Jtam.dvi JOURNAL OF THEORETICAL AND APPLIED MECHANICS 49, 3, pp. 765-789, Warsaw 2011 DYNAMIC INVESTIGATION OF TWIST-BEND COUPLING IN A WIND TURBINE BLADE Marcin Luczak Institute of Fluid-Flow Machinery Polish Academy of Sciences, Gdańsk, Poland e-mail: marcin.luczak@imp.gda.pl Simone Manzato, Bart Peeters LMS International, Leuven, Belgium e-mail: simone.manzato@lmsintl.com Kim Branner, Peter Berring Wind Energy Division, Risø DTU, Denmark e-mail: kibr@risoe.dtu.dk Maciej Kahsin Gdansk University of Technology, Gdańsk, Poland e-mail: mkahsin@pg.gda.pl This paper presents some results and aspects of the multidisciplinary and interdisciplinary research oriented for the experimental and numerical stu- dy in static and dynamic domains on the bend-twist coupling in the full scale section of a wind turbine blade structure. The main goal of the con- ducted research is to confirm experimentally the numerical prediction of modification of the dynamic and static properties of a wind turbine blade. The bend-twist coupling was implemented by adding angled UD (UniDi- rectional) layers on the suction and pressure side of the blade. Static and dynamic tests were performed on a section of the full scale wind turbine bladeprovidedbyVestasWindSystemsA/S.The results arepresentedand comparedwith themeasurements of the original andmodified blade. Com- parisonanalysis confirmed thatUD layers introducemeasurablebend-twist couplings, which was not present in the original blade. Key words: wind turbine blade, bending, torsion, composite materials 1. Introduction Wind turbine blades must be designed to resist the extreme load cases and fatigue loads from normal operation. Sudden wind gusts are often too qu- ick for the active pitch control system to react and may shorten the fati- 766 M. Luczak et al. gue life substantially. This problem may be overcome by an aero-elastic ta- iloring of the blades. Particular implementation of an anisotropic composi- te material can introduce bend-twist coupling in the blade (Deilmann, 2009; Lobitz and Laino, 1998; Lobitz and Veers, 1999; Lobitz et al., 2001; Loc- ke and Valencia, 2004; Ong and Tsai, 1999; Walsh, 2009; Zuteck, 2002). The coupling causes the feathering blade to twist under the bending load, and as a result decreases the angle of attack. This paper presents the pro- gress and results of a comprehensive long-term scientific research focused on bend-twist coupling analysis, design and implementation in a wind tur- bine blade made from composite materials. First three parts of the paper briefly recall the research activity carried out (Berring et al., 2007; Branner et al., 2007; Fedorov et al., 2010) while the main focus is put on the me- asuring and modeling of the dynamic behavior described within the fourth part. The first part of the paper reports on the experimental and numerical stu- dies of a standard wind turbine blade section. Thewind turbine blade section madeof a compositematerial was statically tested andmodeled.Different load configurations were applied at the tip of the blade section to assess the twist and bend behavior (Berring et al., 2007). The second part of the paper presents the structural dynamics identifi- cation, which was performed by means of experimental modal analysis. The Finite Element Method model (FEM) (Branner et al., 2007) was developed, updated and validated against the static measurement results. A part of this work has been published in Luczak et al. (2011). Based on the validated mo- del the modified design of existing blade was studied. The baseline concept of the modification was implementation for bend-twist coupling by means of application of additional compositematerial layers. The original blade section was modified with four layers of UD1200 (glass fibre, 1200g/m2), which were laminated on thepressure and suction side of the blade,with an angle of 25 de- grees to the blade axis, in order to create a measurable flapwise bend-twist coupling. In the third part, the static experimental and numerical analysis of the modified blade section is exposed to verify the design correctness of the bend- twist coupling. Finally, in the fourth part the dynamic behavior of the modified blade section is experimentally identified with the assessment of the coupling. Mo- reover, the influence of the support structure dynamics on the test specimen is discussed. Dynamic investigation of twist-bend coupling... 767 2. Static investigation of original blade section Static loads in bending, torsion and combined bending and torsion configura- tions were introduced with different loading force levels (Berring et al., 2007; Branner et al., 2007; Fedorov et al., 2010). 2.1. Object of the investigations The object of investigation is an 8 meter long section cut from a 23 me- ter wind turbine blade. This section is mounted in two root clamps with an additional clamp at the tip for hydraulic jack fastening (Fig.1). Fig. 1.Wind turbine blade section under investigation with the coordinate system. For measurements of the original blade, the axis system is rotated of 90◦ about z axis Bending angles are computed considering the rotation of two consecutive measured cross-sections about the x axis. The twist angles are computed as the rotation about the z axis of each cross sectionwith respect to theunloaded configuration.Adetailed description of the calculation can be found inBerring et al. (2007). Figure 2 shows the bend and twist angles of the original blade computed from the static bendingmeasurement. The graph indicates that the bend-twist coupling is equal or close to zero, since the bendingmoment does not result in a twist angle (rotation about the z axis) of the blade section. 768 M. Luczak et al. Fig. 2. Bending slope angle (a) and cross-sections twist angle (b) values under flapwise bending load. The blade section is bending but not twisting. That indicates that the bend-twist coupling is close to zero 3. Dynamic investigation of the original blade section Observations from the static investigations were verified in the structural dy- namics of the tested original blade section (Capellaro, 2006). Measurement points were defined on the leading and trailing edge at thirteen cross-sections. Flapwise and edgewise direction accelerations were measured. Points on the support structure were not measured. As the drawback of this fact, some of the natural frequencies of the support structure were indentified as themodes of the blade while they are modes of the support structure. Natural frequen- cies and corresponding mode shapes were estimated and some examples are presented in Fig.3. Fig. 3. 3D plots of mode shapes of the original blade section. First flap wise bending (a) and first edgewise bending (b) Dynamic investigation of twist-bend coupling... 769 4. Static investigation of the modified blade section One of the primary aims of this long-term research is to study, design and implement the desired bend-twist coupled behavior. The original blade section wasmodifiedwith four layers ofUD1200,whichwere laminatedon thepressure and suction sideof thebladewith thefibersangle of 25◦ to create ameasurable flapwisebend-twist coupling.Theadditional layerswere laminatedas indicated in Fig.4. Fig. 4. Fiber orientations of the extra UD layers To verify the numerical prediction of the bend-twist coupling, the static tests campaign from the original blade was repeated on the modified blade. The twist angles for equidistant cross-sections under flapwise bending load were calculated from themeasurement (Fig.5). Fig. 5. Twist angle under flapwise bending load. This indicates that the blade section has a measurable bend-twist coupling Assuming that the shear center is located in the center of the spar, the rotation angles about the z axis for the modified blade section in Fig.5 show ameasurable bend-twist coupling. 770 M. Luczak et al. 5. Dynamic investigation of modified blade section The flapwise bend-twist coupling was also investigated by means of experi- mental modal analysis. The research was focused on the bend-twist coupling presence in the mode shapes of the blade section. An important aspect was also the analysis of the influence of an additionalmass and stiffness introduced by the extra layers. Finally, the influence of the support structure on the cor- relation analysis between the numerical and the experimental modal models was studied (Luczak et al., 2010; Peeters et al., 2004). The blade section was excited by two electro-dynamic shakers attached at the tip end in the flapwise and the edgewise directions. Frequency Response Functions were measured and stored within 0 and 120Hz frequency range. For adequate identification of the blade dynamic displacement, accelera- tions of the vibrations were measured in 130 points. Thirteen equidistantme- asurement cross-sections were defined along the spanwise direction (Z) every 0.5m. Each cross section contains fivemeasurement points in which the acce- lerations were acquired along the flapwise (X) and edgewise (Y ) directions. These points were located at the leading edge, trailing edge, on the line of the airfoil maximum thickness and in the mid-points between the previous three. The measurement directions were precisely defined based on the CAD (Computer Aided Design) geometry of the blade section. Themodel quality assessment was an integrated part of the investigation. The following conditions have to be fulfilled to satisfy the modal analysis assumptions: time invariance linearity,Maxwell’s reciprocity principle and ob- servability. Possible sources of nonlinearities within the investigated structu- re are material properties, geometrical properties and the existence of bond connections. Verification of the superposition rule is one of the methods of detecting nonlinearities. Linearity check was performed for several levels of driving voltage ranging from 0.5V to 2V with a step of 0.5V. The results are presented in Fig.6. The Frequency Response Function (FRF) between the input signal and the output spectrum defined as the acceleration over force, remains constant independently of the excitation voltage level. This proves that the structure dynamic behavior is linear within the bandwidth of interest. The reciprocity check is basedonMaxwell’s principle,which states that the FRFsobtainedbyapplyingthe force inpoint1andmeasuringthe response in2 and vice versa should be the same. The results for the two checks performed confirmed applicability of the reciprocity rule. Dynamic investigation of twist-bend coupling... 771 Fig. 6. Linearity check for one of the points on the blade. Voltage values = 0.5V, 1V, 1.5V and 2V During the processing of data some significant noise was observed in the acquiredFRFs in the low frequency region.Thedrivingpoints coherence func- tions show a small drop in this region,meaning a non-ideal excitation (Fig.7). Fig. 7. Coherence functions for the two driving points. Measure of the FRF quality is used. Ideally, it should take value equal to 1 Themodal parameter identification technique was not able to clearly sta- bilize the modes in this region, possibly resulting in some local errors in the mode shapes below 7Hz. The estimation provided natural frequencies, mode shapes and correspondingdamping ratios in the frequencybandwidth 0-60Hz. The first five out of 12 identified mode shapes are shown in Fig.8. 772 M. Luczak et al. Fig. 8. Estimated experimental mode shapes of the modified blade section and support structure TheMAC(ModalAssuranceCriterion) can beused to compare twomodal models. TheMAC between twomode shape vectors, φ i and φ i (s), is defined as MAC(s)= |φ⊤ i φ i (s)|2 φ ⊤ i φ i φ ⊤ i (s)φ i (s) (5.1) If a linear relationship exists between the two complex vectors φ i and φ i (s), theMACvaluewill be near to 100. If they are linearly independent, theMAC value will be small (near zero). Figure 9 shows a comparison between the AutoMAC of the modal model obtained by considering only the sensors on the blade and the onewhere also the sensors on the supporting structurewere included. Low valued off-diagonal terms for the bladewithout the support structure model ensure linear independence of the estimated modal vectors. The corre- lation between off-diagonal terms is increased when including the supporting structure in the analysis. This is due to the fact that the clamping is not per- fectly rigid and the support has its own dynamic behavior which influences the measured response of the blade. In Fig.9, the red color corresponds to the MAC value equal to 100. The dark blue color reflects theMAC value 0. Themodes corresponding to frequ- encies 8Hz, 28Hz, 31Hz and 33Hz are related to dynamic properties of the supporting structure. 5.1. Correlation analysis for the simulation and test results Using themodalmodel estimated experimentally and those obtained from the FEMmodels (modeling the original andmodified blades), the correlation Dynamic investigation of twist-bend coupling... 773 Fig. 9. AutoMACmatrices for the experimental modal model with sensors only on the modified blade section (a) and blade section with the support structure (b) analysis can be applied. The FEM models should be characterized by good consistencyof thenatural frequencyvalues andmode shapesobtained fromthe measurement. TheModalAssuranceCriterion is used as the original-modified blade simulation and also test-simulation correlation metrics. The global axis system used to define the testmodel differs from that used in the FEM model. In order to match the models, it is necessary to apply geometric correlation by translation and rotation of the test model (Fig.10). The next step is node mapping. The number of measurement nodes is much 774 M. Luczak et al. smaller than the FEMnodes. Modal vectors are compared only for the nodes from FEM which are located closest to the measurement points. Only the portion of the blade past the clamp is considered. Fig. 10. Test and FEM geometry correlationwith nodemapping For tests described in section 2 to 4, the support structure was not taken into account andmeasured data reduced to obtain a perfectly rigid boundary condition on the clamped section (Berring et al., 2007; Branner et al., 2007; Fedorov et al., 2010; Larsen et al., 2002; Pedersen andKristensen, 2003). Since theFEMmodel of the blade is developedusing the sameassumption, somedif- ferences in the frequency values andmode shapes obtained from experimental results of the modified blade are expected. Theblademodelwas used to computemode shapes in the 0-60Hz frequen- cy bandwidth and computations were performed on a 50Tflop (10 12 FLoating point OPerations per second) cluster in TASK Super Computing Center. The first Modal Assurance Criterions were calculated for the correspon- dingmodes in order to associate the closest numerical and experimentalmode shapes (Fig.11). The procedure accounted for both natural frequency value and themode shape consistency. The following modes were investigated: 1st and 2nd flapwise bending, 1st and 2nd edgewise bending and 1st torsional (Fig.8). The MAC matrix in Fig.11 clearly shows that the off-diagonal terms are low valued, which con- firms the linear independence of estimated modal vectors. The best test and simulationmodal vectors consistency can be observed for the 2ndflapwisemo- Dynamic investigation of twist-bend coupling... 775 Fig. 11.MACmatrix for test and FEM simulationmodal vectors of the modified blade without the support structure de. The consistency of the results can be recognized as satisfactory, however the present differences should be further investigated. Observing the values of the MAC criterion between the test and simulation modes (Fig.9), some differences can be observed. They are caused by the influence of the support structure and not perfectly excited the 1st bendingmode. Comparison of the natural frequencies obtained from experimentalmeasu- rements and simulation for the original and modified blade is presented in Table 1. Table 1. Comparison of the natural frequencies for the experimental and numerical results obtained for the original andmodified blade Original Original Modified Modified blade blade blade blade FEM Test FEM Test 1st bend flap 4.7Hz 4.5Hz 5.01Hz 4.48Hz 1st bend edge 10.85Hz 8.7Hz 12.9Hz 12.08Hz 2nd bend flap 18.56Hz 18.9Hz 20.03Hz 19.24Hz 1st torsion 42.99Hz 39.5Hz 43.75Hz 40.92Hz 776 M. Luczak et al. The difference between the Test and FEM frequencies can be explained by modeling the boundary condition as rigid in the FEM. Moreover, further differences in the frequency values between the original and modified blade results are introduced by the additional mass and stiffness implemented from angled UD layers on the suction and pressure side of the blade. These UD layers introduce measurable bend-twist couplings, not present in the original blade in terms of static anddynamic response of the investigated section of the blade.While in the static response therewas a clear indication of the coupling, themodification of the dynamic stiffness is not fully recognized, which can be observed based on the comparison of the simulation of original and modified blade structural dynamics (Fig.12). Fig. 12.MACmatrix for the original andmodified blade Finite Element models 100%MACmatrix terms for the first threemodesmay indicate that these particular mode shape sensitivity towards the implemented additional layers is not significant. 5.1.1. FEM model of the blade section with FEM model of the support structure The numerical model adopts the previously developed MSC.Patran/Nas- tran blade FEM model comprised of 8-node shell elements (Quad8) and the 20-node solid elements (Hex20).Thismodelhasapproximately 600.000degrees of freedom (Berring et al., 2007). The original FEM model of the blade was developed to study the static response, i.e. cross sectional displacements and rotations caused by different load configurations. A least squares algorithm was developed, which fits a plane through each deformed cross section, and Dynamic investigation of twist-bend coupling... 777 defines a single set of displacements and rotations (three displacements and rotations) per cross section. This least squares algorithm was also used to accommodate problems with a flexible boundary condition by determining the displacements and rotations for a cross section near the boundary. These displacements and rotations are subtracted from all other cross-sections along the blade and therebymaking the blade section fully fixed at the chosen cross section near the boundary (Fig.13). Fig. 13. Original FEMmodel of the blade The boundary conditions which were adequately representing the support structure in static analysis were used in the initial theoretical modal analysis. In the correlation analysis of the test and FEM modal models, it turned out that a non-negligible discrepancy in mode shapes occurs. A relatively light and flexible support (Fig.14a) has significant contribution to the simulated mode shapes of the studied structure which can be observed in the MAC values (Fig.11). It was assumed that introducing into the FEM model the representation of the supporting structure could improve the test-simulation results correlation. The uncertainties concerning the support structure makes the modeling process rather challenging. As the source of uncertainties one can point not only geometrical and material parameters of the support members, but ma- inly their connections. Additionally, analyzing the test modal model it can be seen that these connections are very flexible, therefore very important in the overall dynamical behavior of the analyzed system. As the future activity optimization based updating has been foreseen. The fine-tuning of the FEM model is iterative, therefore, due to the size of the simulated model the main 778 M. Luczak et al. Fig. 14. Supporting structure (a) and it’s FEMmodel (b) assumption prior modification of the original FEMmodel (blade only model) was to keep the additional FEM model (i.e. supporting structure) as simple as possible. The derived model allows correlating the simulation results with measured data in all points used in the tests. As it can be seen in Fig.14a, the real supporting structure comprises of pipes,UNP-profiles (according toEuropeanStandard tables for steel profiles), and support clamps of contour-cut plywood. Basic information about the geo- metry andmaterial properties exploited in the derived additional FEMmodel are presented in Table 2. Table 2. Basic information about geometry andmaterial properties used for modeling of the supporting sructure Pipes C-Shapes I-Shapes Plywood Geometry [mm] Inner radius 170 Standard Two bolted Thickness Outer radius 160 UPN 200 standard 180 UPN 200 Emodulus [GPa] 200 200 200 13.2 Density [kg/m3] 7890 7890 7890 736 Poisson’s ratio 0.3 0.3 0.3 0.01 Dynamic investigation of twist-bend coupling... 779 In themodified FEMmodel, the steel members (i.e. pipes andUNPprofi- les) aremodeledbybeamelements (CBEAMinNastrannotation).Theplywo- od in clamps ismodeled by shell elements (CQUAD4). Allmountings between the support members are represented by elastic spring elements (CELAS1). It should be noted that beam elements representing the frame are placed at the centerline of the frame members. As a consequence, discrepancy between the nodes of the FEM model and geometrical positions of measuring points exists. For the correlation, additional nodes were inserted at the measuring points. Connections between the beams and the additional nodes were reali- zed by the use of rigid bar elements (RBAR). Due to the same fact, there were gaps between the beams and plywood in the FEMmodel representation of the clamps (Fig.12). Nodes of the beams and shells elements constituting the clampwere connected by the use of rigid bar elements (RBE2). The rigid connection between the plywood and I shapes is justified because of the large difference in E modules of bothmaterials. Representation of FE-to-test mat- ching with the rigid bars does not introduce additional stiffness to the system and is acceptable as long as the global mode shapes of the support are of interest only. After preparation of the support FEmodel, both the additional and the original FEmodels were merged (Fig.15). The nodes at the interface Fig. 15. Updated FEMmodel of the blade with the flexible supporting structure between the blade and supporting structure, i.e. between the plywood and the outer surface of the blade, have restrained rotational DOFs (Degree Of Fre- edom). Such an approachwas taken because in the real structure the interface 780 M. Luczak et al. between the profile-cut plywood and the blade is realized on approximately 200mm of width, while in the numerical model a single row of nodes is used only. An initial correlation analysis of updated FEM model is presented in Fig.16. Fig. 16. Geometry correlation of the TEST and FEMmodels of the modified blade section with the support structure. 124 out of 132measurement points are now coupled with the FEM nodes The initial correlation results of the updated FEMmodel show an impro- vement in comparisonwith the results from the original FEMmodel (Fig.17). 5.2. Computation of twist and bend angles Using the modal vectors identified from the experimental modal analysis results or computed from theFEMmodel, the twisting andbending angles for particular mode shapes can be computed. For the experimental mode shapes, a fitting is applied to smooth out some odd local behavior due to inherent errors in the measuring process and excitation limitations shown in Fig.7. For the twisting angles, a similar approach to the one used in the static computation is applied (Berring et al., 2007). For each cross-section, displace- ments from the undeformed configuration are associated with the estimated modal amplitudes. Only the leading and trailing edge results are used in the computation. Both relative angles for each section and the sumof these angles along the blade were computed. Dynamic investigation of twist-bend coupling... 781 Fig. 17. MACmatrix, Test vs. updated FEMmodel of the blade with the flexible support Fig. 18.Mountings of the supporting structure (a), drilled UPN (b), two bolted UPN (c) Bending angles were computed in a slightly different way than for static measurements. Themodal displacements obtained from the sensors located in the point of themaximum thickness of the airfoil for each section are conside- red. By considering two consecutive cross-sections, relative bending is evalu- ated as the difference between themodal displacements in the x direction (see Fig.1). The tangent of the relative angle for the section is then computed by dividing the difference for the length of the section. On each cross-section, the global bending angle is computed by summing all the relative angles for the previous cross-sections. Computed angles are modal angles, so they depend 782 M. Luczak et al. on the scaling applied to the modal vectors. Moreover, torsional angles are computed assuming that themass and shear centers are in the same locations. 5.3. Numerical and experimental twist and bend angles Themethodology described in Section 5.2 is applied both to numerical and experimental modal analysis results. In the present paper, only a limited part of the experimental results is presented in details. The results for the first and secondflapmodeswill bediscussed.Theoriginal blade simulationflapbending translation values (Fig.19a,e) are plotted with the dark blue, edge bending with red and rotation with light blue. The original blade experimental results are presented in Fig.19c,g. Deflection in the flap direction is plotted with the red line, edgewise direction with blue and rotation around the blade axis is marked with green. The angles values are assumed to be 0 in the clamped section. For themodifiedblade section bendingangle, the values are plotted indark blue and twist in red. In order to have an overview of the overall behavior, the absolute angles are presented. The results are obtained by processing the fitted mode shapes. By graphically comparing the original and modified blade plots (Fig.19), it can be immediately observed how the computed torsional angle is higher, as expected, for the modified blade than for the original one. Moreover, the trend of bending is the same in all plots confirming that observation. Even if Fig.19b,d,f,h show bending modal angles instead of displacements, they can be directly related. By comparingFig.19bwith 19d, andFig.19fwith 19h, somedifference can be observed. The modal angle values depend on the scaling applied and the specific values are not important.Moreover, themodeling of clamping as rigid in the FEMmodel introduces a different than the measured boundary condi- tion, whichmodifies themode shapes. Finally, fitting applied to experimental results gives a smoother behavior but can also introduce some numerical er- ror. Despite these local problems, a good agreement between the overall trend of the twist and bend angle for the measured and simulated results can be observed. This observation would confirm some discrepancies between the test and simulations, which were already spotted for the 1st mode (Fig.11). For the original blade, the bending was decoupled from twisting which was measured (Fig.19g) and simulated (Fig.19e). The introduced coupling is clearly visible in the increased torsional response for the simulated (Fig.19f) andmeasured results (Fig.19h). Dynamic investigation of twist-bend coupling... 783 1st bend flap wise Original blade Modified blade (a) Finite Element (b) Finite Element (c) Experimental (d) Experimental 2nd bend flap wise Original blade Modified blade (e) Finite Element (f) Finite Element (g) Experimental (h) Experimental Fig. 19. Twist and bend angle computation from numerical and experimental results for the original (left column) andmodified (right column) blade section for 1st and 2nd flapwise bending mode shapes. The blade segment between 2 clamps is not plotted 784 M. Luczak et al. 5.4. Bend-twist coupling index One of themain objectives is to investigate the amount of twisting introduced by the additional UD layers implemented on the modified blade. To obtain a quantitative measure of the coupling between twisting and bending angle, an index is introduced. For each considered blade cross-section, the ratio between the computed relative twisting and bending angles is evaluated. A coupling index value close to zero means that the twisting is negligible with respect to the bending for the considered mode. On the contrary, a high index value means that twisting is dominant. If the index is close to one, twisting and bending are of the same order of magnitude. Figure 20 shows the computed coupling index for the 1st and 2nd flapwise modes both for the experimental and numerical results. Fig. 20. Bend-twist coupling index for 1st (a) and 2nd (b) flapwise mode 5.5. Discussion of results, problems and proposed improvements Comparing the numerical and experimental results, certain differences can be observed. These differences are directly related to the observation made for Fig.11. For the firstmode shape, theMACvalue between the experiment and numerical model is 65.5. The explanation of the lower MAC value for the 1st mode can be found in the coherence plot presented inFig.7. It is a ratio of the maximum energy in a combined output signal due to its various components, and the total amount of energy in the output signal. The coherence is used as a measure of the power in the output channel that is caused by the power in the input or a reference channel. As such, it is useful in assessing the accura- cy of the frequency response function measurements. The coherence function can take values that range between 0 and 1. A high value (near 1) indicates that the output is due almost entirely to the input and one can feel confident in the frequency response function measurements. A low value (near 0) indi- Dynamic investigation of twist-bend coupling... 785 cates problems such as extraneous input signals not being measured, noise, nonlinearities or time delays in the system. The first mode is located in the frequency range of relatively poor coherence leading to the decreased quality of its estimation. The shakers which were used in the measurement have a low frequency limit around 2Hz. The excitation signal was random that pro- vides homogenous distribution of injected energy over the excited bandwidth. This could lead to the insufficient energy exciting the 1st mode. Moreover, the shakers were hung from support cables. In such a case, at a very low fre- quency in the sub-10Hz range there is a problem to provide more inertia to push against the structure being excited for improved performance. To impro- ve the excitation of the 1st mode shape, an additional mass was applied to increase the inertia of the shakers apparently bringingnotmuch improvement. For the 2nd mode located in the higher frequency range of the investigated bandwidth, the vibration amplitudes become lower and the consistency be- tween the test and simulation is much better confirmed by the MAC value of 88.5. Bend and twist angles calculations are based on the experimental and nu- mericalmodal vectors. Thedifferencebetween the experimental andnumerical models bend-twist indexes is caused by the relatively weak excitation of the 1st mode due to the above mentioned reasons. The supporting structureFEMmodel shouldbe furthermoreupdated.The comparison of deformation in the simulated model and the test model shows that the real structure is more flexible. This situation can be caused by un- certainties related to unknown properties of the beam mountings (Fig.18a), the fact that the beams are drilled (Fig.18b), the use of I shape clamp beams, while in the real clamp the beams consist of two bolted UPNs (Fig.18c), and the assumed properties values of plywood, unknown properties of the frame- to-groundmounting. Further research should introduce two clamps in the FEMmodel in order to get realistic structural behavior of the clamps. In detail, the plywood plates and steel profiles should be included and contact elements should be applied tomodel the contact between the clamps and the blade section. It is expected that amore sophisticated support structure FEM representation will improve the consistency between the test and simulations. The coupling index for the numerical 1st flapwise bending has approxima- tely a constant value throughout theblade,whiledifferentbehavior is observed from the experiments. In addition to some odd behavior from local mode sha- peswhich could influence the results, it shouldalsobenoted that themeasured boundary conditions are different than those modeled. 786 M. Luczak et al. The latter observation can be applied also for the 2nd flap wise bending mode, which on the other hand shows a very similar trend between the nume- rical and experimental results. To be able to improve the correlation between the presented results, the same boundary conditionsmust be used both for the experimental and nume- rical models. 6. Conclusions This paper presents some results and aspects of multidisciplinary and interdi- sciplinary research oriented towards the experimental and numerical study in static and dynamic domains of the bend-twist coupling in the full scale section of a wind turbine blade structure. An extensive test campaign performed on the original and modified wind turbine blade section was presented. The test setups included different load configurations, excitation andmeasurement techniques of the contact andnon- contact type. Experimental test data examples were shown and used for two purposes. Firstly, to evaluate the ability of different test methods to measure the bend-twist coupling and, secondly, to use the test results for FEMmodels updating. The common observation from displayed comparisons is that the original blade section did not have ameasurable bend-twist coupling. Because one of the primary aims of this work is to develop and use a FEM-model capable of modeling correctly the bend-twist coupling behavior, the original blade section was modified. For this purpose, four UD layers were laminated on the pressure and suction side of the blade section to introduce ameasurable flapwise bend-twist coupling. The successful implementationof thebend-twist couplingwas confirmedby extensive static and dynamic measurement campaigns. In both experimental methods, the comparisonof original andmodifiedbladeproperties clearly show the presence of the bend-twist coupling. Acknowledgements VestasWind SystemsA/Shas providedandmodified the blade sections presented in this study. The work is partly supported by the Danish EnergyAuthority through the 2007 Energy Research Programme (EFP 2007). The supported EFP-project is titled “Anisotropic beam model for analysis and design of passive controlled wind turbine blades” and has journal no. 33033-0075.The support is gratefully acknowled- ged and highly appreciated. Authors would like to acknowledge the assistance ofMr. Dynamic investigation of twist-bend coupling... 787 PhilippHaselbachduring the dynamic test campaign.Researchpresented in Section 5 was conducted in the context of the FP7 project PROND Ref No. 239191. Compu- tations were performed on a 50Tflop cluster in TASKAcademic Computer Centre in Gdansk, Poland, within the framework of PL-Grid Project. References 1. Berring P., Branner K., Berggreen C., Knudsen H.W., 2007, Torsio- nal performance of wind turbine blades – Part 1: Experimental investigation, International Conference on Composite Materials (ICCM-16), Kyoto, Japan 2. BrannerK.,BerringP.,BerggreenC.,KnudsenH.W., 2007,Torsional performance of wind turbine blades – Part II: Numerical validation, Interna- tional Conference on Composite Materials (ICCM-16), Kyoto, Japan 3. CapellaroM., 2006,ModelingBendTwist CoupledBlades,MScThesis,Tech- nical University of Denmark 4. Deilmann C., 2009, Passive Aeroelastic Tailoring of Wind Turbine Blades – a Numerical Analysis, MsC Thesis Thesis (S.M.) – Massachusetts Institute of Technology, Dept. ofMechanical Engineering 5. Fedorov V.A., Dimitrov N., Berggreen C., Krenk S., Branner K., Berring P., 2010, Investigation of structural behaviour due to bend-twist co- uplings in wind turbine blades,NAFEMS Nordic Seminar of Simulating Com- posite Materials and Structures, Esbjerg, Denmark 6. Griffith D.T., Smith G., Casias M., Reese S., Simmermacher T.W., 2006, Modal Testing of the TX-100 Wind Turbine Blade, Sandia Report SAND2005-6454 7. Larsen G.C., HansenM.H., Baumgart A., Carlén I., 2002,Modal Ana- lysis ofWindTurbineBlades, Risø-R-1181,RisøNationalLaboratory,Roskilde, Denmark 8. Lobitz D.W., Laino D.J., 1998, Loadmitigation with twist-coupled HAWT blades,Proceedings of 1998 ASME Wind Energy Symposium, January 9. LobitzD.W.,VeersP.S., 1999,Aeroelasticbehaviorof twist-coupledHAWT blades,Proceedings of 1999 ASME Wind Energy Symposium, January 10. LobitzD.W.,VeersP.S., EislerG.R., LainoD.J.,MiglioreP.G.,Bir G., 2001,The Use of Twisted-Coupled Blades to Enhance the Performance of Horizontal Axis Wind Turbines, Sandia Report SAND01-1303 788 M. Luczak et al. 11. LockeJ.,Hildago I.C., 2002,The Implementation ofBraidedCompositeMa- terials in the Design of a Bend-Twist Coupled Blade, Sandia Report SAND02- -2425 12. Locke J., ValenciaU., 2004,Design Studies for Twisted-CoupledWindTur- bine Blades, Sandia Report, SAND04-0522 13. Luczak M., Branner K., Berring P.,Manzato S., Haselbach P., Pe- eters B., 2011, Experimental verification of the implementation of bend-twist coupling in a wind turbine blade, [In:] Scientific Proceedings of EWEA2011 European Wind Energy Conference 2011, Brussels, Belgium 14. Luczak M., Dziedziech K., Vivolo M., Desmet W., Peeters B., Van der Auweraer H., 2010, Contact versus non-contactmeasurement of a heli- copter main rotor composite blade, 9th International Conference on Vibration Measurements By Laser and Noncontact Techniques, AIVELA 2010, Ancona, Italy 15. Ong C.-H., Tsai S.W., 1999, Design, Manufacture and Testing of a Bend- Twist D-spar, Sandia Report SAND99-1324 16. Pedersen H.B, Kristensen O.J.D., 2003,Applied Modal Analysis of Wind Turbine Blades, Risø-R-1388(EN), Risø National Laboratory, Roskilde, Den- mark 17. Peeters B., Van Der Auweraer H., Guillaume P., Leuridan J., 2004, The polymax frequency-domainmethod: a new standard for modal parameter estimation? Shock And Vibration, Special Issue dedicated to Professor Bruno Piombo, 11, 395-409 18. Walsh J.M., 2009, Composite Material Bend-Twist Coupling for Wind Tur- bine Blade Applications, MsC Thesis, University of Wyoming, Department of Mechanical Engineering 19. Zuteck M., 2002,Adaptive Blade Concept Assessment: Curved Planform In- duced Twist Investigation, Sandia Report SAND02-2996 Badania eksperymentalne i symulacyjne sprzężenia typu zginanie-skręcanie w dynamice łopaty turbiny wiatrowej Streszczenie W artykule przedstawiono wyniki multidyscyplinarnych prac badawczych prowa- dzonych na sekcji dużej łopaty turbiny wiatrowej. Prace obejmowały badania ekspe- rymentlane oraz symulacje numeryczne sprzężenia typu zginanie-skręcanie w staty- ce i dynamice badanej łopaty. Podstawowym celem badawczym było potwierdzenie Dynamic investigation of twist-bend coupling... 789 w drodze eksperymentu wyników symulacji numerycznej własności statycznych i dy- namicznych łopaty. Sprzężenie typu zginanie-skręcanie zostałowprowadzonewkonstrukcji łopaty po- przez dodtkowe zalaminowanie warstw kompozytów włóknistych jednokierunkowych po obu stronach łopaty pod dobranym kątem. Testy statyczne i dynamiczne zostały zrealizowane na sekcji dużej łopaty turbiny wiatrowej dostarczonej przez producenta Vestas Wind Systems A/S. W pracy przedstawiono i porównano wyniki badań ory- ginalnej sekcji łopaty oraz sekcji zmodyfikowanej. Analiza porównawczawynikówpo- twierdziła wyniki symulacji numerycznych.Wprowadznie dodatkowychwarstw kom- pozytu włóknistego spowodowało powstanie w badanej łopacie mierzalnego sprzęże- nia typu zginanie-skręcanie. Sprzężenie to nie było obserwowaneworyginalnej łopacie przedmodyfikacją. Manuscript received March 7, 2011; accepted for print May 19, 2011