Guidelines for Design of Wind Turbines A publication fromDNV/Ris Second Edition Guidelines for Design of Wind Turbines 2nd Edition DetNorskeVeritas,Copenhagen(Wind.Turbine.Certification@dnv.com)andWind Energy Department, Ris National Laboratory (Certification@risoe.dk) 2002. Allrightsreserved.Nopartofthispublicationmaybereproduced,storedinaretrieval system, or transmitted, in any form or by any means, electronical, mechanical, photocopying, recording and/or otherwise without the prior written permission of the publishers. This book may not be lent, resold, hired out or otherwise disposed of by way of trade in any form of binding or cover other than that in which it is published, without the prior consent of the publishers. The front-page picture is from Microsoft Clipart Gallery ver. 2.0. Printed by Jydsk Centraltrykkeri, Denmark 2002 ISBN 87-550-2870-5 Guidelines for Design of Wind Turbines DNV/Ris PrefaceiPreface Theguidelinesfordesignofwindturbines have been developed with an aim to compile into one book much of the knowledge about designandconstructionofwindturbines that has been gained over the past few years. Thisappliestoknowledgeachievedfrom researchprojectsaswellastoknowledge resultingfrompracticaldesignexperience. Inaddition,thevariousrulesandmethods requiredfortypeapprovalwithinthemajor markets for the wind turbine industry form a basisfortheguidelines,withemphasison the international standards for wind turbines givenbytheInternationalElectrotechnical Commission, IEC. Theobjectiveistoprovideguidelines, whichcanbeusedfordesignofdifferent typesofwindturbinesinthefuture.The guidelinesproviderecommendationsand guidancefordesigntogetherwithapplica-tion-orientedsolutionstocommonlyen-countered design problems. Theguidelinescanbeusedbywindturbine manufacturers,certifyingauthorities,and windturbineowners.Theguidelineswill also be useful as an introduction and tutorial fornewtechnicalpersonnelandasarefer-ence for experienced engineers. Theguidelinesareavailableasaprinted book in a handy format as well as electroni-cally in pdf format on a CD-ROM. Thedevelopmentoftheguidelinesisthe resultofajointeffortbetweenDetNorske VeritasandRisNationalLaboratory.The developmenthasbeenfoundedbyDanish EnergyAgency,DetNorskeVeritasand Ris National Laboratory. These guidelines for design of wind turbines havebeenthoroughlyreviewedbyinternal and external experts. However, no warranty, expressed or implied, is made by Det Norske VeritasandRisNationalLaboratory,asto theaccuracyorfunctionalityoftheguide-lines,andnoresponsibilityisassumedin connection therewith. Guidelines for Design of Wind Turbines 2002 Det Norske Veritas and Ris National Laboratory Guidelines for Design of Wind Turbines DNV/Ris Contentsii Contents 1.WIND TURBINE CONCEPTS ......... 1 1.1INTRODUCTION................................. 1 1.2CONCEPTUAL ASPECTS..................... 1 1.2.1Vertical axis turbines .................... 2 1.2.2Horizontal axis turbines................ 2 1.2.3Number of rotor blades................. 3 1.2.4Power control aspects ................... 3 1.3ECONOMICAL ASPECTS..................... 5 1.4POWER PRODUCTION........................ 5 1.4.1Power curve.................................. 6 1.4.2Annual energy production ............ 7 1.5CONFIGURATIONS AND SIZES ........... 7 1.6FUTURE CONCEPTS........................... 8 REFERENCES........................................... 9 2.SAFETY AND RELIABILITY........ 10 2.1SAFETY PHILOSOPHY...................... 10 2.2SYSTEM SAFETY ANDOPERATIONAL RELIABILITY............ 11 2.2.1Control system............................ 12 2.2.2Protection system........................ 13 2.2.3Brake system .............................. 14 2.2.4Failure mode and effectsanalysis ....................................... 15 2.2.5Fault tree analysis ....................... 16 2.3STRUCTURAL SAFETY..................... 18 2.3.1Limit states ................................. 18 2.3.2Failure probability and other measures of structural reliability. 18 2.3.3Structural reliability methods ..... 19 2.3.4Code format, characteristicvalues, and partial safety factors. 19 2.3.5Code calibration.......................... 20 2.3.6Example axially loadedsteel tower................................... 22 2.3.7Example fatigue of FRPblade root in bending .................. 24 2.3.8Tests and calculations for verification.................................. 26 2.3.9Inspection andinspection intervals ..................... 26 2.4MECHANICAL SAFETY.................... 27 2.5LABOUR SAFETY ............................ 28 2.5.1Transportation, installationand commissioning..................... 28 2.5.2Normal operation........................ 29 2.5.3Service, maintenance and repair . 29 2.6CODES AND STANDARDS ................ 30 REFERENCES......................................... 30 3.EXTERNAL CONDITIONS............ 32 3.1WIND CONDITIONS ......................... 32 3.1.110-minute mean wind speed....... 32 3.1.2Standard deviation ofwind speed.................................. 34 3.1.3Turbulence intensity ................... 36 3.1.4Lateral and vertical turbulence ... 37 3.1.5Stochastic turbulence models ..... 37 3.1.6Wind shear.................................. 40 3.1.7Wind direction............................ 42 3.1.8Transient wind conditions .......... 43 3.1.9Extreme winds gusts................ 43 3.1.10Site assessment ........................... 46 3.2OTHER EXTERNAL CONDITIONS...... 48 3.2.1Temperatures .............................. 48 3.2.2Density of air .............................. 49 3.2.3Humidity..................................... 50 3.2.4Radiation and ultraviolet light .... 50 3.2.5Ice ............................................... 50 3.2.6Rain, snow and hail .................... 50 3.2.7Atmospheric corrosion and abrasion................................ 51 3.2.8Earthquake.................................. 51 3.2.9Lightning .................................... 53 REFERENCES......................................... 53 4.LOADS............................................... 55 4.1LOAD CASES................................... 55 4.1.1Design situations ........................ 55 4.1.2Wind events................................ 55 4.1.3Design load cases ....................... 55 4.2LOAD TYPES................................... 58 4.2.1Inertia and gravity loads ............. 58 4.2.2Aerodynamic loads..................... 59 4.2.3Functional loads.......................... 60 Guidelines for Design of Wind Turbines DNV/Ris Contentsiii4.2.4Other loads.................................. 60 4.3AEROELASTIC LOADCALCULATIONS .............................. 60 4.3.1Model elements........................... 61 4.3.2Aeroelastic models for load prediction.................................... 70 4.3.3Aerodynamic data assessment .... 70 4.3.4Special considerations ................ 72 4.4LOAD ANALYSIS AND SYNTHESIS ... 76 4.4.1Fatigue loads............................... 76 4.4.2Ultimate loads............................. 82 4.5SIMPLIFIED LOAD CALCULATIONS .. 86 4.5.1Parametrised empirical models... 86 4.5.2The simple load basis ................. 86 4.5.3Quasi-static method .................... 87 4.5.4Peak factor approach forextreme loads.............................. 88 4.5.5Parametrised load spectra ........... 89 4.6SITE-SPECIFIC DESIGN LOADS......... 92 4.7LOADS FROM OTHER SOURCESTHAN WIND..................................... 93 4.7.1Wave loads ................................. 93 4.7.2Current loads .............................. 99 4.7.3Ice loads...................................... 99 4.7.4Earthquake loads......................... 99 4.8LOAD COMBINATION...................... 99 REFERENCES....................................... 101 5.ROTOR............................................ 104 5.1BLADES........................................ 104 5.1.1Blade geometry......................... 104 5.1.2Design loads ............................. 105 5.1.3Blade materials ......................... 105 5.1.4Manufacturing techniques ........ 108 5.1.5Quality assurance for bladedesign and manufacture ............ 109 5.1.6Strength analyses ...................... 110 5.1.7Tip deflections .......................... 113 5.1.8Lightning protection ................. 113 5.1.9Blade testing ............................. 114 5.1.10Maintenance ............................. 116 5.2HUB ............................................. 116 5.2.1Determination of design loads .. 117 5.2.2Strength Analysis...................... 117 5.2.3Analysis of bolt connections..... 118 5.2.4Hub enclosure........................... 118 5.2.5Materials................................... 118 5.2.6Standards .................................. 119 REFERENCES....................................... 119 6.NACELLE....................................... 120 6.1MAIN SHAFT................................. 120 6.1.1Determination of design loads.. 120 6.1.2Strength analysis....................... 120 6.1.3Fatigue strength ........................ 121 6.1.4Ultimate strength ...................... 125 6.1.5Main shaft-gear connection ...... 126 6.1.6Materials................................... 126 6.1.7Standards .................................. 127 6.2MAIN BEARING............................ 127 6.2.1Determination of design loads.. 129 6.2.2Selection of bearing types......... 130 6.2.3Operational andenvironmental conditions.......... 130 6.2.4Seals, lubricationand temperatures....................... 130 6.2.5Rating life calculations ............. 132 6.2.6Connection to main shaft .......... 133 6.2.7Bearing housing........................ 133 6.2.8Connection to machine frame... 133 6.2.9Standards .................................. 133 6.3MAIN GEAR.................................. 133 6.3.1Gear types................................. 134 6.3.2Loads and capacity ................... 137 6.3.3Codes and standards ................. 141 6.3.4Lubrication ............................... 141 6.3.5Materials and testing................. 142 6.4COUPLINGS .................................. 145 6.4.1Flange couplings....................... 145 6.4.2Shrink fit couplings .................. 146 6.4.3Key connections ....................... 146 6.4.4Torsionally elastic couplings. ... 146 6.4.5Tooth couplings ........................ 146 6.5MECHANICAL BRAKE ................... 146 6.5.1Types of brakes......................... 146 6.5.2Brake discs and brake pads....... 148 6.5.3Brake torque sequence.............. 148 6.6HYDRAULIC SYSTEMS .................. 149 6.6.1Arrangement ............................. 149 6.6.2Accumulators............................ 149 6.6.3Valves....................................... 149 Guidelines for Design of Wind Turbines DNV/Ris Contentsiv6.6.4Application inprotection systems .................... 150 6.6.5Additional provisions ............... 151 6.6.6Codes and standards ................. 151 6.7GENERATOR................................. 151 6.7.1Types of generators .................. 151 6.7.2Climate aspects......................... 153 6.7.3Safety aspects ........................... 153 6.7.4Cooling and degree of sealing .. 155 6.7.5Vibrations ................................. 155 6.7.6Overspeed................................. 155 6.7.7Overloading .............................. 155 6.7.8Materials ................................... 156 6.7.9Generator braking..................... 156 6.7.10Lifetime .................................... 157 6.7.11Testing of generators ................ 157 6.8MACHINE SUPPORT FRAME........... 157 6.9NACELLE ENCLOSURE .................. 158 6.10YAW SYSTEM............................... 158 6.10.1Determination of design loads .. 160 6.10.2Yaw drive ................................. 161 6.10.3Yaw ring................................... 162 6.10.4Yaw brake................................. 162 6.10.5Yaw bearing.............................. 163 6.10.6Yaw error and control............... 166 6.10.7Cable twist ................................ 166 6.10.8Special design considerations ... 166 REFERENCES....................................... 167 7.TOWER........................................... 169 7.1LOAD CASES................................. 170 7.2DESIGN LOADS ............................. 170 7.3GENERAL VERIFICATIONS FOR TOWERS........................................ 171 7.3.1Dynamic responseand resonance .............................. 171 7.3.2Critical bladedeflection analysis ....................... 172 7.4TUBULAR TOWERS ....................... 173 7.4.1Loads and responses ................. 173 7.4.2Extreme loads ........................... 174 7.4.3Fatigue loads............................. 174 7.4.4Vortex induced vibrations......... 174 7.4.5Welded joints............................ 175 7.4.6Stress concentrations nearhatches and doors...................... 176 7.4.7Stability analysis....................... 177 7.4.8Flange connections ................... 178 7.4.9Corrosion protection................. 179 7.4.10Tolerances and specifications... 179 7.5ACCESS AND WORKING ENVIRONMENT ............................. 180 7.6EXAMPLE OF TOWER LOAD CALCULATION.............................. 180 7.6.1Loads and responses ................. 180 7.6.2Occurrence of extreme loads during normal power production.......... 181 7.6.3Extreme loads parked turbine 182 7.6.4Fatigue loading ......................... 183 REFERENCES....................................... 186 8.FOUNDATIONS............................. 187 8.1SOIL INVESTIGATIONS .................. 187 8.1.1General ..................................... 187 8.1.2Recommendations for gravitybased foundations ..................... 188 8.1.3Recommendations for pile foundations ............................... 189 8.2GRAVITY-BASED FOUNDATIONS... 189 8.2.1Bearing capacity formulas ........ 190 8.3PILE-SUPPORTED FOUNDATIONS... 193 8.3.1Pile groups................................ 194 8.3.2Axial pile resistance.................. 195 8.3.3Laterally loaded piles................ 197 8.3.4Soil resistance for embeddedpile caps.................................... 200 8.4FOUNDATION STIFFNESS............... 201 8.5PROPERTIES OF REINFORCED CONCRETE ................................... 206 8.5.1Fatigue...................................... 206 8.5.2Crack-width.............................. 207 8.5.3Execution.................................. 208 8.6SELECTED FOUNDATIONSTRUCTURE CONCEPTS FOROFFSHORE APPLICATIONS ............. 208 8.6.1Introduction to concepts ........... 208 8.6.2Monopile .................................. 209 8.6.3Tripod....................................... 215 REFERENCES....................................... 221 Guidelines for Design of Wind Turbines DNV/Ris Contentsv9.ELECTRICAL SYSTEM............... 223 9.1ELECTRICAL COMPONENTS........... 223 9.1.1Generators................................. 223 9.1.2Softstarter ................................. 225 9.1.3Capacitor bank.......................... 225 9.1.4Frequency converter ................. 226 9.2WIND TURBINE CONFIGURATIONS 227 9.3POWER QUALITY AND GRID CONNECTION................................ 229 9.4ELECTRICAL SAFETY.................... 230 9.5WIND FARM INTEGRATION ........... 231 REFERENCES....................................... 232 10. MANUALS ...................................... 233 10.1USER MANUAL ............................. 233 10.2SERVICE AND MAINTENANCEMANUAL....................................... 233 10.3INSTALLATION MANUAL............... 233 REFERENCE ......................................... 233 11. TESTS AND MEASUREMENTS.. 234 11.1POWER PERFORMANCE MEASUREMENTS........................... 234 11.2LOAD MEASUREMENTS................. 236 11.3TEST OF CONTROL ANDPROTECTION SYSTEM.................... 237 11.4POWER QUALITY MEASUREMENT . 237 11.5BLADE TESTING............................ 237 11.6NOISE MEASUREMENTS ................ 237 REFERENCES....................................... 237 A.BOLT CONNECTIONS................. 239 A.1BOLT STANDARDIZATION............. 239 A.2STRENGTH.................................... 239 A.3IMPACT STRENGTH....................... 239 A.4SURFACE TREATMENT.................. 239 A.5S-N CURVES................................. 240 A.5.1S-N curves in structuralsteel codes................................. 241 A.5.2Allowable surface pressure....... 242 A.6PRETENSION................................. 242 A.6.1Safety against loosening ........... 244 A.7MINIMUM DEPTH OFTHREADED HOLES......................... 244 A.8BOLT FORCE ANALYSIS ................ 245 A.8.1Stiffness of bolts ....................... 245 A.8.2Stiffness of the mating parts ..... 246 A.8.3Force triangle............................ 246 A.9CONNECTIONS SUBJECTEDTO SHEAR ..................................... 247 A.10BOLTS SUBJECTEDTO TENSILE LOAD ......................... 248 A.11BOLTS SUBJECTED TO TENSILELOAD AND SHEAR......................... 249 A.12EXECUTION OF BOLTCONNECTIONS .............................. 249 A.13CODES AND STANDARDS.............. 249 REFERENCES....................................... 249 B.RULES OF THUMB....................... 250 B.1LOADS.......................................... 250 B.1.1Rotor loads................................ 250 B.1.2Fatigue loads............................. 250 B.2ROTOR ......................................... 250 B.3NACELLE...................................... 251 B.3.1Main shaft................................. 251 B.4NOISE........................................... 251 REFERENCES....................................... 251 C.FATIGUE CALCULATIONS ....... 252 C.1STRESS RANGES............................ 252 C.2FRACTURE MECHANICS ................ 252 C.3S-N CURVES ................................. 253 C.4THE PALMGREN-MINER RULE...... 254 C.5FATIGUE IN WELDED STRUCTURES 255 C.6CHARACTERISTIC S-N CURVESFOR STRUCTURAL STEEL............... 256 C.7CHARACTERISTIC S-N CURVESFOR FORGED OR ROLLED STEEL .... 256 C.8S-N CURVES FOR COMPOSITES...... 257 C.9OTHER TYPES OF FATIGUE ASSESSMENT ................................ 258 REFERENCES....................................... 258 Guidelines for Design of Wind Turbines DNV/Ris ContentsviD.FEM CALCULATIONS................. 260 D.1TYPES OF ANALYSIS ..................... 260 D.2MODELLING ................................. 261 D.2.1Model........................................ 261 D.2.2Elements ................................... 262 D.2.3Boundary conditions................. 264 D.2.4Loads ........................................ 265 D.3DOCUMENTATION ........................ 265 D.3.1Model........................................ 265 D.3.2Results ...................................... 267 E.MATERIAL PROPERTIES .......... 268 E.1STEEL........................................... 268 E.1.1Structural steel .......................... 268 E.1.2Alloy steel................................. 269 E.2CAST IRON.................................... 269 E.3FIBRE REINFORCED PLASTICS ...... 269 E.3.1Glass fibre reinforced plastics .. 269 E.4CONCRETE ................................... 270 E.4.1Mechanical properties............... 270 REFERENCES....................................... 270 F.TERMS AND DEFINITIONS ....... 271 REFERENCES....................................... 276 G.TABLES AND CONVERSIONS... 277 G.1ENGLISH/METRIC CONVERSION .... 277 G.2AIR DENSITY VS. TEMPERATURE .. 277 G.3AIR DENSITY VS. HEIGHT.............. 277 G.4RAYLEIGH WIND DISTRIBUTION.... 277 Guidelines for Design of Wind Turbines DNV/Ris 1 Wind Turbine Concepts11.Wind Turbine Concepts 1.1Introduction Wind-poweredships,grainmills,water pumps,andthreshingmachinesall exemplifythatextractionofpowerfrom windisanancientendeavour.Withthe evolutionofmechanicalinsightand technology,thelastdecadesofthe20th century,inparticular,sawthedevelopment of machines which efficiently extract power fromwind.Windturbinesisnowbeing usedasagenerictermformachineswith rotatingbladesthatconvertthekinetic energy of wind into useful power. Inthe20thcentury,earlywindturbine designsweredrivenbythreebasic philosophiesforhandlingloads:(1) withstanding loads, (2) shedding or avoiding loads, and (3) managing loads mechanically, electrically,orboth.Inthemidstofthis evolution,manywindturbinedesignssaw thelightofday,includinghorizontalaxis and vertical axis turbines. Turbinesthat spin abouthorizontalandverticalaxes, respectively,andareequippedwithone, two, three or multiple blades. Modernturbinesevolvedfromtheearly designsandcanbeclassifiedastwoor three-bladedturbineswithhorizontalaxes andupwindrotors.Today,thechoice betweentwoorthree-bladedwindturbines ismerelyamatterofatrade-offbetween aerodynamicefficiency,complexity,cost, noise and aesthetics. Additional key turbine design considerations includewindclimate,rotortype,generator type,loadandnoiseminimisation,and controlapproach.Moreover,currenttrends, drivenbytheoperatingregimeandthe marketenvironment,involvedevelopment oflow-cost,megawatt-scaleturbinesand lightweightturbineconcepts.Whereas turbinesoperatingatconstantrotorspeed havebeendominatinguptonow,turbines withvariablerotorspeedarebecoming increasinglymorecommoninanattemptto optimise the energy capture, lower the loads, obtain better power quality, and enable more advanced power control aspects. 1.2Conceptual aspects Someearlywindturbinedesignsinclude multiple-bladed concepts. These turbines are all characterised by rotors with high solidity, i.e.theexposedareaofthebladesis relativelylargecomparedtothesweptarea of the rotor. Figure1-1.Multiple-bladedwindturbinesofvarious designs. A disadvantage of such a high-solidity rotor istheexcessiveforcesthatitwillattract duringextremewindspeedssuchasin hurricanes.Tolimitthisundesirableeffect of extreme winds and to increase efficiency, modernwindturbinesarebuiltwithfewer, longer,andmoreslenderblades,i.e.witha Guidelines for Design of Wind Turbines DNV/Ris 1 Wind Turbine Concepts2muchsmallersolidity.Tocompensatefor theslendernessoftheblades,modern turbines operate at high tip speeds. 1.2.1Vertical axis turbines Verticalaxiswindturbines(VAWTs),such astheoneshowninFigure1-2withC-shapedblades,areamongthetypesof turbine that have seen the light of day in the past century. Figure1-2.EoleC,a4200kWverticalaxisDarrieus windturbinewith100mrotordiameteratCapChat, Qubec,Canada.Themachine,whichistheworld's largestwindturbine,isnolongeroperational.From www.windpower.org(2000),DanishWindTurbine Manufacturers Association. Classicalwaterwheelsallowthewaterto arrivetangentiallytothewaterwheelata rightangletotherotationalaxisofthe wheel.Verticalaxiswindturbinesare designed to act correspondingly towards air. Though,suchadesignwould,inprinciple, work with a horizontal axis as well, it would requireamorecomplexdesign,which wouldhardlybeabletobeattheefficiency ofapropeller-typeturbine.Themajor advantages of a vertical axis wind turbine, as the one illustrated in Figure 1-2, are that the generatorandgearboxareplacedonthe groundandarethuseasilyaccessible,and thatnoyawmechanismisneeded.Among the disadvantages are an overall much lower levelofefficiency,thefactthattheturbine needstotaldismantlingjusttoreplacethe mainbearing,andthattherotorisplaced relatively close to the ground where there is not much wind. 1.2.2Horizontal axis turbines Horizontalaxiswindturbines(HAWTs), suchastheonesshowninFigure1-3, constitutethemostcommontypeofwind turbineinusetoday.Infactallgrid-connectedcommercialwindturbinesare todaydesignedwithpropeller-typerotors mountedonahorizontalaxisontopofa verticaltower.Incontrasttothemodeof operationoftheverticalaxisturbines,the horizontalaxisturbinesneedtobealigned withthedirectionofthewind,thereby allowing the wind to flow parallel to the axis of rotation. Figure1-3.Three-bladedupwindturbinesbeingtested at Ris, August 1986. Insofarasconcernshorizontalaxiswind turbines,adistinctionismadebetween upwindanddownwindrotors.Upwind rotorsfacethewindinfrontofthevertical towerandhavetheadvantageofsomewhat avoidingthewindshadeeffectfromthe presence of the tower. Upwind rotors need a yawmechanismtokeeptherotoraxis Guidelines for Design of Wind Turbines DNV/Ris 1 Wind Turbine Concepts3alignedwiththedirectionofthewind. Downwindrotorsareplacedontheleeside ofthetower.Agreatdisadvantageinthis design is the fluctuations in the wind power duetotherotorpassingthroughthewind shade of the tower which gives rise to more fatigueloads.Theoretically,downwind rotorscanbebuiltwithoutayaw mechanism,providedthattherotorand nacellecanbedesignedinsuchawaythat thenacellewillfollowthewindpassively. This may, however, induce gyroscopic loads and hamper the possibility of unwinding the cableswhentherotorhasbeenyawing passivelyinthesamedirectionforalong time,therebycausingthepowercablesto twist.Asregardslargewindturbines,itis ratherdifficulttouseslipringsor mechanicalcollectorstocircumventthis problem. Whereas, upwind rotors need to be ratherinflexibletokeeptherotorblades clearofthetower,downwindrotorscanbe mademoreflexible.Thelatterimplies possiblesavingswithrespecttoweightand maycontributetoreducingtheloadsonthe tower. The vast majority of wind turbines in operation today have upwind rotors. 1.2.3Number of rotor blades Thethree-bladedconceptisthemost commonconceptformodernwindturbines. Aturbinewithanupwindrotor,an asynchronousgeneratorandanactiveyaw systemisusuallyreferredtoastheDanish concept. This is a concept, which tends to be astandardagainstwhichotherconceptsare evaluated. Relative to the three-bladed concept, the two and one-bladed concepts have the advantage ofrepresentingapossiblesavinginrelation to the cost and weight of the rotor. However, their use of fewer rotor blades implies that a higherrotationalspeedoralargerchordis needed to yield the same energy output as a three-bladedturbineofasimilarsize.The useofoneortwobladeswillalsoresultin morefluctuatingloadsbecauseofthe variationoftheinertia,dependingonthe bladesbeinginhorizontalorvertical positionandonthevariationofwindspeed whenthebladeispointingupwardand downward.Therefore,thetwoandone-bladedconceptsusuallyhaveso-called teeteringhubs,implyingthattheyhavethe rotorhingedtothemainshaft.Thisdesign allowstherotortoteeterinorderto eliminatesomeoftheunbalancedloads. One-bladedwindturbinesareless widespread than two-bladed turbines. This is duetothefactthatthey,inadditiontoa higherrotationalspeed,morenoiseand visualintrusionproblems,needacounter-weight to balance the rotor blade. Figure1-4.Three,twoandone-bladedwindturbine concepts. From www.windpower.org (2000), Danish Wind Turbine Manufacturers Association. 1.2.4Power control aspects Windturbinesaredesignedtoproduce electricityascheapaspossible.Forthis purpose,windturbinesareingeneral designedtoyieldamaximumpoweroutput atwindspeedsaround15m/s.Itwouldnot paytodesignturbinestomaximisetheir poweroutputatstrongerwinds,because suchstrongwindsareusuallytoorare. However,incaseofstrongerwinds,itis necessary to waste part of the excess energy toavoiddamageonthewindturbine.Thus, thewindturbineneedssomesortofpower control.Thepowercontrolisdividedinto two regimes with different concepts: power optimisation for low wind speeds power limitation for high wind speeds Guidelines for Design of Wind Turbines DNV/Ris 1 Wind Turbine Concepts4Theseregimesareseparatedbythewind speedatwhichthemaximumpoweroutput is achieved, typically about 15 m/sec. Basically,therearethreeapproachesto power control: stall control pitch control active stall control Stall-controlledwindturbineshavetheir rotorbladesboltedtothehubatafixed angle. The stall phenomenon is used to limit thepoweroutputwhenthewindspeed becomestoohigh.Thisisachievedby designing the geometry of the rotor blade in such a way that flow separation is created on thedownwindsideofthebladewhenthe windspeedexceedssomechosencritical value. Stall control of wind turbines requires correcttrimmingoftherotorbladesand correctsettingofthebladeanglerelativeto therotorplane.Somedrawbacksofthis methodare:lowerefficiencyatlowwind speeds, no assisted start and variations in the maximumsteadystatepowerdueto variationintheairdensityandgrid frequencies. Pitch-controlledwindturbineshaveblades thatcanbepitchedoutofthewindtoan angle where the blade chord is parallel to the winddirection.Thepoweroutputis monitoredandwheneveritbecomestoo high,thebladeswillbepitchedslightlyout ofthewindtoreducetheproducedpower. Thebladeswillbepitchedbackagainonce thewindspeeddrops.Pitch control of wind turbinesrequiresadesignthatensuresthat thebladesarepitchedattheexactangle requiredinordertooptimisethepower outputatallwindspeeds.Nowadays,pitch controlofwindturbinesisonlyusedin conjunctionwithvariablerotorspeed.An advantageofthistypeofcontrolisthatit has a good power control, i.e. that the mean value of the power output is kept close to the ratedpowerofthegeneratorathighwind speeds.Disadvantagesencompassextra complexityduetothepitchmechanismand high power fluctuations at high wind speeds. Activestall-controlledturbinesresemble pitch-controlled turbines by having pitchable blades.Atlowwindspeeds,activestall turbineswilloperatelikepitch-controlled turbines.Athighwindspeeds,theywill pitchthebladesintheoppositedirectionof whatapitch-controlledturbinewoulddo and force the blades into stall. This enables a ratheraccuratecontrolofthepoweroutput, andmakesitpossibletoruntheturbineat the rated power at all high wind speeds. This control type has the advantage of having the abilitytocompensateforthevariationsin the air density. Figure1-5showsiso-powercurvesfora wind turbine as a function of the blade angle andthemeanwindspeed.Therangesfor pitchcontrolandactivestallcontrolare separatedatabladeangleof0withthe rotor plane. At low wind speeds, the optimal operationofthewindturbineisachievedat abladeanglecloseto0.Athigherwind speeds,theturbinewilloverproduceifthe bladeangleisnotadjustedaccordingly. Withpitchcontrol,thebladeispitched positively with its leading edge being turned towardsthewind.Withactivestallcontrol, thebladeispitchednegativelywithits trailingedgeturnedtowardsthewind.The powercontroland,inparticular,thepower limitationathigherwindspeedsare indicatedinanidealisedmannerforboth controlapproachesbythedashedcurvesin Figure 1-5. The dashed curves illustrate how thetransitionbetweenoperationwith0 bladeangleatlowwindspeedsandpower-limitingoperationalonganiso-powercurve athighwindspeedscanbeachievedfora three-bladedrotorataratedpowerof400 kW. In this example, the rated power is reached at a wind speed of about 12 m/sec. Guidelines for Design of Wind Turbines DNV/Ris 1 Wind Turbine Concepts51.3Economical aspects The ideal wind turbine design is not dictated bytechnologyalone,butbyacombination oftechnologyandeconomy.Windturbine manufacturerswishtooptimisetheir machines,sothattheydeliverelectricityat thelowestpossiblecostperunitofenergy. Inthiscontext,itisnotnecessarilyoptimal to maximise the annual energy production, if thatwouldrequireaveryexpensivewind turbine. Since the energy input (the wind) is free,theoptimalturbinedesignisonewith low production costs per produced kWh. Thechoiceofrotorsizeandgeneratorsize dependsheavilyonthedistributionofthe wind speed and the wind energy potential at aprospectivelocation.Alargerotorfitted withasmallgeneratorwillproduce electricityduringmanyhoursoftheyear, butitwillonlycaptureasmallpartofthe windenergypotential.Alargegenerator willbeveryefficientathighwindspeeds, butinefficientatlowwindspeeds. Sometimes it will be beneficial to fit a wind turbinewithtwogeneratorswithdifferent rated powers. Figure 1-5. Iso-power curves for a wind turbine at 26.88 rpm vs. blade angle and mean wind speed 1.4Power production AstudyofdifferentDanishwindturbine designsshowsthatthespecificpower performanceintermsofproducedenergy perm2rotorareaperyear(kWh/m2/year)is almostindependentoftherotorsize.See Petersen, 1998. Hence,themainconsiderationinthe evaluationofthecostoftheturbineisthe specificrotorpower(kW/m2)andthe specificcost(cost/m2rotor)togetherwith Guidelines for Design of Wind Turbines DNV/Ris 1 Wind Turbine Concepts6expectedservicelifeandcostand availability.Anavailabilityfactor,i.e.the amount of time that the turbine is producing energy,orisreadyforproduction,of98% is common for commercial turbines. Thustheprimaryfactoraffectingthepower performanceistherotorsize.Secondary factors are the control principle such as stall- or pitch control and single- dual- or variable speed. 1.4.1Power curve Thepowerbeingproducedbyanytypeof wind turbine can be expressed as P = V3ACP Poutput power air density Vfree wind speed Arotor area CPefficiency factor The power coefficient CP is a product of the mechanicalefficiencym,theelectrical efficiencye,andoftheaerodynamic efficiency.Allthreefactorsaredependent onthewindspeedandtheproducedpower, respectively.Themechanicalefficiencym ismainlydeterminedbylossesinthe gearboxandistypically0.95to0.97atfull load.Theelectricalefficiencycoverslosses inthegeneratorandelectricalcircuits.At fullloade=0.97-0.98iscommonfor configurationswithaninductiongenerator. Itcanbeshownthatthemaximumpossible valueoftheaerodynamicefficiencyis16/27=0.59,whichisachievedwhenthe turbinereducesthewindspeedtoone-third of the free wind speed (Betz' law). Power [kW] Wind speed Efficiency [%] Figure 1-6. Power- and efficiency curve Theproducedpowervarieswiththewind speed as can be seen from the blue graph in Figure1-6.Theformofthegraphvaries slightlyfromdifferentconcepts.Assuming constantefficiency(e.g.constanttipspeed ratio)thegraphbasicallyconsistsofathird degreepolynomialuptotheratedwind speedatwhichthenominalpoweris reached.Atthispointthepowerregulation setsin,eitherbythebladesstallingorby pitchingthebladestoattainan approximatelyconstantpower.Thepower curveandthepowerefficiencycurveare oftenpresentedinthesamegraph,withthe power and the efficiency scales on each side of the graph as shown in Figure 1-6. Figure1-7illustratesthecontrolledpower curveofawindturbine,inthecaseof1) stallcontrolled,fixedspeedconfiguration, and2)pitchcontrolled,variablespeed configuration. Power Wind speed5 15 25 2010 Rated power Wind power resource Stall regulated & fixed speed Pitch control & variable speed [Kw] [m/s] Power Wind speed5 15 25 2010 Rated power Wind power resource Stall regulated & fixed speed Pitch control & variable speed Wind power resource Stall regulation & fixed speed Pitch control & variable speed [Kw] [m/s] Figure 1-7. Examples of power curves for two types of wind turbines. Guidelines for Design of Wind Turbines DNV/Ris 1 Wind Turbine Concepts7TheefficiencyfactorCPtypicallyreachesa maximum at a wind speed of 7-9 m/sec and, normally,itdoesnotexceed50%.The electricpowertypicallyreachestherated poweroftheturbineatawindspeedof14-16 m/sec. In the Danish approval scheme as well as in theIECwindturbineclassificationsystem, the power curve is required to be determined from measurements. 1.4.2Annual energy production Thewindspeeddistributionisoften representedbyaWeibulldistribution.The densityfunctionoftheWeibullprobability can be expressed as shown in Figure 1-8. ) ) ( exp( ) (110k kkUAuuAku f = Figure 1-8. Weibull density function. Hence,foraspecificsitethewindclimate canbedescribedintermsoftheparameters Aandk,andfromthesetheaveragewind speedVave,asdefinedinIEC61400-1,can be calculated =+ =2 , 2 /) / 1 1 (k if Ak AVave in which is the gamma function. IntheDanishstandarditiscommonto describethewindclimateintermsofthe roughnessclass,andanexplicitcorrelation existsbetweentheroughnessclassandthe average wind speed. Foraparticularwindturbine,thepower production is defined by the power curve as describedabove.Bycombiningthepower curvewiththewinddistributiontheactual energyproductionisyielded,often expressedintermsoftheannualenergy production Eyear. =stopstartVVyeardu u f u P N E ) ( ) (0 P(u)power curve function f(u)wind distribution function Vstartcut-in wind speed Vstopcut-out wind speed N0 = 8765 hours/year DividingEyearbytherotorareayieldsthe specificpowerperformance,whichis anothercommonwaytoexpresstheturbine efficiency. Yet another way of expressing the efficiency ofaparticularturbineisgivenbyusingthe capacity factor, which is defined as the ratio ofactualaveragepowertotheratedpower measuredoveraperiodoftime(average kW/ratedkW).Thetotalenergythatwould be produced by a wind turbine during a one-yearperiod,assumingacertaindistribution ofwindspeedprobabilitydensityand assuming100percentavailability,is referredtoasthepotentialannualenergy output. 1.5Configurations and sizes Windturbinesareerectedasstand-alone turbines,inclustersofmultipleturbines,or onalargerscaleinpark configurations. Beforethe1980s,windenergy development focused on the individual wind turbine.Bythelate1980s,thisperspective begantochangeasattentionshiftedto collectivegenerationofelectricpowerfrom Guidelines for Design of Wind Turbines DNV/Ris 1 Wind Turbine Concepts8anarrayofwindturbineslocatedinthe vicinityofeachotherandcommonly referredtoaswindparksorwindfarms.In theearly1980s,thetypicalsizeofawind turbinewasabout55kWintermsofrated power,whereasturbinesizestodayhave exceeded2MW.Table1-1givesexamples oftypicalcombinationsofrotordiameter andratedpowerforanumberofdifferent tower heights. Table 1-1. Typical wind turbine sizes Tower height (m) Rotordiameter (m) Rated power (kW)222155 3130225 3535450 35-4041-44500 4443600 5048750 50541000 60581500 64-8072-762000 851155000 Figure1-9.VindebyOffshoreWindFarm.From www.windpower.org (1997) Bonus. Whilewindenergyisalreadyeconomicin termsofgoodonshorelocations,itis currentlyabouttocrosstheeconomic frontiersetbyshorelines:offshorewind energyisbecomingcompetitivewithother power-generatingtechnologies.Offshore windenergyisapromisingapplicationof windpower,inparticular,incountrieswith highpopulationdensityandthuswith difficultiesinfindingsuitablesitesonland. Constructioncostsaremuchhigheratsea, butenergyproductionisalsomuchhigher. Currently,windenergyfromturbines erectedonfixedfoundationsinupto15m waterdepthisconsideredeconomically feasible. Figure 1-9 shows an example of an early offshore wind farm. 1.6 Future concepts Figure1-10.Aconceptionofaflexiblewindturbine. Duringstand-still,thebladesdeflectasarushinthe wind. Untilnowthedevelopmentoflargewind turbineshasmainlybeenpromptedby upscaling the dominating concepts described inthissection.Still,newconceptsforwind turbinedesignsandcomponentsarebeing developedinanattempttoanticipatethe demandsforthecontinuinggrowthofwind turbines. Oneproposalforafutureconceptis characterisedbymoreflexiblewindturbine concepts. One element in this is an expected increaseinthestructuralflexibilityofwind turbines.Figure1-10exemplifieshowthe lattercanbeconceived.Anotherelementis an expected increase in the flexibility of the drivetrain,e.g.intermsofgearlessdesigns withvariablerotationalspeeds,andamore Guidelines for Design of Wind Turbines DNV/Ris 1 Wind Turbine Concepts9extensiveuseofpowerelectronicscanalso be expected. Also,moreflexiblecontrolsystemscanbe foreseenasanincreasingnumberof computersandsensorsareincorporatedto allow for adaptive operation. In this context, it is possible that a shift may take place from focusing on wind turbine control to focusing on wind farm control. A development which willinevitably,putotherrequirementson the individual wind turbine in addition to the ones we are experiencing today. REFERENCES DanishEnergyAgency,TechnicalCriteria fortheDanishApprovalSchemeforWind Turbines, 2000. IEC61400-12Windturbinegenerator systems,part12:Windturbinepower performance testing, 1st edition, 1999. IECWT0,IECSystemforConformity TestingandCertificationofWindturbines, RulesandProcedures,International ElectrotechnicalCommission,1stedition, 2001-04. Petersen,H.,Comparisonofwindturbines basedonpowercurveanalysis,Helge PetersenConsult,DarupAssociatesLtd., 1998. Guidelines for Design of Wind Turbines DNV/Ris 2 Safety and Reliability102.Safety and Reliability 2.1Safety philosophy Awindturbineshouldbedesigned, dimensionedandmanufacturedinsucha way that it, if correctly used and maintained overitsanticipatedservicelife,can withstandtheassumedloadswithinthe prescribedlevelofsafetyandpossessa sufficientdegreeofdurabilityand robustness. Calculation, or a combination of calculationandtesting,canbeusedto demonstrate that the structural elements of a windturbinemeettheprescribedlevelof safety. Theprescribedlevelofsafetyfora structural element can be expressed in terms ofarequirementfortheprobabilityof failure and be determined on the basis of so-called risk acceptance criteria. It depends on thetypeandconsequenceoffailure.The typeoffailurecanbecharacterisedbythe degree of ductility and the amount of reserve capacityorstructuralredundancy.The consequenceoffailurecanbecharacterised intermsofthefatalitiesandsocietal consequences involved. The more severe the consequenceis,andthemorelimitedthe reservecapacityis,thesmalleristhe acceptablefailureprobability.The prescribed safety is standardised in terms of safetyclassesasdescribedinmoredetailin Section2.3.Adistinctionismadebetween low,normalandhighsafetyclass.The higherthesafetyclassis,theheavieristhe requirementforthelevelofsafety,i.e.the smaller is the acceptable failure probability. Awindturbineisequippedwithacontrol andprotectionsystem,whichdefinesan envelopeofpossibledesignsituationsthat thewindturbinewillexperience.Tokeep thewindturbinewithinthisenvelope,itis partofthesafetyphilosophythatthe protection system shall possess a sufficiently highlevelofreliabilitytorenderthejoint probabilitynegligiblethatafailureshould occurduringanextremeeventandthatthe protectionsystemshouldbeunabletofulfil itstask.Usually,non-redundantstructural partsoftheprotectionsystemaretherefore designed to high safety class. Theprescribedlevelofsafetyorthechoice ofsafetyclassmaydifferfordifferentparts ofthewindturbine.Therotorisusually designedtoatleastnormalsafetyclass. Otherstructuralpartssuchastowerand foundationareusuallyassignedtosafety classesaccordingtothepossible consequencesofafailure.Sincefailureof thefoundationwillhaveconsequencesfor thetowerandtherotor,whilefailureofthe rotororthetowermaynotnecessarilyhave consequencesforthefoundation,an attractiveapproachtothechoiceofsafety classesforvariousstructuralpartscouldbe toattemptaso-calledfail-gracesequence, i.e.asequenceoffailuresinwhichthe foundationwillbethelaststructuralpartto fail.Thissequenceisbasedonafeasibility studyoftheconsequencesoffailureforthe windturbinestructureanditsfoundation only. However, usually the rotor is designed tonormalorhighsafetyclass,whichisnot necessarilyinaccordancewiththeabove fail-gracephilosophy.Requirementsfor designingawindturbinerotortonormalor highsafetyclassderivefromthehazards thattherotorposesonitssurroundings, whentheturbinerunsaway,orwhenthe rotorfails.Insuchevents,partsoftherotor may be shed in distances up to one kilometre orevenfartherawayfromtheturbine location. Itshouldbenotedthatthelevelofsafetyis usuallytheresultofatrade-offwith economy.InDS472,emphasisisplacedon safety.However,intermsofoffshore turbinesitisinevitablethateconomical aspects will becomemore predominant than Guidelines for Design of Wind Turbines DNV/Ris 2 Safety and Reliability11theyareonshore,andthatmoreemphasis will eventually be placed on financial factors whenitcomestosafetyissuesand determination of acceptable safety levels. Limitstatedesignisusedtoachievethe prescribedlevelofsafety.Itiscommonto verifythesafetyofawindturbinewith respect to the following limit states: ultimate limit state serviceability limit state accidental limit state Forthispurpose,designloadsarederived from multiplying characteristic loads by one ormorepartialsafetyfactors,anddesign capacitiesarederivedfromdividing characteristiccapacitiesbyoneorseveral otherpartialsafetyfactors.Partialsafety factorsareappliedtoloadsandmaterial strengthstoaccountforuncertaintiesinthe characteristicvalues.Verificationofthe structural safety is achieved by ensuring that thedesignload,orthecombinationofaset ofdesignloads,doesnotexceedthedesign capacity.Incaseacombinationofloadsis used,itshouldbenotedthatitiscommon alwaystocombineoneextremeloadwith oneorseveralnormalloads.Twoor severalextremeloadsareusuallynot combined,unlesstheyhavesome correlation. Characteristicloadsandcharacteristic capacities constitute important parameters in thedesignprocess.Characteristicloadsfor assessmentoftheultimatelimitstateare usually determined as load values with a 50-yearrecurrenceperiod,andtheyare thereforeofteninterpretedasthe98% quantileinthedistributionoftheannual maximumload.Thischoicedoesnot necessarilyimplythatadesignlifetimeof exactly50yearsisconsidered.Itismorea matteroftraditionandconvenience.Nor shoulditbetakenasa50-yearguarantee withinwhichfailureswillnotoccur.For assessmentoffatigue,adesignlifetimeis needed,andinthiscontextitiscommonto considera20-yeardesignlifetimeforwind turbines.Characteristiccapacitiesare usuallychosenaslowquantilesinthe associated capacity distributions. The partial safetyfactorsthatareappliedinthedesign accountforthepossiblemoreunfavourable realisationoftheloadsandcapacitiesthan thoseassumedbythechoiceof characteristicvalues.Noteinthiscontext that some of the partial safety factors, which arespecifiedinstandards,arenotsafety factors in the true sense, but rather reduction factorswhichaccountfordegradation effects,scaleeffects,temperatureeffects, etc.andwhichhappentoappearinthe designexpressionsinexactlythesame manner as true partial safety factors. With structural safety being amajor goal of thedesign,itisimportanttomakesurethat the characteristic values of load and material quantities, which have been assumed for the design,areachievedinpractice.Non-destructivetestingofcompletedstructural parts plays a role in this context, and control ofworkmanshipanother.Material certificatesalsocomeinhandyinthis context.Ingeneral,onemaysaythat inspectionisanimportantpartofthesafety philosophy.Notleast,asitwillallowfor verification of assumptions made during the designandfortakingremedialactionsif averseconditionsaredetectedduringthe service life of the wind turbine. Fordetailsaboutstructuralsafetyandlimit statedesign,referenceismadetoSection 2.3.Fordetailsaboutcombinationsof designsituationsandexternalconditions,as wellasadefinitionofloadcases,reference is made to Chapter 4. 2.2Systemsafetyandoperational reliability Awindturbineistobeequippedwith controlandprotectionsystemswhichare Guidelines for Design of Wind Turbines DNV/Ris 2 Safety and Reliability12meanttogovernthesafeoperationofthe wind turbine and to protect the wind turbine fromillconditions.Somecomponentswill actinboththecontrolandtheprotection function,butdistinctionismadeinthatthe controlsystemmonitorsandregulatesthe essentialoperatingparameterstokeepthe turbinewithindefinedoperatingrange whereastheprotectionsystemensuresthat theturbineiskeptwithinthedesignlimits. Theprotectionsystemmusttakes precedence over the control system. 2.2.1Control system Controlsareusedforthefollowing functions: to enable automatic operation tokeeptheturbineinalignmentwith the wind to engage and disengage the generator to govern the rotor speed to protect the turbine from overspeed or damage caused by very strong winds tosensemalfunctionsandwarn operatorsoftheneedformaintenance or repair Thecontrolsystemismeanttocontrolthe operationofthewindturbinebyactiveor passivemeansandtokeepoperating parameterswithintheirnormallimits. Passivecontrolsusetheirownsensingand areexercisedbyuseofnaturalforces,e.g. centrifugal stalling or centrifugal feathering. Activecontrolsuseelectrical,mechanical, hydraulicorpneumaticmeansandrequire transducerstosensethevariablesthatwill determine the control action needed. Typical variablesandfeaturestobemonitoredin this respect include:rotor speed wind speed vibration external temperature generator temperature voltage and frequency at mains connec-tionconnection of the electrical load power output cable twist yaw error brake wear Thecontrolsystemismeanttokeepthe windturbinewithinitsnormaloperating range.Asaminimum,thenormaloperating rangeshouldbecharacterisedbythe following properties and requirements: amaximum10-minutemeanwind speedathubheight,Vmax,i.e.thestop windspeedbelowwhichthewind turbine may be in operationamaximumlong-termmeannominal powerPnom,interpretedasthehighest poweronthepowercurveofthewind turbineinthewindspeedinterval [Vmin;Vmax], where Vmin denotes the start wind speed for the turbine a maximum nominal power Pmax, which on average over 10 minutes may not be exceeded for a wind speed at hub height of V10min,hub < Vmax amaximumoperatingfrequencyof rotation nr,max for the wind turbine amaximumtransientfrequencyof rotation nmax for the wind turbine awindspeedbelowwhichthewind turbine may be stopped Thewindturbineiskeptwithinitsnormal operatingrangebymeansofthecontrol system,whichactivatesand/ordeactivates the necessary controls, e.g.: yaw (alignment with the wind) blade angle regulationactivation of the brake systempower network connectionpower limitation shutdownatlossofelectricalnetwork or electrical load Guidelines for Design of Wind Turbines DNV/Ris 2 Safety and Reliability13Inaddition,itmustbepossibletostopthe windturbine,e.g.forthepurposeof inspectionandrepairs,orincaseof emergencies.Monitoringofthecontrol systemanditsfunctionsmustbeadaptedto the actual design of the wind turbine. Design ofawindturbinecontrolsystemrequiresa backgroundinservotheory,i.e.theory dealing with control of continuous systems. Thecontrolsystemisofparticular importanceinareaswhereweakgridsare encountered.Weakgridscan,forexample, befoundinsparselypopulatedareaswhere thecapacityofthegridscanoftenbea limitingfactorfortheexploitationofthe windresourceinquestion.Twoproblems are identified in this context: increase of the steady-state voltage level of the grid above the limit where power consumptionislowandwindpower input is high voltagefluctuationsabovetheflicker limitmayresultfromfluctuatingwind power input caused by fluctuating wind and wind turbine cut-ins. The solution to the above problems is to use aso-calledpowercontrolaspartofthe controlsystem.Thepowercontrolconcept impliesbufferingthewindturbinepowerin periodswherethevoltagelimitsmaybe violated and releasing it when the voltage is lower.Thismethodiscombinedwitha smootheningofthepoweroutput,suchthat fluctuations are removed, in particular those that would exceed the flicker limit. 2.2.2Protection system The protection system is sometimes referred toasthesafetysystem.Mechanical, electricalandaerodynamicprotection systems are available. The protection system is to be activated when, as a result of control system failure or of the effects of some other failureevent,thewindturbineisnotkept withinitsnormaloperationrange.The protectionsystemshallthenbringthewind turbinetoasafeconditionandmaintainthe turbineinthiscondition.Itisusually requiredthattheprotectionsystemshallbe capable of bringing the rotor to rest or to an idling state from any operating condition. In theIECstandardanadditionalrequirement isthatmeansshallbeprovidedforbringing therotortoacompletestopfroma hazardous idling state in any wind speed less thantheannualextremewindspeed.The activationlevelsfortheprotectionsystem havetobesetinsuchawaythatdesign limits are not exceeded. Situationswhichcallforactivationofthe protectionsysteminclude,butarenot necessarily limited to: overspeed generator overload or fault excessive vibration failuretoshutdownfollowingnetwork loss, disconnection from the network, or loss of electrical load abnormalcabletwistowingtonacelle rotation by yawing Theprotectionsystemshouldthereforeasa minimumcovermonitoringofthe following: rotational speed or rotational frequency overloadofageneratororotherenergy conversion system/load extreme vibrations in the nacelle safety-related functioning of the control system Asoverspeedisbyfarthemostcritical error,rotationalspeedmonitorsforma crucial element of the protection system. A protection system consists of: a registering unit an activating unit a brake unit Theprotectionsystemshallincludeoneor moresystems(mechanical,electricalor aerodynamic)capableofbringingtherotor torestortoanidlingstate.IntheDanish Guidelines for Design of Wind Turbines DNV/Ris 2 Safety and Reliability14standard at least two brake systemsmust be included, and at least one of these must have an aerodynamically operated brake unit. See Section 2.3.3. Toensureimmediatemachineshutdownin caseofpersonalrisk,anemergencystop button,whichwilloverruleboththecontrol andnormalprotectionsystem,shallbe provided at all work places. Inadditiontowhatisstatedabove,the protectionsystemis,asaminimum,tobe subjected to the following requirements: theprotectionsystemmusttake precedence over the control system theprotectionsystemmustbefail-safe in the event that the power supply fails structural components in mechanisms of theprotectionsystemshallbedesigned to high safety class theprotectionsystemmustbeableto registerafaultandtobringthewind turbinetoastandstillortocontrolled freewheelinginallsituationsinwhich the rotor speed is less than nmax. theprotectionsystemmustbetolerant towards a single fault in a sensor, in the electronicandelectricalaswellasthe hydraulicsystemsorinactive mechanicaldevices,i.e.anundetected fault in the system must not prevent the systemfromdetectingafaultcondition and carrying out its function thereliabilityoftheprotectionsystem mustensurethatsituationscausedby failuresintheprotectionsystem, wherebytheextremeoperating rangeis exceeded, can be neglected The reliability of the protection systemmay beensured by meansofeither(1)theentire protection system being of a fail-safe design, or(2)redundancyofthepartsofthe protectionsystemwhereitcannotbemade fail-safe,or(3)frequentinspectionsofthe functioningoftheprotectionsystem,in whichriskassessmentisusedtodetermine the interval between inspections. Fail-safeisadesignphilosophy,which throughredundancyoradequacyofdesign in the structure, ensures that in the event of a failure of a component or power source, the wind turbine will remain in a non-hazardous condition. 2.2.3Brake system Thebrakesystemistheactivepartofthe protectionsystem.Examplesofbrake systems are: mechanical brake aerodynamic brake generator brake Anaerodynamicbrakesystemusually consistsofturningthebladetipor,as commonlyseenonactivestall-andpitch-controlledturbines,ofturningtheentire blade90aboutthelongitudinalaxisofthe blade.Thisresultsinaerodynamicforces thatcounteracttherotortorque.Also, spoilersandparachuteshavebeenusedas aerodynamic brakes. Figure2-1.Tipbrake,fromwww.windpower.org (2000),DanishWindTurbineManufacturers' Association. Thereliabilityofabrakesystemisofthe utmost importance to ensure that the system willserveitspurposeadequately.Inthis respect,itisimportanttobeawareof possibledependenciesbetweendifferent brakesordifferentbrakecomponents.For example,ifallthreebladesareequipped withtipbrakes,somedependencybetween thethreetipbrakescanbeexpected,cf.the commoncausefailuresthatcanbeforeseen forthesebrakes.Thiswillinfluencethe overallreliabilityagainstfailureofthe systemofthethreetipbrakesandneedsto Guidelines for Design of Wind Turbines DNV/Ris 2 Safety and Reliability15betakenintoaccountifthefailure probability exceeds 0.0002 per year. Brakes or components of brake systems will be subject to wear. Thus, current monitoring and maintenance are required. IEC61400-1requiresthattheprotection systemshallincludeoneormoresystems, i.e.mechanical,electrical,oraerodynamic brakes,capableofbringingtherotortorest ortoanidlingstatefromanyoperating condition. At least one of these systems shall act on the low-speed shaft or on the rotor of thewindturbine.Theideabehindthisisto haveabrakesystemwhichensuresthata faultwillnotleadtoacompletefailureof the wind turbine. DS472ismorestrictbyrequiringatleast twofail-safebrakesystems.Ifthetwo systems are not independent, i.e. if they have some parts in common, then the turbine shall automaticallybe brought to a complete stop ortocontrolledidlingintheeventofa failureinthecommonparts.Atleastone brakesystemisrequiredtohavean aerodynamic brake unit. 2.2.4Failuremodeandeffects analysis Afailuremodeandeffectsanalysisisa qualitativereliabilitytechniquefor systematicanalysisofmechanicalor electricalsystems,suchasawindturbine protectionsystem.Theanalysisincludes examinationofeachindividualcomponent ofthesystemfordeterminationofpossible failuremodesandidentificationoftheir effectsonthesystem.Theanalysisisbased onaworksheetthatsystematicallylistsall components in the system, including: component name function of component possible failure modes causes of failure how failures are detected effectsoffailureonprimarysystem function effects of failure on other components necessary preventative/repair measures The failure mode and effects analysis can be supplemented by a criticality analysis, which isaprocedurethatratesthefailuremodes accordingtotheirfrequencyorprobability ofoccurrenceandaccordingtotheir consequences.Theassignedratingscanbe used to rank the components with respect to theircriticalityforthesafetyofthesystem. An example of a worksheet is given in Table 2-1. Afailuremodeandeffectsanalysiscanbe conductedatvariouslevels.Before commencing,itisthusimportanttodecide whatlevelshouldbeadoptedassomeareas mayotherwisebeexaminedingreatdetail, while others will be examined at the system levelonlywithoutexaminationofthe individualcomponents.Ifconductedattoo detailedalevel,theanalysiscanberather time-consumingandtedious,butwill undoubtedlyleadtoathoroughunder-standing of the system. Table 2-1. Example of worksheet. COMPONENTFAILURE MODE FAILURE CAUSE FAILUREEFFECT FAILUREDETECTION FREQUENCY RATING SEVERITY RATING ValveLeak past stem Deteriorated seal Oil leakVisual by ROV LowLow Fails to close on command Control system failure Valve will not shut off flow Flow does not shut off MediumLow Guidelines for Design of Wind Turbines DNV/Ris 2 Safety and Reliability16Thefailuremodeandeffectsanalysisis primarilyariskmanagementtool.The strengthofthefailuremodeandeffects analysisisthat,ifcarriedoutcorrectly,it identifies safety-critical components where a singlefailurewillbecriticalfortheentire system.Itisaweakness,however,thatit depends on the experience of the analyst and thatitcannoteasilybeappliedtocover multiple failures. 2.2.5Fault tree analysis A fault tree is a logical representation of the manyeventsandcomponentfailuresthat maycombinetocauseonecriticalevent suchasasystemfailure.Ituseslogic gates (mainly AND and OR gates) to show howbasiceventsmaycombinetocause the critical top event. Application Fault tree analysis has several potential uses inrelationtowindturbineprotection systems: Infrequencyanalysis,itiscommonto quantify the probability of the top event occurringbasedonestimatesofthe failure rates of each component. The top eventmaycompriseanindividual failurecase,orabranchprobabilityin an event tree. In risk presentation, it may also be used toshowhowthevariousrisk contributorscombinetoproducethe overall risk. Inhazardidentification,itmaybeused qualitatively to identify combinations of basiceventsthataresufficienttocause the top event, also known as cut sets. Construction of a fault tree Constructionofafaulttreeusually commenceswiththetopeventandthen worksitswaydownwardstothebasic events.Foreachevent,itconsiderswhat conditionsarenecessarytoproducethe event,anditthenrepresentstheseasevents atthenextlevel.Ifoneofseveralevents causesthehigherevent,itisjoinedwithan ORgate.Iftwoormoreeventsmustoccur incombination,theyarejoinedwithan AND gate. Figure 2-2. Fault tree symbols. Ifquantificationofthefaulttreeisthe objective,downwarddevelopmentshould stop once all branches have been reduced to eventsthatcanbequantifiedintermsof probabilities or frequencies of occurrence. Variousstandardsforsymbolsareused themosttypicalonesareshowninFigure 2-2.Anexampleofafaulttreeisshownin Figure 2-3. Sometypesofevents,forexampleafireor power failure, may affect many components inthesystematonce.Theseareknownas common-causefailuresandmaybe representedbyhavingthesamebasicevent occurringateachappropriateplaceinthe fault tree. Guidelines for Design of Wind Turbines DNV/Ris 2 Safety and Reliability17Combinationoffrequenciesandproba-bilitiesBothfrequenciesandprobabilitiescanbe combined in a fault tree, providing the rules in Table 2-2 are followed. Table2-2.RulesforCombiningFrequenciesand Probabilities GateInputsOutputs ORProbability + ProbabilityProbability ANDFrequency + FrequencyFrequency Frequency + ProbabilityNot permitted Probability ProbabilityProbability Frequency FrequencyNot permitted Frequency ProbabilityFrequency Figure 2-3. Example of fault tree. Minimal cut set analysis Cutsetscomprisecombinationsofevents thataresufficienttocausethetopevent. Minimalcutsetscontaintheminimumsets ofeventsnecessarytocausethetopevent, i.e.aftereliminatinganyeventsthatoccur morethanonceinthefaulttree.Incaseof simple fault trees, with each basic event only occurringonce,theminimalcutsetscanbe identified by meansofinspection. Formore complextrees,formalmethodssuchasthe Booleananalysisarerequired.Mostoften, computerprogramsareusedtoidentify minimal cut sets. Minimalcutsetscanbeusedinhazard identificationtodescribecombinationsof events necessary to cause the top event. Minimalcutsetscan,moreover,beusedto rankandscreenhazardsaccordingtothe numberofeventsthatmustoccur simultaneously. In principle, single event cut sets are of concern because only one failure canleadtothetopevent.Inreality,larger cutsetsmayhaveahigherfrequencyof occurrence. Nevertheless, the method can be usefulforhazardscreeningandfor suggesting where additional safeguards may be needed. Quantification of a fault tree Simple fault trees may be analysed by using a gate-by-gate approach to determine the top eventprobability,providedthatallevents areindependentandthatthereareno commoncausefailures.Thisgate-by-gate approachisusefulforQRA(Quantitative RiskAnalysis),becauseitquantifiesall intermediateeventsinthefaulttreeand providesagoodinsightintothemain contributorstothetopeventandthe effectivenessofsafeguardsrepresentedin thetree.However,becauseitcannot representrepeatedeventsordependencies correctly,itisnormallynotusedforformal reliabilityanalysis.Reliabilityanalysisof morecomplexfaulttreesrequiresminimal cut set analysis to remove repeated events. Strengths and weaknesses Preparationandexecutionofafaulttree analysisisadvantageousinthesensethatit forcesthewindturbinemanufacturerto examinetheprotectionsystemofhiswind turbinesystematically.Whenprobabilities canbeassignedtoevents,thefaulttree methodologycanbeappliedtoassessthe overallreliabilityofthewindturbine Guidelines for Design of Wind Turbines DNV/Ris 2 Safety and Reliability18protectionsystem.Furthermore,itcanbe usedtodeterminethemostcriticalevents and parts of the protection system. Beware, however, that it is often hard to find outwhetherafaulttreeanalysishasbeen carriedoutproperly.Faulttreeanalyses becomecomplicated,time-consumingand difficulttofollowforlargesystems,andit becomes easy to overlook failure modes and common cause failures. It is a weakness that thediagrammaticformatdiscourages analystsfromstatingassumptionsand conditionalprobabilitiesforeachgate explicitly.Thiscanbeovercomebycareful back-uptextdocumentation.Itisa limitationthatalleventsareassumedtobe independent.Faulttreeanalyseslosetheir claritywhenappliedtosystemsthatdonot fall into simple failed or working states such as human error, adverse weather conditions, etc. 2.3Structural safety 2.3.1Limit states Duringthelifetimeofastructure,the structureissubjectedtoloadsoractions. Theloadsmaycauseachangeofthe conditionorstateofthestructurefroman undamagedorintactstatetoastateof deterioration,damage,orfailure.Structural malfunction can occur in a number of modes coveringallfailurepossibilitiesthatcanbe imaginedforthestructure.Althoughthe transitionfromanintactstatetoastateof malfunctioncanindeedbecontinuous,itis commontoassumethatallstateswith respecttoaparticularmodeofmalfunction canbedividedintotwosets:1)statesthat have failed, and 2) states that have not failed oraresafe.Theboundarybetweenthesafe statesandthefailedstatesisreferredtoas thesetoflimitstates.Thesafetyor reliabilityofastructureisconcernedwith how likely it is that the structure will reach a limit state and enter a state of failure. Thereareseveraltypesoflimitstates.Two typesarecommon,ultimatelimitstatesand serviceabilitylimitstates.Ultimatelimit statescorrespondtothelimitoftheload-carrying capacity of a structure or structural component, e.g. plastic yield, brittle fracture, fatiguefracture,instability,buckling,and overturning. Serviceability limit states imply thattherearedeformationsinexcessof tolerancewithoutexceedingtheload-carryingcapacity.Examplesarecracks, wear,corrosion,permanentdeflectionsand vibrations. Fatigue is sometimes treated as a separatetypeoflimitstate.Othertypesof limit states are possible, e.g. accidental limit states and progressive limit states. 2.3.2Failureprobabilityandother measures of structural reliability Instructuraldesign,thereliabilityofa structuralcomponentisevaluatedwith respecttooneormorefailuremodes.One suchfailuremodeisassumedinthe following.Thestructuralcomponentis describedbyasetofstochasticbasic variablesgroupedintoonevectorX, including,e.g.itsstrength,stiffness, geometry,andloading.Eachofthese variables are stochastic in the sense that they owingtonaturalvariabilityandother possible uncertainties may take on random degreesofrealisationaccordingtosome probabilitydistribution.Fortheconsidered failuremode,thepossiblerealisationofX canbeseparatedintotwosets:1)thesetfor which the structural component will be safe, and2)thesetforwhichitwillfail.The surfacebetweenthesafesetandthefailure set in the space of basic variables is denoted thelimitstatesurface,andthereliability problemisconvenientlydescribedbyaso-calledlimitstatefunctiong(X),whichis defined such that Guidelines for Design of Wind Turbines DNV/Ris 2 Safety and Reliability19 setfailure infor 0surface state limitonfor 0 set safe infor 0) (XXXX g Thelimitstatefunctionisusuallybasedon somemathematicalengineeringmodelfor theconsideredlimitstate,basedonthe underlyingphysics,andexpressedinterms ofthegoverningloadandresistance variables. Thefailureprobabilityistheprobability content in the failure set [ ]= =0 ) () ( 0 ) (Xx x XXgd f g P PF wherefX(x)isthejointprobabilitydensity function for X and represents the uncertainty andnaturalvariabilityinthegoverning variablesX.ThecomplementPS=1PFis referredtoasthereliabilityandis sometimesalsodenotedtheprobabilityof survival. The reliability may be expressed in terms of the reliability index, = 1(PF), whereisthestandardisednormal distribution function. The failure probability, thereliability,andthereliabilityindexare all suitable measures of structural safety. ForthesimpleexamplethatXconsistsof twovariables,theloadLandtheresistance R,andthelimitstatefunctioncanbe specifiedasg(X)=RL,thefailure probabilitybecomesasimpleconvolution integral [ ] < === < =dl l f l Fdrdl l f r fdrdl l f r f L R P PL RlL RL RL R F) ( ) () ( ) () ( ) ( 00 wherefRandfLaretheprobabilitydensity functionsofRandL,respectively,andfR(r) = dFR(r)/dr, where FR is the cumulative distribution function of R. 2.3.3Structural reliability methods Thereliabilityindexcanbesolvedina structuralreliabilityanalysisbymeansofa reliability method which can be any amongst severalavailablemethods,including numericalintegration,analyticalfirst-and second-orderreliabilitymethods,and simulationmethods.Referenceismadeto Madsenetal.(1986).Someofthese methodsareapproximatemethods,which willleadtoapproximateresultsforthe reliabilityindex.Numericalintegrationis usuallyonlyfeasiblewhenXconsistsof veryfewstochasticvariablessuchasinthe example above. Analytical first- and second-ordersolutionstothefailureprobabilityare oftensufficientlyaccurate,andtheyare advantageoustosimulationresultswhen failure probabilities are small. A useful by-product of a structural reliability analysisbythesemethodsistheso-called designpointx*.Thisisapointonthelimit statesurfaceandisthemostlikely realisationofthestochasticvariablesXat failure. 2.3.4Codeformat,characteristic values, and partial safety factors Astructuraldesigncodespecifiesdesign rulesthataretobefulfilledduringthe designofastructureorstructural component. The general layout of the design Guidelines for Design of Wind Turbines DNV/Ris 2 Safety and Reliability20rulesinadesigncodeisknownasthecode format. Thecodeformatmostfrequentlyusedin designcodestodayisaformatwhichis expressedintermsofdesignvaluesof governingloadandresistancevariables. Thesedesignvaluesaredefinedas characteristicvaluesoftheloadand resistancevariables,factoredbypartial safetyfactors.Suchaformofcodeformat resultsfromrequirementsforaneasyand yeteconomicaldesignandisknownasa designvalueformat.Designaccordingtoa design value format is sometimes referred to as load and resistance factor design (LRFD). In its simplest form, a code requirement can be expressed as a design rule in terms of an inequality LD < RD in which LD is the design load effect and RD isthedesignresistance.Thedesignload effect is calculated as LD = f LC where LC is the characteristic load effect and fisaloadfactor.Similarly,thedesign resistance is calculated as RD = mCR where RC is the characteristic resistance and m is a material factor. Usually,anumberofdifferentloadeffects havetobecombinedintoaresultingload effect,andoftenthelargestofseveral different such load combinations is used for theloadeffectinthedesignrule.An exampleofsuchaloadcombinationisthe combination of gravitational loads and wind loads. Thecharacteristicvaluesareusuallytaken asspecificquantilesintheloadand resistancedistributions,respectively.The characteristicvaluesareoftenthemean value for dead load, the 98% quantile in the distributionoftheannualmaximafor variable (environmental) load, and the 2% or 5% quantile for strength. Reference is made to Figure 2-4. The code also specifies values to be used for the partial safety factors f and m . fL, fR L R L, R LCRC Figure2-4.Probabilitydensityfunctionsand characteristic values for load L and resistance R. Thedesignequationisaspecialcaseofthe designruleobtainedbyturningthe inequalityintoanequality,i.e.,RD =LDin the example. 2.3.5Code calibration Fulfilment of the design rules, as required by astructuraldesigncode,ismeanttoensure thataparticularprescribedstructuralsafety levelisachieved.Thepurposeofacode calibrationistodeterminethesetofpartial safetyfactorstobeusedwiththechosen codeformat,suchthatstructuraldesigns accordingtothecodewillmeetthis prescribed level of safety. Structuralreliabilityanalysisresults, obtainedasoutlinedabove,playan importantroleincodifiedpracticeand design.Theirapplicationtocalibrationof partialsafetyfactorsforuseinstructural designcodesisofparticularinterest. Guidelines for Design of Wind Turbines DNV/Ris 2 Safety and Reliability21Structuralreliabilityanalysisdirectly capitalisesonthevariabilityand uncertaintiesinloadandstrengthand thereforeproducesareliabilityestimate, whichisadirectmeasureofthestructural safety.Becauseofthis,structuralreliability analysisformstherationalbasisfor calibrationofpartialsafetyfactors,which are used in code checks during conventional deterministicdesign.Withstructural reliabilitymethodavailable, itispossibleto determinesetsofequivalentpartialsafety factorswhich,whenappliedwithdesign rulesinstructuraldesigncodes,willleadto designs with the prescribed reliability. Asafirststep,atargetreliabilityindexT mustbeselected.Thechoiceofthetarget reliabilityindexcanbederivedfroma utility-basedfeasibilitystudyinadecision analysis, or by requiring that the safety level asresultingfromthedesignbyastructural reliabilityanalysisshallbethesameasthe oneresultingfromcurrentdeterministic design practice. The latter approach is based ontheassumptionthatestablisheddesign practice is optimal with respect to safety and economyor,atleast,leadstoasafetylevel acceptable by society. WhenthetargetreliabilityindexTcannot beestablishedbycalibrationagainst establisheddesignpractice,orotherwise, then its value may be taken from Table 2-3, dependingonthetypeandconsequenceof failure. Note that the numbers given in Table 2-3aregivenforareferenceperiodofone year, i.e. they refer to annual probabilities of failure and corresponding reliability indices. Reference is made to NKB (1978). Inthecaseofaprescribedreliabilityindex, whichisdifferentfromtheonethatresults from an actually executed reliability analysis ofastructuralcomponent,thegeometrical quantitiesofthiscomponentmustbe adjusted.Theadjustmentismadeinsucha waythattherequiredreliabilityindexwill resultfromanewreliabilityanalysisofthe modifiedcomponent.Thegeometrical quantities, which can be adjusted to achieve aspecifiedreliabilityindex,aresometimes denoteddesignparameters.Itismost practicabletooperateonjustonesuch designparameterwhenadjustingthedesign inordertoreachthespecifiedreliability index. Adesigncaseisformedasaspecific combinationofenvironmentalloading regime, type of material, and type and shape ofstructure.Foraparticulardesigncase, whichcanbeanalysedonthebasisofa structuralreliabilitymethod,asetofpartial safetyfactorscanthusbedeterminedthat will lead to a design which exactly meets the prescribedlevelofreliability.Differentsets ofpartialsafetyfactorsmayresultfrom differentdesigncases. Asimpleexampleof a calibration of partial safety factors is given belowforanaxiallyloadedsteeltruss.A structuraldesigncodeusuallyhasascope thatcoversanentireclassofdesigncases, formedbycombinationsamongmultiple environmentalloadingregimes,different structuralmaterials,andseveraltypesand shapesofstructures.Thedesigncodewill usuallyspecifyonecommonsetofpartial safetyfactors,whichistobeapplied regardlessofwhichdesigncaseisbeing analysed.Thispracticalsimplification implies that the prescribed level of reliability willusuallynotbemetinfull,butonly Guidelines for Design of Wind Turbines DNV/Ris 2 Safety and Reliability22Table 2-3. Target annual failure probabilities PFT and corresponding reliability indices T. Failure consequence Failure typeLess serious LOW SAFETY CLASS (small possibility for personal injuries and pollution, small economic consequences, negligible risk to life) Serious NORMAL SAFETY CLASS (possibilities for personal injuries, fatalities, pollution, and significant economic consequences) Very serious HIGH SAFETY CLASS (large possibilities for personal injuries, fatalities, significant pollution, and very large economic consequences) Ductile failure with reserve capacity (redundant structure) PF = 103 T = 3.09 PF = 104 T = 3.72 PF = 105 T = 4.26 Ductile failure with no reserve capacity (significant warning before occurrence of failure in non-redundant structure) PF = 104 T = 3.72 PF = 105 T = 4.26 PF = 106 T = 4.75 Brittle failure(no warning before occurrence of failure in non-redundant structure) PF = 105 T = 4.26 PF = 106 T = 4.75 PF = 107 T = 5.20 approximately, when designs are carried out accordingtothecode.Hence,thegoalofa reliability-basedcodecalibrationisto determinetheparticularcommonsetof partial safety factors that reduces the scatter ofthereliabilities,achievedbydesigns accordingtothecode,toaminimumover thescope.Thiscanbeaccomplishedby meansofanoptimisationtechnique,oncea closenessmeasurefortheachieved reliabilityhasbeendefined,e.g.expressed in terms of a penalty function that penalises deviationsfromtheprescribedtarget reliability.Forprinciplesandexamplesof such code optimisation, reference is made to Haugeetal.(1992),Ronold(1999),and Ronold and Christensen (2001). 2.3.6Exampleaxiallyloadedsteel tower The example given below deals with design ofanaxiallyloadedsteeltoweragainst failure in ultimate loading. The probabilistic modellingrequiredforrepresentationof loadandcapacityispresented.Astructural reliabilityanalysisofthetoweriscarried out, and a simple calibration of partial safety factors is performed. Thedesignoftheaxiallyloadedtoweris governed by the maximum axial force Q in a Guidelines for Design of Wind Turbines DNV/Ris 2 Safety and Reliability23one-yearreferenceperiod.Themaximum axial force Q follows a Gumbel distribution FQ(q) = exp(exp(a(qb))) inwhicha=0.4275andb=48.65 correspondtoameanvalueE[Q]=50MN and a standard deviation D[Q] = 3 MN. The yieldstrengthofsteelFfollowsanormal distributionwiththemeanvalueE[F]=400 MPa and the standard deviation D[F] = 24MPa.Thecross-sectionalareaofthe towerisA.Failureoccurswhentheaxial forceQexceedsthecapacityFA,thusa naturalformatofthedesignruleisF,DA qD.ThesubscriptDdenotesdesignvalue. Thelimitstatefunctioniscorrespondingly chosen as g = FAQ andtheareaAisusedasthedesign parameter. Analysisbyafirst-orderreliabilitymethod leadstodeterminationofA=0.2007m2in ordertomeetatargetreliabilityindex = 4.2648, which corresponds to an annual failureprobabilityPF=105.The characteristicvalueoftheaxialforceis taken as the 98% quantile in the distribution oftheannualmaximumforce,qC=q98%= 57.78MN.Thecharacteristicvalueofthe yield strength is taken as the 5% quantile in the strength distribution, F,C = F,5% = 360.5 MPa.Onepartialsafetyfactor,1,is introducedasafactoronthecharacteristic force, and another one, 2, is introduced as a factoronthecharacteristiccapacity. Substitutionoftheexpressionsforthe designforceandthedesigncapacityinthe design equation yields 2F,CA1qC = 2360.50.2007157.78 = 0 which gives a requirement to the ratio of the partialsafetyfactors1/2=1.252.Thereis thusaninfinitenumberofpairs(1,2)that willleadtotherequiredreliability.This impliesanarbitrarinessinselectingthe partialsafetyfactorset(1,2)forthecode. Thereliabilityanalysisgivesthedesign pointvaluesq*=70.89MNfortheforce and F* = 353.2 MPa for the strength. These arethemostlikelyvaluesofthegoverning variablesatfailure.Arobustchoiceforthe partialsafetyfactors(1,2)canbeachieved by designing to the design point values from thereliabilityanalysis.Thepartialsafety factors are therefore selected as 227 . 1*1= =Cqqand980 . 0*,2= =C FF Accordingtocurrent design practice,aload factor is used as a factor on the characteristic loadtogivethedesignload,andamaterial factorisusedasadivisoronthe characteristicresistancetogivethedesign resistance. Hence, these factors become 227 . 11= = f and021 . 112= =m, respectively. Notethattheexampleispurelytutorialto explainaprinciple.Inreality,thecapacity maybemoreuncertainthanassumedhere, e.g.owingtomodeluncertaintynot accountedfor.Moreover,alargerdegreeof variabilityintheaxialforcemayalsobe expected,dependingonthesourceandtype of loading and the amount of data available. The resulting partial safety factors may then become larger than the ones found here. Guidelines for Design of Wind Turbines DNV/Ris 2 Safety and Reliability242.3.7Example fatigue of FRP blade root in bending The example given here deals with design of anFRPbladerootagainstfatiguefailure duringitsdesignlifeof20years.The probabilisticmodellingrequiredfor representationofloadandresistanceis presented. A structural reliability analysis of thebladerootiscarriedout,andasimple safety factor calibration is performed. The design of the blade root is governed by thelong-termdistributionofthebending momentrangeX.Inthelongterm,the bendingmomentrangesareassumedto follow an exponential distribution ) exp( 1 ) (0xxx FX = inwhichx0=50kNmisrecognisedasa Weibullscaleparameter.Thetotalnumber ofbendingmomentrangesoverthedesign lifeTL=20yearsisntot=0.9109.The bendingmomentrangesXgiveriseto bendingstressrangesS=X/W,whereW denotesthesectionmodulusoftheblade root.Hence,thebendingstressrange distribution becomes ) exp( 1 ) (0xsWs FS = ForagivenstressrangeS,thenumberof bendingstresscyclesNtofailureis generallyexpressedthroughanSNcurve, which on logarithmic form reads + = S m K N ln ln ln wherethepair(lnK,m)=(114.7,8.0) describesthe expected behaviour.Thezero-meantermrepresentsthenatural variability about the expectation and follows anormaldistributionwithastandard deviation = 0.86. The cumulative damage is calculated as the Miners sum m totmStoti iiWxmKns KdsdsdFns Ns nD) )( 1 () exp() exp() () (00+ === where denotes the gamma function. AccordingtoMinersrule,fatiguefailure occurs when the cumulative damage exceeds athresholdof1.0.Anaturalformatofthe design rule is DD 1, where DD is the design damage. The limit state function is chosen as g = 1D andthesectionmodulusWisusedasthe design parameter. Analysisbyafirst-orderreliabilitymethod leads to determination of W = 0.00209 m3 in ordertomeetatargetreliabilityindex=3.29,whichcorrespondstoatarget failureprobabilityPF=0.5103overthe design life of 20 years. Fordesignagainstfatiguefailureithas hardlyanymeaningtochoosethe98% quantile of the annual maximum load, or any otherquantileforthatmatter,asthe characteristicloadvalue.Acharacteristic loaddistributionisneededratherthana characteristicloadvalue.Thelong-term stressrangedistribution,FS(s),ischosenas thecharacteristicstressrangedistribution, and a load factor f = 1.0 on all stress ranges according to this distribution is prescribed. Guidelines for Design of Wind Turbines DNV/Ris 2 Safety and Reliability25Thisisbasedontheassumptionthatthe long-term stress range distribution is known. Thevalidityofthisassumptioninthe context of wind turbines is discussed later. It isalsoassumedthatvariabilityinthe individualdamagecontributionsfromthe individualstressrangesaveragesoutover themanycontributingstressrangesinthe long-termdistribution.Thisassumption wouldnotholdifthecumulativedamage wasdominatedbydamagecontributions fromonlyoneoraveryfewlargestress ranges,whichcouldbethecaseforvery large m values, say m>10. ThecharacteristicSNcurveistakenasthe expectedSNcurveminustwostandard deviations.Apartialsafetyfactor,m,is applied as a divisor on all stress range values of the characteristic SN curve. The design damage DD is then obtained as m mCtotDWxmKnD ) )( 1 (0+ = inwhichKC=Kexp(2)reflectsthe chosencharacteristicSNcurve. Substitutionofnumbersintothedesign equationDD=1.0leadstothefollowing requirement to the material factor m = 1.149 Discussion Notethatthisexampleispurelytutorialto explainaprinciple.Inreality,theresistance maybemoreuncertainthanassumedhere, e.g.whenthereisalimitedamountof materialdataavailable,suchthatthe estimatedvaluesofKandmareuncertain. Moreover,modeluncertaintymaybe associatedwiththeapplicationofMiners rule.Withsuchuncertaintiesproperly accountedfor,thevalueoftheresulting materialfactor,m,willbecomelargerthan the one found here.Notealsothatwithf=1.0prescribed,all levelsofuncertaintyandvariability associatedwithafatigueproblemsuchas thepresent,areaccountedforbyonesingle safetyfactor,m.Thisfactoristhusapplied as a safety factor on resistance, regardless of whethersomeoftheuncertaintyis associatedwithloadratherthanwith resistance.Thisisinaccordancewithmost standards.Note,however,thatinthenew DanishstandardDS409/DS410,apartial safety factor, f, on load is introduced which, undercertainconditions,istobetakenasa valuegreaterthan1.0.Thisappliesto situationswheretheloadscausingfatigue damageareencumberedwithuncertaintyor ambiguity,suchasiftheyaretrafficloads, orifthevariousquantilesofthelong-term stressdistributionoverthedesignlifeare statistically uncertain. Asregardsdesignofwindturbinesagainst fatigue,theloadscausingfatiguedamage aredominatedbyloadsgeneratedbythe wind.Whereasthedistributionofthe10-minutemeanwindspeedonalocationmay bewell-known,thedistributionofthe turbulenceintensityisusuallynotwell-determined,owingtolocalconditionsand influencefromthepresenceoftheturbine. Noristhetransferfunctiontostress responseinthewindturbinealwaysclear. Thedistributionofwind-generatedloadsin awindturbinecanthereforebeexpectedto be known only with some uncertainty, and a load factor f greater than 1.0 would thus be required.However,inpractice,onewould account for such uncertainty or ambiguity in theloaddistributionbychoosingaload distributiononthesafeside,a conservativeenvelopeloadspectrum,so tospeak.Inthepresentedexample,this wouldimplythechoiceofaconservatively high value of the Weibull scale parameter x0, whichcouldthenbeusedinconjunction with f = 1.0. Guidelines for Design of Wind Turbines DNV/Ris 2 Safety and Reliability262.3.8Testsandcalculationsfor verification Itisimportanttodemonstratethatthe structuralstrengthsorcapacitiesofthe various components that constitute the wind turbinestructurearesufficient.Thiscanbe donebyundertakingcalculationsaccording to some theory or calculation method, and it canbesupportedbycarryingoutfull-scale tests of the component in question. Note that afull-scaletestofastructuralcomponent maygiveamoreaccurateestimateofthe componentstrengththantheoretical calculations,becausemodeluncertaintyand bias,owingtosimplificationsand limitationsassociatedwiththeapplied calculationmethod,willbereducedor removed. Whencarryingoutafull-scaletestofa structuralcomponentsuchasablade,itis importanttoacknowledgethattheloads used in the test are generated and controlled inacompletelydifferentmannerthanthe variablenaturalloads(suchaswindloads) whichthecomponentwillexperiencein realityandwhichitshouldbedesignedfor. For selection of loads to be used in full-scale testsforverification,itisthereforenot relevanttoapplypartialsafetyfactorsfor loadsthatareprescribedfordesignin accordancewithparticularcodesand standards. The loads to be used in tests need to be chosen after thorough consideration of thevariabilityanduncertaintyinthe strength,giventhedegreeofknowledge about the strength or capacity available prior to the test. Inthiscontextitshouldbenotedthatthe effectofproofloadingrepresentsan increaseinconfidenceinthestructural strengthorcapacity,resultingfromprior successful loading of the component. 2.3.9Inspectionandinspection intervals Thedesignprocessisonlyoneelementin ensuringsafeandreliablestructures.In fabricationandservice,othersafety elementscanbeintroducedsuchasquality control, alignment control, visual inspection, instrumented monitoring, and proof loading. Eachoftheseitemsprovideinformation aboutthestructure,additionaltothe informationpresentatthedesignstage,and mayhencereducetheoveralluncertainty associatedwiththestructure.The probabilisticmodelusedindesigncanthen beupdatedandcalibratedagainstrealityby including the additional information. Theadditionalinformationobtainedduring fabricationandinservicemaybeobtained either directly as information about some of thegoverningvariablesthemselves,e.g. strength,orindirectlybyobserving substitutevariables,whicharefunctionsof thegoverningvariables,e.g.cracksor deformations. Itisofinteresttoupdatethefailure probability from its value in the design stage toavaluewhichreflectstheadditional informationgainedbyinspection. Probabilityupdatingbyinspectionisbased onthedefinitionofconditionalprobability. LetFdenotetheeventofstructuralfailure. Inthedesignprocesstheprobabilityof failurePF=P[F]issolvedaccordingtothe procedures described above. Let I denote an event such as the observation of a governing variableortheobservationofafunctionof oneormoreofthegoverningvariables, obtained by inspection when the structure is in service. The updated probability of failure istheprobabilityoffailureconditionedon the inspection event I, [ ][ ][ ] I PI F PI F P= |Guidelines for Design of Wind Turbines DNV/Ris 2 Safety and Reliability27Theprobabilityinthenumeratorcanbe solvedbyareliabilityanalysisofaparallel system,forwhichsolutionsareavailable. Theprobabilityinthedenominatorcanbe solved by a reliability analysis of a structural componentasdescribedabove,oncea suitablelimitstatefunctionhasbeen defined,andwithdueaccountfor measurementuncertaintyandprobabilityof detection,whicharetwocontributing uncertaintysourcesofimportanceinthis context. For some limit states, e.g. crack growth and fatiguefailure,thefailureprobabilityPF increasesasafunctionoftime.Forsuch limitstates,predictionofthefailure probability as a function of time can be used topredictthetimewhenthefailure probabilitywillexceedsomecritical threshold, e.g. a maximum acceptable failure probability. This predicted point in time is a naturalchoiceforexecutionofan inspection.Thefailureprobabilitycanthen beupdatedas outlinedabove,depending on thefindingsfromtheinspectionand includingimprovementsfromapossible repairfollowingtheinspection.Thetime untilthefailureprobabilitywillagain exceedthecriticalthresholdandtriggera newinspectioncanbepredictedandthus formsaninspectioninterval.Thiscanbe used to establish an inspection plan. Note in this context that for some limit states suchasthefatiguelimitstate,itmay actually be a prerequisite for maintaining the required safety level over the design life that inspectionsarecarriedoutatspecified intervals. 2.4Mechanical safety Thereareseveralmechanicalsystemsina wind turbine: transmission:hub,shaft,gear, couplings,brakes,bearingsand generator mechanicalcontrolsystems:pitch system, teeter mechanism, yaw system, hydraulic system and pneumatic system Thesafetyofmechanicalcomponentswill usuallybedeterminedbytheirstructural safetyasdescribedinprevioussections. However,asthecomponentsarepartof mechanicalsystems,thereareseveral aspectstobeconsideredinadditiontothe structuralsafetywhenthesafetyof mechanical systems is to be evaluated. Thestructuralstrengthwillinmanycases becomelimitedbysurfacedamagesdueto wearsuchasfrettingcorrosionin connectionsduetomicromovements,or graystainingofgearteethduetopoor lubricationconditions.Hence,aspectslike friction and lubrication conditions as well as surfacetreatmentareessentialforthe mechanical safety. Mechanicalcomponentsareoftenmadeof ratherbrittlehigh-strengthmaterialsuchas case-hardened steel for gears and induction-hardenedrollerbearingsteel.Furthermore, the strength of mechanical components often