Simulation of fatigue failure in a full composite wind turbine blade

Please download to get full document.

View again

of 11
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
Document Description
Simulation of fatigue failure in a full composite wind turbine blade
Document Share
Document Tags
Document Transcript
  Simulation of fatigue failure in a full composite wind turbine blade Mahmood M. Shokrieh  * , Roham Rafiee Composites Research Laboratory, Mechanical Engineering Department, Iran University of Science and Technology, Narmak, Tehran 16844, Iran Available online 13 June 2005 Abstract Lifetime prediction of a horizontal axis wind turbine composite blade is considered. Load cases are identified, calculated andevaluated. Static analysis is performed with a full 3-D finite element method and the critical zone where fatigue failure begins isextracted. Accumulated fatigue damage modeling is employed as a damage estimation rule based on generalized material propertydegradation. Since wind flow (loading) is random, a stochastic approach is employed to develop a computer code in order to sim-ulate wind flow with randomness in its nature on the blade and subsequently each load case is weighted by its rate of occurrenceusing a Weibull wind speed distribution.   2005 Elsevier Ltd. All rights reserved. Keywords:  Composites; Finite element; Stochastic; Wind turbine blade; Damage; Failure analysis 1. Introduction Pollution free electricity generation, fast installationand commissioning capability, low operation and main-tenance cost and taking advantage of using free andrenewable energies are all advantages of using wind tur-bines as an electricity generators. Along with theseadvantages, the main disadvantage of this industry isthe temporary nature of wind flow. Therefore, using reli-able and efficient equipment is necessary in order to getas much as energy from wind during the limited periodof time that it flows strongly.The blade is the most important component in a windturbine which nowadays is designed according to a re-fined aerodynamic science in order to capture the maxi-mum energy from the wind flow. Blades of horizontalaxis are now completely made of composite materials.Composite materials satisfy complex design constraintssuch as lower weight and proper stiffness, while provid-ing good resistance to the static and fatigue loading.Generally, wind turbines are fatigue critical machinesand the design of many of their components (especiallyblades) are dictated by fatigue considerations. Severalfactors expose wind turbine blades to the fatigue phe-nomena which can be summarized as shown below [1]:1. Long and flexible structures2. Vibrations in its resonant mode3. Randomness in the load spectra due to the nature of the wind4. Continuous operation under different conditions5. Low maintenance during lifetimeA wind turbine blade expects to sustain its missionfor about 20–30 years. Fig. 1 shows a comparison be-tween different industrial components and their expectedlife cycles.The above mentioned reasons and extensive expectedlifetime cause design constraints for wind turbine struc-tures that fall into either extreme load or fatigue catego-ries. For the case of extreme load design, the loadestimation problem is limited to finding a single maxi-mum load level that the structure can tolerate. For de-sign against fatigue, however, loads must be defined 0263-8223/$ - see front matter    2005 Elsevier Ltd. All rights reserved.doi:10.1016/j.compstruct.2005.04.027 * Corresponding author. Tel./fax: +98 21 749 1206. E-mail address: (M.M. Shokrieh).Composite Structures 74 (2006) 332–  for all input conditions and then summed over the distri-bution of input conditions weighted by the relative fre-quency of occurrence.Most ongoing research on fatigue phenomena usesload spectra obtained by digital sampling of a specificconfiguration of strain gauges which read the strain ata specific location near the root of blade. Then, the rep-resentative sample of each load spectrum is weighted byits rate of occurrence that can be obtained from the sta-tistical study of the wind pattern. Finally all weightedload spectra are summed and total load spectrum is de-rived. This load spectrum is utilized to estimate fatiguedamage in the blade using Miner  s rule [2]. One of themain shortcomings of this method is the linear natureof Miner  s rule. Not only is Miner  s rule not a properrule for fatigue consideration in metals, but also it hasbeen proven that this rule is not suitable for composites[3]. Another problem with using Miner  s rule is theweakness of this model to simulate the load sequenceand history of load events. This shortcoming can clearlybe seen in the difference of predicted lifetimes of bladeswith two orders of magnitudes for two load cases withdifferent load sequences [2]. Furthermore, most investi-gations in fatigue simulation of composite blades arelimited to the deterministic approach. In addition, iden-tification of a place to install the strain gauges in orderto extract the load spectrum is also questionable. Usinga massive and high cost material fatigue database [4], isanother problem with these methods. It forces research-ers to characterize each configuration of lay-up sepa-rately and is the main reason for publishing newversions of these databases each year due to introduc-tion of new lay-up configurations in blade structure byindustry. 2. State of the art In this paper, a model to study the fatigue phenom-ena for wind turbine composite blades is presented inorder to overcome the aforementioned shortcomingsof current methods. As a case study, a 23 m blade of aV47-660 wind turbine, manufactured by the VestasCompany, was selected. Firstly, loading on the bladeis considered carefully using the finite element methodand the critical zone where catastrophic fatigue failureinitiates is determined. Then, each load case is weightedand finally fatigue is studied using a developed stiffnessdegradation method. The main advantages of this meth-od can be expressed in its ability to simulate the load se-quence and load history. Due to randomness of windflow; stochastic analysis will be employed instead of deterministic analysis. Based on the capability of themethod for fatigue modeling, performing a large quan-tity of experiments in order to characterize complete fa-tigue behavior of materials is avoided. 3. Modeling The investigated blade consists of three main partscalled shell, spar and root-joint (see Fig. 2).The shell is responsible to help create the requiredpressure distribution on the blade. The cross sectionsof the blade are different airfoils based on aerodynamicconsiderations. The shell also twists about 15   due toaerodynamic reasons and also has a tapered shape fromroot to tip.A brief summary of pertinent data related to theinvestigated blade is shown in Table 1.The spar, which is also called the main beam, has tosupport loads on the blade that arise from differentsources. The shell structure carries only 20% of totalloads while the rest has to be carried by the spar. Thecross section of the spar has a box shape.The root joint is the only metallic part in the currentblade that connects whole blade structure to the hub byscrews. This metallic joint is covered by composite lam-inates internally and externally. Fig. 2. Depiction of root-joint, shell and spar. Number of Fatigue Cycles to Failure    A   l   l  o  w  a   b   l  e   C  y  c   l   i  c   S   t  r  e  s  s Commercial AircraftBridges Hydrofoils Helicopter Wind Turbine Blade 20-30 year Life 10 5 10 6 10 7 10 8 10 9 Fig. 1. Schematic  S   –  N   curves for different industrial components [1]. M.M. Shokrieh, R. Rafiee / Composite Structures 74 (2006) 332–342  333  In order to provide data for the finite element model,a geometrical model was created based on cross sectionprofiles of shell and spar using Auto-CAD [5] software.Then the wire frame model, which is the fundamentalgeometrical model, was transferred to ANSYS finite ele-ment software [6]. After that, geometrical modeling wascompleted by creating surfaces using the loft method. Inthe meshing process, second order shell elements wereemployed to increase accuracy of the modeling. In addi-tion, the selected element type is compatible with com-posites and, in order to not having any triangularelements, a manual meshing method was employed.Therefore all elements have quadratic shape with 8nodes and acceptable aspect ratio. Elements of theroot-joint segment are second order cubic solidelements.Convergence criteria should be considered to evalu-ate the results. Convergence analysis is performed on ametallic model of the blade. By improving mesh densitystep-by-step a suitable number of elements is obtained.The stabilization of tip deflection and Von-Misses equiv-alent stress at a location far from the applied loadsare the criteria of convergence. Also in order to examinethe case with both bending and torsion in the structure,the loads were applied on the trailing edge and bound-ary conditions consisted of fixing all 6 degrees of free-dom of nodes that are placed at the root. Thedepiction of the FE model is shown in Fig. 3.Fig. 4 shows the results of convergence analysis.From Fig. 4, it is clear that convergence is obtained withthe use of about 10,000 or more elements. 4. Material characterization The investigated blade consists of three main types of pre-preg glass/epoxy composites: Unidirectional plies,bi-axial and tri-axial materials.Bi-axial and tri-axial plies contain two and threesame unidirectional fabrics, respectively, which arestitched together (thus are not woven). The main reasonfor not using woven form of composites is due to disad-vantages of using them based on their fatigue properties.In the woven form, due to out-of-plane curvature of fabrics, stress concentrations happen and consequentlyfatigue performance of these materials decreases dra-matically [3].Tri-axial and bi-axial fabrics are used in the shellstructure and unidirectional and bi-axial are used inthe spar structure. The configuration of the bi-axial lam-ina is [0/90] T  and the configuration of the tri-axial lam-inate is [0/+45/  45] T . Two kinds of foam (PMI andPVC) are used respectively at the spar and shell loca-tions, respectively, in order to construct the sandwichpanel. Table 1General specifications of investigated bladeLength 22,900 mmMaximum chord 2087 mmStation of maximum chord R4500Minimum chord 282.5 mmTwist 15.17  Station of CG R8100Weight of blade 1250 kgTip to tower distance 4.5 mSurface area 28 m 2 Airfoil cross-section types FFA-W3, NACA-63-xxx, MIXFig. 3. Finite element model of turbine blade showing loads andboundary conditions for convergence consideration. Convergency Based on Stress 600109266042371221312170161541813232102848508766457284358263434281876 Number of Elements    V  o  n   M   i  s  s  e  s   S   t  r  e  s  s Convergency Based on Deflection 6001092187626346604572843583428850810284132321541817016766421312 23712 Number of Elements    T   i  p   D  e   f   l  e  c   t   i  o  n Fig. 4. Convergence graphs of FE model.334  M.M. Shokrieh, R. Rafiee / Composite Structures 74 (2006) 332–342  The required properties for analyzing the structureare mechanical and strength properties where the firstgroup is used for stress analysis and the second one isused for failure analysis [7]. Material characterizationis used to extract the aforementioned properties. Fullmaterial properties of the U-D fabric are availableexperimentally [4] but for the bi-axial and tri-axial com-posites, the mechanical properties are not complete. Theavailable data of these materials are limited to the elasticmodulus in the 0   and 45   directions [8]. The bi-axialand tri-axial laminates and the direction where theexperimental results are available are shown in Fig. 5.The available experimental data are summarized inTable 2.As it can be seen from Table 2, elastic modulus  E  1 and  E  2  are equal in both [+45/  45] T  and [0/90] T  and[0/90/  45] T  due to symmetry. 4.1. Extracting non-available mechanical properties First, stiffness matrices of tri-axial and bi-axial mate-rials are calculated by considering the proportion of fab-ric in each direction and then the matrices were inversed.In parallel, compliance matrices based on available datawere made and finally the recently obtained compliancematrices were compared to the previously extractedcompliance matrices from the stiffness matrices. Finally,after omitting redundant equations, the remaining onesform 14 sets of equations with 14 unknowns enablingthe solution for the mechanical properties of constructedunidirectional plies.Those equations are not reported here due to spacelimitations and complete list of the governing equationscan be found in Ref. [9]. Due to this fact that those 14equations are highly dependent and nonlinear, solvingthem by conventional method is impossible and it needsa start point. The best suggestion for a starting point isthe Poisson  s ratio for the bi-axial material in configura-tion of [0/90] T . Experiments show that the Poisson  sratio of [0/90] T  configuration is somewhere between0.05 and 0.1 and these values also are verified by theory[10]. In Table 3, the results obtained by using different amount of Poisson  s ratio are summarized.As can be seen, the best results are achieved whenPoisson  s ratio is set to 0.06 because of its  E  Y  . Therefore,mechanical properties of unidirectional plies are avail-able from experiments [4] and mechanical properties of bi-axial and tri-axial laminates can be calculated basedon the technique in this research and using mechanicalproperties of their constructed U-D ply which is calcu-lated and inserted in Table 3. The mechanical propertiesare summarized in Table 4.If we calculate the flexural modulus using the ob-tained data, there is very good agreement to the experi-mentally available value for this parameter as shown inTable 5.One of the main advantages of using the inversemethod can be realized in the fatigue modeling methodthat is employed in this paper and will be discussed laterin detail. 45˚ 0˚ 0˚ 45˚  Biax LaminateTriax Laminate Fig. 5. Experiment directions of bi-axial and tri-axial laminates.Table 2Available and non-available mechanical properties [4,8]Composites Configuration  E  1 [GPa] E  2 [GPa] m 12  E  6 [GPa]Unidirectional – 43 9.77 0.32 3.31Bi-axial [±45] T  6.8 6.8 N/A a N/ABi-axial [0/90] T  16.7 16.7 N/A N/ATri-axial [0/±45] T  20.7 ± 3.1 N/A N/A N/ATri-axial [0/90/  45] T  15.1 ± 2.3 15.1 ± 2.3 N/A N/A a N/A: Not available.Table 3Mechanical properties of constructed U-D ply varied by change inmajor Poisson  s ratio of bi-axial laminateMPB a E  X   [GPa]  E  Y   [GPa]  m  G   [GPa]0.05 28.549 4.5643 .21 2.1070.06 28.892 4.3624 .26 2.0960.07 29.010 4.2215 .32 2.0900.08 29.127 4.100 .37 2.0850.09 29.197 3.8240 .40 2.0800.10 29.234 3.6942 .43 2.077 a MPB: Major Poisson  s ratio of bi-axial [0/90] T .Table 4Mechanical properties of involved materials in the blade E  1 [GPa] E  2 [GPa]MajorPoisson  s ratio G  XY  [GPa]Unidirectional ply 43 9.77 0.32 3.31Bi-axial [0/90] T  16.7 16.7 0.06 2.01Tri-axial [0/+45/  45] T  17.6 7.01 0.52 5.07Table 5Experimental and theoretical comparison of flexural modulusTri-axial [0/90/  45] Tri-axial [0/45/  45]FlexuralmodulusExperimental[8]Theoretical Experimental[8]Theoretical15.1 ± 2.3 15.23 16.7 ± 2.5 17.85 M.M. Shokrieh, R. Rafiee / Composite Structures 74 (2006) 332–342  335  The last step of this stage is devoted to extracting theweight of the finite element model and comparing it withrealistic data. These data are shown in Table 6 and de-scribe the health of model from lay-up configurationand prove that the model is in a good accordance withactual structures. 5. Loading The load cases are determined by the combination of specific operating and external conditions. Operatingand external conditions are assumed to be statisticallyindependent [11]. Operating conditions divided into fourgroups known as normal, fault, after occurrence of thefault and transportation, erection and maintenance(see Fig. 6). External conditions are divided into twogroups called normal and extreme. The load cases tobe used for fatigue analysis generally arise only fromthe combination of normal external and operating con-ditions. It is assumed in this context that because theyoccur so rarely, the other combinations will not haveany significant effect on fatigue strength [11].The involved subgroups of normal operating andnormal external conditions are shown in Fig. 6.All of the load cases that occur during the mentionedevents can be categorized as follows:1. Aerodynamic loads on the blade2. Weight of the blade3. Annual gust4. Changes in the wind direction5. Centrifugal force6. Force that arise from start/stop angular acceleration7. Gyroscopic forces due to yaw movements8. Activation of mechanical brake9. Thermal effectAll of the above load cases were calculated and it wasevident that gyroscopic forces and forces from mechan-ical brake activation can be ignored in comparison withthe other loads [12]. A temperature range was consid-ered between   10   and +40   according to the surround-ing ambient environment. 6. Static analysis At the first step of static analysis, in order to insurethat the model is compatible with real structures, a freevibration analysis of model is performed [12]. The result-ing data are in a good agreement with experimental data[12]. The results of free vibration analysis show thatmodel is in accordance with real structures from differ-ent aspects such as dimension, material properties andlay-up sequence.In the second step, calculated forces from the previ-ous section are applied and the response of the structureis studied. In some cases, performing non-linear analysiswas necessary due to large rotation effects as a source of non-linearity. The results of these analyses are summa-rized in Table 7. Table 6Mass comparison of the bladeMass [kg] CG location [mm]FE model 1268.45 7575.23 (from the root)Realistic 1250.00 7500.00 (from the root) Stand-byStart-upPower ProductionNormal Shut-down Condition Operating ConditionsExternal Conditions Normal OperatingFault After the Occurrence of the Fault Transport, Erection & Maintenance ExtremeNormalNormal Operating GustTemperature RangesChanges in the Wind Direction Fig. 6. General view of operating and external conditions of a wind turbine.336  M.M. Shokrieh, R. Rafiee / Composite Structures 74 (2006) 332–342
Similar documents
View more...
Search Related
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks