part
working environment during blade inspec- 1-1: General rules and rules for buildings
tion using an inspection hatch in the tower DS/ENV1993-1-1
side must be dealt with. Turbines erected in
Denmark should comply with the require- Forskningscenter Risø, Rekommandation til
ments specified by the Danish Arbejdstil- opfyldelse af krav i teknisk grundlag, Ener-
synet. gistyrelsens regeludvalg for godkendelse af
vindmøller, Godkendelsessekretariatet, Prø-
REFERENCES vestationen for Vindmøller, Roskilde, Den-
mark.
Krohn, S., Danish Wind Turbine Manufac-
turers Association, www.windpower.org Jørgensen, E., “Notat om undersøgelse af
spændingsforhold ved luge på rørtårn,”
Arbejdstilsynet, anvisning nr. 2.2.0.1. Prøvestationen for vindmøller, Risø, Den-
mark, July 1990.
Danish Energy Agency, “Rekommandation
Mørk, K., Vindlast på svingningsfølsomme
for Teknisk Godkendelse af Vindmøller på
konstruktioner, AUC, Aalborg, Denmark.
Havet,” Bilag til Teknisk Grundlag for Ty-
pegodkendelse og Certificering af Vindmøl- Thomsen, K., The Statistical Variation of
ler i Danmark, Copenhagen, Denmark, Wind Turbine Fatigue Loads, Risø National
2000. Laboratory, Risø-R-1063(EN), 1998.
Dansk Ingeniørforening, The Danish Code Thomsen, K., P.P. Madsen, E. Jørgensen,
of Practice for the Design and Construction “Status og perspektiv for forskning i aero-
of Pile Supported Offshore Steel Structures, elasticitet – Lastgrundlag of sikkerhed,”
DS 449, 1983. Risø-R-964(DA), Forskningscenter Risø,
Roskilde, Denmark, February 1997.
Dansk Ingeniørforening, The Danish Code
of Practice for Loads for the Design of
Structures DS410, 1998.
7 – Tower 185
186 7 – Tower
8. Foundations weight is achieved by placing a heavy bal-

last such as crushed olivin inside the cavities
of the steel structure. The tripod foundation
8.1 Introduction is a steel frame structure with three legs. The
Land-based wind turbines are usually sup- wind turbine tower is mounted on top of the
ported by either a slab foundation or a pile tripod, while each leg is supported by either
foundation. The soil conditions at the loca- a driven pile or a suction bucket for transfer
tion usually govern whether a slab founda- of loads to the supporting soils.
tion or a pile foundation is chosen. A slab
foundation is usually preferred when the top The choice of foundation type is much de-
soil is strong enough to support the loads pendent on the soil conditions prevailing at
from the wind turbine, while a pile- the planned site of a wind turbine. Once a
supported foundation is attractive when the foundation concept has been selected and a
top soil is of a softer quality and the loads foundation design is to be carried out, the
need to be transferred to larger depths where following geotechnical issues need to be
stronger soils are present to absorb the loads. addressed:
For assessment of whether the top soil is • bearing capacity, i.e. geotechnical sta-
strong enough to carry the foundation loads, bility, e.g. against sliding and over-
it is important to consider how far below the turning
foundation base the water table is located. • degradation of soil strength in cyclic
loading
For offshore wind turbines, the foundation is • consolidation settlements
a more comprehensive structure in that it • differential settlements
includes a separate structure to transfer loads • scour and erosion
from the bottom of the wind turbine tower
through the water to the supporting soils. In
addition to the loads from the wind turbine, 8.2 Soil investigations
such a foundation structure will experience
loads from current, waves and ice owing to 8.2.1 General
its placement in a marine environment. The soil investigations should provide all
Three basically different foundation struc- necessary soil data for a detailed design of a
ture concepts exist for offshore wind tur- specific foundation structure at a specific
bines, viz. location. The soil investigations may be di-
• monopile vided into the following parts
• gravity base • geological studies
• tripod • geophysical surveys
The monopile is in principle a vertical pipe, • geotechnical investigations
driven or bored into the soil like any other which are briefly dealt with in the following.
pile, and onto which the wind turbine tower
is then mounted. The gravity base founda- A geological study should be based on in-
tion rests on the sea floor or on an excavated formation about the geological history of the
bottom by means of its own weight and is area where the wind turbine is to be in-
usually constructed from reinforced con- stalled. The purpose of the this study is to
crete. The wind turbine tower is mounted on establish a basis for selection of methods
top of this concrete structure. The gravity and extent of the site investigation.
base foundation can also be constructed
from steel in which case the necessary A geophysical survey can be used to extend
8 – Foundations 187
the localized information from single bor- Results from such in-situ tests can be used to
ings and in-situ testing in order to get an interpret parameters such as the undrained
understanding of the soil stratification shear strength of clay. The extent to which
within a considered area, and – for offshore the various types of in-situ tests and labora-
locations – of the seabed topography within tory tests are required depends much on the
that area. Such a survey can provide guide- foundation type in question, for example,
lines for selection of a suitable foundation whether it is a piled foundation or a gravity-
site within the area, if not already decided. based foundation.
Geophysical surveys are carried out by
means of seismic methods. The recommendations to the Danish techni-
cal criteria for type approval of wind tur-
A geotechnical investigation consists of bines (Danish Energy Agency, 1998) specify
• soil sampling for laboratory testing that a geotechnical report from the geologi-
• in-situ testing of soil cal and geotechnical surveys shall be pre-
The soil investigations should be tailored to pared. This geotechnical report should con-
the geotechnical design methods used. The tain sufficient information about the site and
field and laboratory investigations should its soils, e.g. in terms of soil strength and
establish the detailed soil stratigraphy across deformation properties, to allow for design
the site providing the following types of of the foundation with respect to
geotechnical data for all important soil lay- • bearing capacity
ers • stability against sliding
• data for classification and description of • settlements
the soil, such as • foundation stiffness
• unit weight of sample • need for and possibility of drainage
• unit weight of solid particle • static and dynamic coefficients of com-
• water content pressibility
• liquid and plastic limits • sensitivity to dynamic loading
• grain size distribution The geotechnical report is further required to
• parameters required for a detailed and contain identification of soil type at founda-
complete foundation design, such as tion level, classification of environment, and
• permeability tests estimation of highest possible water table.
• consolidation tests For further details, reference is made to
• static tests for determination of shear Danish Energy Agency (1998).
strength parameters such as friction an-
gle φ for sand and undrained shear 8.2.2 Recommendations for gravity base
strength cu for clay (triaxial tests, direct foundations
simple shear tests) For a gravity base foundation, an extensive
• cyclic tests for determination of strength investigation of the shallow soil deposits
and stiffness parameters (triaxial tests, should be performed. This investigation
direct simple shear tests, resonant col- should cover the soil deposits to a depth,
umn tests) which is deeper than the depth of any possi-
ble critical shear surface. Further, all soil
Sampling can be carried out with and with- layers influenced by the structure from a
out drilling. The cone penetrometer test settlement point of view should be thor-
(CPT) and various vane tests form the most oughly investigated. This also holds for all
commonly used in-situ testing methods. soil layers contributing to the foundation
188 8 – Foundations
stiffness. The foundation stiffness is of im- For axial pile analysis, at least one down-
portance for the design of the structure sup- the-hole CPT boring giving a continuous
ported by the foundation. The depth to be CPT profile should be carried out, and one
covered by the thorough investigation nearby boring with sampling should be car-
should at least equal the largest base dimen- ried out. for the axial pile capacity analysis.
sion of the structure. The minimum depth should be the antici-
pated penetration of the pile plus a zone of
The extent of shallow borings with sampling influence sufficient for evaluation of the risk
should be determined based on type and size of punch-through failure. The sampling in-
of structure as well as on general knowledge terval should be determined from the CPT
about the soil conditions in the area consid- results, but is recommended not to exceed 3
ered for installation. Emphasis should be m.
given to the upper layers and potentially
weaker layers further down. The sampling For offshore installations, a number of sea-
interval is recommended not to exceed 1.0- bed samples (gravity cores or equivalent)
1.5 m. A number of seabed samples (gravity evenly distributed over the area considered
cores or the equivalent) evenly distributed for installation of the foundation should be
over the area should be taken for evaluation taken for evaluation of the scour potential.
of the scour potential.
Special attention should be paid when po-
Shallow CPTs distributed across the instal- tential end bearing layers or other dense
lation area should be carried out in addition layers are found. Here additional CPT and
to the borings. The number of CPTs depends sampling should be carried out in order to
on the soil conditions and on the type and determine the thickness and lateral extension
size of structure. If the soil conditions are of such layers within the area considered for
very irregular across the foundation site, the the foundation.
number of CPTs will have to be increased.
The shallow CPTs should provide coninuous
graphs from the soil surface to the maximum 8.3 Gravity-based foundations
depth of interest.
8.3.1 General
Special tests such as plate loading tests, Requirements to foundation stability are
pressuremeter tests and shear wave velocity often the most decisive factor for determi-
measurements should be considered where nation of foundation area, foundation em-
relevant. bedment and necessary weight for a struc-
ture with a gravity type foundation. It is
8.2.3 Recommendations for pile founda- therefore essential in an optimal design pro-
tions cess to give high emphasis to foundation
stability calculations.
For lateral pile analysis, shallow cone pene-
tration tests should be carried out from the
The foundation stability is most commonly
surface to 20-30m depth. In addition, shal-
solved by by limiting equilibrium methods,
low borings with sampling should be con-
ensuring equilibrium between driving and
sidered for better determination of charac-
resisting forces. Using limiting equilibrium
teristics of the individual layers identified by
methods, several trial failure surfaces will
the cone penetration tests. The sampling
have to be analysed in order to find the one,
interval is recommended not to exceed 1.0-
1.5m.
which is the most critical with respect to the foundation. Reference is made to Figure
stability. 8-1, and the eccentricity is calculated as
However, as foundations of wind turbines Md

usually have relatively small areas, bearing e=
Vd
capacity formulas for idealised conditions
will normally suffice and be acceptable for
design. Bearing capacity formulas are given where Md denotes the resulting design
in the following. overturning moment about the foundation-
soil interface.
8.3.2 Bearing capacity formulas
Correction for torque
Forces When a torque MZ is applied to the founda-
All forces acting on the foundation, includ- tion in addition to the forces H and V, the
ing forces transferred from the wind turbine, interaction between the torque and these
are transferred to the foundation base and forces can be accounted for by replacing H
combined into resultant forces H and V in and MZ with an equivalent horizontal force
the horizontal and vertical direction, respec- H’. The bearing capacity of the foundation is
tively, at the foundation-soil interface. then to be evaluated for the force set (H’,V)
instead of the force set (H,V). According to a
method by Hansen (1978), the equivalent
horizontal force can be calculated as
2
2⋅Mz æ 2⋅Mz ö
LC =
H' + H +ç ÷
l eff ç l ÷
V è eff ø
f [kN/m2]
H in which leff is the length of the effective

area as determined in the following.
e [m]
Effective foundation area
rupture 2 rupture 1 For use in bearing capacity analysis an ef-
fective foundation area Aeff is needed. The
effective foundation area is constructed such
Figure 8-1 Loading under idealised condi- that its geometrical centre coincides with the
tions load centre, and such that it follows as
closely as possible the nearest contour of the
In the following, it is assumed that H and V true area of the foundation base. For a quad-
are design forces, i.e., they are characteristic ratic area of width b, the effective area Aeff
forces that have been multiplied by their can be defined as
relevant partial load factor γf. This is indi-
cated by index d in the bearing capacity Aeff = beff ⋅ l eff
formulas, hence Hd and Vd. The load centre,
denoted LC, is the point where the resultant in which the effective dimensions beff and leff
of H and V intersects the foundation-soil depend on which of two idealised loading
interface, and implies an eccentricity e of the scenarios leads to the most critical bearing
vertical force V relative to the centre line of capacity for the actual foundation.
e Aeff Reference is made to Figure 8-2. The effec-

tive area representation that leads to the
poorest or most critical result for the bearing
capacity of the foundation is the effective
area representation to be chosen.
LC1
For a circular foundation area with radius R,
leff
an elliptical effective foundation area Aeff
can be defined as
é e ù
Aeff = 2ê R 2 arccos( ) − e R 2 − e2 ú
beff
ë R û
with major axes

Aeff
be = 2(R − e )
e
and
2
æ b ö
LC2 le = 2 R 1 − ç1 − ÷
leff
è 2R ø
beff Aeff
Figure 8-2 Quadratic footing with two ap-

proaches to how to make up the effective R
foundation area
LC
leff
le
Scenario 1 corresponds to load eccentricity
with respect to one of the two symmetry
axes of the foundation. By this scenario, the
following effective dimensions are used:
beff
beff = b − 2 ⋅ e , l eff = b
be
Scenario 2 corresponds to load eccentricity
with respect to both symmetry axes of the Figure 8-3 Circular and octangular footings
foundation. By this scenario, the following with effective foundation area marked out
effective dimensions are used:
The effective foundation area Aeff can now
be represented by a rectangle with the fol-
beff = leff = b − e 2 lowing dimensions
le l iγ iq ic inclination factors, dimensionless

leff = Aeff and beff = eff be Reference is made to DS415.
be le
In principle, the quoted formulas apply to
For an area shaped as a double symmetrical foundations, which are not embedded. How-
polygon (octagonal or more), the above ever, the formulas may also be applied to
formulas for the circular foundation area can embedded foundations, for which they will
be used provided that a radius equal to the lead to results, which will be on the conser-
radius of the inscribed circle of the polygon vative side. Alternatively, depth effects as-
is used for the calculations. sociated with embedded foundations can be
calculated according to formulas given in
Bearing capacity DNV (1992).
For fully drained conditions and failure ac-
cording to Rupture 1 as indicated in Figure The calculations are to be based on design
8-1, the following general formula can be shear strength parameters:
applied for the bearing capacity of a foun-
dation with a horizontal base, resting on the c tan(φ )
soil surface: c ud = and φ d = arctan( )
γc γφ
1
qd = γ 'beff N γ s γ iγ + p 0'N q s q i q +c d N c s c i c The material factors γf an γφ must be those
2
associated with the actual design code and
the type of analysis, i.e. whether drained or
For undrained conditions, which imply φ=0,
undrained conditions apply.
the following formula for the bearing capac-
ity applies:
The dimensionless factors N, s and i can be
determined by means of formulas given in
q d =c ud ⋅N c0 ⋅ s c0 ⋅ ic0 + p 0 the following.
The symbols used have the following expla- Drained conditions:

nations
Bearing capacity factors N:
qd design bearing capacity [kN/m2]
γ' effective (submerged ) unit weight 1 + sin φ d
of soil [kN/m3] N q = e π tan φd ⋅
1 − sin φ d
p'0 effective overburden pressure at the
level of the foundation-soil inter-
face [kN/m2] N c = ( N q − 1) ⋅ cot φ d
cd design cohesion or design
undrained shear strength assessed 1 3
on the basis of the actual shear N γ = ⋅ (( N q − 1) ⋅ cos φ d ) 2

4
strength profile, load configuration
and estimated depth of potential
According to Hansen (1970), Nγ may alter-
failure surface [kN/m2]
natively be calculated according to
Nγ Nq Nc bearing capacity factors, dimen-
sionless
3
sγ sq sc shape factors, dimensionless N γ = ⋅ ( N q − 1) ⋅ tan φ d
2
Shape factors s: iγ = i q
2
beff
s γ = 1 − 0 .4 ⋅ H
l eff ic0 = 0.5 + 0.5 ⋅ 1 +
Aeff ⋅ cud
beff
s q = s c = 1 + 0.2 ⋅
l eff
The bearing capacity is to be taken as the
smallest of the values for qd resulting from
Inclination factors i: the calculations for Rupture 1 and Rupture
2.
2
æ Hd ö
i q = ic = ç1 − ÷ Sliding resistance of soil
ç V + A ⋅ c ⋅ cot φ ÷
è d eff d d ø Foundations subjected to horizontal loading
iγ = i q
2
must also be investigated for sufficient slid-
ing resistance. The following criterion ap-
plies in the case of drained conditions:
Undrained conditions, φ=0:
H < A eff ⋅c + V ⋅ tan φ
N c0 = π + 2
For undrained conditions in clay, φ=0, the

s c0 = s c following criterion applies:
H H < A eff ⋅cud

ic0 = 0.5 + 0.5 ⋅ 1 −
Aeff ⋅ cud and it must in addition be verified that
H
Extremely eccentric loading < 0 .4
In the case of extremely eccentric loading, V
i.e., an eccentricity in excess of 0.3 times the
foundation width, e>0.3b, an additional
bearing capacity calculation needs to be car- 8.4 Pile-supported foundations
ried out, corresponding to the possibility of a 8.4.1 General
failure according to Rupture 2 in Figure 8-1.
This failure mode involves failure of the soil A pile foundation consists of one or more
also under the unloaded part of the founda- piles that transfer loads from a superstruc-
tion area, i.e., under the heel of the founda- ture, such as a wind turbine tower or a dis-
tion. For this failure mode, the following tinctive structure supporting the tower, to
formula for the bearing capacity applies the supporting soils. The loads applied to a
pile at its head are transferred down the pile
q d = γ 'beff N γ s γ iγ +c d N c s c i c (1.05 + tan 3 φ ) and absorbed by the soil through axial and
lateral pile resistance. The axial and lateral
with inclination factors pile resistance arises from soil resistance
mobilised against the pile when the pile,
H
i q = ic = 1 + subjected to its loading, is displaced relative
V + Aeff ⋅ c ⋅ cot φ to the soil.
For design of piles, it is common to disre- skin friction against the pile surface. The
gard a possible interaction between the axial axial, lateral and torsional pile displace-
pile resistance and the lateral pile resistance ments and corresponding soil reactions for a
locally at any point along the pile and treat single pile subjected to external loading at
these two resistances as independent of each its head are illustrated in Figure 8-4.
other. The argument for this is that the soil
near the soil surface principally determines Note that for design of piles, it is important
the lateral resistance without contributing to consider effects of the installation proce-
much to the axial resistance, while the soil dure. For example, the stress history during
further down along the pile toward the pile pile driving contributes significantly to fa-
tip principally determines the axial resis- tigue loading and needs to be considered for
tance without contributing much to the lat- design against fatigue failure in the pile
eral capacity. The axial and lateral resistance wall.
models presented in the following conform
to this assumption of independence between 8.4.2 Pile groups
local axial resistance and local lateral resis-
For foundations consisting of pile groups,
tance.
i.e. clusters of two or more piles spaced
closely together, pile group effects need to
be considered when the axial and lateral
resistance of the piles is to be evaluated.
There are two types of group effects:
• The total capacity of the pile group is
less than the sum of the capacities of the
individual piles in the group, because of
overlap between plastified soil zones
around the individual piles. A lower
limit for the axial pile group capacity is
the axial capacity of the envelope “pier”
that encloses all the piles in the group
and the soil between them.
• Larger pile displacements for a given
Figure 8-4 Single loaded pile with dis- load result for an individual pile when it
placements and reactions, from Reese et al. is located in a pile group than when it is
(1996), an only pile, because its supporting soils
will have displacements caused by loads
Note, however, that on a global level the transferred to the soil from adjacent
lateral pile capacity and the axial pile ca- piles in the group. This type of group
pacity, resulting from the local resistances effect is also known as pile-soil-pile
integrated along the length of the pile, may interaction. For practical purposes, such
interact, because second-order effects may pile-soil-pile interaction can often be
cause the axial loading to influence the lat- reasonably well represented by means
eral behaviour of the pile. When the pile is of Mindlin’s point force solutions for an
subjected to torque, a torsional resistance elastic halfspace.
will be set up and come in addition to the The knowledge of the behaviour of a pile
axial and lateral resistances. Note that the group relative to the behaviour of individual
local axial and torsional resistances are in- piles in the same group is limited, and con-
terdependent, since they both arise from the servative assumptions are therefore recom-
mended for the prediction of pile group re- ì 1

sistance. Note, when dealing with pile ï for c u / p 0 '≤ 1.0
ï2 cu p0 '
groups, that in addition to the above pile α =í
1
group effects there is also an effect of the ï for c u / p 0 '> 1.0
presence of the superstructure and the way ïî 24 cu p 0 '
in which it is connected to the pile heads.
The interconnection of the piles through
where cu is the undrained shear strength of
their attachment to the common superstruc-
the soil and p0’ is the effective overburden
ture influences the distribution of the loads
pressure at the point in question.
between the piles in the group. The piles can
often be assumed to be fixed to the super-
(2) effective stress methods, e.g. the β
structure, and the superstructure can often be
method, which yields
assumed as a rigid structure or a rigid cap
when the responses in the piles are to be
determined. f Si = βp 0 '
8.4.3 Axial pile resistance in which β values in the range 0.1-0.25 are
suggested for pile lengths exceeding 15 m.
Axial pile resistance is composed of two
parts
(3) semi-empirical λ method, by which the
• accumulated skin resistance soil deposit is taken as one single layer, for
• tip resistance which the average skin friction is calculated
as
For a pile in a stratified soil deposit of N soil
layers, the pile resistance R can be expressed f S = λ ( p0m '
+2c um )
as
N where p0m’ is the average effective overbur-

R = R S + RT = å f Si ASi + q T AT den pressure between the pile head and the
i =1
pile tip, cum is the average undrained shear
strength along the pile shaft, and λ is the
where fSi is the average unit skin friction dimensionless coefficient, which depends on
along the pile shaft in layer i, ASi is the shaft the pile length as shown in Figure 8-5.
area of the pile in layer i, qT is the unit end Hence, by this method, the total shaft resis-
resistance, and AT is the gross tip area of the tance becomes RS=fSAS, where AS is the pile
pile. shaft area.
Clay. For piles in mainly cohesive soils, the For long flexible piles, failure between pile
average unit skin friction fS may be calcu- and soil may occur close to the seabed even
lated according to before the soil resistance near the pile tip has
been mobilized at all. This is a result of the
(1) total stress methods, e.g. the α method, flexibility of the pile and the associated dif-
which yields ferences in relative pile-soil displacement
along the length of the pile. This is a length
f Si = αc u effect, which for a strain-softening soil will
imply that the static capacity of the pile will
in which be less than that of a rigid pile.
qp=Nqp0’≤ql
0,5
in which the bearing factor Nq can be taken
0,45
from Table 8-1 and ql is a limiting tip resis-
0,4
tance, see Table 8-1 for guidance.
0,35
0,3
The unit tip resistance of piles in cohesive
λ 0,25
0,2
soils can be calculated as
0,15
0,1 qp=Nccu
0,05
0 where Nc=9 and cu is the undrained shear

0 10 20 30 40 50 60 70
strength of the soil at the pile tip.
Pile length (m)
Table 8-1 Design parameters for axial
resistance of driven piles in cohesionless
silicious soil (extracted from API (1987))
Figure 8-5 Coefficient λ vs. pile length
For deformation and stress analysis of an

axially loaded flexible pile, the pile can be
modelled as a number of consecutive col-
umn elements supported by nonlinear
springs applied at the nodal points between
the elements. The nonlinear springs are de-
noted t-z curves and represents the axial
load-displacement relationship between the
pile and the soil. The stress t is the inte-
grated axial skin friction over and z is the
relative axial pile-soil displacement neces-
sary to mobilize this skin friction.
Sand. For piles in mainly cohesionless soils

(sand), the average unit skin friction may be
calculated according to
fS=Kp0’tanδ≤fl t-z curves

The t-z curves can be generated according to
in which K=0.8 for open-ended piles and a method by Kraft et al. (1981). By this
K=1.0 for closed-ended piles, p0’ is the ef- method, a nonlinear relation applies between
fective overburden pressure, δ is the angle of the origin and the point where the maximum
soil friction on the pile wall as given in Ta- skin resistance tmax is reached,
ble 8-1, and fl is a limiting unit skin friction,
t
see Table 8-1 for guidance. z IF − r f
R t
z =t ln max
for 0≤t≤tmax
Tip resistance G0 t
The unit tip resistance of plugged piles in 1 − rf
t max
cohesionless soils can be calculated as
in which R denotes the radius of the pile, G0 G 0 = 600c u − 170c u OCR − 1

is the initial shear modulus of the soil, ZIF is
a dimensionless zone of influence, defined
where su is the undrained shear strength of
as the radius of the zone of influence around
the pile divided by R, and rf is a curve fitting the clay, and OCR is the overconsolidation
factor. For displacements z beyond the dis- ratio. For sands, the initial shear modulus of
the soil to be used for generation of t-z
placement where tmax is reached, the skin
curves is to be taken as
resistance t decreases in linear manner with
z until a residual skin resistance tres is
reached. For further displacements beyond m σ aσ v
G0 = with m=1000tanφ
this point, the skin resistance t stays con- 2(1 + ν )
stant. An example of t-z curves generated
according to this method is given in Figure
8-6. The maximum skin resistance can be in which σa=100 kPa is the atmospheric
calculated according to one of the methods pressure and σv is the vertical effective
for prediction of unit skin friction given stress, ν is the Poisson’s ratio of the soil, and
above. φ is the friction angle of the soil.
8.4.4 Laterally loaded piles

The most common method for analysis of
laterally loaded piles is based on the use of
so-called p-y curves. The p-y curves gives
the relation between the integral value p of
the mobilized resistance from the surround-
ing soil when the pile deflects a distance y
laterally. The pile is modelled as a number
of consecutive beam-column elements, sup-
ported by nonlinear springs applied at the
nodal points between the elements. The
nonlinear support springs are characterized
by one p-y curve at each nodal point, see
Figure 8-7.
The solution of pile displacements and pile

Figure 8-6 Example of t-z curves generated stresses in any point along the pile for any
by model applied load at the pile head results as the
solution to the differential equation of the
For clays, the initial shear modulus of the pile
soil to be used for generation of t-z curves
can be taken as d4y d2y
EI + Q A
− p( y) + q = 0
dx 4 dx 2
G0 = 2600 cu
However, Eide and Andersen (1984) suggest with

a somewhat softer value according to the
formula d3y dy d2y
EI + Q A
= Q L
and EI =M
dx 3 dx dx 2
where x denotes the position along the pile the t-z curve approach presented above is
axis, y is the lateral displacement of the pile, included. Some of the available programs
EI is the flexural rigidity of the pile, QA is can be used to analyze not only single piles
the axial force in the pile, QL is the lateral but also pile groups, including possible pile-
force in the pile, p(y) is the lateral soil reac- soil-pile interaction and allowing for proper
tion, q is a distributed load along the pile, representation of a superstructure attached at
and M is the bending moment in the pile, all the pile heads, either as a rigid cap or as a
at the position x. Reference is made to Fig- structure of finite stiffness.
ure 8-8.
Figure 8-8 Beam column model with de-

tails, from Reese et al. (1996),
Several methods are available for represen-

tation of the p-y curves that are essential in
solving the differential equation for a later-
Figure 8-7 p-y curves applied at nodal ally loaded pile. For construction of p-y
points in beam column representation of curves, the type of soil, the type of loading,
pile, from Reese et al. (1996) the remoulding due to pile installation and
the effect of scour should be considered. The
A finite difference method usually forms the most commonly applied procedures for con-
most feasible approach to achieve the struction of p-y curves are those given by
sought-after solution of the differential DNV (1992) and API (1987). For piles in
equation of the pile. A number of commer- clay, the procedure according to DNV is
cial computer programs are available for this presented below. For sand, the procedures
purpose. These programs usually provide according to DNV and API are identical and
full solutions of pile stresses and displace- are presented below.
ments for a combination of axial force, lat-
eral force and bending moment at the pile The lateral resistance per unit length of pile
head, i.e., also the gradual transfer of axial for a lateral pile deflection y is denoted p.
load to the soil along the pile according to The static ultimate lateral resistance per unit
length is denoted pu. This is the maximum Here, yc=2.5εcD, in which D is the pile di-
value that p can take on when the pile is ameter and εc is the strain which occurs at
deflected laterally. one-half the maximum stress in laboratory
undrained compression tests of undisturbed
Clay. For piles in cohesive soils, the static soil samples. For further details, reference is
ultimate lateral resistance is recommended made to DNV (1992).
to be calculated as
Sand. For piles in cohesionless soils, the
ì(3c + γ 'X ) D + Jc u X for 0 < X ≤ X R static ultimate lateral resistance is recom-
pu = í u mended to be calculated as
î9c u D for X > X R
ì(C X + C 2 D )γ 'X for 0 < X ≤ X R
where X is the depth below soil surface and pu = í 1
XR is a transition depth, below which the îC 3 Dγ 'X for X > X R
value of (3cu+γ’X)D+JcuX exceeds 9cuD.
Further, D is the pile diameter, cu is the where the coefficients C1, C2 and C3 depend
undrained shear strength of the soil, γ’ is the on the friction angle φ as shown in Figure 8-
effective unit weight of soil, and J is a di- 9, and where X is the depth below soil sur-
mensionless empirical constant whose value face and XR is a transition depth, below
is in the range 0.25-0.50 with 0.50 recom- which the value of (C1X+C2D)γ’X exceeds
mended for soft normally consolidated clay. C3Dγ’X. Further, D is the pile diameter, and
γ’ is the submerged unit weight of soil.
For static loading, the p-y curve can be gen-
erated according to The p-y curve can be generated according to
ì pu y 1 / 3 kX
for y ≤ 8y c p = Ap u tanh( y)
ï ( )
p = í 2 yc Ap u
ïî p u for y > 8y c
For cyclic loading and X>XR, the p-y curve

can be generated according to
ì pu y 1 / 3
ï ( ) for y ≤ 3y c
p = í 2 yc
ïî0.72 p u for y > 3y c
For cyclic loading and X≤XR, the p-y curve

can be generated according to
ì pu y 1 / 3
ï ( ) for y ≤ 3y c
ï 2 yc
ï X y − 3yc
p = í0.72 p u (1 − (1 − ) ) for 3y c < y ≤ 15 y c
ï X R
12 y c
ï X
ï0.72 p u X for y > 15 y c Figure 8-9 Coefficients as functions of fric-
î R
tion angle, from DNV Class. Notes 30.4
in which k is the initial modulus of subgrade against the embedded part of the cap. This
reaction and depends on the friction angle φ effect is similar to the soil resistance, repre-
as given in Figure 8-10, and A is a factor to sented by p-y curves, on piles under lateral
account for static or cyclic loading condi- loading. When assessing such an effect of
tions as follows soil in front of the pile cap, it is important
also to assess the possibility that this soil
ì0.9 for cyclic loading may be removed due to erosion or other
ï
A=í H natural actions.
ïî(3 − 0.8 D ) ≥ 0.9 for static loading
For further details, reference is made to

DNV (1992).
Figure 8-11 Distribution of soil resistance

on the front side of pile cap, from Reese et
all (1996)
The pile cap as dealt with here essentially is

considered and modelled as a rigid structure.
Sometimes, however, the pile heads are not
interconnected by such a rigid superstruc-
ture, but are connected to a superstructure
with some finite stiffness, such as a frame or
Figure 8-10 Initial modulus of subgrade a lattice tower. This will influence the distri-
reaction k as function of friction angle φ, bution of forces between the piles relative to
from DNV Class. Notes 30.4 that of the rigid-cap structure and will in
turn have an impact on the overall resulting
8.4.5 Soil resistance for embedded pile foundation stiffness.
caps
Some piled foundations include an embed-
ded pile cap as indicated in Figure 8-11. 8.5 Foundation stiffness
Whereas the rotational capacity and stiffness
of a pile cap is primarily governed by the The overall foundation stiffness is depend-
axial capacity and stiffness of the piles, it is ent on the strength and stiffness of the soil
important to consider the effect on the lat- as well as on the structural foundation ele-
eral capacity and stiffness of the soil acting ments. The foundation stiffness needs to be
determined as a basis for predicting the dy-
namic structural response to wind, wave and • Earthquakes: Large strains up to 10−2 to
earthquake loading. The foundation stiffness 10−1
is in general frequency dependent. This is • Rotating machines: Small strains usu-
particularly important when predicting dy- ally less than 10−5
namic response to earthquake. • Wind and ocean waves: Moderate
strains up to 10−2, typically 10−3
The soil that supports a foundation structure
usually has finite stiffness. It can therefore
usually not be justified to model the soil as a
rigid mass. In other words, the foundation
structure cannot be assumed to have a fixed
support. In any analysis of a foundation
structure and of the wind turbine structure
that it supports, it is therefore important to
model the actual boundary conditions
formed by the supporting soils properly.
To represent the finite stiffness of the sup-

porting soils in such analyses, it is common Figure 8-12 Shear modulus and damping
to model a set of so-called foundation ratio vs. strain level
springs to be applied in one or more mod-
elled support points on the structure to be The following empirical relation can be used
analysed. The set of foundation springs may to establish the initial shear modulus G0 in a
include the following springs associated soil:
with various modes of motion
• Vertical spring stiffness (3 − e) 2
G0 = A σ0'
(OCR) k
• Horizontal spring stiffness 1+ e
• Rocking spring stiffness
• Torsional spring stiffness in which σ0’ andG0 are both to be given in
units of kPa and A=3000±1000 depending
Soil behaves in nonlinear manner. Founda- on material (size, angularity of grains etc.).
tion springs are therefore nonlinear. It is OCR denotes the overconsolidation ratio for
common to apply linear spring stiffnesses, in clays and has to be set equal to 1.0 for sands,
which case the stiffness values are chosen the exponent k is a function of the plasticity
dependent on the strain level that the soil index IP as given in Table 8-2, and e is the
will experience for the load case under con- void ratio.
sideration. It is common to deal with the
shear modulus of the soil, G. This equivalent Table 8-2 Exponent k
shear modulus relates to the initial shear Plasticity k
modulus G0 as a function of the shear strain index IP
γ as indicated in Figure 8-12. Figure 8-12 0 0
also gives damping ratios ξ for the soil as a 20 0.18
40 0.30
function of the shear strain γ.
60 0.41
The following shear strain levels can be ex- 80 0.48
pected for the three most important sources >100 0.50
of dynamic loading of soils
The confining effective stress σ0’ is defined 5. calculate G0 from one of the formulas
as the average of the three principal effec- given above, chosen dependent on soil
tive stresses type
1 6. calculate G as the product of G0 and
σ 0 '= (σ 1 ' +σ 2 '
+σ 3 ') G/G0
3
Example: Wind loading on a foundation in
clay with undrained shear strength su=200
Note that in geotechnics, the effective stress
kPa:
is defined as the total stress minus the pore
1. The source of loading is wind.
pressure.
2. The expected strain level is typically
Alternatively, the following relation for sand γ=10−3.
can be applied 3. The damping ratio is about ξ=0.10-0.15.
4. The shear modulus ratio is G/G0=0.35.
5. The initial shear modulus is G0=2600su
G 0 = 1000 K σ 0 ' =520 MPa.
6. The shear modulus becomes G
in which σ0’ andG0 are both to be given in =0.35⋅520=180 MPa.
units of kPa, and K takes on values accord-
ing to Table 8-3. For clay, one can use the Table 8-4 provides guidance for assessment
following relation as an alternative to the of Poisson’s ratio.
above formula
Table 8-4 Poisson’s ratioν
G0=2600su Soil type ν
Dense sands 0.25-0.30
in which su is the undrained shear strength of Loose sands, stiff clays 0.35-0.45
the soil. Saturated clays ≈0.50
Table 8-3 Factor K Once the equivalent shear modulus G has
Soil type K been established from G0 and Figure 8-12,
Loose sand 8 and the Poisson’s ratioν has been assessed,
Dense sand 12 the foundation stiffnesses can be derived.
Very dense sand 16 The following four stiffnesses are consid-
Very dense sand and gravel 30-40 ered
• vertical stiffness, KV=V/δV, which ex-
The steps in establishing the shear modulus presses the ratio between the vertical
and the damping ratio can be listed as fol- force V and the vertical displacement δV
lows • horizontal stiffness, KH=H/δH, which
1. determine source of dynamic loading expresses the ratio between the hori-
(earthquake, wind, waves, machine vi- zontal force H and the horizontal dis-
brations)
placement δH
2. find expected strain level γ for loading
• rotational stiffness, KR=M/θ, which ex-
from list below Figure 8-12
presses the ratio between the overturn-
3. read off damping ratio ξ from Figure 8-
ing moment M and the rotation angle θ
12
in rocking
4. read off shear modulus ratio G/G0 from
Figure 8-12
• torsional stiffness, KT=MT/θT, which

expresses the ratio between the torque The slenderness ratio for a pile is defined as
MT and the twist θT L/D where L is the length and D is the di-
ameter of the pile. For L/D>10, most piles
Formulas for spring stiffnesses are given in are flexible, i.e., the active length of the pile
Tables 8-5 through 8-7 for various types of is less than L, and the pile head response is
foundations which cover independent of the length of the pile. Spring
• Circular footing on stratum over bed- stiffnesses at the pile head of flexible piles
rock are given in Table 8-7 for three idealised soil
• Circular footing on stratum over half- profiles and the following modes of motion:
space • horizontal
• Circular footing embedded in stratum • rocking
over bedrock • coupled horizontal-rocking
• Piled foundations Reference is made to the Young’s modulus
For the footings it is assumed that they are of the soil E, which relates to the shear
rigid relative to the soil and always in full modulus G through
contact with the soil. The piled foundations
are assumed to be flexible as dealt with be- E = 2G (1 + ν )
low.
ES is the value of E at a depth z equal to the
The spring stiffnesses given in Tables 8-5 pile diameter D, and EP is the Young’s
through 8-7 are all static stiffnesses, i.e., modulus of the pile material.
they are stiffnesses for frequencies ap-
proaching zero. The dynamic stiffnesses The formulas given for foundation stiff-
may deviate from the static stiffnesses in nesses in the tables in this section can be
particular in case of high-frequent vibra- used to calculate spring stiffnesses to sup-
tions. However, for wind and wave loading port the tower in aeroelastic wind turbine
of wind turbine foundations, on land as well analyses. As a rule of thumb, the eigenfre-
as offshore, the induced vibrations will be of quency of the tower will be reduced by be-
such a nature that the static stiffnesses will tween 0 and 5%, when the assumption of a
be representative for the dynamic stiffnesses rigid foundation (fixed-ended tower) is re-
that are required in structural analyses. For placed by a realistic finite foundation stiff-
earthquake loading, however, frequency- ness.
dependent reductions of the static stiffnesses
to get appropriate dynamic stiffness values
may be necessary and should be considered.
Table 8-5 Circular footing on stratum over bedrock or on stratum over halfspace
On stratum over bedrock On stratum over halfspace
Mode of motion Foundation stiffness Foundation stiffness

Vertical 4GR R R
KV = (1 + 1.28 ) 1 + 1.28
1 −ν H 4G1 R H ; 1≤H/R≤5
KV =
1 −ν 1 R G1
1 + 1.28
H G2
Horizontal 4GR R R
KH = (1 + 1.28 ) 1+
1 −ν H 8G R
KH = 1 2 H ; 1≤H/R≤4
1 −ν 1 R G1
1+
2H G2
Rocking 8GR 3 R R
KR = (1 + ) 1+
3(1 − ν ) 6H 8G1 R 3 6 H ; 0.75≤ H/R ≤2
KR =
3(1 − ν 1 ) R G1
1+
6H G 2
Torsion 16GR 3 Not given
KT =
3
Table 8-6 Circular footing embedded in stratum over bedrock
Mode of mo- Foundation stiffness

tion
Vertical 4GR R D D D/H
KV = (1 + 1.28 )(1 + )(1 + (0.85 − 0.28 ) )
1 −ν H 2R R 1− D / H
Horizontal 8GR R 2D 5 D
KH = (1 + )(1 + )(1 + )
1 −ν 2H 3 R 4H
Rocking 8GR 3 R D D
KR = (1 + )(1 + 2 )(1 + 0.7 )
3(1 − ν ) 6H R H
Torsion 16GR 3 8D
KT = (1 + )
3 3R
Table 8-7 Flexible pile

Standardised springs at pile head
Soil profile K K K H ,R
Horizontal H Rocking 3 R Coupled 2
DE S D ES D ES
Linear increase æE ö
0.35
æE ö
0.80
æE ö
0.60
with depth 0.6çç P ÷

÷ 0.14çç P ÷
÷ − 0.17çç P ÷
÷
E=ESz/D è ES ø è ES ø è ES ø
Increase with æE ö
0.28
æE ö
0.77
æE ö
0.53
square-root of 0.8çç P ÷
÷ 0.15çç P ÷
÷ − 0.24çç P ÷
÷
è ES ø è ES ø è ES ø
depth E=ES z / D
Homogeneous æE ö
0.21
æE ö
0.75
æE ö
0.50
E = ES 1.08çç P ÷
÷ 0.16çç P ÷
÷ − 0.22çç P ÷
÷
è ES ø è ES ø è ES ø
8.6 Selected foundation structure limited duration during extreme loading is

concepts for offshore applications allowable.
Each of the three offshore foundation con-

8.6.1 Introduction to concepts
cepts has its pros and cons. The monopile is
Three basically different foundation struc- usually an attractive foundation solution for
ture concepts exist for offshore wind tur- wind turbines to be placed in shallow waters
bines, viz. and smooth seas. This concept is attractive
• monopile also because it offers a fast installation. The
• gravity base tripod is preferred over the monopile in
• tripod harsher climates and less shallow waters,
These three foundation concepts are differ- however, the solution with suction buckets
ent in the manner by which they transfer the under the three legs is only feasible for
loads to the supporting soils. The monopile foundations in clay and will not work in
and the tripod both transfer vertical forces as sands or other permeable soils. The tripod
axial shear forces at the pile-soil interfaces. concept is a light-weight structure, which is
The gravity base foundation transfers such usually cost-efficient. It is an attractive fea-
forces as a vertical contact pressure at the ture of gravity base foundations that they
foundation-soil interface. Both the monopile can be equipped with more or less shallow
and the tripod primarily transfer horizontal skirts under their bottoms. This will move
forces by way of horizontal earth pressures the depth of load transfer deeper below the
against the piles at some depth. The gravity sea floor and it will contribute to prevent
base foundation primarily transfers such scour from occurring around and in particu-
forces directly to horizontal shear forces in lar underneath the foundation, especially in
the soil at the foundation-soil interface. The sand. Gravity base foundations may often be
monopile also transfers bending moments feasible from a technical point of view, but
about the seabed as horizontal earth pres- may not always come out economically fea-
sures against the pile. The tripod primarily sible. They form a well-tested concept,
decomposes such overturning moments into which is, however, also a time- and re-
axial forces in the three legs, and these source-demanding concept. With these pros
forces are then transferred to the soil as axial and cons of the various foundation types in
shear forces at the pile-soil interfaces. The mind, it should be noted that the type of
gravity base foundation transfers overturn- foundation to be chosen also much depends
ing moments about the seabed by a varying on the actual soil profile and the soil proper-
vertical contact pressure over the founda- ties on the site.
tion-soil interface. During design, it is im-
portant to make sure that this contact pres- Two of the three distinct foundation struc-
sure in extreme load situations never takes tures for offshore wind turbine installations
on negative values. Any significant, nega- are dealt with in the following sections, viz.
tive contact pressure would correspond to a the monopile and the tripod structure.
tension, which can hardly be transferred
unless purely undrained conditions with 8.6.2 Monopile
development of a suction can be counted on,
and which may lead to an undesirable sepa- General Description
ration between foundation and soil. How- Monopiles have traditionally been used as
ever, a limited “lift” of the foundation of the preferred foundation solution for some
types of marine structures. Piles with di-
ameters up to about 1 m have been used to turbine. The conical transition is welded to
support lighthouses and moorings. Equip- the pile after completion of the pile driving.
ment to install piles of diameters up to 3-4 m
to large penetration depths is available and Ice loads have a very large impact on the
makes monopiles a feasible foundation al- necessary dimensions of a monopile foun-
ternative for offshore wind turbines. An ex- dation. If only small variations in the water
ample of a monopile foundation is depicted level are expected, it is possible to design an
in Figure 8-13. ice cone, which will reduce the ice load sig-
nificantly and contribute to eliminate large
oscillations of the wind turbine tower. When
large variations in water level are present,
such an ice cone will not be feasible. This
may leave a very large dynamic ice load
acting on the monopile, and large oscilla-
tions of the wind turbine tower may result.
Determination of monopile dimensions

(design approach)
Three basic dimensions characterize a
monopile:
• length of pile in soil, L
• outer diameter of pile, D
• wall thickness of pile, t
The values of these geometrical parameters
are partly interdependent and have to be
selected in an iterative process, because
some of the loads, in particular the ice loads
Figure 8-13 Monopile Foundation, and the wave loads, will increase with in-
from LIC Engineering (1997) creasing pile diameter. At the same time,
however, the bending stress ranges causing
The advantages of a monopile foundation fatigue will decrease with increasing di-
include ameter.
• a fast and highly automated installation
with no prior preparation of the seabed The pile diameter D and the wall thickness t
• simple fabrication are determined from requirements expressed
The monopile foundation for a wind turbine in terms of design rules for the pile as a steel
consists of three basic parts: the bare pile, a structure. The design rules to be fulfilled
conical transition to the tower that it sup- consist of four criteria
ports, and a boat landing. By alteration of • ultimate limit state criterion
the conical transition, the foundation can • fatigue limit state criterion
easily be adjusted to fit different tower di- • local buckling criterion
ameters. • hard driving criteria
Once the pile diameter and the wall thick-
The boat landing is clamped to the pile and ness have been determined by making sure
provides a basis for the J-tube that carries that these criteria are fulfilled, geotechnical
the power cable from the seabed to the wind bearing capacity considerations for the sup-
porting soil are used to determine the neces-
sary length L of the pile. The axial load is wave load and from pile driving shall be
typically a compressive load. Usually the below the acceptable level according to the
axial bearing capacity of the pile will be Palmgren-Miner’s rule.
much larger than required, and the require-
ments to the lateral capacity, needed to Local Buckling Criterion
counteract the horizontal shear force and the When a tubular with a large diameter-to-
overturning moment set up by wind, wave wall-thickness ratio is exposed to bending or
and ice loads, will govern the design. The axial loading, a risk of local buckling is pre-
most economical design comprises a combi- sent. Local buckling can be checked ac-
nation of D, t and L which will minimise the cording to DS 449 (DS 449, 1983).
total pile mass. Once D, t and L have been
determined, a final step in the design of the Hard Driving Criterion
monopile consists of a dynamic analysis of During driving, the pile is exposed to very
the pile subjected to ice loading, and t is large axial stresses, which may cause local
adjusted if necessary. The design rules for buckling of the pile during so-called hard
determination of D and t are listed in the driving. Hard driving is defined as more
following for the respective limit states con- than 820 blows per meter penetration into
sidered: the soil. Hard driving shall be avoided. The
criterion is taken according to API (API,
Ultimate Limit State Criterion 1993).
In the ULS situation, the maximum design
Von Mises stress in the pile wall shall be The loads to be considered are:
smaller than or equal to the design yield • Loads due to wind acting on the wind
stress. The design von Mises stress is the turbine.
characteristic von Mises stress multiplied by • Loads due to waves and current acting
a load factor. The design yield stress is the on the foundation. These loads are in-
characteristic yield stress divided by a mate- cluded in the Ultimate Limit State
rial factor. (ULS) as well as in the Fatigue Limit
State (FLS) and may amount to 30-40%
Fatigue Limit State Criterion of the wind load acting on the wind tur-
The fatigue damage is calculated as a bine.
Miner’s sum. For this purpose, a design • Ice loads are very important. The result
long-term distribution of stress ranges shall of the analysis is a Dynamic Amplifica-
be applied together with a design S−N curve. tion Factor (DAF factor), by which the
The design stress range distribution is de- stresses found from a quasi-static analy-
rived from the characteristic stress range sis shall be multiplied.
distribution by multiplication of all stress For load calculations, reference is made to
range values according to this distribution Chapter 4.
by a load factor. At wall thickness transi-
tions and at the offshore on-site weld be- Analysis of laterally loaded piles
tween the pile and the conical transition, The minimum length of the pile is the small-
stress concentration factors are to be in- est length for which there will be no toe-kick
cluded. The design S−N curve is derived when the pile head is subjected to a lateral
from the characteristic S−N curve by divi- displacement at mudline. This is the length,
sion of all characteristic stress range values at which no further reduction of the lateral
according to this curve by a material factor. deflection at mudline can be obtained when
The total fatigue damage from wind and the length of the pile is increased. Reference
is made to Figure 8-14 and Barltrop (1991). Two alternative concepts are considered
For details about methods for analysis of feasible:
laterally loaded piles, reference is made to • Flange Connection (fast installation)
Section 8.4. • Welded Connection (strong connection)
Reference is made to Lyngesen and Brend-
strup (1997) for more details.
Corrosion Protection
As the surface of the pile is exposed to a
harsh environment due to waves, abrasion
from suspended sediments and a high salin-
ity, a large corrosion rate is expected. For
unprotected steel, the extreme surface corro-
sion of an exposed steel pile in ocean water
can be estimated to (DS 464/R, 1988):
0 years after installation 0 mm
The design life of a wind turbine foundation

usually varies between 30 and 50 years. The
maximum corrosion is located in the splash
Figure 8-14 Lateral deflections of a pile zone. The maximum bending stresses in the
exposed to horizontal loading. (Barltrop, pile during extreme loading are found below
1991). this zone, and therefore extreme stability
Connection between Monopile and Tower will be maintained even if a rather large cor-
The monopile foundation concept includes a rosion depth is found in the splash zone.
conical connection piece between the mono-
pile and the wind turbine tower. The con- However, a corroded surface generates areas
nection piece comprises the access platform with very large stress concentrations and
as an integral part. At the top of the conical cracks will propagate faster in corroded steel
connection piece, a flange for connection of than in uncorroded steel. Therefore, it is
the wind turbine tower is located. At the recommended to use corrosion protection in
bottom of the conical connection piece, an corrosive environments, even if the structure
in-situ connection to the monopile founda- is capable of accommodating the stresses
tion is to be established. over the extent of the corroded cross section
(DS 412, 1983).
The connection between the pile and the
conical connection piece is an essential part With a view to the above numbers, this calls
of the monopile foundation concept. The for implementation of some corrosion pro-
basic requirements to this connection are: tection system.
• Sufficient strength against extreme and
Three primary options exist for corrosion
fatigue loading
protection of a monopile
• Easy installation
• coating (vulnerable to damage during
• Low cost
installation)
• cathodic protection by sacrificial anodes hammers. The technique capitalises on the

(a large number of anodes are required shock wave which is generated in the pile by
for a 50-year design life) the impact from the hammer. The transfer of
• impressed current (facilitated by pres- the shock wave to the soil generates plastic
ence of electrical power supply) deformations of the soil and thereby moves
For a design life in excess of 30 years, the the pile downwards.
preferred option for monopile foundations is
protection by use of impressed current. So-called pile refusal occurs when the shock
wave amplitude is not large enough to create
The pile has a large surface, but normally plastic soil deformations. The penetration of
only a part of this requires protection. The the pile then ceases. Damping of the shock
surface exposed to wave action including the waves caused by soil adhesion along the pile
splash zone shall be protected. Depending surface may contribute to pile refusal, in
on the type of sediment, the potential for particular in cohesive soils. In this context it
scour etc., also a part of the pile below is important to be aware that the smaller the
mudline may need protection. Typically, the ratio between cross-sectional pile area and
area to be protected extends down to a level ram area, the longer is the shock wave
2 to 10 m below mudline. Below this limit- transmission time, the smaller is the shock
ing level, protection of the pile surface is wave amplitude, and the larger is the possi-
usually not required. bility for pile refusal.
Inside the hollow pile water will be present, Driving using vibrators:
but there will only be little exchange of wa- In frictional soils, the pile may be vibrated
ter and therefore corrosion on the inside of into the soil. Vibration is a very fast instal-
the pile is expected to be limited. However, lation method even compared to driving by
it is recommended that internal zones are hammer. Furthermore, it does not generate
protected either with coating or with ca- very large shock waves in the soil that could
thodic protection (DNV, 1998). If the pile is otherwise have caused damage of adjacent
filled with sand, corrosion will be mini- structures. However, only a few very large
mized or even prevented (DS 464/R, 1988). vibration hammers are available world-wide.
The vibrations generate a liquefaction of the
Installation soil locally around the pile, thus reducing
Driving of large diameter piles is well- the side adhesion to a minimum. Then the
known from the offshore industry. Gener- pile penetrates into the soil mainly due to its
ally, at least three different pile-driving own weight.
methods are available. These methods are
described briefly in the following. Driving using drilling or excavation:
For installation of hollow piles, drilling or
Driving using piling hammer: excavation may be applied. The soil in the
Pile driving by means of a hammer is the core and below the pile is excavated, gener-
oldest technique. In its simplest version, it ating a hole with a diameter usually slightly
consists of a ram dropped on the top of the larger than that of the pile. Then the pile
pile from a certain height. Today, hydraulic penetrates into the soil due to the mass of
fluid is widely used for lifting the ram and the pile. If side adhesion or lateral resistance
further accelerating it during the downward is required, grout may be injected in the an-
stroke. Hydraulic piling hammers are the nulus between the pile surface and the soil.
most efficient compared to diesel and steam
Notes on pile refusal Inspection of Cathodic Protection

The pile has to be driven to a particular The monopile foundation is protected
penetration depth, which is required to en- against corrosion by means of an impressed
able the pile to carry its design lateral load. current system. A reference control point
However, pile refusal may be encountered (the point with minimum potential) on the
before the final penetration depth is reached. monopile is selected. The electrical potential
This may happen if the piling hammer does of this point is measured and is reported
not have sufficient excess capacity com- through the wind turbine monitoring system.
pared to the theoretically calculated driving In case the anode disappears or the area of
resistance. the anode becomes too small, the potential at
the control point will increase, and the anode
In case of pile refusal, a decrease of up to needs to be replaced. Further, it is recom-
50% in the driving resistance can be ob- mended to perform a visual check of the
tained by removing the soil plug inside the cathodic protection system whenever an
pile. In sandy and silty soils, this can be inspection on the wind turbine tower is car-
achieved by jetting. In clays and tills, this ried out.
can be achieved by using an auger. The
driving resistance can also be reduced by Inspection of Scour Protection
installation of an external “driving shoe” on Scour will occur around the monopile. To
the pile tip, however, this may also lead to a protect from local scour around the pile,
reduced lateral capacity and may therefore scour protection is installed. Throughout the
not always be recommended. lifetime of the wind turbine the scour holes
around the pile are to be measured from time
Maintenance to time. The holes can be measured by
Inspection and maintenance are to be carried means of a 3-D echo sounder. The scour
out to ensure a sufficient structural strength. protection is not absolutely stable and needs
to be maintained. When an unacceptable
Crack Inspection scour occurs, more scour protection is to be
Cracks may be initiated at the weld between installed.
the monopile and the conical transition sec-
tion. Therefore the weld is to inspected for Spring Inspection
fatigue cracks. The inspection will consist of When the foundation is located in an envi-
Visual Inspection and/or Non-Destructive ronment where ice loading may occur, it is
Testing (NDT): recommended to perform an annual inspec-
• Magnetic Powder Inspection (MPI) tion every spring to verify whether ice
• Eddy current loading has caused damage to any structural
• Ultrasonic examination members or coating during the past winter.
In case of damage, repair is to be carried
Depending on the fabrication method of the out.
monopile, welds may also be located in the
submerged zone above mudline. These Dismantling
welds also need to be inspected. The extent In the following, two different methods are
of each inspection will be determined as part outlined for removing the monopile founda-
of the inspection planning. The interval be- tion from the seabed after its service life has
tween each inspection is to be determined terminated.
based on requirements to lifetime and on the
predicted accumulated fatigue damage.
1. Cutting off the piles is a method that can pressive loads in three hollow steel piles that
be used in general. By this method, the are driven into the seabed. From below the
monopile is cut below the mudline. The tie-in flange, a large diameter tubular, re-
method is based on using a cutting tool low- ferred to as the centre column, extends
ered down inside the pile. The subsea cut- downwards. On the upper section immedi-
ting is performed by a high pressure jet, ately below the flange, the dimension of the
consisting of water mixed with a non-pollute centre column is identical to that of the
grinding compound (sand). tower. Below this section, a transition sec-
tion reduces the diameter and increases the
The procedure consists of the following wall thickness. Reference is made to Figure
steps: 8-15. The tripod concept is still under devel-
• The soil core inside the tubular is re- opment, so the presentation here is brief
moved to the elevation of the cut line. only.
• The cutting tool is lowered to the eleva-
tion of the cut line, inside the pile.
• The pile is cut at elevation of cut line
• The upper part of the pile is lifted off
and transported to shore on barge.
The remaining part of the pile will be left

below the seabed, and will not cause any
damage to the environment.
2. Complete removal of the pile is another

method of dismantling. A vibration hammer
can be used for this purpose. The vibration
hammer is typically used for removal of
piles in non-cohesive soils. The hammer is
placed on top of the pile, and the induced
vibrations in the pile will cause local lique-
faction in the soil around the outer surface of
the pile. When the soil is thus liquefied, no Figure 8-15 Tripod with piles, from Ram-
skin friction between soil and pile will be bøll (1997)
present, and the pile can then be pulled up This concept is advantageous for several
using a crane vessel. reasons:
• The steel volume applied as corrosion
The advantage of this method is that the
appliance is reduced since the diameter
entire pile is removed. The disadvantage is
is reduced.
that the method is not likely to work for re-
• The layout of the joints connecting up-
moval of piles in cohesive soils.
per leg bracing to centre column is im-
proved with less difference in diame-
ters. The punching shear capacity of the
8.6.3 Tripod
connections between the upper braces
General description and the centre column are increased,
The tripod foundation concept consists of a and the stiffness of the joint is im-
steel frame that transfers the sectional forces proved. This is a paramount for
from the tower to primarily tensile and com- achieving the desired fatigue life.
• The conical section will decrease the as strut members, i.e., transferring compres-
design governing ice loading somewhat, sive and tensile forces only. Reference is
although the cone would have to be made to Figure 8-16.
higher to provide full reduction for all
sea levels depending on the environ-
mental conditions.
• The hydrodynamic loading is reduced.
Each leg frame consists of a pile, a pile

sleeve and two braces. At the bottom of each
pile sleeve, a mudmat is located to support
the structure on the seabed prior to the pile
driving. After completion of the pile driving,
the annulus between the pile and the sleeve
is filled with grout to form a rigid connec-
tion. It is recommended to batter the pile
from the vertical, such that the extension of
the centre line of the pile intersects with the
point of impact of the dominating load on
the tripod structure. Note that excessive
batter requires the hammer to be supported
laterally during pile driving and should
therefore be avoided.
As an alternative to using three piles to sup-

port the tripod structure and transfer the Figure 8-16 Tripod with suction buckets,
loads to the soil, so-called skirt foundations from Rambøll (1998)
may be applied, e.g. in the form of suction The tripod concept makes use of the conical
buckets. One suction bucket would then transition between the tower and the sup-
have to support each of the three tripod legs. porting column in the tripod structure as an
When suction buckets are used for the foun- ice cone. When solid ice is forced against
dation, a variant of the tripod structure is the structure, the conical transition causes
feasible, in which the central column sup- the ice to be pressed downwards, and
porting the tower is extended down to the crushing takes place at a level with a smaller
seabed and further supported by a suction diameter than that of the tower, and the ice
bucket. With this configuration of the cen- load on the structure is correspondingly
tral column with a bucket support, only two smaller. A complete elimination of ice
more supports with suction buckets are crushing is not feasible, since this would
needed. These are placed such that their re- pose requirements to the inclination of the
spective frames are perpendicular to each conical transition far beyond what would be
other. This implies that access of service feasible considering the diameters of the
vessels is possible within a 270 degrees an- involved tubulars. The disadvantage associ-
gle. By this design, the centre column will ated with the use of the conical transition as
participate more actively in the transfer of an ice cone is that it implies an upwards
forces from the tower to the seabed. All vertical ice load will be acting on the centre
supporting elements including the centre column. However, the vertical capacity is
column are designed as circular tubes acting usually quite large, so this disadvantage is
usually not critical, yet it needs to be duly pletely provided the tripod as a whole can
accounted for in the design. absorb the design loads.
The tripod structure is not compatible with The plastic design moment shall be calcu-
too shallow water depths. This is due to: lated by the formula:
• A sufficient water depth above all parts
of the structure is required to allow M P (N ) = σ γ ⋅ WP ⋅ cos[π2 ⋅ σ N⋅ A ] ,
Y
service vessels to approach the struc-

ture. WP = D 3 − ( D − 2 ⋅t ) 3
• At very low water depths, the distance

6
between the upper and lower braces be- Where:

comes too short for the forces to be MP(N): Design plastic moment capacity as
transferred in any reasonably sized a function of the axial force
cross section. N: Axial force in pile
• A solution with the braces protruding σy: Nominal steel yield stress
the water surface is not desirable due to WP: Plastic section modulus
the possible interlocking or packing of D: Pile diameter
ice and the possibility of collision with t: Pile wall thickness
service vessels. A: Steel area of pile
The tripod causes relatively little blocking
for wave action and for current flows which As a minimum, the result of the analysis
is of importance for environmentally sensi- must contain the pile utilization ratio and a
tive areas. This is so even when a large verification that the penetration depth is
number of foundations are installed as in an sufficient. All the piles are allowed to reach
offshore wind farm. the design plastic moment. The forces and
moments at the pile heads shall be in equi-
Geotechnical analysis – Piles librium with the design loading acting upon
In the ultimate limit state, the foundation the structure and shall be smaller or equal to
analysis shall consist of an elastic and a the allowable design values.
plastic analysis check. The two analysis
checks can be described as follows: In both elastic and plastic analysis, the fol-
lowing features must be taken into account:
In the elastic analysis, a check is carried out • Linear-elastic superstructure that inter-
of the stresses in the piles and the structure connects the piles
with design loads (ULS) and design material • Non-linear behavior of the piles later-
properties. One pile only is allowed to reach ally and axially
its design maximum allowable stress. The • Pile group effects
allowable design value of the von Mises’
• Second-order moments in the piles
stress is the yield stress fy divided by the
• Soil plug length taken as 0.9 times
partial safety factor for steel strength. The
length of pile embedded in soil
result of the analysis must as a minimum
contain the utilisation ratio of the pile steel.
The procedures for development of t-z and
In the plastic analysis, the entire tripod
q-w curves for the axial bearing capacity and
structure is checked using design loads
p-y for the lateral bearing capacity can be
(ULS) and design material properties. All
done using the API approach (API, 1993).
the piles are permitted to yield or fail com-
When characteristic values for the skin fric- ally can be assumed and modeled as spring
tion and tip resistance are given, the t-z constants. The true pile-soil interaction rela-
curves for axial pile-soil interaction can be tionship is usually a smooth nonlinear curve.
established by means of any relevant recog- The applied spring constants can be taken as
nised computer program. The shape of the the initial slope of the smooth curve.
curves is fully described by
Geotechnical analysis – Suction Buckets
2 éz π G ù The buckets work by principle of restricting
t = ⋅ t max ⋅ arctan ê ⋅ ⋅ ú flow of water from outside the bucket to
π ë 3 D t max û inside the bucket. A sudden application of
load yields a remarkable resistance towards
Where: pullout, gradually tapering off if the loading
G: Shear modulus is sustained. This resistance is formed by
z: Deflection pile/soil suction in the porewater inside the bucket,
D: Diameter of pile mobilised as the immediate reaction to the
tmax: Maximum skin friction, calculated suddenly applied load. The bucket founda-
in accordance with API procedure tion is thus primarily designed to absorb
E: Young’s Modulus load peaks. A static load in excess of a cer-
ν: Poisson’s ratio tain limit and acting over some length of
Reference is made to Clausen et al. (1982). time will gradually pull out the bucket as
water is allowed to flow from the outside to
The shape of the q-w curves for the tip load- the inside of the bucket and thereby neutral-
displacement relationship can be assumed to ises and eventually eliminates the suction
be bilinear with a required relative dis- condition inside the bucket.
placement between pile and soil of 5% of The concept is new only in relation to the
the pile diameter to cause yield. application of suction technology to the de-
sign of offshore wind turbine foundations.
The lateral bearing capacity can be based on The suction (or differential pressure) tech-
p-y data developed by the API-RP2A-WSD nology has been suggested and applied as
procedures using cycling soil strength. Soil both anchoring device for ships (and tension
parameters needed in the modeling of the leg platforms) and as foundation for fixed
curves are: leg platforms as an alternative to piles.
Clay: uc: Undrained shear strength However model testing and analyses are still
γ’: Submerged unit weight required in order to obtain a sufficient de-
J: Empirical constant sign basis against hydraulic instability of
ε50: Strain at one-half the max. skirted foundation of offshore wind turbines.
stress in laboratory un-
drained compression test Structural analysis (Limit states)
Sand γ’: Submerged unit weight The structure and all structural members
Φ: Angle of internal friction shall be verified according to three limit
(plane) states. Different sets of partial safety factors
The curves and the pile group effects can be for load and resistance apply to the different
modeled by means of a relevant recognised limit states. The normal safety class ac-
computer program. cording to DS449 (DS 449, 1983) can be
adopted, as the consequences of failure are
In a natural frequency and fatigue analysis limited (loss of investment but no loss of
linear behavior of the piles laterally and axi- human life).
In the ultimate limit state (ULS) the structions. The areas which are most likely to be
ture is checked against extreme loads. The dimensioned by fatigue loads is the centre
structure above seabed should be able to column transition to the ice cone and the
sustain these loads without collapse or per- upper joint at the centre column where the
manent deformation. For piles, a fully de- upper leg framing attaches to the structure.
veloped plastic failure mode is accepted. Consider the distribution of stresses due to
Horizontal ice load is based on static ice pure bending of the tower. This will lead to
load, multiplied by a dynamic amplification a maximum stress above the upper joint at
factor (DAF). Vertical ice load may destabi- two points on a line perpendicular to the axis
lize the structure, because it is applied as an of the applied moment. The stresses will
upward load. vary linearly between these points, but as the
wind direction changes, also the locations of
In the serviceability limit state (SLS) the these points move. Thus the loading (both in
maximum deflections are checked with spe- terms of magnitude and in terms of number
cial emphasis on the allowable tilt of the of stress cycles) shall be applied for a fixed
foundation due to differential settlements. In but critical point in the cross section, i.e., the
this state, no hydrodynamic loads on the fatigue loads are to be applied from one par-
foundation need to be included, since these ticular direction only, and this direction is to
loads will contribute only marginally to the be taken as the most critical direction. The
overall moment on the tripod. All partial conical transition induces an additional
load factors are to be set equal to unity, i.e. stress, which has to be accounted for. The
characteristic loads are to be used in calcu- designer may find it feasible to locate inter-
lations. Loads from the wind turbine shall be nal ringstiffeners, where the cone connects
taken as the worst damaging load case to the tubular, and thereby lower the overall
among all dynamic load cases. The maxi- dimensions of the conical section. Several
mum allowable tilt under this characteristic layouts of the stiffener(s) are possible:
loading is 0.5° off vertical. • A number of internal ringstiffeners
• Bulkheads
In the fatigue limit state (FLS) the structure • The leg framing may protrude into the
is checked against failure due to fatigue interior of the centre column to be
damage. The cumulative fatigue damage for joined at the centre line, thereby dis-
all load situations during the design life tributing the loads directly between the
must be taken into account and the inte- leg frames
grated effect should be investigated, e.g. by The most advantageous option may be cho-
using the rain-flow counting scheme for sen, taking into consideration the require-
stress cycles. Fatigue loads from the wind ments of internal clearance for risers, con-
turbine shall be considered in conjunction ductors, power cables etc.
with wave-induced fatigue loading, and it is
to be verified whether it is acceptable to Natural frequency analysis
disregard contributions to the fatigue dam- A complete natural frequency analysis shall
age from ice crushing. be performed for the combined structure
consisting of turbine, tower, tripod and piles.
As fatigue is considered critical for the For this purpose, the non-linear soil must be
structure, it is of importance to verify that linearised. It is to be verified that the lowest
the fatigue life does not fall short of the de- frequencies differ from at least ±10% of the
sign life, based on detailed information of 1P and 3P rotor frequencies at nominal
waves from different geographical direc- power.
Grouted connections take into account the presence of both local

Axially loaded pile-to-sleeve connections and global scour around the tripod.
are checked using the formulae given by The centre column must be sealed with a
DNV (1977), including a safety factor of flange in order to avoid resonance between
3.0. The connections can be established by a water inside the column and the surrounding
system widely used within the oil/gas in- sea.
dustry. This system consists of flexible grout
lines (OD 50 mm) attached to the tubes and Installation
connecting a simple manifold above water Besides the geophysical and geotechnical
level to an outlet in the upper part of the pilesurveys as described in Chapter 8.2, the only
sleeves. preparation of the seabed necessary before
installation can take place is the establish-
Corrosion protection ment of the mudmats. The purpose of the
Two options are found advantageous as pri- mudmats is to secure the on-bottom stability
mary protection. They are both of the ca- prior to final driving of the three piles. They
thodic type: shall be designed to satisfy the following
• Sacrificial anodes of the ZN-Al-In type, requirements:
placed as slender anodes in “handles” • axial and lateral capacity of the soil
attached to the structure below sea level • lateral resistance due to stabbing pile
• Impressed current system • horizontal level adjusted installation
The disadvantage of the sacrificial anode tolerances
system is the need for inspection. A sacrifi- The installation procedure is to a large ex-
cial anode system may have to be renewed tent based on the choice made by the con-
during a design life, depending on water tractor and is therefore not taken into con-
salinity. sideration here.
Due to the on-line monitoring system al-
ready present for the turbine, measuring Maintenance
potential differences easily monitors an im- The maintenance of the steel structure in-
pressed current system. The current re- cludes activities that are not specific to the
quirements may, however, pose a problem tripod but apply to all steel foundation
for remote and isolated offshore steel struc- structures.
tures.
Prior to coating, the entire structure must be Crack monitoring
sandblasted to comply with at least Sa 2½ Crack monitoring of welded joints will have
according to ISO 8501-1n. to be performed at regular intervals, at least
The areas above the lower limit of the splash until a general experience with the structure
zone is to be coated with 2.5 mm of solvent- has been established. Crack monitoring con-
free epoxy coating. The remaining structure sists of ultrasound or X-ray inspection at
is to be coated with a shop primer to resist critical time intervals. For the tripod, the
the environment at the construction site. critical details are considered to be the cone
sections and the joint between the centre
Scour protection column and the upper leg bracing. To fa-
No scour protection is needed for the tripod cilitate NDT inspection of the tripod, the
structure. The small overall dimensions of centre column can be sealed off at some
all members close to the seabed yield rather elevation below the joint and thereby allow
small increases in velocities, and thus only for dry testing. The structure may be
little local scour. However, the design shall
equipped with attach points for the scanning Clausen, C.J.F., P.M. Aas, and I.B. Al-
equipment to facilitate positioning. meland, “Analysis of the Pile Foundation
System for a North Sea Drilling Platform”,
Specifying the allowed fatigue utilization Proceedings, BOSS, 1982.
the designer can allow for a trade-off be-
tween initial manufacturing costs and subse- Danish Energy Agency, “Recommendation
quent maintenance costs. to Comply with the Requirements in the
Technical Criteria for Danish Approval
Surface protection system Scheme for Wind Turbines, Foundations,”
Maintenance consists of repair of coating in August 1998.
case of damage and monitoring/maintenance
of the impressed current system. The im- Dansk Ingeniørforening, “Dansk Ingeniør-
pressed current system is believed to have a forenings Norm for Stålkonstruktioner,”
lifetime of 15 to 25 years after which the DS412 (in Danish), Teknisk Forlag, Norm-
anode must be replaced. styrelsens Publikationer, Copenhagen,
Denmark, April 1983.
Dismantling
The tripod can be moved from the seabed by Dansk Ingeniørforening, “Dansk Ingeniør-
cutting off the piles at or preferably below forenings Norm for Fundering” DS415 (in
mudline. The pile sections thus left behind Danish), Teknisk Forlag, Normstyrelsens
pose no ecological threat. Total removal of Publikationer, Copenhagen, Denmark, 1998.
the pile can be accomplished by an inverse
driving procedure and/or by vibration Dansk Ingeniørforening, “Dansk Ingeniør-
equipment. forenings Norm for Pælefunderede Offshore
If suction buckets are used for the founda- Stålkonstruktioner,” DS449 (in Danish),
tion, the tripod can be removed completely. Teknisk Forlag, Normstyrelsens Publika-
tioner, Copenhagen, Denmark, April 1983.
REFERENCES Dansk Ingeniørforening, “Dansk Ingeniør-

forenings Anvisning for Korrosionsbeskyt-
API, “Recommended Practice for Planning,
telse af Stålkonstruktioner i Marine Omgiv-
Designing and Constructing Fixed Offshore
elser,” DS 464/R, Teknisk Forlag, Normsty-
Platforms – Working Stress Design”, API
relsens Publikationer, Copenhagen, Den-
Recommended Practice 2A-WSD(RP2A-
mark, June 1988.
WSD) 20th Edition, American Petroleum
Institute, July 1, 1993.
DNV, “Rules for the Design, Construction
and Inspection of Offshore Structures,” Det
API, “Recommended Practice for Planning,
Norske Veritas, Høvik, Norway, 1977. (Re-
Designing and Constructing Fixed Offshore
print 1978.)
Platforms,” API RP2A, 17th Edition,
American Petroleum Institute, 1987.
DNV, “Foundations,” Classification Notes,
No. 30.4, Det Norske Veritas, Høvik, Nor-
Barltrop, N.D.P., and A.J. Adams, “Dy-
way, 1992.
namics of Fixed Marine Structures,” The
Marine Technology Directorate Limited,
DNV, “Rules for Classification of Fixed
Butterworth-Heinemann Ltd., Third Edition,
Offshore Installations,” Det Norske Veritas,
1991.
Høvik, Norway, 1998.
Eide, O., and K.H. Andersen, “Foundation Lyngesen, S., and C. Brendstrup “Vindmøl-
Engineering for Gravity Structures in the lefundamenter i Havet, EFP-96, J.nr.
Northern North Sea,” Norwegian Geotech- 1363/96-0006, Final Report Mono Pile
nical Institute, Publication No. 154, Oslo, Foundation,” LIC Engineering A/S, Febru-
Norway, 1984. ary 1997.
Hansen, B., “Limit Design of Pile Founda- Rambøll, “Vindmøllefundamenter i Havet,

tions”, Bygningsstatiske Meddelser, No. 2, EFP-96, J.nr.1363/96-0006, Tripod Founda-
September, 1959. tion Volume 1 – Report”, Rambøll, February
1997.
Hansen, B., ”Geoteknik og Fundering, Del II
Forelæsningsnotater til Kursus 5821 − Geo- Rambøll, “Havmøllefundamenter med sug,
teknik 2”, Notat nr. 16, Den private in- UVE-98 J.nr. 51171/97-0047, Fase 1-
geniørfond ved Danmarks tekniske Høj- rapport,” SEAS, Niras, Rambøll, DGI, Risø,
skole, Lyngby, Denmark, 1978. September 1998.
Hansen, J.B., “A Revised and Extended Reese, L.C. and H. Matlock, “Numerical
Formula for Bearing Capacity,” Danish Analysis of Laterally Loaded Piles”, Pro-
Geotechnical Institute, Bulletin No. 28, pp. ceedings, Second Structural Division Con-
5-11, Copenhagen, Denmark, 1970. ference on Electronic Computation, Ameri-
can Society of Civil Engineers, Pittsburgh,
Kraft, L.M., R.P. Ray and T. Kagawa, Pennsylvania, pp. 657, 1960.
“Theoretical t-z curves,”Journal of Geo-
technical Engineering, ASCE, Vol. 107, No. Reese, L.C. and S.-T. Wang, “Documenta-
11, pp. 1543-1561, 1981. tion of Computer Program Group 4.0”, En-
soft, Inc., Austin, Texas, 1996.
9. Electrical Installations separate disconnection devices, such that

these circuits may remain energized while
all other circuits are deenergized. It is rec-
9.1 Safety Systems ommended that any electrical system oper-
Reference is made to Chapter 2. ating above 1000 V AC or 1500 V DC shall
be able to be earthed for maintenance.
9.2 Electrical Safety
The IEC61400-1 requires that all electrical
The electrical system is usually considered components and systems shall meet the re-
to comprise all electrical equipment installed quirements of IEC60204-1, and compliance
in an individual wind turbine up to and in- with the requirements of IEC60364 for de-
cluding the wind turbine terminals. The sign of the electrical system of a wind tur-
power collection system is usually not con- bine is also required.
sidered included in the electrical system.
9.3 Grid Connection
The design of the electrical system for a
wind turbine shall ensure minimal hazards to It is recommended to make provisions to
people and livestock as well as minimal po- enable the wind turbine to be isolated from
tential damage to the wind turbine and the the public grid in a safe manner. This ap-
external electrical system during operation plies to normal situations, abnormal and/or
and periods of maintenance of the wind tur- faulty conditions as well as during mainte-
bine under all normal and extreme condi- nance and repair.
tions. It is usually required that the design of
the electrical system shall take into account 9.3.1 Power factor correction
the fluctuating nature of the power genera- An asynchronous generator contains no
tion from the wind turbine. It is usually also permanent magnets and is not separately
required that the electrical system shall in- excited. This means that an asynchronous
clude suitable devices that ensure protection generator has to get its exciting current from
against malfunctioning of both the wind the grid, and the magnetic field of the gen-
turbine and the external electrical system erator is only established when the generator
which may lead to an unsafe condition. is connected to the grid. When the generator
gets its exciting current from the grid, the
A wind turbine should be equipped with an generator consumes reactive power from the
earthing system both for conducting light- grid. The current in the cable to the genera-
ning currents and for earthing the electrical tor will consist of two parts
system of the turbine. Connection to the • active current corresponding to the ac-
earthing network should be possible at every tive power production (unit: kW)
location where electrical systems or equip- • reactive current corresponding to the
ment are located. reactive power consumption (unit:
kVA)
It is recommended that disconnection of the Because of the exciting current, there is a
electrical system from all electrical sources phase shift between the current and the volt-
as required for maintenance or testing shall age. The current is delayed by a phase angle
be possible. Semiconductor devices should
φ. The active delivered electrical power can
not be used alone as disconnection devices.
be expressed as
Where lighting or other electrical systems
are necessary for safety during maintenance,
auxiliary circuits should be provided with Pnet = 3U eff I eff cos φ
9 – Electrical Installations 221

in which Ueff and Ieff are the line-to-line root- open, a bypass contactor is closed, such that
mean-square voltage and the line-to-line the thyristors are closed.
root-mean-square current, respectively, and
cosφ is referred to as the power factor. The When disconnecting the generator from the
maximum voltage Umax is defined as the grid, the thyristors are completely opened
line-to-line peak voltage, and the maximum before the bypass contactor is disconnected.
current Imax is defined as the line-to-line The thyristors can then be trigged down
peak current. The root-mean-square voltage during approximately 80 msec, until the
and the root-mean-square current are ex- current is zero, and then the generator is
pressed in terms of the maximum voltage disconnected. This procedure ensures that
and the maximum current as follows the contactor does not have to break any
current, and there will be no arc inside the
U max I max contactor.
U eff = and I eff =
2 2 The thyristor cut-in and cut-out result in a
smooth connection and disconnection to the
A part of the generator’s exciting current or grid and contribute to increasing the lifetime
reactive power is normally delivered from of the contactor.
capacitors. This part is known as the power
factor correction. The capacitors are usually 9.4 Wind Turbine Controller
connected to the grid a little later than the
generator and disconnected again before the The purpose of the wind turbine controller is
generator is disconnected from the grid. The to monitor and control all functions in the
advantage of the power factor correction is turbine in order to ensure that the perform-
that the losses in the grid decreases, because ance of the turbine is optimal at any wind
the grid current decreases. At no-load, the speed. It is common to continuously collect
grid current is about 0 A, because the gen- data about the performance of the wind tur-
erator’s no-load current, i.e. only reactive bine, e.g.
current, is delivered from the capacitors. • rotor and generator speed
• wind speed
9.3.2 Frequency converter • hydraulic pressure
A frequency converter is required for tur- • temperatures
bines, which are being design to run at vari- • power and energy production
able rotational speeds. • pitch
A more detailed list of signals and parame-
9.3.3 Thyristor cut-in ters, which may be monitored by the con-
troller is given below
When an asynchronous generator is con- • electrical grid (voltage, current, main
nected directly to the grid, there will be a circuit breaker, overvoltage protection)
large cut-in current. The cut-in current can • yaw system (feedback yaw, twist, yaw
be up to 7-8 times the rated current, and can pulses, thermorelay yaw motor)
cause disturbances in the grid. To avoid such • hydraulics (oil level, differential pres-
disturbances, the generator can be cut in at sure filter, pressure in brake, hydraulic
the synchronous speed through thyristors. pressure, feedback hydraulic pump
Thyristor cut-in implies that thyristors motor, thermo-relay hydraulic pump
slowly ”open up” for the current. Such thy- motor, temperature hydraulic oil, ther-
ristor cut-in usually takes approximately 1.5
secs. When the thyristors are completely
222 9 – Electrical Installations

mistor brake, feedback brake, hydraulic • thyristors (feedback bypass contactor,

pump motor, emergency pitch, brake) voltage, thermistor thyristors, bypass
• ambience (ambient temperature, ane- contactor, trigging pulses)
mometer, wind direction, thermostat top • pitch system (pitch position, pitch ve-
and bottom, nacelle temperature, feed- locity, software debug)
back nacelle fan, thermorelay nacelle • computer (temperature – the tempera-
fan, busbar fan start temperature, top- ture in the computer that is used to
controller temperature, busbar tem- monitor the wind turbine needs to stay
perature, feedback nacelle fan, thermo- above 0°C to ensure correct data proc-
relay nacelle fan, nacelle fan, start/stop essing)
controller fan, contactor nacelle fan, top • emergency stop and miscellaneous
controller fan) (emergency circuit top processor, emer-
• rotation mechanics (rotor, generator, gency circuit ground processor, watch-
gear oil temperature, vibration sensor, dog)
gear bearing temperature, flow/filter er-
ror oil cooler, gear oil heater) 9.5 Remote Control
• generator (feedback generator, tem-
perature, contactor, generator heating) Remote control of the wind turbine must not
• power factor correction (feedback sen- take place whenever personnel is present in
sors, contactors) the tower or in the nacelle.
9 – Electrical Installations 223

224 10 – Manuals
10. Manuals will normally prescribe in detail how to

maintain:
10.1 User manual • Hydraulic systems

The user manual and/or operating instruc- • Main gear
tions are issued to assist the wind turbine • Yaw system
operator and will normally include the fol- • Blade pitch or tip brake mechanism
lowing information: • Sensors
• Lubrication and cooling systems
• General description of the wind turbine • Inspection for corrosion, wear and
• Personal safety instructions cracks including remedial action
• Service/Maintenance schedule • List of spare parts
• Instructions for use of the controller • Trouble shooting instructions
including settings and list describing • Relevant drawings, diagrams, specifi-
alarms cations and descriptions
• Trouble shooting incl. instructions at
dangerous events like run away and fire 10.3 Installation manual
• Description of locking devices for rotor,
The installation manual will describe the
yaw and pitch system
various installation activities as well as the
• List of grease, lubrication oil, hydraulic
limitation for these with respect to, for ex-
oil etc.
ample, wind speed. The activities should be
• List defining bolt pretension for the described specifying a.o.:
main connections
• General description of the wind turbine
In the EU, there are also minimum require-
• Personal safety instructions
ments to the manual through the Machinery
Directive, see European Union (1998). • Transportation and handling
• Procedures for bolt connections in-
10.2 Service and maintenance man- cluding pretension procedures
ual • Tests and checks for various parts and
systems of the turbine
This manual shall assist qualified engineers • Commisioning test normally including
in their work and will normally include the start, stop, yawing and test of protection
same information as that which is given in system
the user manual in addition to specific in-
structions for the service and maintenance REFERENCE
work. The level of detail depends on the
education of the service staff. The manual European Union, Machinery Directive,
European Directive No. 98/37/EC, 1998.
10 – Manuals 225
226 10 – Manuals
A. Bolt Connections −N curves

A.2 S−
In wind turbine structures, two major types For design of bolt connections, the single
of bolt connections are used, viz. most important issue to consider is the
• prestressed flange connections proper choice of an S−N curve for fatigue
• scarf joints predictions. This S−N curve is much de-
pendent on how the bolt is manufactured.
Scarf joints transfer forces by means of fric- Several possibilities exist for this manufac-
tion, and they can be one-sided and two- turing, including
sided, i.e., with one or two scarf plates, re- • thread form-rolled and then heat-treated
spectively. One-sided scarf joints are asym- • material of near nominal diameter fully
metric and should therefore be avoided. heat-treated before form-rolling of
thread
A.1 Bolt qualities • material fully heat-treated, then surface
ground before form-rolling of thread
Bolt strength is standardised according to
DS/ISO898/1. The different qualities are
characterised in terms of property classes by The general form of the S−N curve is
two numbers separated by a dot. The first
number is 1/100 of the minimum ultimate log10N=log10a−m⋅log10∆σr
strength measured in units of MPa. The
product of the two numbers is 1/10 of the in which N is the number of cycles to failure
minimum yield strength, also measured in at stress range ∆σr, and log10a and m are the
units of MPa. parameters of the S−N curve.
For example, bolt quality 10.9 has a mini- For stay bolts of Quality 8.8 in prestressed
mum ultimate strength of 100⋅10=1000 MPa flange connections, for which the thread is
and a minimum yield strength of rolled after heat treatment of the stay bolt
10⋅10⋅9=900 MPa. material, the following S−N curve can be
used for design:
For clamping bolts, only qualities 8.8, 10.9
and 12.9 are relevant. Caution should be For N<1.5⋅106:m=4, log10a=14.95
exercised when using quality 12.9 because For 1.5⋅106≤N<108:m=7, log10a=21.53
of the risk of brittle failure associated with For N≥108:∆σr=86 kPa (constant cutoff)
this quality.
Quality 8.8 refers to ISO898/Property Class
It is common to use bolts with rolled thread. 8.8.
The strength of bolts with cut thread is usu-
ally 85% of the strength of bolts with rolled For bolts whose manufacturing procedure is
thread. not known, the following S−N curve can be
used for design:
Bolts must not be tightened to a tensile
stress in excess of 70% of the ultimate For N<107:m=3, log10a=12.16
strength of the bolt. For N≥107:m=5, log10a=15.61
A – Bolt Connections 227

228 A – Bolt Connections

C. Fatigue Calculations ten be located in areas with stress concen-

trations. Depending on definitions of failure
criteria, the crack growth period until unsta-
C.1 Methods for structural steel ble fracture occurs may also be different for
To ensure that a structure will fulfil its in- the two prediction methods.
tended function, fatigue assessment, sup-
ported where appropriate by a detailed fa- The long-term stress range distribution is a
tigue analysis, should be carried out for each fundamental prerequisite for fatigue analy-
type of structural detail, which is subjected sis. This distribution can be determined in
to extensive dynamic loading. It should be various ways. Two approaches to arrive at
noted that every welded joint and attachment the stress range distributions are given here,
or other form of stress concentration is po- • a postulated generic form of the long-
tentially a source of fatigue cracking and term stress range distribution. It is
should be individually considered. common to assume a Weibull distribu-
tion, which has the advantage of leading
Fatigue design can be carried out by meth- to a simple formula for prediction of fa-
ods based on S−N curves from fatigue tests tigue damage. In case the Weibull dis-
in the laboratory and/or methods based on tribution is for nominal stresses, the
fracture mechanics. For design purposes, nominal stresses have to be multiplied
fatigue analysis based on S−N curves from by relevant stress concentration factors
fatigue tests is normally the most suitable to give local notch stresses before the
method. It is then used in conjunction with cumulative damage can be calculated.
Palmgren-Miner’s rule for prediction of cu- • a spectral method for estimation of
mulative damage. For assessment of in- long-term stress ranges. This method
service cracks with respect to maximum implies that the long-term stress range
inspection intervals or time before repair, distribution is predicted from given or
the crack propagation must be evaluated by assumed external conditions such as the
means of fracture mechanics. wind climate and, for offshore wind
turbines, the wave climate. This can be
The fatigue life consists of three stages, viz. combined with different levels of re-
• crack initiation finement of structural analysis.
• crack propagation
• final fracture Note that true fatigue lives are a function of
workmanship related to fabrication and cor-
A fatigue analysis based on S−N curves
rosion protection. Therefore, it is important
from fatigue tests may be considered to in-
that the fabrication is performed according
clude the two first stages. A fracture me-
to good practice with acceptance criteria
chanics analysis may be used to predict the
fulfilled as assumed for the prediction.
number of load cycles in the crack propaga-
tion stage of an actual structure. The extent,
The fatigue life – or in other terms the cu-
to which a fracture mechanics calculation
mulative damage – under varying loading
can provide comparable information on fa-
tigue life with that derived from S-N curves, can be predicted based on the S−N curve
approach under the assumption of linear
will depend on the number of load cycles in
cumulative damage by Palmgren-Miner’s
the initiation stage. Frequently, however, the
rule. The total damage that a structure will
initiation stage for welded joints is almost
negligible, because a fatigue crack will de- experience during its design life may be
expressed as the cumulative damage from
velop from existing defects, which may of-
C – Fatigue 229
each load cycle at different stress levels, detail, the state of stress considered and the
independent of the sequence in which the surrounding environment. The S−N curve
stress cycles occur, i.e. no sequence- should take into account possible material
dependency or so-called ”load cycle effect” thickness effects.
is present.
When the long-term stress range distribution
According to Palmgren-Miner’s rule, the is a Weibull distribution with scale parame-
accumulated damage D can be predicted as ter s0 and shape parameter h,
follows
s
k
∆n ( S i ) FS ( s ) = 1 − exp(−( ) h ) ,
D=å s 0
i =1 N ( S i )
and the total number of stress cycles in the

in which ∆n denotes the number of stress design life is ntot, then the cumulative dam-
cycles of stress range S in the lifetime of the age D can be calculated as
structure, and N is the number of cycles to
failure at this stress range. ∆n is determined n m m
from the long-term distribution of the stress D = tot Γ(1 + ) s 0
K h
ranges, and N is determined from the ex-
pression for the S−N curve. The sum is over in which Γ() denotes the gamma function.
all stress ranges Si in a sufficiently fine dis-
cretization of the stress range space into k The criterion for design is usually formu-
blocks of constant-range stress cycles. lated as
It is noted that the expression for the accu-
D≤1
mulated damage D can be interpreted as a
sum of partial damages owing to load cycles Caution should be exercised in cases for
at various stress ranges, regardless of the which load cycle effects are present, and for
sequence in which the load cycles occur.
which the sequence in which the load cycles
arrive is important. In such cases, the expre-
The S−N curve can usually be expressed as sion for D by the Palmgren-Miner’s sum
may then underpredict the true accumulated
N = KS − m damage.
or on logarithmic form An example of application is given in Sec-

tion 2.3.7 and demonstrates how the Palm-
lnN=lnK−mlnS gren-Miner’s sum can be used to predict the
probability of failure by assigning probabil-
which implies a linear relationship between ity distributions to the various parameters
lnS and lnN, in which lnK appears as an in- that enter the expression for the Palmgren-
tercept value. For design, the value of K has Miner’s sum for a structural detail.
to be selected as the value of K that corre-
sponds to the intercept of the design S−N Fatigue data consist of pairs of stress range S
curve on logarithmic form. The S−N curve and number of stress cycles N to failure at
valid for analysis of a structural detail is to this stress range. They provide the basis for
be applicable for the material, the structural calibration of a linear model on log-log
230 C – Fatigue
scale, lnN=lnK−mlnS, where K and m are C.2 Fatigue in welded structures

material constants. On a plot, this gives the
Welds in welded structures usually possess
S−N curve. defects, which may form the basis for crack
growth and eventually lead to fatigue fail-
The material will exhibit a natural variability ure. For welded joints involving potential
of lnN about the mean as given by this fatigue cracking from the weld toe, an im-
curve. Usually, the mean curve is specified, provement in strength by a factor of 2 on the
such that in 50% of the cases the true value fatigue life can be obtained by controlled
of lnN will be larger than given by the curve local machining or grinding of the weld toe.
and in 50% of the cases smaller than. For This can be carried out by means of a rotary
design of wind turbine structures, it is re- burr. This treatment should produce a small
quired to use a characteristic S−N curve, concave profile at the weld toe with the
which is to be chosen in conservative man- depth of the depression penetrating into the
ner such that lnN in a large fraction, 1−p, of plate surface at least 0.5 mm below the bot-
the cases exceeds the value given by the tom of any visible undercut. The undercut
characteristic curve. Usually 1−p is taken as should be small enough to allow the maxi-
97-98%. When the mean S−N curve is mum required grinding depth to remain
given, the characteristic S−N curve can be within 2 mm or 5% of the plate thickness,
found by keeping the values of lnN fixed whichever is the smaller. Reference is made
and changing the corresponding S values to Figure C-1.
according to the following formula
σ ln N
S1− p = S 50% exp(−k )
m
in which S50% is the stress range corre-

sponding to the mean S−N curve for given
N, S1−p is the corresponding stress range
corresponding to the characteristic S−N
curve with 1−p probability of exceedance,
σlnN is the standard deviation of lnN repre-
senting the natural variability in lnN for Figure C-1. Grinding a weld toe tangentially
given S, and k depends on p and can be to the plate surface (A) will produce little
taken from Table C-1. As a basis for design, improvement in strength. Grinding must
the characteristic S−N curve for p=2% is to extend below the plate surface (B) in order
be used. to remove toe defects, from DNV Class.
Notes 30.2,
Table C-1 Quantile coefficient k vs. frac-
tion p By grinding the weld profile such that a cir-
Frac- 20 10 5 2 1 0.1 cular transition between base material and
tion p weld surface is achieved, the stress concen-
(%)
k 0.84 1.28 1.65 2.05 2.33 3.09
tration becomes reduced as the grinding
radius is increased. This may lead to addi-
tional improvement of the welded joint.
When σlnN is not known, it can be set equal
to 0.46 for plain steels free from welds, see
DNV (1984).
C – Fatigue 231
The benefit of grinding may be claimed only

for welded joints, which are adequately The curves for base material assume that
protected from seawater corrosion. The flame-cut surfaces and otherwise rough sur-
benefit of grinding should usually not be faces are ground. The curves for welded
taken into account during design, as grind- joints assume welds proven free from sig-
ing is considered as the only improvement, nificant defects.
which can be carried out during fabrication,
to increase the fatigue strength. Table C-2 Characteristic bilinear S−N
curves for structural steel
Exter- Material N≤107 N>107
C.3 Characteristic S−N curves for nal
logKC m logKC m
condi-
structural steel tions
Air Welded joint 12.65 3.0 16.42 5.0
Characteristic S−N curves applicable for Base 12.89 3.0 16.81 5.0
normal and high strength steels have been material
defined by DNV (1998), based on interpre- Corro- Welded joint 12.38 3.0 12.38 3.0
tation of fatigue test data and an assumption sive Base 12.62 3.0 12.62 3.0
of 97.7% survival probability. The curves envi- material
ron-
are defined as bilinear curves in a logS-logN ment
diagram with a change in slope at N=107
stress cycles. Note that these curves are de- Table C-3 Characteristic one-slope S−N
fined in terms of log base 10 rather than the curves for structural steel
natural logarithm. External Material logKC m
conditions
The characteristic S−N curves given here
correspond to test results from smooth Air Welded joint 12.76 3.0
Base material 13.00 3.0
specimens with a stress concentration factor Corrosive Welded joint 12.38 3.0
1.0. For fatigue analysis of details with environ- Base material 12.62 3.0
stress concentration factors different from ment
1.0, the stress range must incorporate the
strain concentration factor. The characteristic S-N curves for steel given
in Tables C-2 and C-3 are represented
The basic equation for the characteristic S−N graphically in Figure C-2. The S-N curves
curve is given in Tables C-2 and C-3 can be used for
design of hull and casing structures, e.g.
logN=logKC−mlogS towers.
and values for logKC and m for base material Note that an S-N curve slope m=3 for base
and welded joint are given in Table C-2. material as given in Tables C-2 and C-3 may
be questionable especially for high strength
The use of a bilinear S−N curve complicates steel.
the damage calculations, cf. the closed-form
solution for the cumulative damage for Recommended characteristic S−N curves for
Weibull-distributed stress ranges quoted steel bolts are given in Appendix A.
above. In order to reduce the computational
efforts, simplified one-slope S−N curves are To account for effects of possible non-zero
given in Table C-3. mean stresses, reference is made to DNV
(1998).
232 C – Fatigue
σB Ry
C.4 Characteristic S−N curves for f (R y , σ B ) = 1 − log 10 ( )
4000 10
forged or rolled steel
For forged and rolled steel subject to fully Ry is the peak-to-peak roughness, Ry≈6Ra
reversed stress cycles about a zero-mean
stress, i.e. R=−1, where R is the ratio be- Ra=mean roughness
tween the compressive stress amplitude and
the tensile stress amplitude, the following The S−N curve slope m is expressed as
applies:
12
m= +3
The fatigue strength limit for rotating bend- β2
ing with 50% survival probability at 10 6
cycles can be calculated as
where β is the notch factor,
σD=1.25⋅(0.436σy+77)⋅f(Ry,σB)⋅f(d,r)
β = 1 + η (α − 1)
where σD and σy both have to be in units of
MPa, and in which the following definitions in which α is the stress concentration factor,
apply and the notch sensitivity η is given by
σy is the yield strength or 0.2% proof stress η = 0.62 + 0.2 log 10 r + σ y 10 − 4 log10 (
400
)
for the shaft material, related to the actual r
dimension.
The static notch factor is defined as
σB is the ultimate strength for the shaft ma-
terial, related to the actual dimension. σy 2
β m = 1 + (α − 1)( )
1000
f(d,r) is a size factor, which depends on the
shaft diameter d and the fillet radius r. For
A standard S−N curve, in principle applica-
unnotched parts, the size factor is ble to all non-welded machine steels, can
now be constructed, expressed in terms of
2
f (d , r ) = 0.8 + the above quantities, as shown in Figure C-
d 3. This standard S−N curve, which repre-
sents 50% survival probability (i.e., the 50%
and for notched shafts it is quantile in the realisations of fatigue tests),
is to be considered as a characteristic S−N
ì0.9 + 1 / r for r ≥ 10 mm curve for the material. A materials factor γm
f (d , r ) = í
î 1.0 for r < 10 mm should be applied to all stress values of this
characteristic S−N curve to get a design S−N
These expressions for the size factor require curve for use in design. For critical compo-
d and r to be given in units of mm. nents such as the main shaft, γm=1.8 should
be used. Note that the standard S−N curve in
f(Ry,σB) is a surface roughness factor, which Figure C-3 cannot be applied to case-
is calculated as hardened steels.
C – Fatigue 233
Note that the standard S−N curve gives fibre reinforced composites follow the same
number of cycles to failure N at stress am- strain-life curve, denoted ε−N curve. Stan-
plitude σ. This is in contrast to structural dard ε−N curves for design, established from
engineering, where S−N curves gives num- fatigue tests in the laboratory on a number of
ber of cycles to failure N at stress range S, different materials, can be found in Mayer
and caution should be exercised to avoid (1992).
confusion and errors. Note also that in
structural engineering, the characteristic References
S−N curve for structural steel is usually de- DNV, Fatigue Strength Analysis for Mobile
fined differently as the S−N curve, which Offshore Units, Classification Notes No.
represents 97.7% survival probability (i.e., 30.2, Det Norske Veritas, Høvik, Norway,
the 2.3% quantile in the realisations of fa- 1984.
tigue tests), and it is therefore usually com-
bined with a smaller requirement to the ma- DNV, Fatigue Assessment of Ship Struc-
terials factor γm than the 1.8 quoted above tures, Classification Notes No. 30.7, Det
for non-welded machine steel. Norske Veritas, Høvik, Norway, 1998.
Mayer, R.M., ed., Design of Composite

C.5 S−N curves for composites Structures against Fatigue, Applications to
S−N curves for composite materials such as Wind Turbine Blades, Mechanical Engi-
fibre-reinforced plastics may vary from case neering Publications Ltd., Bury St. Ed-
to case, depending on the composition of the munds, Suffolk, U.K., 1996.
material. However, many different glass
Figure C-2 Characteristic S−N curves for structural steel in hull and casing structures, from DNV Class.
Notes 30.7
234 C – Fatigue
Figure C-3 Characteristic S-N curve for non-welded, forged or rolled machine steel
C – Fatigue 235
236 C – Fatigue
D. FEM calculations Frequency analysis

Frequency analyses are used to determine
the eigen frequencies and normal modes of a
D.1 General structural part.
D.1.1 The use of FEM analyses The FEM program will normally perform an
analysis for the lowest frequencies. How-
A Finite Element analysis should be carried ever, by specifying a shift value, it is possi-
out for a structural component, when no ble to obtain results also for a set of higher
simple calculations can be performed to frequencies around a user-defined fre-
document the strength and stiffness of the quency.
component.
Note that the normal modes resulting from
The model to be included in the analysis and such analyses only represent the shape of the
the type of analysis should be chosen with deflection profiles, not the actual deflec-
due consideration of interaction with the rest tions.
of the structure.
Dynamic analysis
Because a FEM analysis is thus normally Dynamic FEM analyses can be used to de-
used when simple calculations are insuffi- termine the time dependent response of a
cient or impossible, care must be taken to structural part, e.g. as a transfer function.
assure that the model and analyses reflect The analyses are normally based on modal
the physical reality. This must be done by superposition, as this type of analysis is
evaluation of the input to as well as the re- much less time consuming than a ‘real’ time
sults from the analysis. Guidelines for such dependent analysis.
an evaluation are given below.
Stability/Buckling analysis
D.1.2 What type of analysis Stability/Buckling analysis is relevant for
Different types of analysis can be performed slender structural parts or sub-parts, as the
by means of FEM analysis, however, most load causing local or global buckling may be
analyses are usually static analyses for de- lower than the loads causing strength prob-
termination of the strength and stiffness of lems.
structures or structural components. FEM
analyses are usually computer-based analy- The analysis is normally performed by ap-
ses which make use of some FEM computer plying a set of static loads, and then the
program. factor by which this set of loads has to be
multiplied for stability problems to occur, is
Static analysis found by the program.
For static analyses the structural parts are
Thermal analysis
commonly examined with respect to which
By thermal analysis, the temperature distri-
extreme loads govern the extreme stress,
bution in structural parts are determined,
strain and deflection responses.
based on the initial temperature, heat in-
As the analysis is linear, unit loads can be
put/output, convection etc. This is normally
applied, and the response caused by single
a time-dependent analysis, however, it is
loads can be calculated. Then the actual ex-
usually not very time consuming as only one
treme load cases can be examined by linear
degree of freedom is present at each mod-
combinations - superposition.
elled node.
D – FEM calculations 237

Note that a thermal analysis set-up as men- tions, and hence the coordinate system valid
tioned here can be used to analyse analogous for the elements and boundary conditions
types of problems involving other time- should be checked, e.g. by plots. This is
dependent quantities than temperature. This important for beam elements, as it is not
applies to problems governed by the same always logical which axes are used to define
differential equation as the one which gov- the sectional properties.
erns heat transfer. An example of such an
application can be found in foundation engi- Similarly, the boundary conditions should be
neering for analysis of the temporal evolu- checked, as a wrong coordinate system for
tion of settlements in foundation soils. symmetry conditions may seriously corrupt
the results.
Various analyses
The analyses mentioned above are only For laminate elements, the default coordi-
some of the types of analyses that can be nate system is often an element coordinate
performed by FEM analysis. Other types of system, which may have as a consequence
analyses are plastic analyses and analyses that the fibre directions are distributed ran-
including geometric non-linearities. domly across a model.
Combinations of analyses can also be per- Material properties

formed. As examples of this, the results of Several different material properties may be
an initial frequency analysis can be used as a used across a model, and plots should be
basis for subsequent dynamic analyses. Also checked to verify that the material is distrib-
the result of a thermal analysis may be used uted correctly.
to form a load case in a subsequent static
analysis. Drawings are often made using units of mm,
to get appropriate values, and when the
D.2 Modelling model is transferred to the FEM program the
dimensions are maintained. In this case care
The results of a FEM analysis is normally
should be taken in setting the material prop-
documented by plots and printouts of se-
erties (and loads) correctly, as kg-mm-N-s is
lected extreme response values. However,
not a consistent set of units. It is advisable to
as the structural FEM model used can be
use SI-units (kg-m-N-s).
very complex, it is important also to docu-
ment the model itself. Even minor devia-
Material models
tions from the intention may give results that
The material model used is usually a model
do not properly reflect reality.
for isotropic material, i.e. the same proper-
ties in all directions. Note, however, that for
composite materials an orthotropic material
D.2.1 Model
model has to be used to reflect the different
The input for an FEM model must be docu- material properties in the different direc-
mented thoroughly by relevant printouts and tions. For this model, material properties are
plots. The printed data may preferably be defined for three orthogonal directions. By
stored or supplied as files on a CD-rom. definition of this material, the choice of co-
ordinate system for the elements has to be
Coordinate systems made carefully.
Different coordinate systems may be used to
define the model and the boundary condi-
238 D – FEM calculations

D.2.2 Elements some programs the thickness can be shown

by 3D views.
For a specific structural part, several differ-
ent element types and element distributions
The stresses at connections such as welds
may be relevant depending on the type of
cannot be found directly by these elements
analysis to be carried out. Usually, one par-
either.
ticular element type is used for creation of a
FEM model.
3D
By the use of solid elements the correct ge-
However, different element types may be
ometry can be modelled to the degree of
combined within the same FEM model. For
detail wanted. However this may imply that
such a combination special considerations
the model will include a very large number
may be necessary.
of nodes and elements, and hence the solu-
tion time will be very long. Furthermore, the
Element types
mesh for a solid model may need to be more
dense than for a beam or shell element
1D
model, as most solid element types only
Models with beam elements are quite simple
have three degrees of freedom at each node.
to create and provide good results for frame
like structures.
Combinations
Combinations of the three types of elements
One difficulty may be that the sectional
are possible, but as the elements may not
properties are not visible and hence the input
have the same number of degrees of free-
should be checked carefully for the direction
dom at each node, care should be taken not
of the section and the numerical values of
to create unintended hinges in the model.
the sectional properties. Some FEM pro-
grams can generate 3D views showing the
Beam elements have 6 degrees of freedom
dimensions of the sections, and this facility
in each node – three translations and three
should be used if present.
rotations, while solid elements normally
only have 3 – the three translations. Shell
Naturally, the stresses in the connections
elements normally have 5 degrees of free-
cannot be calculated accurately by the use of
dom – the rotation around the surface nor-
beam elements only.
mal is missing. However, these elements
may have 6 degrees of freedom, but then the
2D
stiffness for the last rotation is fictive.
Shell and plate elements should be used for
parts consisting of plates or constant thick-
Connecting beam or shell elements to solid
ness sub-parts. The wall thickness need not
elements in a point, respectively a line, in-
to be very thin to obtain a good representa-
troduces a hinge. This problem may be
tion by such elements, as shell elements
solved by adding additional ‘dummy’ ele-
suitable for thick plates exist, and these ele-
ments to get the correct connection. Alter-
ments include the desired behaviour through
natively constraints may be set up between
the thickness of the plate.
the surrounding nodal displacements and
The same problems as for beam elements
rotations. Some FEM programs can set up
are present for shell elements as the thick-
such constraints automatically.
nesses of the plates are not shown. The
thickness can, however, for most FEM pro-
grams be shown by colour codes, and for

Element size/distribution try to improve the element quality by redis-

The size, number and distribution of ele- tribution of the nodes.
ments required in an actual FEM model de-
pend on the type of analysis to be performed The quality of the elements should always
and the type of elements used. be checked for an automatically generated
mesh, in particular for the internal nodes and
Generally, as beam and shell element have 5 elements. For manually generated mesh it is
or 6 degrees of freedom in each node, good usually possible to generate good quality
results can be obtained with a small number elements.
of elements. As solid elements only have 3
degrees of freedom in each node, they tend For high-order elements generated automati-
to be more stiff, and hence more elements cally, care should be taken to check that the
are needed. nodes on the element sides are placed at the
surfaces of the model and not just on the
Furthermore, the shape and order of the linear connection between the corner nodes.
elements influence the required number of This problem often arises when linear ele-
elements. Triangular elements are more stiff ments are used in the initial calculations, and
than quadrilateral elements, and first-order the elements are then changed into higher
elements are more stiff than second-order order elements for a final calculation.
elements.
Benchmark tests to check the element qual-
This can be seen from the example below, in ity for different element distributions and
which a cantilever is modelled by beam, load cases are given by NAFEMS. These
shell/in-plane and solid elements. tests include both beam, shell and solid ele-
ments, as well as static and dynamic loads.
Element quality
The results achieved by a certain type and D.2.3 Boundary conditions
number of elements depend on the quality of
Definition of boundary conditions
the elements.
The boundary conditions applied to the
model should as a matter of course be as
Several measures for the quality of elements
realistic as possible. This may require that
can be used, however, the most commonly
the FEM model becomes extended to in-
used are aspect ratio and element warping.
clude element models of structural parts
other than the particular one to be investi-
The aspect ratio is the ratio between the side
gated. One situation where this comes about
lengths of the element. This should ideally
is when the true supports of a considered
be equal to 1, but aspect ratios of up to 3-5
structure have stiffnesses which cannot be
do usually not influence the results and are
too well defined unless they are modelled by
thus acceptable.
means of elements that are included in the
FEM model.
Element warping is the term used for non-
flatness or twist of the elements. Even a
When such an extended FEM model is
slight warping of the elements may influ-
adopted, the deviations from the true stiff-
ence the results significantly.
nesses at the boundary of the structural part
in question may then become minor only.
Most available FEM programs can perform
This as the nonrealistic effects due to inade-
checks of the element quality, and may even
quately modelled boundary conditions are

transferred further away to the neighbouring these types of symmetry can normally be
structural parts or sub-parts, which are now defined in an easy manner in most FEM
represented by elements in the extended programs by using appropriate coordinate
FEM model. systems.
Types of restraint The loads for a symmetric model may be a

The types of constraints normally used are combination of a symmetric and an anti-
constrained or free displacements/rotations metric load. This can be considered by cal-
or supporting springs. Also a fixed non-zero culating the response from the symmetric
displacement/rotation may be specified, or loads for a model with symmetric boundary
contact, i.e. the displacement is restrained in conditions, and adding the response from the
one direction but not in the opposite direc- antimetric loads for a model with antimetric
tion. boundary conditions.
The way that a FEM program handles the If both model and loads have rotational
fixed boundary condition may vary from one symmetry, a sectional model is sufficient for
program to another. One approach is to re- calculating the response.
move the actual degree of freedom from the
model, another is to apply a spring with a Some FEM programs offer the possibility to
large stiffness at the actual degree of free- calculate the response of a model with rota-
dom. The latter approach may lead to sin- tional symmetry by a sectional model even
gularities if the stiffness of the spring is if the load is not rotational-symmetric, as the
much larger than the stiffness of the element program can model the load in terms of Fou-
model. Of course the stiffness can also be rier series.
too small, which may also result in singu-
larities. D.2.4 Loads
The loads applied for the FEM calculation
An appropriate value for the stiffness of
are usually structural loads, however, cen-
such a stiff spring may be approximately 106
trifugal and temperature loads are also rele-
times the largest stiffness of the model.
vant.
The contact boundary condition requires a
Structural loads consist of nodal forces and
non-linear calculation, as the program must
moments and surface pressures. Nodal
first identify if the displacement has to be
forces and moments are easily applied, but
constrained or free.
may result in unrealistic results locally, be-
cause no true loads act in a single point.
Symmetry/Antimetry
Therefore application of loads as pressure
Another type of boundary conditions are
loads will in most cases form the most real-
symmetric and antimetric conditions, which
istic way of load application.
may be applied if the model and the loads
possess some kind of symmetry. Taking
Load application
such symmetry into account may reduce the
The loading normally consists of several
size of the FEM model significantly.
load components, and all of these compo-
nents may be applied at the same time. This
The two types of symmetry that are most
is, however, not vere rational, as a slightly
frequently used are planar and rotational
different load combination in a new analysis
symmetries. The boundary conditions for
then will require a full new calculation.

Instead, each of the load components should gravity should correspond with the expected
be applied separately as a single load case, value.
and the results found from each of the corre-
sponding analyses should then be combined. Material
In this way a large range of load combina- Several different materials can be used in
tions can be considered. To facilitate this the same FEM model. Some of these may be
procedure, unit loads should be used in the fictitious, and this should be checked by
single load cases, and the actual loads then plots showing which material is assigned to
used in the linear combinations. each element, and by listing the material
properties. Here, care should be taken to
As only one or more parts of the total struc- check that the material properties are given
ture is modelled, care should be taken to according to a consistent set of units.
apply the loads as they are experienced by
the actual part. To facilitate such load appli- Element type
cation, ‘dummy’ elements may be added, i.e. Also, several different element types can be
elements with a stiffness representative of used, and here also plots and listing of the
the parts which are not modelled – these are element types should be presented.
often beam elements. The loads can then be
applied at the geometrically correct points, Local coordinate system
and then transferred via the beam elements For beam and composite elements, the local
to the structural part being considered. coordinate systems should be checked, pref-
erably by plotting the element coordinate
D.3 Documentation systems.
D.3.1 Model
Loads – Boundary Conditions
The result of a FEM analysis can be docu- The loads and boundary conditions should
mented by a large number of plots and be plotted to check the directions of these,
printouts, which can make it an over- and from listings the actual numbers should
whelming task to find out what has actually be checked. Here check of the correspon-
been calculated, and how the calculations dence between plots and listings may require
have been carried out. documentation of node/element numbers
and coordinates.
The documentation for the analysis should
clearly document which model is consid- The above listed aspects can and should be
ered, and the relevant results should be checked prior to performing the analyses.
documented by plots and printouts. Possibilities of checking the results – be-
sides comparison with hand calculations –
Geometry control also exist in the various available programs.
A check of the geometric model by checking
the dimensions is an important and often Reactions
rather simple task. This simple check may The reaction forces and moments are nor-
reveal if numbers are unintentionally input mally calculated by the FEM programs and
wrongly. should be checked, at least it should be
checked that the total reaction corresponds
Mass – Volume – Centre of Gravity with the applied loads. This is especially
The mass/volume of the model should al- relevant when loads are applied to areas and
ways be checked. Similarly, the centre of volumes, and not as discrete point loads.

For some programs it is possible to plot the same interpolation functions as are used for
nodal reactions, which can be very illustra- displacements in defining the element stiff-
tive. nesses.
A major reason for choosing a FEM analysis Another way of getting an indication of the
as the analysis tool for a structure or struc- error in the stresses, is comparison of the
tural part is that no simple calculation can be nodal stresses calculated at a node for each
applied for the purpose. This implies that of the elements that are connected to that
there is no simple way to check the results. node. Large variations indicate that the mesh
Instead checks can be carried out to make should be more dense.
probable that the results from the FEM
analysis are correct. Load combinations
If the results are found as linear combina-
Mesh refinement tions of the result from single load cases, the
The simplest way of finding out if the pres- load combination factors should of course be
ent model or mesh is dense enough is to clearly stated.
remesh the model with a more dense mesh,
and then calculate the differences between Displacement
analysis results from use of the two meshes. The global deflection of the structure should
This procedure can, however, be very time be plotted with appropriately scaled deflec-
consuming, as several meshes may have to tions. For further evaluation deflection com-
be created and tried out. Also, this procedure ponents could be plotted as contour plots to
may in some cases result in too dense se the absolute deflections.
meshes, as modelling simplifications can
induce unrealistic behaviour locally. For models with rotational symmetry a plot
of the deflection relative to a polar coordi-
Instead an indication of whether the model nate system may be more relevant for
or mesh is sufficient would be preferable. evaluation of the results.
D.3.2 Results Stress

All components of the stresses are calculated
Unrealistic results
and it is should be possible to plot each
Initially, the results should be checked to see
component separately to evaluate the calcu-
if they seem realistic. A simple check is the
lated stress distribution.
evaluation of the deflection of the compo-
nent, which of cause should reflect the load
The principle stresses could also be plotted
and boundary conditions applied and the
with an indication of the direction of the
stiffness of the component. Also, the stresses
stress component, and these directions
on a free surface should be zero.
should be evaluated in relation to the ex-
pected distribution.
Error estimates
Most commercial FEM programs have some
Strain
means for calculation of error estimates.
Similarily to the stresses the components of
Such estimates can be defined in several
the strains and the principle strain could be
ways, but one of the most commonly used
plotted for evaluation of the calculated re-
estimates is an estimate of the error in the
sults.
stress. The estimated ‘correct’ stress is
found by interpolating the stresses by the


H. Tables and conversions Fahrenheit temperature = Celsius tempera-

ture times 1.8 plus 32
H.1 English/metric conversion H.2 Air density vs. temperature
Conversion between US units and metric The density ρ of air in units of kg/m3 can be
units expressed as
Metric units US units 3.53 ⋅10 4
LENGTH ρ=
0.3048 m 1 foot 273.15 + T
1m 3.281 feet
in which T is the temperature in Celsius.
1m 39.37 inches
2.540 cm 1 inch
H.3 Air density vs. height
1 km 0.621 mile
1.609 km 1 mile The change ∆ρ in air density for an increase
AREA ∆h in the elevation above sea level can be
1 km2 0.3861 square mile expressed as
2.590 km2 1 square mile
1 m2 10.76 square feet ∆ρ
= −1.194 ⋅10 − 4 kg/m3.
0.093 m2 1 square foot ∆h
VOLUME
1 m3 353 cubic feet H.4 Rayleigh wind distribution
0.093 m3 1 cubic foot
CAPACITY The long-term distribution of the 10-minute
1 liter 0.0353 cubic feet mean wind speed U10 at a site is usually well
represented by a Weibull distribution
28.33 liter 1 cubic foot
1 liter 61.02 cubic inches
u k
0.01639 liter 1 cubic inch FU (u ) = 1 − exp(−( ) )
1 liter 0.2642 US gallon
10
u0
3.7854 liter 1 US gallon
MASS and WEIGHT in which u0 and k are distribution parame-
1 metric ton 1.1 tons ters. When k=2, which is often a good ap-
1 kg 2.2046 lbs proximation for wind speed data, the distri-
0.4536 kg 1 lb bution becomes a Rayleigh distribution,
1 kN 0.2247 kips
4.45 kN 1 kip u 2
FU (u ) = 1 − exp(−( ) )
1 kPa 0.000147 ksi 10
u0
6804 kPa 1 ksi
1 kPa 0.147 psi and the following relationship exists be-
6.804 kPa 1 psi tween u0 and the mean value E[U10] in the
long term:
Temperature conversion:
E [U 10 ]
2
Celsius temperature = Fahrenheit tempera- u0 =
ture times 5/9 minus 32 π
H – Tables and conversions 245

Let the cumulative probability be denoted p.

The corresponding value of the wind speed,
u, can then be found by reading as
u = u0 ⋅ x
where x is tabulated in Table H-1.
Table H-1 Quantiles of stan-

dard Rayleigh distribution
Probability p Quantile x
0.0001 0.0100
0.001 0.0316
0.002 0.0447
0.005 0.0708
0.01 0.1003
0.1 0.3246
0.2 0.4724
0.3 0.5972
0.4 0.7147
0.5 0.8326
0.6 0.9572
0.7 1.0973
0.8 1.2686
0.9 1.5174
0.99 2.1460
0.995 2.3018
0.998 2.4929
0.999 2.6283
0.9999 3.0349
246 H – Tables and conversions

Guidelines for Design of Wind Turbines - DNV/Risø
Index
leading edge ..................4, 48, 61, 62, 73
Accessibility .......2, 27, 110, 150, 154, 159 loads ........................................... 89, 118
Acetal (POM) ...................................... 166 manufacturing .................................. 104
Active brakes ....................................... 147 material 56, 78, 103, 104, 110, 111, 112,
Aerodynamic 115
coefficients ....................... 58, 63, 68, 69 mortars ............................................. 109
damping ..................56, 72, 89, 116, 180 natural frequency ................71, 113, 116
data extrapolation ............................... 69 number ....................................... 1, 3, 57
modelling ..................................... 59, 61 pitch angle ...............................4, 67, 103
Aeroelastic calculation 58, 68, 72, 171, 178 pressure side ............................ 103, 151
Air principal axes ............................. 66, 104
density ........................................ 46, 245 profile .....................56, 65, 66, 103, 104
kinematic viscosity .......................... 179 suction side ...................................... 103
Alloyed steel ........................ 125, 139, 144 tests .....24, 105, 106, 107, 113, 115, 116
Amplitude decay ratio ............................ 70 tip angle ................................... 118, 119
Annual energy production ....................... 4 trailing edge ......................4, 61, 64, 111
Asynchronous generator 3, 153, 155, 156, twist ......................................56, 72, 103
221, 222 web ...................................103, 104, 111
Atmospheric conditions ................... 31, 38 Blocking mechanisms .............27, 163, 225
Atmospheric pressure ............................ 46 Bolts
Autocorrelation function .................. 37, 38 connections 110, 118, 119, 184, 225, 227
fatigue ...................................... 180, 227
Beam theory 66, 103, 131, 178, 183, 198, pretension ............................26, 127, 225
239 quality classes .................................. 227
Bearings Boundary layer theory ..................... 34, 62
alignment ......................................... 130 Bowen ratio ........................................... 39
basic dynamic capacity .................... 137 Brakes
clearance .......................................... 134 aerodynamic ..........................13, 58, 110
friction torque moment .................... 165 callipers .................................... 147, 163
gear .................................................. 136 disc ........................................... 147, 163
lubrication .........132, 137, 157, 164, 166 electrical .................................. 157, 163
main bearing .................... 117, 124, 128 mechanical ............13, 87, 147, 149, 155
rating life .................................. 133, 136 pads .......................................... 147, 149
rolling bearings .133, 137, 164, 166, 167 Braking loads ......................................... 58
seals ................................. 131, 132, 166 Braking systems ..11, 12, 13, 147, 151, 157
slide bearings ................... 164, 166, 167 Brittle fracture 17, 20, 45, 107, 166, 184,
spherical roller bearing .... 128, 131, 136 227
tight fits .................................... 123, 124 Bulk temperature ................................. 141
Bevel gears .................................. 135, 136
Blades Cable twist ....................3, 11, 12, 159, 167
adhesives .................................. 105, 109 Cathodic protection ...................... 210, 211
aspect ratio ......................................... 63 Caughey series ....................................... 70
buckling ....................110, 111, 112, 115 Centrifugal forces ...................56, 118, 153
damping ..56, 69, 70, 71, 73, 74, 89, 116 Ceramic brake pads ............................. 149
element momentum method ... 61, 64, 68 Certification ..................................... 26, 28
geometry ...................................... 4, 103 Characteristic chord length .............. 73, 89
Index 247
Characteristic values ........................ 10, 18 Cyclones .......................................... 31, 42

Charnock’s formula ............................... 32
Clay 188, 193, 195, 197, 198, 199, 201, Damage accumulation ................75, 81, 82
202, 206, 211, 215 Damping
Coating 48, 105, 106, 110, 157, 209, 210, aerodynamic ............56, 72, 89, 116, 180
211, 217, 218 negative ........................................ 58, 71
Coefficient of variance ................... 80, 174 structural ...............56, 69, 70, 71, 72, 89
Coherence model ............................. 38, 60 Darrieus ................................................... 2
Combination of loads Davenport ...........................37, 60, 85, 173
extreme ...........10, 19, 99, 179, 237, 243 Delamination ................107, 113, 115, 116
fatigue .......................... 79, 91, 121, 179 Design codes ........................See Standards
Common-cause failures ............. 13, 15, 16 Design lifetime .......10, 112, 158, 181, 182
Computational fluid dynamics ............... 63 Design parameters ................................. 20
Concepts Design situations .......................... 9, 10, 53
downwind ............................................ 2 Detail categories ...................180, 181, 182
horizontal axis .............2, 3, 28, 159, 164 Deterministic method .................... 19, 100
upwind ............................................. 1, 2 Diffraction theory .................................. 98
vertical axis .......................................... 2 Dismantling ..............2, 127, 211, 212, 218
Conductor cable ........................... 114, 115 Displacement transducers .................... 115
Cone penetrometer test (CPT) ............. 188 Drag 48, 56, 58, 62, 63, 69, 72, 73, 74, 88,
Configurations 95, 96, 97, 173
clusters ......................................... 6, 194 Drivetrain .........................See transmission
parks ..................................................... 6 Ductility ..................................20, 119, 121
stand-alone ........................................... 6 Duhamel’s integral ................................ 49
wind farms ............................... 6, 35, 54 Dynamic amplification factor ...... 208, 216
Control system ....................... 7, 10, 25, 27 Dynamic load rating ...............26, 133, 137
modelling ........................................... 67 Dynamic response ...................69, 149, 175
parameters .................................... 11, 67 Dynamic stall ................................... 72, 74
Controlled idling ...................... 12, 13, 156
Conversions ......................................... 245 Earthquake ..........48, 49, 99, 201, 202, 203
Cooling system .............150, 155, 156, 225 Eccentric loading ................................. 193
Coordinate system ............... 131, 161, 238 Economy .....................................4, 6, 9, 19
Coriolis forces ........................................ 56 Edgewise vibrations ............58, 71, 72, 116
Coriolis parameter ................................. 41 Effective foundation area ..................... 190
Corrosion protection 48, 110, 132, 166, Efficiency .................1, 2, 5, 6, 67, 87, 177
184, 209, 217, 229 Eigen frequency ...... See Natural frequency
Couplings 121, 123, 124, 127, 142, 146, Embedded foundations ........................ 192
147 Emergency stop ..................................... 28
Coupon test .................................. 145, 146 Endurance limit .................................... 140
Crack Environmental conditions 28, 110, 131, 213
growth .......................... 25, 75, 229, 231 EP additives ..........................132, 133, 138
inspection ................................. 146, 211 Epicyclic gears ..................................... 135
monitoring ........................................ 217 Equivalent loads .....................80, 164, 165
Cumulative damage 22, 23, 75, 81, 113, Euler force ........................................... 183
118, 229, 230, 232 External conditions .......27, 31, 45, 53, 229
Cut-in current ....................................... 222 Extreme value analysis .........41, 75, 83, 85
Cut-in wind speed .................................. 79 Extreme value distribution ..........42, 83, 84
248 Index
Extreme winds ....................................... 41 Frictional velocity .......................31, 32, 39
Fabrication ..............24, 207, 211, 229, 232 Gamma function ...............22, 34, 111, 230
Fail-grace ................................................. 9 Gear
Fail-safe ....................12, 13, 150, 155, 157 design contact stress ........................ 139
Failure mode analysis .......... 14, 16, 17, 53 gearless ................................................ 7
Failure probability ..........9, 17, 18, 25, 137 gray staining .............................. 25, 141
Fatigue heat treatment .......................... 138, 144
design life ................10, 25, 79, 137, 229 helical .............................................. 135
equivalent loads ..........80, 137, 165, 173 housing ............................................ 142
load spectrum ............................... 80, 89 hypoid .............................................. 135
loads ............................. 75, 79, 126, 170 installation ....................................... 143
S-N curve 22, 75, 80, 112, 125, 208, 227, materials ................................... 138, 143
229 micro pitting .....................138, 142, 144
Fault tree analysis .................................. 14 scuffing .................................... 141, 145
Feasibility assessment ............................ 19 surface durability ..............138, 139, 147
FEM ................ See Finite element Method tooth form ................................ 135, 140
Fibre reinforcements ......103, 105, 107, 234 Generator
Filtering ............................... 133, 151, 157 asynchronous 3, 153, 155, 156, 221, 222
Finite Element Method ............ 58, 68, 237 multiple-poled .................................. 157
Fire ......................................... 15, 150, 225 overload ......................12, 116, 156, 158
First-order reliability method ........... 21, 23 reactive power .......................... 221, 222
Flange connections .......118, 180, 184, 227 size ....................................................... 4
Flange coupling ................................... 146 slip ................................................... 153
Flapwise vibrations ................................ 71 synchronous ......................153, 156, 157
Flash temperature criterion .................. 141 variable speed .....................67, 155, 172
Flex4 ...................................................... 68 Geological study .................................. 187
Flicker .................................................... 13 Geometrical size effect ................ 122, 123
Flutter ............................................... 67, 73 Geophysical survey .............................. 187
Foundation Geostrophic wind speed ................... 38, 41
bearing capacity factor ..................... 192 Geotechnical analysis ...187, 188, 214, 215
friction angle .............188, 197, 199, 200 Germanischer Lloyd .............................. 28
gravity base ...................... 187, 188, 206 Glauert’s correction ............................... 65
inclination factors .................... 192, 193 Goodman formulation ............................ 75
monopile ............................ 98, 187, 206 Gravity loads ..............56, 87, 90, 113, 118
shape factor ...................................... 193 Grid connection ............................... 13, 79
shear strength parameters ......... 188, 192 Grid loss ............................................... 149
skin friction .......194, 196, 197, 212, 215 Gumbel distribution ..21, 42, 83, 85, 86, 89
spring stiffness ......................... 201, 203 Gust .................40, 41, 53, 55, 88, 153, 179
stability ............................................ 189 Gyroscopic effects ....................56, 57, 162
stiffness .....177, 188, 189, 200, 202, 203
tripod ................................ 187, 206, 212 Hail ........................................................ 48
Four-point ball bearing ................ 164, 166 Harris spectrum ...................................... 36
Frechet distribution ................................ 34 HawC ............................................... 59, 68
Frequency analysis ..........14, 216, 237, 238 High-cycle fatigue ........................... 75, 77
Frequency converter .....155, 156, 157, 222 Homogeneous terrain 31, 34, 35, 37, 59, 60
Fretting corrosion ................................... 25 Humidity ...........................40, 47, 110, 154
Index 249
Hurricanes .............................................. 42 Markovian cumulative damage model ... 81

Material certificates ................10, 105, 144
Ice ............................................ 31, 47, 211 Mean stress 75, 77, 78, 80, 112, 113, 118,
cone .................................. 207, 213, 216 124, 125, 232
loads 47, 58, 99, 207, 208, 211, 213, 216 Measurements ........................................ 46
Idling .....12, 13, 14, 53, 54, 55, 78, 79, 156 damping ....................................... 69, 71
Inertia full-scale test ...................................... 24
coefficient .......................................... 96 load .................................................. 170
loads ....................................... 56, 96, 97 model verification .................60, 68, 175
Inflow angle ............62, 63, 64, 69, 74, 167 uncertainty ................................... 25, 43
Inspection ................................. 10, 11, 181 wind speed ..............................35, 43, 44
intervals ............................. 24, 182, 229 Micro pitting ................................See Gear
Installation 27, 53, 143, 180, 194, 198, 206, Middelgrunden ................................. 97, 98
209, 210, 217, 225 Miner .........................See Palmgren Miner
Insulation ........................26, 154, 157, 158 Minimal cut set analysis ........................ 16
Integral temperature criterion .............. 141 Modal analysis ..........................58, 68, 171
Monin-Obukhov length ................... 39, 40
JONSWAP spectrum ............................. 94 Monitoring 4, 11, 12, 13, 133, 142, 151,
217, 222
Kaimal spectrum .................. 36, 37, 60, 89 Morison’s equation ................................ 95
Labour safety ................... 26, 27, 185, 225 NACA profiles ....................................... 68
Labyrinth seals ............................. 131, 132 Natural frequency 71, 88, 91, 113, 116, 171,
Lift 178, 203, 215, 216, 237
coefficients ..........62, 63, 64, 69, 74, 159 Near-coastal locations ............................ 32
curve ...................................... 63, 68, 73 Network connection ............11, 12, 55, 221
forces ............................. 58, 61, 62, 103 Nitrided Gears ..................................... 146
Light conditions ..................................... 27 Noise .................................1, 3, 6, 132, 159
Lightning ............26, 49, 50, 114, 115, 221 Nominal power ..........................See Power
Limit states 10, 16, 17, 18, 21, 23, 25, 54, Non destructive testing ................ 128, 211
118, 207, 208, 214, 215, 216 Notch sensitivity factor ........................ 123
Load distribution .................. 23, 24, 89, 90
Load verification .............. 93, 96, 170, 175 Offshore turbines 6, 31, 47, 48, 184, 187,
Logarithmic decrement .............. 69, 70, 71 189, 206
Longuet-Higgins distribution ................. 95 Operation range ..................................... 11
Low-cycle fatigue .................... 75, 77, 127 Operation temperature ................... 45, 121
Lubrication 25, 131, 137, 138, 142, 166, Operational conditions ................53, 58, 78
225 Operational procedures .................. 27, 225
grease ....................................... 132, 166 Overconsolidation ratio ............... 197, 201
oil ..................................... 133, 138, 142 Overloading ......................................... 156
relubrication interval ................ 133, 166 Overspeed .........................11, 12, 156, 158
viscosity ratio ........................... 137, 138
Palmgren-Miner rule 22, 75, 113, 126, 181,
Main shaft ...............87, 121, 131, 134, 233 182, 208, 229, 230
Mann model ..................................... 60, 61 Parametrized load spectra ...................... 89
Manual operation ........................... 54, 160 Parks .................................See Wind farms
Manuals ......................................... 27, 225
250 Index
Partial safety factors 10, 18, 19, 20, 21, 22, Rain .........................................47, 116, 159
23, 24, 100, 110, 111, 112, 121, 162, 164, Rain-flow counting method ....75, 112, 179
170, 173, 177, 181, 182, 214, 215 Range counting method ................... 76, 77
calibration .............................. 19, 20, 21 Rayleigh damping model ................. 70, 71
Passive brakes ...................................... 147 Rayleigh distribution ..78, 94, 95, 245, 246
Peak counting method ...................... 76, 77 Recurrence period .........10, 41, 45, 89, 170
Peak-over-threshold method .................. 89 Redundant design ...13, 149, 151, 152, 167
Periodical maximum method ................. 89 Relative humidity .......................... 47, 154
Personal safety ...............See Labour safety Reliability
Pile foundation 187, 189, 193, 203, 205, analysis .......................................... 9, 16
206 index .....................17, 18, 19, 20, 21, 23
Pile refusal ................................... 210, 211 Resins ........................................... 106, 109
Pile-driving ...194, 207, 208, 210, 213, 217 Return period ......................................... 42
Pile-soil-pile interaction ............... 194, 198 Reynolds number ................................... 62
Pitch angle ................................. 4, 67, 103 Richardson number ................................ 39
Pitch control ............. See Power regulation RIX number ..................................... 44, 45
Planetary gears ..................................... 135 Rotating bending .................................. 233
Polyamide (PA) ................................... 166 Rotor coning .................................. 56, 159
Polyethylene Terephtalate (PET) ......... 166 Rotor loads ..................................57, 87, 91
Polyurethane (PUR) ............................. 166 Rotor speed
Power fixed ........................................1, 87, 171
electrical ............................... 5, 152, 221 variable ...................1, 67, 171, 176, 222
iso-power curve ................................... 4 Roughness parameter 31, 32, 33, 34, 36, 41
mechanical ......................................... 67 Rødsand ........................................... 97, 98
nominal ...................................... 11, 156
performance ............................... 72, 222 S−N curves 22, 23, 75, 78, 80, 112, 208,
rated ......................................... 4, 6, 153 227, 229
wind ..................................................... 5 Safety
Power regulation .................. 3, 13, 67, 153 classes ...............................9, 12, 20, 215
active stall ............................................ 4 electrical .......................................... 221
pitch ............................................. 4, 155 philosophy ..................................... 9, 28
stall ......................................... 4, 61, 153 requirements ...............9, 10, 19, 26, 127
Power spectral density ....36, 37, 38, 72, 94 system .............9, 10, 150, 151, 154, 222
Prandtl’s tip loss factor .......................... 65 Safety factors ...... See Partial safety factors
Probabilistic method 21, 22, 24, 26, 81, 99, Saffir-simpson ....................................... 42
171 Sand ......150, 196, 197, 199, 202, 210, 215
Protection system ................. 9, 11, 13, 154 Scarf joints ........................................... 227
P-y curves ............................ 197, 198, 200 Scour protection ................................... 211
Scuffing ............................................... 141
Quality Seals ......................................131, 132, 166
assurance ............................ 28, 121, 128 Sensors ................7, 12, 153, 160, 167, 223
control .........24, 105, 127, 140, 143, 144 anemometer ............................... 60, 223
Quarter chord point ................................ 73 cable twist ........................................ 167
Quasi-static method ............................... 88 direction ........................................... 160
Quasi-steady analysis ............................. 72 Service schedule .................................. 225
Serviceability limit state ...........10, 17, 216
Radiation ................................................ 47 Shear modulus ..............197, 201, 202, 203
Index 251
Short-circuit ................................. 156, 158 standards .......19, 45, 128, 180, 182, 184
Shrink fits .............121, 123, 124, 127, 147 Student’s t distribution ..................... 81, 84
Shutdown ..........12, 14, 28, 54, 55, 79, 142 Surface conditions ....................... 118, 122
Site assessment ...................................... 42 Surface roughness ................................ 122
Slab foundation .................................... 187
Slewing bearing ................................... 164 Tale vane ............................................. 159
service life curve .............................. 165 Technological size effect ..................... 122
Slide bearing ........................................ 166 Teetering Hub .......................................... 3
Slip rings .................................................. 3 Temperatures ..............................10, 45, 46
Snow ................................................ 47, 48 Terrain
Soil conditions ................. 48, 93, 187, 189 complex ....................35, 37, 44, 93, 178
Poisson’s ratio .................................. 202 homogeneous ........31, 34, 35, 37, 59, 60
skin friction .............................. 195, 196 roughness parameter .................... 31, 33
sliding resistance .............................. 193 types ................................................... 31
stratigraphy ...................................... 188 Tests ..............................See Measurements
Solar radiation ........................................ 47 The Danish approval scheme ............... 155
Spoilers .................................................. 13 Thermal analysis .................................. 237
Spur gears ............................................ 135 Tip angle ...................................... 118, 119
Stall angle .............................................. 62 Tip brake ................................... See Brakes
Stall regulation ......... See Power regulation Tip speed ratio ................................... 6, 67
Stall strips ........................................ 73, 74 Tolerances 26, 114, 119, 124, 127, 134,
Stall-induced vibrations ......................... 71 158, 163, 164, 178, 184, 185, 217
Standard components ........................... 160 Topography .........................40, 44, 50, 188
Standards 18, 19, 20, 28, 111, 128, 142, Tower
152, 192 blade clearance .....................3, 114, 177
DIBt Richtlinien ................................. 28 buckling ..................................... 49, 182
DS 412 ........................45, 180, 182, 184 connections .......................184, 187, 209
DS 472 9, 13, 28, 34, 36, 45, 53, 54, 89, cost ........................................... 169, 170
90, 111, 112, 160, 173, 175, 176 damping ..................................... 69, 177
IEC61400-1 13, 28, 34, 36, 53, 55, 177, door .................................................. 182
221 erection .............................170, 180, 187
NVN11400-0 ..................................... 28 guided .............................................. 170
Start-up .......54, 78, 79, 113, 133, 155, 177 height ....................6, 169, 180, 183, 184
Stop ring .............................................. 134 imperfection ............................. 183, 184
Storms .............................................. 42, 98 lattice ................................169, 170, 200
Strain gauges ........................................ 115 loads ..................21, 28, 58, 88, 170, 178
Stress concentration factors 118, 122, 123, natural frequency ..............175, 177, 203
124, 126, 181, 182, 208, 209, 229, 232 relative slenderness ratio .................. 183
Stress range distribution 22, 23, 77, 78, 112, shadow ..............................2, 54, 58, 171
208, 229, 230 tubular ...................................... 169, 178
Strouhal number ................................... 180 vortex shedding .......................... 58, 180
Structural wake ..................................................... 3
damping ................56, 69, 70, 71, 72, 89 welds ................................................ 180
modelling ........................................... 66 Transient wind conditions ................ 40, 54
redundancy ........................................... 9 Transmission system 25, 67, 117, 130, 146,
reliability analysis .17, 18, 19, 20, 21, 22 152
safety .....................10, 16, 17, 19, 25, 26 Transportation .....26, 27, 53, 169, 170, 225
252 Index
Trends ............................................ 1, 7, 93 climate 1, 31, 35, 78, 79, 80, 99, 100, 229
Turbulence ................24, 32, 33, 60, 67, 88 data .............................................. 41, 44
intensity ........................... 34, 35, 60, 78 resource ...................................... 13, 169
length scale .................................. 60, 61 shear ..........................38, 40, 53, 55, 178
model ........................................... 36, 60 synthetic field .................................... 60
scale parameter .................................. 36 tunnel ......................................62, 68, 74
simulation .......................................... 41 Wind farms .................................. 6, 35, 54
wake ................................................... 35 Wind speed
Turkstra’s rule ...................................... 100 10-minute mean 31, 35, 41, 42, 43, 44,
Two-dimensional flow ........................... 61 245
Type approval ...................................... 105 50-year ....................................32, 41, 83
T-z curves .................................... 196, 215 distribution ....................................... 173
extreme .............................41, 44, 55, 69
US units ............................................... 245 gust .............40, 41, 53, 55, 88, 153, 179
nominal stall ...................................... 90
Veers model ..................................... 60, 61 profile 31, 32, 38, 39, 40, 44, 53, 55, 59,
Velocity pressure ............................. 41, 42 178
Viscosity classes .......................... 142, 143 standard deviation ...................32, 43, 60
Von Karman’s constant ..31, 32, 34, 41, 42 Wings ........................................See Blades
Vortex generators ............................. 73, 74 Working areas ........................................ 27
Vortex-induced vibrations ................... 180 Worm gears ......................................... 135
Wöhler curve ....................... See S-N curve
Wake effect ...................................... 35, 54
Wasp Engineering ............................ 44, 93 Yaw
Wave loads .................93, 98, 99, 207, 208 active .......................3, 87, 159, 160, 162
Waves .....31, 32, 48, 93, 99, 201, 203, 209 brake ...................................27, 147, 163
Weak grids ............................................. 13 error 11, 54, 69, 79, 113, 159, 160, 161,
Wear .........................13, 25, 132, 133, 158 162, 163, 167
Weibull distribution 22, 24, 31, 34, 44, 78, loads ........................................... 58, 160
79, 91, 93, 99, 111, 229, 230, 245 passive ..................................3, 159, 162
scale parameter ...22, 24, 31, 32, 91, 174 system .............2, 57, 147, 158, 159, 225
shape parameter ........................... 31, 91 Yield strength 21, 123, 125, 127, 147, 227,
Welds ............180, 182, 211, 231, 232, 239 233
Wind Young’s modulus ..................139, 166, 203
Index 253

GUIDELINES For Design of Wind Turbines PDF

Transféré par

Informations du document

Titre original

Copyright

Formats disponibles

Partager ce document

Partager ou intégrer le document

Options de partage

Avez-vous trouvé ce document utile ?

Ce contenu est-il inapproprié ?

Droits d'auteur :

Formats disponibles

GUIDELINES For Design of Wind Turbines PDF

Transféré par

Droits d'auteur :

Formats disponibles

Guidelines for

The front-page picture is from Microsoft Clipart Gallery ver. 2.0.

Printed by DSI Grafisk Service, Denmark 2001

Preface quirements that need to be fulfilled in order

Guidelines for Design of Wind Turbines

0. Contents 2.5.3 Normal operation .................. 27

6.4.4 Torsionally Elastic Couplings. 6.10.8 .. Special design considerations

8.3 GRAVITY-BASED C. FATIGUE CALCULATIONS ....... 229

A. BOLT CONNECTIONS................. 227

1. Wind Turbine Concepts operating at constant rotor speed have been

Early wind turbine designs in the 20th cen-

Modern turbines evolved from the early

1 – Wind Turbine Concepts 1

Figure 1-2 Eole C, a 4200 kW vertical axis

2 1 – Wind Turbine Concepts

1.2.4 Number of rotor blades 1.2.5 Power control aspects

1 – Wind Turbine Concepts 3

Basically, there are two approaches to power

4 1 – Wind Turbine Concepts

1.4 Efficiency aspects in which Pwind=0.5ρAV03 is the power of the

1 – Wind Turbine Concepts 5

6 1 – Wind Turbine Concepts

Also more flexible control systems can be

1 – Wind Turbine Concepts 7

8 1 – Wind Turbine Concepts

2. Safety and Reliability fill its task. Non-redundant structural parts

2 – Safety and Reliability 9

10 2 – Safety and Reliability

2 – Safety and Reliability 11

12 2 – Safety and Reliability

2 – Safety and Reliability 13

each individual component of the system for

14 2 – Safety and Reliability

Table 2-2 Rules for Combining Frequencies

Figure 2-2 Fault Tree Symbols

If quantification of the fault tree is the ob-

2 – Safety and Reliability 15

Minimal Cut Set Analysis Strongholds and Weaknesses

16 2 – Safety and Reliability

occur in a number of modes covering all on random realizations according to some

of limit state. Other types of limit states are

2 – Safety and Reliability 17

A useful byproduct of a structural reliability RC

18 2 – Safety and Reliability

Usually, a number of different load effects prescribed safety.

L R As a first step, a target reliability index βT

2 – Safety and Reliability 19

result from a new reliability analysis of the

20 2 – Safety and Reliability

2 – Safety and Reliability 21

designing to the design point values from

22 2 – Safety and Reliability

and the section modulus W is used as the

2 – Safety and Reliability 23

24 2 – Safety and Reliability

2 – Safety and Reliability 25

hardened roller bearing steel. Furthermore, quirement to structural safety shall be a

26 2 – Safety and Reliability

2 – Safety and Reliability 27

other disadvantageous lighting conditions nisms, internal electrical systems, mechani-

28 2 – Safety and Reliability

2 – Safety and Reliability 29

30 2 – Safety and Reliability