Académique Documents
Professionnel Documents
Culture Documents
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.
ISBN 87-550-2870-5
2
Guidelines for Design of Wind Turbines − DNV/Risø
Preface i
Guidelines for Design of Wind Turbines − DNV/Risø
ii Preface
Guidelines for Design of Wind Turbines − DNV/Risø
0 – Contents iii
Guidelines for Design of Wind Turbines − DNV/Risø
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
Guidelines for Design of Wind Turbines − DNV/Risø
0 – Contents v
Guidelines for Design of Wind Turbines − DNV/Risø
vi 0 – Contents
Guidelines for Design of Wind Turbines − DNV/Risø
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.
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
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 tur- bines need a counter-weight to balance the
bines in operation today have upwind rotors. rotor blade.
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
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.
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.
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
• 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
• 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.
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.
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
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
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-
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
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 31
Guidelines for Design of Wind Turbines − DNV/Risø
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
Guidelines for Design of Wind Turbines − DNV/Risø
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
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-
3 – External Conditions 33
Guidelines for Design of Wind Turbines − DNV/Risø
34 3 – External Conditions
Guidelines for Design of Wind Turbines − DNV/Risø
3 – External Conditions 35
Guidelines for Design of Wind Turbines − DNV/Risø
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
Another frequently used model for the For design it is common to relate calcula-
power spectral density is the Kaimal spec- tions 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
36 3 – External Conditions
Guidelines for Design of Wind Turbines − DNV/Risø
m j = ò ω j SU (ω )dω
with Lk=8.1λ.
0
0,1
fSu/
3 – External Conditions 37
Guidelines for Design of Wind Turbines − DNV/Risø
N eutral
which reflects the correlation length of the 1000 U nsta ble
38 3 – External Conditions
Guidelines for Design of Wind Turbines − DNV/Risø
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)
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
3 – External Conditions 39
Guidelines for Design of Wind Turbines − DNV/Risø
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).
100
downwind
upwind
ridge crest
Figure 3-9 Example of strong wind shear.
height (m)
1
0 5 10 15 20
wind speed (m/s)
40 3 – External Conditions
Guidelines for Design of Wind Turbines − DNV/Risø
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
3 – External Conditions 41
Guidelines for Design of Wind Turbines − DNV/Risø
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-
42 3 – External Conditions
Guidelines for Design of Wind Turbines − DNV/Risø
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
3 – External Conditions 43
Guidelines for Design of Wind Turbines − DNV/Risø
44 3 – External Conditions
Guidelines for Design of Wind Turbines − DNV/Risø
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-
3 – External Conditions 45
Guidelines for Design of Wind Turbines − DNV/Risø
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-
46 3 – External Conditions
Guidelines for Design of Wind Turbines − DNV/Risø
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
3 – External Conditions 47
Guidelines for Design of Wind Turbines − DNV/Risø
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.
48 3 – External Conditions
Guidelines for Design of Wind Turbines − DNV/Risø
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.
3 – External Conditions 49
Guidelines for Design of Wind Turbines − DNV/Risø
50 3 – External Conditions
Guidelines for Design of Wind Turbines − DNV/Risø
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.
3 – External Conditions 51
Guidelines for Design of Wind Turbines − DNV/Risø
52 3 – External Conditions
Guidelines for Design of Wind Turbines − DNV/Risø
4 – Loads 53
Guidelines for Design of Wind Turbines − DNV/Risø
54 4 – Loads
Guidelines for Design of Wind Turbines − DNV/Risø
4 – Loads 55
Guidelines for Design of Wind Turbines − DNV/Risø
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
56 4 – Loads
Guidelines for Design of Wind Turbines − DNV/Risø
n
M 0 = 2ω K ω å m i ri
2
i =1
4 – Loads 57
Guidelines for Design of Wind Turbines − DNV/Risø
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
58 4 – Loads
Guidelines for Design of Wind Turbines − DNV/Risø
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
Guidelines for Design of Wind Turbines − DNV/Risø
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
Guidelines for Design of Wind Turbines − DNV/Risø
4 – Loads 61
Guidelines for Design of Wind Turbines − DNV/Risø
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
Guidelines for Design of Wind Turbines − DNV/Risø
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 = 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
4 – Loads 63
Guidelines for Design of Wind Turbines − DNV/Risø
hind the trailing edge of the blade, and the dT = 4πrρV 0 a (1 − a)dr
2
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
Guidelines for Design of Wind Turbines − DNV/Risø
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
4 – Loads 65
Guidelines for Design of Wind Turbines − DNV/Risø
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
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
Guidelines for Design of Wind Turbines − DNV/Risø
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
Guidelines for Design of Wind Turbines − DNV/Risø
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
Guidelines for Design of Wind Turbines − DNV/Risø
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
Guidelines for Design of Wind Turbines − DNV/Risø
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
Guidelines for Design of Wind Turbines − DNV/Risø
4 – Loads 71
Guidelines for Design of Wind Turbines − DNV/Risø
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.
72 4 – Loads
Guidelines for Design of Wind Turbines − DNV/Risø
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).
4 – Loads 73
Guidelines for Design of Wind Turbines − DNV/Risø
Dynamic stall
The onset of stall becomes delayed when the
inflow angle is varying periodically, e.g.
according to
74 4 – Loads
Guidelines for Design of Wind Turbines − DNV/Risø
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
Guidelines for Design of Wind Turbines − DNV/Risø
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)
76 4 – Loads
Guidelines for Design of Wind Turbines − DNV/Risø
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
Guidelines for Design of Wind Turbines − DNV/Risø
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)
78 4 – Loads
Guidelines for Design of Wind Turbines − DNV/Risø
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.
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
Guidelines for Design of Wind Turbines − DNV/Risø
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
Guidelines for Design of Wind Turbines − DNV/Risø
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
4 – Loads 81
Guidelines for Design of Wind Turbines − DNV/Risø
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
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
å p (i ) = 1
i =1
t
Var [Wb ] = 2å t (1 − FW (t , b)) + E [Wb ] − E [Wb ]
b
2
t =1
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
Guidelines for Design of Wind Turbines − DNV/Risø
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
Guidelines for Design of Wind Turbines − DNV/Risø
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
84 4 – Loads
Guidelines for Design of Wind Turbines − DNV/Risø
4 – Loads 85
Guidelines for Design of Wind Turbines − DNV/Risø
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
Guidelines for Design of Wind Turbines − DNV/Risø
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
Guidelines for Design of Wind Turbines − DNV/Risø
expression is to be calculated as U 10
88 4 – Loads
Guidelines for Design of Wind Turbines − DNV/Risø
4 – Loads 89
Guidelines for Design of Wind Turbines − DNV/Risø
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
Guidelines for Design of Wind Turbines − DNV/Risø
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.
4 – Loads 91
Guidelines for Design of Wind Turbines − DNV/Risø
92 4 – Loads
Guidelines for Design of Wind Turbines − DNV/Risø
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
Guidelines for Design of Wind Turbines − DNV/Risø
(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
Guidelines for Design of Wind Turbines − DNV/Risø
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
Guidelines for Design of Wind Turbines − DNV/Risø
96 4 – Loads
Guidelines for Design of Wind Turbines − DNV/Risø
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
Guidelines for Design of Wind Turbines − DNV/Risø
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
98 4 – Loads
Guidelines for Design of Wind Turbines − DNV/Risø
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
Guidelines for Design of Wind Turbines − DNV/Risø
100 4 – Loads
Guidelines for Design of Wind Turbines − DNV/Risø
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
Guidelines for Design of Wind Turbines − DNV/Risø
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.
102 4 – Loads
Guidelines for Design of Wind Turbines − DNV/Risø
5 – Rotor 103
Guidelines for Design of Wind Turbines − DNV/Risø
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
Guidelines for Design of Wind Turbines − DNV/Risø
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
Guidelines for Design of Wind Turbines − DNV/Risø
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
Guidelines for Design of Wind Turbines − DNV/Risø
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
5 – Rotor 107
Guidelines for Design of Wind Turbines − DNV/Risø
Table 5-4 Testing of mechanical properties of the cured end product (laminate of
fibre-reinforced plastics)
Property Test standard Number of tests
108 5 – Rotor
Guidelines for Design of Wind Turbines − DNV/Risø
5 – Rotor 109
Guidelines for Design of Wind Turbines − DNV/Risø
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
Guidelines for Design of Wind Turbines − DNV/Risø
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
Guidelines for Design of Wind Turbines − DNV/Risø
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
Guidelines for Design of Wind Turbines − DNV/Risø
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
Guidelines for Design of Wind Turbines − DNV/Risø
114 5 – Rotor
Guidelines for Design of Wind Turbines − DNV/Risø
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
Guidelines for Design of Wind Turbines − DNV/Risø
116 5 – Rotor
Guidelines for Design of Wind Turbines − DNV/Risø
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 – Rotor 117
Guidelines for Design of Wind Turbines − DNV/Risø
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
Guidelines for Design of Wind Turbines − DNV/Risø
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 – Rotor 119
Guidelines for Design of Wind Turbines − DNV/Risø
120 5 – Rotor
Guidelines for Design of Wind Turbines – DNV/Risø
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
Guidelines for Design of Wind Turbines – DNV/Risø
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]
122 6 – Nacelle
Guidelines for Design of Wind Turbines – DNV/Risø
6 – Nacelle 123
Guidelines for Design of Wind Turbines – DNV/Risø
124 6 – Nacelle
Guidelines for Design of Wind Turbines – DNV/Risø
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:
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
Guidelines for Design of Wind Turbines – DNV/Risø
126 6 – Nacelle
Guidelines for Design of Wind Turbines – DNV/Risø
6 – Nacelle 127
Guidelines for Design of Wind Turbines – DNV/Risø
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.
128 6 – Nacelle
Guidelines for Design of Wind Turbines – DNV/Risø
Figure 6-11 Bearing frame with main shaft and bearing arrangement for wind turbine with two main bear-
ings; courtesy Vestas
6 – Nacelle 129
Guidelines for Design of Wind Turbines − DNV/Risø
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.
yg
ys
yr
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
Guidelines for Design of Wind Turbines − DNV/Risø
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
Guidelines for Design of Wind Turbines − DNV/Risø
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
Guidelines for Design of Wind Turbines − DNV/Risø
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
Guidelines for Design of Wind Turbines − DNV/Risø
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).
134 6 – Nacelle
Guidelines for Design of Wind Turbines − DNV/Risø
6 – Nacelle 135
Guidelines for Design of Wind Turbines − DNV/Risø
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
136 6 – Nacelle
Guidelines for Design of Wind Turbines − DNV/Risø
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
Guidelines for Design of Wind Turbines − DNV/Risø
138 6 – Nacelle
Guidelines for Design of Wind Turbines − DNV/Risø
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
Guidelines for Design of Wind Turbines − DNV/Risø
140 6 – Nacelle
Guidelines for Design of Wind Turbines − DNV/Risø
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
Guidelines for Design of Wind Turbines − DNV/Risø
142 6 – Nacelle
Guidelines for Design of Wind Turbines − DNV/Risø
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
Guidelines for Design of Wind Turbines − DNV/Risø
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-
144 6 – Nacelle
Guidelines for Design of Wind Turbines − DNV/Risø
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
Guidelines for Design of Wind Turbines − DNV/Risø
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
Guidelines for Design of Wind Turbines − DNV/Risø
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
Guidelines for Design of Wind Turbines − DNV/Risø
148 6 – Nacelle
Guidelines for Design of Wind Turbines − DNV/Risø
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.
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
Guidelines for Design of Wind Turbines − DNV/Risø
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
Guidelines for Design of Wind Turbines − DNV/Risø
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
Guidelines for Design of Wind Turbines − DNV/Risø
152 6 – Nacelle
Guidelines for Design of Wind Turbines − DNV/Risø
6 – Nacelle 153
Guidelines for Design of Wind Turbines − DNV/Risø
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
Guidelines for Design of Wind Turbines − DNV/Risø
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
Guidelines for Design of Wind Turbines − DNV/Risø
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
Guidelines for Design of Wind Turbines − DNV/Risø
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
Guidelines for Design of Wind Turbines − DNV/Risø
158 6 – Nacelle
Guidelines for Design of Wind Turbines − DNV/Risø
• 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
Guidelines for Design of Wind Turbines − DNV/Risø
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
160 6 – Nacelle
Guidelines for Design of Wind Turbines − DNV/Risø
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
Guidelines for Design of Wind Turbines − DNV/Risø
Figure 6-27. Relation between wind speed, yaw error and yaw moment.
162 6 – Nacelle
Guidelines for Design of Wind Turbines − DNV/Risø
6 – Nacelle 163
Guidelines for Design of Wind Turbines − DNV/Risø
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.
164 6 – Nacelle
Guidelines for Design of Wind Turbines − DNV/Risø
Feq
åP ⋅O
i
p
i
P= p
i
,
åO i
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
Guidelines for Design of Wind Turbines − DNV/Risø
166 6 – Nacelle
Guidelines for Design of Wind Turbines − DNV/Risø
6 – Nacelle 167
Guidelines for Design of Wind Turbines − DNV/Risø
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
Guidelines for Design of Wind Turbines − DNV/Risø
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
Guidelines for Design of Wind Turbines − DNV/Risø
170 7 – Tower
Guidelines for Design of Wind Turbines − DNV/Risø
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
Guidelines for Design of Wind Turbines − DNV/Risø
172 7 – Tower
Guidelines for Design of Wind Turbines − DNV/Risø
7 – Tower 173
Guidelines for Design of Wind Turbines − DNV/Risø
174 7 – Tower
Guidelines for Design of Wind Turbines − DNV/Risø
7 – Tower 175
Guidelines for Design of Wind Turbines − DNV/Risø
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
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
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
Guidelines for Design of Wind Turbines − DNV/Risø
7 – Tower 177
Guidelines for Design of Wind Turbines − DNV/Risø
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
178 7 – Tower
Guidelines for Design of Wind Turbines − DNV/Risø
7 – Tower 179
Guidelines for Design of Wind Turbines − DNV/Risø
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.
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
180 7 – Tower
Guidelines for Design of Wind Turbines − DNV/Risø
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
Guidelines for Design of Wind Turbines − DNV/Risø
182 7 – Tower
Guidelines for Design of Wind Turbines − DNV/Risø
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
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
Guidelines for Design of Wind Turbines − DNV/Risø
184 7 – Tower
Guidelines for Design of Wind Turbines − DNV/Risø
7 – Tower 185
Guidelines for Design of Wind Turbines − DNV/Risø
186 7 – Tower
Guidelines for Design of Wind Turbines − DNV/Risø
8 – Foundations 187
Guidelines for Design of Wind Turbines − DNV/Risø
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
Guidelines for Design of Wind Turbines − DNV/Risø
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.
8 – Foundations 189
Guidelines for Design of Wind Turbines − DNV/Risø
which is the most critical with respect to the foundation. Reference is made to Figure
stability. 8-1, and the eccentricity is calculated as
2
2⋅Mz æ 2⋅Mz ö
LC =
H' + H +ç ÷
l eff ç l ÷
V è eff ø
f [kN/m2]
190 8 – Foundations
Guidelines for Design of Wind Turbines − DNV/Risø
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
be = 2(R − e )
e
and
2
æ b ö
LC2 le = 2 R 1 − ç1 − ÷
leff
è 2R ø
beff Aeff
8 – Foundations 191
Guidelines for Design of Wind Turbines − DNV/Risø
192 8 – Foundations
Guidelines for Design of Wind Turbines − DNV/Risø
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
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.
8 – Foundations 193
Guidelines for Design of Wind Turbines − DNV/Risø
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-
194 8 – Foundations
Guidelines for Design of Wind Turbines − DNV/Risø
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
8 – Foundations 195
Guidelines for Design of Wind Turbines − DNV/Risø
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
196 8 – Foundations
Guidelines for Design of Wind Turbines − DNV/Risø
8 – Foundations 197
Guidelines for Design of Wind Turbines − DNV/Risø
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.
198 8 – Foundations
Guidelines for Design of Wind Turbines − DNV/Risø
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
ì pu y 1 / 3
ï ( ) for y ≤ 3y c
p = í 2 yc
ïî0.72 p u for y > 3y c
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
8 – Foundations 199
Guidelines for Design of Wind Turbines − DNV/Risø
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
200 8 – Foundations
Guidelines for Design of Wind Turbines − DNV/Risø
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.
8 – Foundations 201
Guidelines for Design of Wind Turbines − DNV/Risø
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
202 8 – Foundations
Guidelines for Design of Wind Turbines − DNV/Risø
8 – Foundations 203
Guidelines for Design of Wind Turbines − DNV/Risø
Table 8-5 Circular footing on stratum over bedrock or on stratum over halfspace
On stratum over bedrock On stratum over halfspace
204 8 – Foundations
Guidelines for Design of Wind Turbines − DNV/Risø
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 – Foundations 205
Guidelines for Design of Wind Turbines − DNV/Risø
206 8 – Foundations
Guidelines for Design of Wind Turbines − DNV/Risø
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.
8 – Foundations 207
Guidelines for Design of Wind Turbines − DNV/Risø
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
208 8 – Foundations
Guidelines for Design of Wind Turbines − DNV/Risø
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
10 years after installation 8 mm
20 years after installation 11 mm
30 years after installation 14 mm
8 – Foundations 209
Guidelines for Design of Wind Turbines − DNV/Risø
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
210 8 – Foundations
Guidelines for Design of Wind Turbines − DNV/Risø
8 – Foundations 211
Guidelines for Design of Wind Turbines − DNV/Risø
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.
212 8 – Foundations
Guidelines for Design of Wind Turbines − DNV/Risø
• 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.
8 – Foundations 213
Guidelines for Design of Wind Turbines − DNV/Risø
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
214 8 – Foundations
Guidelines for Design of Wind Turbines − DNV/Risø
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).
8 – Foundations 215
Guidelines for Design of Wind Turbines − DNV/Risø
In the ultimate limit state (ULS) the struc- tions. 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.
216 8 – Foundations
Guidelines for Design of Wind Turbines − DNV/Risø
8 – Foundations 217
Guidelines for Design of Wind Turbines − DNV/Risø
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.
218 8 – Foundations
Guidelines for Design of Wind Turbines − DNV/Risø
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, 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.
8 – Foundations 219
Guidelines for Design of Wind Turbines − DNV/Risø
220 8 – Foundations
Guidelines for Design of Wind Turbines − DNV/Risø
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
224 10 – Manuals
Guidelines for Design of Wind Turbines − DNV/Risø
10 – Manuals 225
Guidelines for Design of Wind Turbines − DNV/Risø
226 10 – Manuals
Guidelines for Design of Wind Turbines − DNV/Risø
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
C – Fatigue 229
Guidelines for Design of Wind Turbines − DNV/Risø
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 )
230 C – Fatigue
Guidelines for Design of Wind Turbines − DNV/Risø
σ ln N
S1− p = S 50% exp(−k )
m
C – Fatigue 231
Guidelines for Design of Wind Turbines − DNV/Risø
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
Guidelines for Design of Wind Turbines − DNV/Risø
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
Guidelines for Design of Wind Turbines − DNV/Risø
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.
Figure C-2 Characteristic S−N curves for structural steel in hull and casing structures, from DNV Class.
Notes 30.7
234 C – Fatigue
Guidelines for Design of Wind Turbines − DNV/Risø
Figure C-3 Characteristic S-N curve for non-welded, forged or rolled machine steel
C – Fatigue 235
Guidelines for Design of Wind Turbines − DNV/Risø
236 C – Fatigue
Guidelines for Design of Wind Turbines − DNV/Risø
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.
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.
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.
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.
E [U 10 ]
2
Celsius temperature = Fahrenheit tempera- u0 =
ture times 5/9 minus 32 π
u = u0 ⋅ x
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
Guidelines for Design of Wind Turbines - DNV/Risø
248 Index
Guidelines for Design of Wind Turbines - DNV/Risø
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
Guidelines for Design of Wind Turbines - DNV/Risø
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
Guidelines for Design of Wind Turbines - DNV/Risø
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
Guidelines for Design of Wind Turbines - DNV/Risø
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
Guidelines for Design of Wind Turbines - DNV/Risø
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