- Email: [email protected]
Wind turbine blade trailing edge failure assessment with sub-component test on static and fatigue load conditions
Accepted Manuscript Wind turbine blade trailing edge failure assessment with sub-component test on static and fatigue load conditions F. Lahuerta, N. Koorn, D. Smissaert PII: DOI: Reference:
S0263-8223(18)31045-6 https://doi.org/10.1016/j.compstruct.2018.07.112 COST 10030
To appear in:
Composite Structures
Please cite this article as: Lahuerta, F., Koorn, N., Smissaert, D., Wind turbine blade trailing edge failure assessment with sub-component test on static and fatigue load conditions, Composite Structures (2018), doi: https://doi.org/ 10.1016/j.compstruct.2018.07.112
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Wind turbine blade trailing edge failure assessment with sub-component test on static and fatigue load conditions F. Lahuerta*, N. Koorn, D. Smissaert Knowledge Centre WMC. Kluisgat 5, 1771 MV Wieringerwerf, The Netherlands.
Abstract Wind turbine blades present different types of failure mechanisms and modes which are associated with specific loading conditions. Trailing edge failure mode has been documented in full-scale blade tests as one of the failure types observed in blades on service. Trailing edge failure is characterized by failure of the trailing edge adhesive joint and the buckling of the trailing edge sandwich panels. This failure is governed by the contribution of edgewise, flapwise and torsion moments, with edgewise moments being the main driver. This paper describes a blade sub-component test setup suitable for studying trailing edge failure on static and fatigue load conditions, which is an improvement in the experimental verification of a trailing edge blade design. The test setup and design drivers are described and studied via FE models. Static and fatigue test results are reported for a full-scale blade section sub-component obtained from a 34 [m] wind turbine blade. Moreover, experimental results are discussed and compared with FE models to describe and study the trailing edge failure mechanism. Keywords: wind turbine blade, adhesive, sub component, buckling, edge moment, trailing, failure
∗ Corresponding
author Email address: [email protected] (F. Lahuerta*)
Preprint submitted to Composite structures
August 4, 2018
1. Introduction Operation and maintenance costs (O&M) are one of the cost drivers in the wind energy sector. Several studies [1, 2, 3] identified the downtime due to blade failures and blade O&M actions to be between 3% to 10%. Especially for 5
offshore turbines these cost and downtime are expected to be even larger [4, 5] . Among the different types of blade failure modes described by Wenxian [6] and Yang [7], Ataya [8] suspected the trailing edge (TE) adhesive failure as one of the main issues concerning the reliability of blade designs leading to failures and downtime.
10
According to IEC 61400-5 test pyramid for rotor blades is divided into fullscale blade test, sub-component test (full-scale structural features), generic element and details (non-full scale generic specimens) and coupon test (where fundamental material properties are obtained). Carrying out full-scale blade tests to study and assess wind turbine blades trailing edge behaviour can be
15
expensive and impractical. Full-scale blade tests are performed at later stages of the design process for qualification purposes [9] at which point there is little room for design improvements. Sub-component tests are therefore more suitable for the study and design of local wind turbine blade failure mechanisms such as trailing edge joint failures. Thick adhesive joint failures have been studied
20
before using generic elements tests, i.e. beam tests [10]. However, very few developments of sub-component test methods to study the trailing edge adhesive failures are available. Several institutions (DTU, IWES, CENER, CRES, WMC) join forces in the IRPWind project to develop sub-component test setups and study trailing edge failure and adhesive joints [11, 12, 13, 14, 15, 16].
25
Studying wind turbine blade structural designs by blade segments [17, 18] via sub-component tests was considered in this project. Trailing edge adhesive joint fracture was described by Eder [19, 20], who reported that the most critical failure mode is related to Mode I contribution (peeling forces) due to the opening of the trailing edge. A method to estimate
30
the SERR (strain energy release rate) of different fracture modes in the trailing
2
edges of large wind turbine rotor blade models was proposed [21]. Blade tests performed by DTU Risø [22] for an SSP 34 [m] blade suggested the occurrence of trailing edge adhesive joint failure due to local buckling in trailing edge panels. In addition, an internal wire reinforcement for the SSP 34 [m] blade was 35
implemented to prevent out-of-plane deflection of the panels, thereby reducing the peeling stresses on the adhered region and impeding the Brazier effect [23]. Moreover, Branner & Haselbach [24] described the blade test trailing edge failure mechanism by an FE model which considered transverse shear stresses (by use of an 8-noded doubly curved thick shell element). It was found that buck-
40
ling led to the trailing edge adhesive joint final failure well below the expected maximum load. It should be pointed out that some of the literature findings might not be totally applicable to the modern MW scale blades. It is known that most blade manufacturers with spar box load carrying structures changed their design to
45
the traditional spar CAP/shear web construction. This has a substantial effect on the trailing edge stiffness (i.e. stiffer constructions). The aim of the work is to describe a test method which allows studying wind turbine blade trailing edge failures due to loading cases dominated by an edgewise moment. The test method allows testing full-scale segments of
50
a blade or scaled sub-components. The test setup design and configuration are discussed with an FE parametric analysis. Experimental results in static and fatigue loading conditions are reported for SSP 34 [m] blade segments, the failure mechanism is described and instrumentation readings are discussed in comparison with FE models.
55
1.1. Trailing edge strength and stability criteria Preliminary results from blade [24] and sub-component tests [16] revealed that three main stages can be observed in the trailing edge instability and failure. This observation agrees with the behaviour of an unstable buckling structure and can be characterized by three main regions.
3
60
The pre-buckling stage. This stage is characterized by a linear edgewise stiffness until the buckling load is reached at point C2 (see figure 1). The trailing edge post-buckling stage. This stage shows a drop of the sub-component edgewise stiffness and a non-linear edgewise stiffness behaviour until the final failure is reached at point C1 (see figure 1).
65
The ultimate failure stage. The failure stage is characterized by the change of edgewise stiffness until suddenly the trailing edge adhesive joint fracture debilitates the structural connection between the suction and pressure side trailing edge panels. At this stage a drop of the edgewise stiffness is observed (point C4 in figure 1).
70
Trailing edge strength and stability criteria were proposed [12] based on Eurocode for the stability of shell structures [25]. Five different criteria are proposed to evaluate the initiation of geometrically non-linear deformation and characterize the sub-component test buckling resistance. Both, stability (C1 to C3) and strength (C4) perspectives are considered. These criteria approach the
75
characterization of the trailing edge resistance from different levels of detail (see figure 1). Criteria C1 and C2 study the sub-component global stiffness curve characteristic features. Criteria C3.1 and C3.2 focus on measurable geometrical variations of the structure. Criterion C4 is based on the classical study of the local material resistance and stress levels.
80
Criterion C1 corresponds to the load-displacement curve (load-factor or lead-lag moment) maximum force at final failure or the maximum bifurcation point if the latter occurs earlier than the final failure. Criterion C2 relates to a linear load bifurcation analysis (LBA). The earlier load bifurcation point from a buckling analysis corresponds to C2 criterion. For
85
experimental results, the C2 criterion corresponds to the earliest point where a change of the stiffness rate is identified and the start of a buckling wave can be observed. It should be noted that criterion C2 load bifurcation analysis differs from global buckling analysis of a blade on the loading conditions. Criteria C3.1 and C3.2 describe measurable geometrical variations related
90
to the buckling of the trailing edge. The β and α angles can be evaluated in 4
both sub-component tests and FE models (see figure 2). The β angle describes the buckling wave degree of development for a given load factor. The α angle is suitable for trailing edge studies according to Eder [19]. Moreover, the α angle is related to the specific trailing edge adhesive joint geometry and the peeling 95
forces that appear with increasing the load factor. In both criteria a maximum ∆αmax or βmax is defined as threshold value. Criterion C4 is based on the design guidelines [26] for stress analysis. A detailed analysis of the adhesive joint is performed and local design stress values are compared against material component resistances. Since this type of analysis
100
is not directly comparable with experimental results, criterion C4 is applicable via modelling techniques. Alternatively, the C4 criterion could be based on the SERR mode-mixity evaluation of the adhesive joint [12].
2. Test setup description Within the IRPWind project framework, WMC developed the following test 105
method. As development framework for the test method segments of an SSP 34 [m] wind turbine blade provided by DTU Risø were used. For the static test a wind turbine blade segment at station Z=24 [m] of 3 [m] long, 2 [m] wide and 400 [mm] of maximum profile height was tested. In the case of the fatigue tests, a second blade segment with similar dimensions of a second blade was tested.
110
Figure 3 shows the internal structure of the sub-component specimen is divided in the thick laminates spar cap and vertical webs with an adhesive joint at the bending neutral line. This assembly forms the spar box which is the load carrier. The sandwich panels of the aerodynamical shells are bonded to the spar box. Suction and pressure side aerodynamical panels are bonded at the trailing
115
and leading edge with an overlap thick adhesive joint. The main motivation for the test setup development is to introduce a primary edgewise moment to study the performance of the trailing edge buckling panels and adhesive joint. The edgewise moment is defined by a rotation axis and an off-axis punctual load (see figure 4). Here the rotation axis position
5
120
defines the neutral axis of the applied edgewise moment and its profile. A symmetry plane in the middle of the sub-component was assumed for the test setup development independently of the specimen non-symmetry along the zR axis. The test setup rig described in figure 4 includes the following parts: The sub-component specimen, the wooden clamp reinforcements and the rotating
125
arms. The segment of the blade, or sub-component specimen, is obtained directly by sectioning a blade in parts of a certain sub-component length Lspecimen . The specimen length is Lspecimen = Lgauge + 2 · Llap . The specimen gauge length Lgauge plays a role on the specimen behaviour during loading (this is further
130
explained in section 3.1). The wooden clamp reinforcements allow introduction of the load exerted by the test frame to the sub-component. The load or moment applied by the loading arms is converted into a shear load by the reinforcement clamps. This shear load is distributed along the surface of the sub-component clamping areas
135
with the help of an adhesive paste with low elastic modulus. These areas are defined by the clamp overlap length parameter Llap . The Llap design required a detailed analysis of the shear stresses at the bond line between the wooden clamps and the sub-component. The wooden clamp manufacturing was based on cross-laminated timber plates cut with the aerodynamical blade profiles. The
140
timber plates were stacked together and bonded to the sub-component sides with a bonding paste, forming the wooden clamp reinforcements. The rotating arms are defined by the position of the rotation axis (or hinge) and the length Larm between the rotation axis and the point where the load is applied by the hydraulic actuator. The Larm length determines the applied
145
edgewise moment, with reference at the hinge position. The Larm length and the hydraulic actuator force determine the moment amplitude and signal form exerted on the sub-component specimen during static and fatigue tests. The moment shape and strain distribution across the subcomponent specimen width can be tuned with the variation of the two parame-
150
ters described in figure 5. These are the hinge rotation axis coordinate xR and 6
the hinge rotation axis twist angle. The hinge rotation axis coordinate xR determines the position of the hinge rotation axis vectors. With this parameter, it is possible to tune the edgewise moment neutral line location and the strain distributions along the sub-component 155
width. Conceivably, this can be used to match a given strain distribution observed in a blade test or from a corresponding loading case. The hinge rotation axis twist angle allows to add a flapwise moment contribution to the edgewise moment. Tuning this parameter allows imposing a combined edgewise and flapwise bending moment where strain distributions
160
along the sub-component width mimic biaxial blade loads cases. These two parameters are further discussed in section 3.2. In the present work, static and fatigue tests were performed using the most simple configuration, avoiding combined edgewise and flapwise bending moments with a zero hinge rotation axis twist angle. The hinge rotation axis xR coordinate was
165
chosen to be at the thick laminate CAP middle width. 2.1. Static and fatigue experiments Figure 6 shows that the test rig was built with structural beam elements and fixed to the lab’s strong concrete floor. The test rig comprises two symmetrical loading arms which can rotate due to a hinge attached between the rotating
170
arms and the test rig frame structure. Both rotating arms are connected at the top by a 250 [kN] hydraulic cylinder actuator which was controlled to perform the static and fatigue tests. The rotating arms were bolted to the wooden clamp reinforcements. The latter were bonded to the sub-component specimen using a combination of epoxy resin and epoxy adhesive for the full Llap wood clamp
175
reinforcement. A width Llap of 320 [mm] was used for the static test specimen, and of 450 [mm] for the fatigue tests. Larger width Llap increased the fatigue resistance of the bondline between the wooden clamp reinforcements and the specimen. In both cases the specimen had a Lspecimen length of 3000 [mm] and a Larm distance of 1920 [mm].
180
The static test was carried out by contracting the actuator at a constant 7
compression displacement speed of 5 [mm/min] until failure. The fatigue test goal was to load the specimen with a sinusoidal load until fatigue failure to investigate the failure mechanism. For this purpose, the buckling load C2 was initially obtained from a static test until a linearity slope deviation larger than 185
5 % was observed in the resistance curve (see figure 7) and the out-of-plane displacement versus the rotation arm angle (see figure 8). Based on the C2 buckling load moment MC2 , fatigue loading was selected according to the following four-step load spectrum, 1.1 · 106 cycles at 45 % · MC2 , 6.9 · 105 cycles at 62 % · MC2 , 5.1 · 105 cycles at 80 % · MC2 and 1.0 · 105 cycles at 110 % · MC2 until
190
failure during fatigue. The fatigue test was carried out at a test frequency of 2 [Hz]. It is well known that especially the trailing edge of a wind turbine rotor blade is subjected to tension/compression fatigue loads. Tension loads present a number of engineering challenges in relation to the clamp design. Larger Llap widths are required for tension than for compression loads. From both, compres-
195
sion loads are responsible for the trailing edge buckling and relative separation of the pressure and suction panels (also denominated as blade breathing), which is a more demanding loading case. Based on these two constraints, the stress ratio was simplified to R=10 with load control. However, it should be noted that the test setup allows tuning the compression/tension load ratio with an
200
educated clamp design. Static and fatigue tests were instrumented with a 250 [kN] load cell and displacement sensor in the actuator. The rotating arm angle was computed by trigonometry based on the measurement of the actuator displacement sensor. The edgewise moment M was computed based on the measurement of the load
205
cell and the Larm distance. The static and fatigue tests were recorded with cameras from different angles to visualize the trailing edge buckling wave and ultimate failure [27]. In addition, a laser was positioned as referencing system to indicate the original position of the trailing edge. The sub-components were instrumented with strain gauges of 5 [mm] gauge
210
section to measure the longitudinal strains and the transverse strains at several positions. Figure 9 shows the position of the strain gauges on the sub-component 8
surface for the static tests. Suction and pressure side surfaces were instrumented with the same layout of strain gauges. Strain gauges were positioned according to a grid of 4 columns and 4 rows and labeled from position 001F000 to 215
014F000 where the last numbers refer to the orientation (longitudinal or transverse strain). The grid instrumentation layout allowed plotting the longitudinal strains versus the xR axis to visualize the evolution of the edge moment at different zR coordinates. In addition, linear displacement sensors (S10 to S14) were placed perpendicular to the specimen at the trailing edge position to record the
220
amplitude of the buckling wave formed due to the edgewise moment. A similar sub-component instrumentation was carried out in the case of the fatigue test with a finer displacement sensor distribution. The instrumentation layout is shown in figure 10.
3. Sub-component FE model description, development and analysis 225
Test setup development and analysis were based on the construction of a subcomponent FE model. Based on this model, a parametric analysis was carried out. The specimen length (sec. 3.1), the multi-axial load introduction (sec. 3.2) and the clamp overlap [28] parameters were considered as design variables. In addition, the FE model behaviour and results of the static tests were compared.
230
For this purpose, a sub-component FE model was built using 2D quad shell elements (Type 75 Quad4) with MSC Marc. A convergence study was carried out for three different methods to obtain the buckling load at point C2. A linear load bifurcation buckling analysis (LBA) was carried out using the Lanczos and the inverse power sweep method. In addition,
235
the buckling load was computed from the transient simulation at 5% linearity stiffness slope deviation of the force versus displacement curve. Figure 11 shows a good agreement for the three methods and convergence of the buckling load at point C2 above 10000 nodes which correspond to a maximum element length of 0.05 [m] and aspect ratios below 0.6.
240
The FE model was divided into different regions with material properties
9
and thicknesses according to the sub-component structure (see figure 12). The thick laminate CAP regions were modelled as a variable thickness of 24 to 38 [mm] distributed along zR axis elements with a [Biax, UD, Biax] layup. The sandwich regions were modelled with thicknesses from 20 to 30 [mm] with a 245
[Biax, PVC foam, Biax] layup. Since the main purpose of the model was to compare the longitudinal strains and the buckling load with the experiments, dissipative modelling energy methods (i.e. cohesive elements) were avoided. Therefore, it was not possible to model the failure load. The regions located at the trailing edge and leading edge adhesive joints were modelled as a single
250
line of nodes connecting the pressure and suction shells. Using shell elements of a thickness of 5 [mm] and a [Biax, Adhesive, Biax] layup for those locations. The regions of the reinforcements clamps were modelled adding an extra thickness of 200 [mm] of wooden material to the layup to mimic the extra stiffness added by the clamps in those locations. Detailed dimensioning of the clamping
255
size region was performed based on an analytical study of the adhesive shear strength between the clamping blocks and the sub-component. Material properties for Biax, UD (unidirectional), Adhesive and PVC foam were collected from properties reported for the original blade [24] from which the sub-components specimens were obtained.
260
Model boundary conditions were applied according to the scheme from figure 4 and are shown in figure 12. The hinge was modelled at the CAP region central node with fixed displacement in the 3 translation degrees of freedom and allowing rotation about the three axes. Since the hinge boundary condition node was located in the region with a high stiffness due to the wood reinforce-
265
ment, local artificial concentrations of stresses at the hinge position were not considered nor transferred to the specimen gauge (Lgauge ). In a similar manner, a point load was applied at the last node of the trailing edge on both sides of the sub-component to create the edgewise moment. The local stress concentrations caused by the point load were neglected since the node was located in the clamp
270
region. The modelling of the static loading cases was performed with a nonlinear analysis applying a linear incremental force. The buckling load at point 10
C2 was solved with a linear load bifurcation buckling analysis (LBA). C3.1 β and C3.2 ∆α criteria angles were computed by trigonometry. In the case of C3.2 ∆α the displacements of inboard and outboard nodes were used. C3.2 β 275
angle was computed by the derivate of the displacements with respect to the zR coordinates of the last nodes row located in the trailing edge, which join the pressure and suction shells (see Eq. 1)):
β=
d∆ yR d(yR − yR,unload ) = dzR dzR
(1)
3.1. Influence of the specimen length The mentioned FE model was used to evaluate the influence of the spec280
imen length. For this purpose, different models were built with a constant Llap = 200 [mm] modifying the total Lspecimen from 2 to 10 [m]. Figure 13 shows a global edgewise stiffness decrease as the sub-component length increases. However, the buckling load C2 remains stable showing a slight increase in the smallest of sub-component lengths, where for very small Lgauge the buckling
285
load tends to infinity. This suggests that there is a minimum Lgauge length for which the C2 buckling load is not representative of the blade trailing edge buckling in that section. Figure 14 shows the buckling wave amplitudes for sub-component lengths between 2 to 10 [m]. Modal changes can be observed for buckling waves of sub-
290
components 2, 4 and 7 [m] long, where non-symmetrical waves are observed. This corresponds to lengths where several harmonic modes overlap. While the sub-component length increases, the buckling wave amplitudes remain stable, and the number of cycles per buckling wave increases with the length. The increase of the number of cycles implies that the period of the wave remains
295
constant and is not dependent on the sub-component length. Figure 15 shows a comparison of the longitudinal strains for each subcomponent length at the central cross section of the sub-component. Independently of the sub-component length, longitudinal strains do not show a change in the slope plotted along the chord xR . Moreover, longitudinal strain variations 11
300
between sub-component lengths are correlated with variations in the buckling wave amplitudes and buckling loads. Sub-components with lengths of 2 meters show the highest deviations, albeit not larger than 5%. The selection of the sub-component length is a trade-off between short subcomponents which require less test setup effort and lengths that are represen-
305
tative of the full blade performance on edgewise loading. According to this analysis, the data suggest Lspecimen > 3 [m]. However, the selection of the subcomponent length depends on the blade geometry and cross-sectional moment of inertia evolution along the blade span. Therefore it is recommended to perform an FE analysis to select the optimal sub-component length. The analysis
310
of longitudinal strains in the middle section or the buckling load evolution and desired buckling wave mode can be considered as indicators. 3.2. Multi-axial test, specimen inclination angle The longitudinal strains resulting from the application of an edgewise moment can be tailored by varying the test setup parameter related with the
315
xR,hinge hinge position and the hinge rotation axis twist angle (see figure 5). Modifying both parameters allows imposing a given longitudinal strain profile which matches the longitudinal strains obtained e.g. from a full-scale blade model or blade test for a given load case. Figure 16 shows the longitudinal strain profiles in the middle section when
320
the xR,hinge hinge position is translated along the xR axis from -0.74 to 0.51 [m], when the xR axis origin is positioned at the middle of the CAP. This figure shows that the longitudinal strain profiles are dependent on the xR,hinge hinge position, where the highest strains will be obtained when xR,hinge = 0 for identical edgewise moments. For higher or lower xR,hinge values, the strain profile mag-
325
nitude decreases due to the fact that the edgewise stiffness is dependent on the boundary condition, and translating the hinge position to the leading edge or the trailing edge increases the model edgewise stiffness. Figure 17 shows the longitudinal strain profiles in the middle section when the hinge rotation axis twist angle γhinge is varied from 78 to -78 ◦ . Here γhinge = 0 corresponds to 12
330
a position where the profile chord line and rotating arms are aligned. Figure 17 shows that increasing or decreasing the hinge angle opens the longitudinal strain profile, promoting a deviation between the pressure and suction side strains. This difference between pressure and suction side strains is related to a flapwise component contribution. Therefore, according to the parametric anal-
335
ysis, varying the xR,hinge hinge position and hinge angle γhinge allows imposing a longitudinal strain profile which matches a combined flapwise and edgewise load case strain profile distribution.
4. Results and discussion 4.1. Sub-component static test results 340
The static subcomponent specimen was loaded with an edgewise moment at a constant speed until failure. During the static test, a characteristic buckling wave was formed along the trailing edge panels above a certain load. The formation of the buckling wave promoted the full debonding of the trailing edge adhesive joint. The static test showed three main stages (see figure 18): the
345
pre-buckling, the post-buckling and the ultimate failure and debonding stage. The pre-buckling stage: This first stage was characterized by a linear edgewise stiffness until the C2 buckling load was reached. No visible buckling wave was formed by the trailing edge panels, and low amplitudes for the out-of-plane XZ displacement sensors were recorded (see figure 19).
350
The post-buckling stage: Once the C2 buckling load has been reached, a stiffness non-linearity developed during the post-buckling stage until the final failure. The inflection point was determined as the C2 buckling point located at 1◦ of rotating arms angle and at an edgewise moment of 95.47 [kN · m]. During the post-buckling stage, off-plane XZ displacement sensors recorded a rapid
355
increase of the trailing edge buckling wave amplitudes, leading to maximum amplitudes of 30 [mm] before the final failure (see figure 19). The ultimate failure stage: The failure was characterized by a sudden trailing edge adhesive joint fracture. The fracture ocurred all along the trailing edge
13
length and at an edgewise moment of 142 [kN · m]. Once the trailing edge adhe360
sive joint failed, a strong decrease in edgewise stiffness was recorded (see figure 18). The failure occurred at the highest buckling wave amplitudes recorded by the out-of-plane XZ displacement sensors. The longitudinal strains plotted across the blade chord xR were function of the applied edgewise moment. Figure 20 shows that longitudinal strains
365
were distributed from a compression state at the trailing edge, to a tension state at the leading edge. Moreover, the longitudinal strain distribution shows zero values for chord positions crossing the hinge position at the central CAP location or neutral axis. In figure 20, longitudinal strains show a less nonlinearity along the chord direction during the pre-buckling stage than during the post-
370
buckling stage. This change in the strain profiles was related to the buckling wave formation recorded by the out-of-plane XZ displacement sensors which altered the strain field. Figure 21 shows the side view of the trailing edge during the test. The images show how the buckling wave is formed at different levels of edgewise
375
moments. While at 20% of the total edgewise capacity the buckling wave is barely visible, higher edgewise moment show higher buckling wave amplitudes until final adhesive failure. The failure mechanism of the trailing edge adhesive joint is driven by the buckling wave formation (also visible in video [27]). The increment of the buckling wave amplitude is associated with a multiaxial load
380
state in the trailing edge adhesive joint, where the pressure and suction trailing panels tends to separate (this is also denominated as the breathing of the blade), leading to higher peeling loads along the adhesive joint. Similar failure mechanisms have been reported for blade tests [20, 24, 22] and trailing edge blade failures in operation [8], where due to the action of combined flapwise and
385
edgewise moments adhesive joint failures occurred. 4.2. Sub-component FE modelling. Static tests The static test setup was modelled according to the model specifications described in section 3. Figure 22 shows the model deformed shape at different 14
rotation arm loading angles where the maximum stress failure index is plot390
ted. Above the point C2 buckling load (in the post-buckling stage), the model deformed shape shows a buckling wave with a third order profile along the trailing edge. In the pre-buckling stage, the deformed shape shows low buckling wave amplitudes with no visible trailing edge panel buckling. The model deformed shape is comparable to the one recorded during the static test and
395
shown in figure 21. Both the model and static test results (see section 4.1) agree in the buckling wave order which is determined by the Lgauge parameter (see section 3.1). Larger failure index values are located on maximum and minimum buckling wave amplitudes where multiaxial strain fields can be expected. The observed specimen loading and failure suggest that due to large
400
deformation the strains and material damage develop together. However, this is not enough to prevent the specimen loading above the pre-buckling region. Once the post-buckling region is reached without significant structural failure, the further loading increases the nonlinear deformation of the trailing edge adhesive joint and sandwich panel, resulting in large stress mode mixity in the
405
bondline and in the skin laminates. Consequently, the trailing edge adhesive joint fails, resulting in a permanent decrease in edgewise stiffness. The edgewise moments and rotation arm angles were computed from the model based on the reaction forces and nodal displacements. Figure 18 shows the edgewise moment versus the rotation angle from the static test and the
410
model. In this figure, the resistance curve slope relates to the edgewise stiffness. Edgewise stiffnesses from models and experiments show a good agreement, and in both it is possible to observe the characteristic change in stiffness which defines the C2 buckling load. The C2 buckling load could be determined by two different methods. The first one, extracting from the experimental and
415
models resistance curve the inflexion point using the linearity slope deviation. The second one, performing a linear load bifurcation analysis (LBA) with the models. Showing differences lower than a 5 % between the mentioned methods. Since the blade profile is asymmetric and the loading vector and boundary conditions do not coincide with the inertia axis of the section, the structure is 15
420
unstable. Therefore, no further artificial boundary conditions were needed to promote the buckling behaviour in the models. Longitudinal strains (in zR direction) recorded during the test and from the FE model are shown in figure 23. Experimental and FE model strain distribution show that an edgewise moment arises across the blade (in xR axis).
425
Negative strains in the trailing edge area indicate compression loading, while positive strains in the leading edge area indicate tension loading. The strain distribution is zero at the neutral line position located in the central CAP location where the hinge is located. This strain distribution coincides with one of an edgewise moment diagram crossing the neutral line, with positive values in
430
the leading edge and negative values in the trailing edge. The model suggests that the trailing edge opening (or breathing of the blade) caused by an edgewise moment is related to the compression loads formed along the trailing edge. Figure 24 shows the criterion C3.2 ∆α opening angle of the trailing edge plotted versus the rotation arm displacements. The figure shows
435
that the trailing edge opening angle increment exhibits a strong dependency on the pre-buckling and post-buckling stages transition. While during the prebuckling stage the opening angle of the trailing edge shows a weak dependency on the edgewise moment, once the post-buckling stage is reached the trailing edge opening angle varies linearly with the load. Similar behaviour can be
440
observed with the C3.1 β criterion in figure 24. The β and ∆α angles can be evaluated independently by instrumenting sub-component tests through numerical models. Therefore these criteria are measurable and comparable. Criteria from section 1.1 and loading curve are shown in figure 25. This figure was normalized based on the LBA load factor from the C2 criterion. The C1
445
criterion corresponds to the maximum load prior to failure from sub-component tests. Criteria C3.1 and C3.2 were determined at maximum values of 0.01 [rad] and 0.5 [◦ ] respectively based on figure 24 data. Criteria C4.1M ax.stress and C4.2T sai−W u were determined based on the maximum stress failure and TsaiWu [29] stress analysis criteria, respectively, without considering safety factors.
450
These are criteria implemented for in-plane stress states. These two criteria 16
do not account for out-of-plane and peeling stresses. Therefore, it can lead to designs above the trailing edge buckling C2 point and not prevent the failure mechanism described by the tests (see figure 25). To account for out-of-plane and peeling stresses; several modelling alternatives exist such as modelling the 455
trailing edge with a detailed 3D mesh or using cohesive elements. When these techniques are too costly, the use of stability criteria allows predicting design regions for multiple load cases. According to the observations, it is recommended that designs aim to be in the trailing edge pre-buckling region where low β and ∆α angles are expected as well as low mode-mixity values.
460
4.3. Sub-component fatigue tests results A fatigue test was carried out with the same test configuration as the static test (see section 2 and video [30]). Inspections of fatigue tests and the static test found that no observable damage occurred at specimen plywood clamps or the adhesive joint with the specimen. A four-step load spectrum was applied until
465
failure occurred during fatigue loading. These four loading steps are shown in figure 27 and figure 28. In these figures, the evolution of the edgewise moments and the rotation arm angles is described and the loading spectrum can be observed. Since the loads were restricted to the pre-buckling region, the first three fatigue loading steps showed no buckling wave. However, the fourth loading
470
step at 110 % of the C2 buckling criterion showed an out-of-plane displacement and the formation of a buckling wave (see figure 26). This buckling wave was of the same order as the buckling wave observed during the static test. The failure occurred in fatigue after 2.4 million cycles and an edgewise compression loading of 110 % of the C2 criterion buckling load (see figure 26). The adhesive
475
joint failure occurred along the Lgauge in a similar manner as the one observed during the static test. During the fatigue test, a stiffness degradation of 20 % was recorded from the beginning of the test to prior to final failure. It should be pointed out that the failure mechanisms obtained from this test are dependent on specimen geometries, material properties, layup and edge-
480
wise / flapwise load contribution. Therefore, advanced numerical models which 17
can predict the variety of failure modes and allow test design and result evaluation in combination with the experimental observations are recommended.
5. Conclusions A sub-component test setup for full-scale wind turbine blade segments was 485
proposed, designed and constructed. Static and fatigue tests were successfully conducted. The test setup loads wind turbine blade sections with a main edgewise moment, allowing to impose a desired strain field which can be obtained from a blade loading case. Based on a numerical analysis, two main parameters were identified in the test setup design. Since the test is dominated by struc-
490
tural buckling and boundary conditions are influencing the structural stability, the specimen length Lspecimen needs to be considered in the test design. It was proposed that by modifying the specimen inclination or hinge rotation angle γhinge an edgewise and flapwise loading ratio can be imposed and tuned. Two full-scale blade sections of 3×2 meters were evaluated applying an edge-
495
wise moment according to the proposed test setup until failure during static and fatigue tests. The experimental and numerical results were evaluated and described based on the trailing edge strength and stability criteria. Static tests showed the formation of a buckling wave along the trailing edge until the trailing edge adhesive joint failure was reached. Pre-buckling and post-buckling regions
500
were identified in the global loading curve response, showing a nonlinear structural response. The edgewise stiffness slope change is defined by the C2 buckling load and the transition between a linear and nonlinear region. This transition point also coincided with a rapid increment of the out-of-plane displacements, the buckling wave and the trailing edge opening angles C3.2 ∆α and C3.1 β.
505
A fatigue test was carried out until failure was reached after 2.4 million cycles. The failure mechanism was similar to the one observed during the static test. A buckling wave formation was observed during the fatigue test once C2 buckling load maximum amplitude was reached. An edgewise stiffness degradation of around 20 % was observed during the fatigue test.
18
510
Acknowledgments The authors acknowledge the contribution of the EU 7th framework programs FP7-ENERGY-2013 IRPWIND Project (under Grant Agreement 609795) and the Dutch TKI-WoZ research SLOWIND project to motivate and partly fund this research. The authors also acknowledge DTU for providing the blade
515
sections and the feedback of IWES, CRES and CENER as IRPWind consortium members.
References [1] M. Wilkinson, B. Hendriks, F. Spinato, E. Gomez, H. Bulacio, P. Tavner, Y. Feng, H. Long, Methodology and Results of the Reliawind Reliability 520
Field Study, European Wind Energy Conference ( EWEC 2010 ) (2010) 7. [2] Report on Wind Turbine Subsystem Reliability - A Survey of Various Databases, NREL/PR-5000-59111,National Renewable Energy Laboratory. [3] M. Reder, E. Gonzalez, J. J. Melero, Wind Turbine Failure Analysis Targeting current problems in Failure Data Analysis, Journal of Physics:
525
Conference Series 753 (072027). doi:10.1088/1742-6596/753/7/072027. [4] S. Faulstich, B. Hahn, P. J. Tavner, Wind turbine downtime and its importance for offshore deployment, Wind Energy 14 (3) (2011) 327–337. doi:10.1002/we.421. [5] M. Junginger, A. Faaij, W. Turkenburg, Cost reduction prospects for off-
530
shore wind farms, Wind engineering 28 (1) (2004) 97–118. doi:10.1260/ 0309524041210847. [6] Y. Wenxian, Testing and condition monitoring of composite wind turbine blades, in: B. Attaf (Ed.), Recent advances in composite materials for wind turbines blades, The World Academic Publishing Co. Ltd., Hong Kong,
535
2013, Ch. 9, pp. 147–169.
19
[7] Progress and trends in nondestructive testing and evaluation for wind turbine composite blade, Renewable and Sustainable Energy Reviews 60 (2016) 1225–1250. doi:10.1016/j.rser.2016.02.026. [8] S. Ataya, M. M. Z. Ahmed, Damages of wind turbine blade trailing 540
edge: Forms, location, and root causes, Engineering Failure Analysisdoi: 10.1016/j.engfailanal.2013.05.011. [9] P. U. Haselbach, K. Branner, Initiation of trailing edge failure in full-scale wind turbine blade test, Engineering Fracture Mechanics 162 (2016) 136– 154. doi:10.1016/j.engfracmech.2016.04.041.
545
[10] F. Sayer, A. Antoniou, A. Van Wingerde, Investigation of structural bond lines in wind turbine blades by sub-component tests, International Journal of Adhesion and Adhesives 37 (2012) 129–135. doi:10.1016/j.ijadhadh. 2012.01.021. [11] M. Rosemeier, F. Lahuerta, K. Branner, A. Antoniou, Subcomponent test-
550
ing of wind turbine blades, in: IRPWind conference (Amsterdam), 2016. doi:10.13140/RG.2.2.30244.88962. [12] K. Branner, P. Berring, P. Haselbach, X. Chen, A. Antoniou, M. Rosemeier, C. Pueyo, A. Makris, G. Roukis, F. Lahuerta, R. Nijssen, Report on validation and improvement of blade design tools. Work Package 7.1:
555
Efficient blade structure Deliverable number 71.3, Tech. rep., IRPWInd, Grant agreement no 609795 (2018). [13] A. Antoniou, M. Rosemeier, C. Pueyo, A. Makris, G. Roukis, F. Lahuerta, R. Nijssen, Testing report on blade subcomponents. Work Package 7.1: Efficient blade structure Deliverable number 7.1.2, Tech. rep., IRPWind,
560
Grant agreement no 609795 (2017). [14] K. Branner, P. Berring, P. U. Haselbach, Subcomponent testing of trailing edge panels in wind turbine blades, in: ECCM17 - 17th European Conference on Composite Materials, no. June, Munich, Germany, 2017. 20
[15] L. Mishnaevsky, K. Branner, H. N. Petersen, J. Beauson, M. McGugan, 565
B. F. Sørensen, Materials for wind turbine blades: An overview, Materials 10 (11) (2017) 1–24. doi:10.3390/ma10111285. [16] F. Lahuerta, M. J. D. Ruiter, L. Espinosa, N. Koorn, D. Smissaert, Assessment of wind turbine blade trailing edge failure with sub-component tests, in: ICCM21 (Ed.), 21st International Conference on Composite Materials
570
(ICCM21, Xi’an), no. August, Xi’an, 2017, pp. 20–25. [17] M. Rosemeier, G. Basters, A. Antoniou, Benefits of sub-component over full-scale blade testing elaborated on a trailing edge bond line design validation, Wind Energy Science 2017 (August) (2017) 1–14. doi:10.5194/ wes-2017-35.
575
[18] M. Rosemeier, M. B¨ atge, A. Antoniou, A novel single actuator test setup for combined loading of wind turbine rotor blade sub-components, in: 2nd International Symposium on Multiscale Experimental Mechanics: Multiscale Fatigue, Lyngby, 2017, pp. 1–10. [19] M. A. Eder, R. D. Bitsche, F. Belloni, Effects of geometric non-linearity on
580
energy release rates in a realistic wind turbine blade cross section, Composite Structures 132 (2015) 1075–1084. doi:10.1016/j.compstruct.2015. 06.050. [20] M. A. Eder, R. D. Bitsche, Fracture analysis of adhesive joints in wind turbine blades, Wind Energy 18 (6) (2015) 1007–1022. arXiv:arXiv:1006.
585
4405v1, doi:10.1002/we.1744. [21] M. Eder, R. Bitsche, M. Nielsen, K. Branner, A practical approach to fracture analysis at the trailing edge of wind turbine rotor blades, Wind Energy 17 (3) (2014) 483–497. doi:10.1002/we.1591. [22] L. T. T. Data, M. Nielsen, F. M. Jensen, P. H. Nielsen, K. Martyniuk,
590
A. Roczek, T. Sieradzan, V. Roudnitski, P. Kucio, R. Bitsche, P. Andresen, T. Lukassen, Z. Andrlov´a, K. Branner, C. Bak, M. Mcgugan, H. Knudsen, 21
A. B. Rasmussen, J. J. Rasmussen, V. J. Wedel-heinen, A. M. Nielsen, H. Per, B. Kallesøe, T. Lindby, F. Sørensen, C. Jensen, U. Uldahl, B. Rasmussen, Full scale test of SSP 34m blade, edgewise loading LTT. Data 595
report 1, Tech. Rep. January, Risø National Laboratory for Sustainable Energy, Technical University of Denmark (2010). [23] Find M. Jensen, A. S. Puri, J. P. Dear, K. Branner, A. Morris, Investigating the impact of non-linear geometrical effects on wind turbine bladesPart 1: Current status of design and test methods and future challenges in design
600
optimization, Wind Energy 14 (2) (2010) 239–254. doi:10.1002/we.415. [24] P. U. Haselbach, Ultimate Strength of Wind Turbine Blades under Multiaxial Loading, Phd, Technical University of Denmark (2015). [25] CEN, Eurocode 3: Design of steel structures - Part 1-6: Strength and stability of shell structures; EN 1993-1-6:2007 + AC:2009, European Committee
605
for Standardization, 2009. [26] GL, Guideline for the Certification of Wind Turbines, Germanischer Lloyd Industrial Services. Hamburg, Germany, 2010. [27] G. de Winkel, YouTube video: Wind turbine rotor blade trailing edge test to failure (2016).
610
URL https://www.youtube.com/watch?v=6pM6SlzNooQ&t=1s [28] F. Lahuerta, Buckling tests on trailing edges. Preliminary FE. Hinge at the middle of the CAP. WMC-2016-007-cc, Tech. rep., Knowledge Centre WMC, Knowledge Centre WMC, Wieringerwerf (2016). [29] S. W. Tsai, E. M. Wu, A General Theory of Strength for Anisotropic
615
Materials, Journal of Composite Materials 5 (1) (1971) 58–80.
doi:
10.1177/002199837100500106. [30] F. Lahuerta, G. de Winkel, YouTube video: Wind turbine rotor blade trailing edge fatigue test to failure (sub-component) (2018). URL https://www.youtube.com/watch?v=U4G9h1h3A04 22
Load factor
C1 C4
Post-buckling stage
C2 C3.1 C3.2
Pre-buckling stage ∆zR
Figure 1: Edgewise stiffness resistance curve. Trailing edge strength and stability criteria.
yR xR
βload
yR
αload = α0 + ∆α
zR C3.1. Load factor at largest tolerable rotation angle C3.2. Load factor at largest tolerable opening & of a buckling wave along the trailing edge closing increment angle between unloaded and loaded βload (zR ) < βmax case along the trailing edge ∆α(zR ) < ∆αmax
Figure 2: Trailing edge strength and stability criteria C3.1 & C3.2.
Trailing edge detail
Trailing edge Adhesive joint
Sandwich panel
Trailing edge panels
Thick laminate CAP Spar CAP Load carrier.
Leading edge Adhesive joint Leading edge panels
Figure 3: Blade segment tested as sub-component specimen. Main parts.
23
Force
Force
xR zR
Llap
Lgauge
Llap Larm
Trailing edge
CAP (Thick laminate)
Hinge
Rotating arm
Wood clamp
Leading edge
Hinge
Wood clamp
Rotating arm
Figure 4: Scheme of the test method. Main parts.
hinge rotation axis γhinge hinge rotation axis twist angle
principal axis chord coordinate xR hinge rotation axis xR chord coordinate
Figure 5: Test setup hinge position parameters.
Figure 6: Test rig setup. Detail of strain gauges and displacement sensors.
24
( G J H Z L V H P R P H Q W 0 > N 1 P @
3 U H E X F N O L Q J 3 R V W E X F N O L Q J % X F N O L Q J O R D G P R P H Q W 0 & D W > N 1 P @
5 R W D W L R Q D U P D Q J O H > @
2 X W S O D Q H G L V S O D F H P H Q W > P P @
Figure 7: Edgewise moment versus rotating arm angle. Fatigue test determination of the buckling load moment MC2 .
3 U H E X F N O L Q J 6 3 R V W E X F N O L Q J 6 % X F N O L Q J O R D G P R P H Q W 0 &
5 R W D W L R Q D U P D Q J O H > @
Figure 8: Out-plane displacement sensor S10 versus rotating arm angle. Fatigue test determination of the buckling load moment MC2 .
25
90 direcction
5,00
004F000
001F000, 001F090
80,00
5.00
S13
007F000 010F000, 010F090 008F000 011F000, 011F090 012F000, 012F090
005F000 002F000, 002F090 006F000
009F000
91 ,2 1
003F000, 003F090
15,00
S14
40,00
S12
S11
S10
40,00
71,80
35.9
80,00
71,80
15.00
*1.21 with spar cap axis
91 ,2 1
0 direction*
(spar cap axis) 013F000
10,00
10,00
014F000
32,00
245,50
32,00
Figure 9: Instrumentation plan for static test. Strain gauges oriented according to axis 0◦ ,90◦ .
90 direction 31.5
S11
S12
5.00
S10
31.5
31.5
S13
S14
005F000, 005F090
58.3
29.1
001F000
31.5
15,00
S16
5.00
31.5
S15
29.1
31.5
009F000
58.3
15.00
0 direction* *1.21 with spar cap axis
006F000, 006F090 002F000
147.5
010F000
91
,2 1
007F000, 007F090 011F000
(spar cap axis)
91 ,2 1
122.5
003F000
012F000 004F000
10
008F000, 008F090
45
219
45
Figure 10: Instrumentation plan for fatigue test. Strain gauges oriented according to axis 0◦ ,90◦ .
26
( G J H Z L V H P R P H Q W 0 0 P D [
/ D Q F ] R V P H W K R G , Q Y H U V H S R Z H U V Z H H S P H W K R G O L Q H D U L W \ V O R S H G H Y L D W L R Q
) ( P R G H O Q X P E H U R I Q R G H V
Figure 11: FE model convergence analysis.
Figure 12: FE model description. Regions division and boundary conditions.
27
( G J H Z L V H P R P H Q W 0 0 P D [
6 X E F R P S O H Q J W K > P @ 6 X E F R P S O H Q J W K > P @ 6 X E F R P S O H Q J W K > P @ 6 X E F R P S O H Q J W K > P @ 6 X E F R P S O H Q J W K > P @ 6 X E F R P S O H Q J W K > P @ % X F N O L Q J O R D G
( G J H Z L V H G L V S O D F H P H Q W
P D [
Figure 13: Edgewise moment versus edgewise displacement (both normalize to a maximum value) for sub-components lengths from 2 to 10 [m].
/ V S H F L P H Q / V S H F L P H Q / V S H F L P H Q / V S H F L P H Q / V S H F L P H Q / V S H F L P H Q / V S H F L P H Q / V S H F L P H Q / V S H F L P H Q
> P P @ V F D O H % X F N O L Q J Z D Y H V D W R I W K H E X F N O L Q J O R D G 6 X E F R P S O H Q J W K ] 5 > P @
Figure 14: Buckling wave amplitudes (displacement in [mm]) along zR trailing edge for specimen lengths from 2 to 10 [m]. Buckling waves shown at 150% of the C2 buckling load.
28
/ R Q J L W X G L Q D O V W U D L Q > V W U D L Q @
/ H Q J W K > P @ / H Q J W K > P @ / H Q J W K > P @ / H Q J W K > P @ / H Q J W K > P @ / H Q J W K > P @ / H Q J W K > P @ / H Q J W K > P @ / H Q J W K > P @
[ 5 F K R U G S R V L W L R Q
Figure 15: Longitudinal strains along the chord xR (at the middle section) for subcomponents lengths from 2 to 10 [m]. Strains shown at 30% of the C2 buckling load.
/ R Q J L W X G L Q D O V W U D L Q > V W U D L Q @
[ 5 K L Q J H [ 5 K L Q J H [ 5 K L Q J H [ 5 K L Q J H [ 5 K L Q J H [ 5 K L Q J H [ 5 K L Q J H
[ 5 D [ L V > P @ R U L J L Q D W P L G G O H & $ 3
Figure 16: Longitudinal strains (in zR direction) along the chord xR (at the middle section) of a 3 [m] sub-component at different xR,hinge hinge position. Strains shown at 30% of the C2 buckling load.
29
/ R Q J L W X G L Q D O V W U D L Q > V W U D L Q @
K L Q J H
K L Q J H K L Q J H K L Q J H K L Q J H K L Q J H K L Q J H
[ 5 D [ L V > P @ R U L J L Q D W P L G G O H & $ 3
Figure 17: Longitudinal strains (in zR direction) along the chord xR (at the middle section) of a 3 [m] sub-component at different hinge rotation axis twist angle γhinge . Hinge xR coordinate at middle of the CAP. Strains shown at 30% of the C2 buckling load.
( G J H Z L V H P R P H Q W 0 > N 1 P @
3 U H E X F N O L Q J 3 R V W E X F N O L Q J ) D L O X U H 8 Q O R D G % X F N O L Q J P R P H Q W > N 1 P @ ) ( P R G H O
5 R W D W L R Q D U P D Q J O H > @
Figure 18: Static test edgewise moment versus rotating arm angle (compression loading). Static test results and FE model curve.
30
' L V S O D F H P H Q W R I I S O D Q H > P P @
6 6 6 6 6
5 R W D W L R Q D U P D Q J O H > @
Figure 19: Static test off-plane XZ displacements versus rotating arm angle. Measurements of liners displacement sensors S10 to S14.
/ R Q J L W X G L Q D O V W U D L Q V > V W U D L Q @
0 H G J H Z L V H > N 1 P @ 0 H G J H Z L V H > N 1 P @ 0 H G J H Z L V H > N 1 P @ 0 H G J H Z L V H > N 1 P @ 0 H G J H Z L V H > N 1 P @
[ 5 D [ L V > P @ R U L J L Q D W P L G G O H & $ 3
Figure 20: Static test longitudinal strain (in zR direction) versus chord position (in xR axis) at different edgewise moments. Strain gauges 004F000,005F000,006F000,013F000.
31
Figure 21: Side view of the trailing edge during the test. Buckling wave formation and adhesive joint failure.
1.13 0.72 0.30 0.01 0.52 0.93
Rotation arm angle 0.22°
1.34
Rotation arm angle 0.56°
1.76 2.17
Max imum stress f ailure index
2.58 3.00 Rotation arm angle 1.0°
Figure 22: FE model at different rotation arm loading angles. Color gradient represent the maximum stress failure index.
32
/ R Q J L W X G L Q D O V W U D L Q V > V W U D L Q @
0 H G J H Z L V H > N 1 P @ 0 H G J H Z L V H > N 1 P @ 0 H G J H Z L V H > N 1 P @ ) ( P R G H O D W > N 1 P @ ) ( P R G H O D W > N 1 P @ ) ( P R G H O D W > N 1 P @
[ 5 D [ L V > P @ R U L J L Q D W P L G G O H & $ 3
* H R P H W U L F D O D Q J O H V D Q G H Y R O X W L R Q > @
Figure 23: Static test longitudinal strain (in zR direction) versus chord position (in xR axis) at different edgewise moments. Comparison between FE model and strain gauges measurements from tests. Strain gauges 004F000,005F000,006F000,013F000.
/ 7 7 O R D G 6 & 7 O H Q J W K > P @ & D Q J O H / 7 7 O R D G 6 & 7 O H Q J W K > P @ & D Q J O H
' L V S O D F H P H Q W
&
Figure 24: C3.1 β and C3.2 ∆α versus rotation arm displacement. Axes normalized based on C2 criterion.
33
/ R D G 3 3 &
/ 7 7 O R D G 6 & 7 O H Q J W K > P @ & P D [ O R D G & H D U O L H U E L I X U F D W L R Q S R L Q W / % $ & > U D G @ & > @ & 0 D [ V W U H V V V W U H V V D Q D O \ V L V & 7 V D L : X V W U H V V D Q D O \ V L V
' L V S O D F H P H Q W &
Figure 25: Buckling resistance criteria applied to specimen loading curve. Axes normalized based on C2 criterion.
34
Figure 26: Side view of the trailing edge during fatigue tests. Buckling wave formation and adhesive joint failure. Four steps load spectrum, 1.1 · 106 cycles at 45 % · MC2 , 6.9 · 105 cycles at 62 % · MC2 , 5.1 · 105 cycles at 80 % · MC2 and 1.0 · 105 cycles at 110 % · MC2 until failure during fatigue.
35
( G J H Z L V H P R P H Q W 0 > N 1 P @
0 D Y J W U D I D W B V 1 P D [ > F \ F O H V @ 0 P D [ W U D I D W B V 0 P L Q W U D I D W B V
& \ F O H V W R I D L O X U H > N F \ F O H V @
Figure 27: Maximum, minimum and average edgewise moment versus cycles to failure. Fatigue tests. Four steps load spectrum, 1.1 · 106 cycles at 45 % · MC2 , 6.9 · 105 cycles at 62 % · MC2 , 5.1 · 105 cycles at 80 % · MC2 and 1.0 · 105 cycles at 110 % · MC2 until failure during fatigue.
5 R W D W L R Q D U P D Q J O H > @
D Y J W U D I D W B V 1 P D [ > F \ F O H V @
P D [ W U D I D W B V P L Q W U D I D W B V
& \ F O H V W R I D L O X U H > N F \ F O H V @
Figure 28: Maximum, minimum and average rotation arm angle versus cycles to failure. Fatigue tests. Four steps load spectrum, 1.1 · 106 cycles at 45 % · MC2 , 6.9 · 105 cycles at 62 % · MC2 , 5.1 · 105 cycles at 80 % · MC2 and 1.0 · 105 cycles at 110 % · MC2 until failure during fatigue.
36
S0263-8223(18)31045-6 https://doi.org/10.1016/j.compstruct.2018.07.112 COST 10030
To appear in:
Composite Structures
Please cite this article as: Lahuerta, F., Koorn, N., Smissaert, D., Wind turbine blade trailing edge failure assessment with sub-component test on static and fatigue load conditions, Composite Structures (2018), doi: https://doi.org/ 10.1016/j.compstruct.2018.07.112
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Wind turbine blade trailing edge failure assessment with sub-component test on static and fatigue load conditions F. Lahuerta*, N. Koorn, D. Smissaert Knowledge Centre WMC. Kluisgat 5, 1771 MV Wieringerwerf, The Netherlands.
Abstract Wind turbine blades present different types of failure mechanisms and modes which are associated with specific loading conditions. Trailing edge failure mode has been documented in full-scale blade tests as one of the failure types observed in blades on service. Trailing edge failure is characterized by failure of the trailing edge adhesive joint and the buckling of the trailing edge sandwich panels. This failure is governed by the contribution of edgewise, flapwise and torsion moments, with edgewise moments being the main driver. This paper describes a blade sub-component test setup suitable for studying trailing edge failure on static and fatigue load conditions, which is an improvement in the experimental verification of a trailing edge blade design. The test setup and design drivers are described and studied via FE models. Static and fatigue test results are reported for a full-scale blade section sub-component obtained from a 34 [m] wind turbine blade. Moreover, experimental results are discussed and compared with FE models to describe and study the trailing edge failure mechanism. Keywords: wind turbine blade, adhesive, sub component, buckling, edge moment, trailing, failure
∗ Corresponding
author Email address: [email protected] (F. Lahuerta*)
Preprint submitted to Composite structures
August 4, 2018
1. Introduction Operation and maintenance costs (O&M) are one of the cost drivers in the wind energy sector. Several studies [1, 2, 3] identified the downtime due to blade failures and blade O&M actions to be between 3% to 10%. Especially for 5
offshore turbines these cost and downtime are expected to be even larger [4, 5] . Among the different types of blade failure modes described by Wenxian [6] and Yang [7], Ataya [8] suspected the trailing edge (TE) adhesive failure as one of the main issues concerning the reliability of blade designs leading to failures and downtime.
10
According to IEC 61400-5 test pyramid for rotor blades is divided into fullscale blade test, sub-component test (full-scale structural features), generic element and details (non-full scale generic specimens) and coupon test (where fundamental material properties are obtained). Carrying out full-scale blade tests to study and assess wind turbine blades trailing edge behaviour can be
15
expensive and impractical. Full-scale blade tests are performed at later stages of the design process for qualification purposes [9] at which point there is little room for design improvements. Sub-component tests are therefore more suitable for the study and design of local wind turbine blade failure mechanisms such as trailing edge joint failures. Thick adhesive joint failures have been studied
20
before using generic elements tests, i.e. beam tests [10]. However, very few developments of sub-component test methods to study the trailing edge adhesive failures are available. Several institutions (DTU, IWES, CENER, CRES, WMC) join forces in the IRPWind project to develop sub-component test setups and study trailing edge failure and adhesive joints [11, 12, 13, 14, 15, 16].
25
Studying wind turbine blade structural designs by blade segments [17, 18] via sub-component tests was considered in this project. Trailing edge adhesive joint fracture was described by Eder [19, 20], who reported that the most critical failure mode is related to Mode I contribution (peeling forces) due to the opening of the trailing edge. A method to estimate
30
the SERR (strain energy release rate) of different fracture modes in the trailing
2
edges of large wind turbine rotor blade models was proposed [21]. Blade tests performed by DTU Risø [22] for an SSP 34 [m] blade suggested the occurrence of trailing edge adhesive joint failure due to local buckling in trailing edge panels. In addition, an internal wire reinforcement for the SSP 34 [m] blade was 35
implemented to prevent out-of-plane deflection of the panels, thereby reducing the peeling stresses on the adhered region and impeding the Brazier effect [23]. Moreover, Branner & Haselbach [24] described the blade test trailing edge failure mechanism by an FE model which considered transverse shear stresses (by use of an 8-noded doubly curved thick shell element). It was found that buck-
40
ling led to the trailing edge adhesive joint final failure well below the expected maximum load. It should be pointed out that some of the literature findings might not be totally applicable to the modern MW scale blades. It is known that most blade manufacturers with spar box load carrying structures changed their design to
45
the traditional spar CAP/shear web construction. This has a substantial effect on the trailing edge stiffness (i.e. stiffer constructions). The aim of the work is to describe a test method which allows studying wind turbine blade trailing edge failures due to loading cases dominated by an edgewise moment. The test method allows testing full-scale segments of
50
a blade or scaled sub-components. The test setup design and configuration are discussed with an FE parametric analysis. Experimental results in static and fatigue loading conditions are reported for SSP 34 [m] blade segments, the failure mechanism is described and instrumentation readings are discussed in comparison with FE models.
55
1.1. Trailing edge strength and stability criteria Preliminary results from blade [24] and sub-component tests [16] revealed that three main stages can be observed in the trailing edge instability and failure. This observation agrees with the behaviour of an unstable buckling structure and can be characterized by three main regions.
3
60
The pre-buckling stage. This stage is characterized by a linear edgewise stiffness until the buckling load is reached at point C2 (see figure 1). The trailing edge post-buckling stage. This stage shows a drop of the sub-component edgewise stiffness and a non-linear edgewise stiffness behaviour until the final failure is reached at point C1 (see figure 1).
65
The ultimate failure stage. The failure stage is characterized by the change of edgewise stiffness until suddenly the trailing edge adhesive joint fracture debilitates the structural connection between the suction and pressure side trailing edge panels. At this stage a drop of the edgewise stiffness is observed (point C4 in figure 1).
70
Trailing edge strength and stability criteria were proposed [12] based on Eurocode for the stability of shell structures [25]. Five different criteria are proposed to evaluate the initiation of geometrically non-linear deformation and characterize the sub-component test buckling resistance. Both, stability (C1 to C3) and strength (C4) perspectives are considered. These criteria approach the
75
characterization of the trailing edge resistance from different levels of detail (see figure 1). Criteria C1 and C2 study the sub-component global stiffness curve characteristic features. Criteria C3.1 and C3.2 focus on measurable geometrical variations of the structure. Criterion C4 is based on the classical study of the local material resistance and stress levels.
80
Criterion C1 corresponds to the load-displacement curve (load-factor or lead-lag moment) maximum force at final failure or the maximum bifurcation point if the latter occurs earlier than the final failure. Criterion C2 relates to a linear load bifurcation analysis (LBA). The earlier load bifurcation point from a buckling analysis corresponds to C2 criterion. For
85
experimental results, the C2 criterion corresponds to the earliest point where a change of the stiffness rate is identified and the start of a buckling wave can be observed. It should be noted that criterion C2 load bifurcation analysis differs from global buckling analysis of a blade on the loading conditions. Criteria C3.1 and C3.2 describe measurable geometrical variations related
90
to the buckling of the trailing edge. The β and α angles can be evaluated in 4
both sub-component tests and FE models (see figure 2). The β angle describes the buckling wave degree of development for a given load factor. The α angle is suitable for trailing edge studies according to Eder [19]. Moreover, the α angle is related to the specific trailing edge adhesive joint geometry and the peeling 95
forces that appear with increasing the load factor. In both criteria a maximum ∆αmax or βmax is defined as threshold value. Criterion C4 is based on the design guidelines [26] for stress analysis. A detailed analysis of the adhesive joint is performed and local design stress values are compared against material component resistances. Since this type of analysis
100
is not directly comparable with experimental results, criterion C4 is applicable via modelling techniques. Alternatively, the C4 criterion could be based on the SERR mode-mixity evaluation of the adhesive joint [12].
2. Test setup description Within the IRPWind project framework, WMC developed the following test 105
method. As development framework for the test method segments of an SSP 34 [m] wind turbine blade provided by DTU Risø were used. For the static test a wind turbine blade segment at station Z=24 [m] of 3 [m] long, 2 [m] wide and 400 [mm] of maximum profile height was tested. In the case of the fatigue tests, a second blade segment with similar dimensions of a second blade was tested.
110
Figure 3 shows the internal structure of the sub-component specimen is divided in the thick laminates spar cap and vertical webs with an adhesive joint at the bending neutral line. This assembly forms the spar box which is the load carrier. The sandwich panels of the aerodynamical shells are bonded to the spar box. Suction and pressure side aerodynamical panels are bonded at the trailing
115
and leading edge with an overlap thick adhesive joint. The main motivation for the test setup development is to introduce a primary edgewise moment to study the performance of the trailing edge buckling panels and adhesive joint. The edgewise moment is defined by a rotation axis and an off-axis punctual load (see figure 4). Here the rotation axis position
5
120
defines the neutral axis of the applied edgewise moment and its profile. A symmetry plane in the middle of the sub-component was assumed for the test setup development independently of the specimen non-symmetry along the zR axis. The test setup rig described in figure 4 includes the following parts: The sub-component specimen, the wooden clamp reinforcements and the rotating
125
arms. The segment of the blade, or sub-component specimen, is obtained directly by sectioning a blade in parts of a certain sub-component length Lspecimen . The specimen length is Lspecimen = Lgauge + 2 · Llap . The specimen gauge length Lgauge plays a role on the specimen behaviour during loading (this is further
130
explained in section 3.1). The wooden clamp reinforcements allow introduction of the load exerted by the test frame to the sub-component. The load or moment applied by the loading arms is converted into a shear load by the reinforcement clamps. This shear load is distributed along the surface of the sub-component clamping areas
135
with the help of an adhesive paste with low elastic modulus. These areas are defined by the clamp overlap length parameter Llap . The Llap design required a detailed analysis of the shear stresses at the bond line between the wooden clamps and the sub-component. The wooden clamp manufacturing was based on cross-laminated timber plates cut with the aerodynamical blade profiles. The
140
timber plates were stacked together and bonded to the sub-component sides with a bonding paste, forming the wooden clamp reinforcements. The rotating arms are defined by the position of the rotation axis (or hinge) and the length Larm between the rotation axis and the point where the load is applied by the hydraulic actuator. The Larm length determines the applied
145
edgewise moment, with reference at the hinge position. The Larm length and the hydraulic actuator force determine the moment amplitude and signal form exerted on the sub-component specimen during static and fatigue tests. The moment shape and strain distribution across the subcomponent specimen width can be tuned with the variation of the two parame-
150
ters described in figure 5. These are the hinge rotation axis coordinate xR and 6
the hinge rotation axis twist angle. The hinge rotation axis coordinate xR determines the position of the hinge rotation axis vectors. With this parameter, it is possible to tune the edgewise moment neutral line location and the strain distributions along the sub-component 155
width. Conceivably, this can be used to match a given strain distribution observed in a blade test or from a corresponding loading case. The hinge rotation axis twist angle allows to add a flapwise moment contribution to the edgewise moment. Tuning this parameter allows imposing a combined edgewise and flapwise bending moment where strain distributions
160
along the sub-component width mimic biaxial blade loads cases. These two parameters are further discussed in section 3.2. In the present work, static and fatigue tests were performed using the most simple configuration, avoiding combined edgewise and flapwise bending moments with a zero hinge rotation axis twist angle. The hinge rotation axis xR coordinate was
165
chosen to be at the thick laminate CAP middle width. 2.1. Static and fatigue experiments Figure 6 shows that the test rig was built with structural beam elements and fixed to the lab’s strong concrete floor. The test rig comprises two symmetrical loading arms which can rotate due to a hinge attached between the rotating
170
arms and the test rig frame structure. Both rotating arms are connected at the top by a 250 [kN] hydraulic cylinder actuator which was controlled to perform the static and fatigue tests. The rotating arms were bolted to the wooden clamp reinforcements. The latter were bonded to the sub-component specimen using a combination of epoxy resin and epoxy adhesive for the full Llap wood clamp
175
reinforcement. A width Llap of 320 [mm] was used for the static test specimen, and of 450 [mm] for the fatigue tests. Larger width Llap increased the fatigue resistance of the bondline between the wooden clamp reinforcements and the specimen. In both cases the specimen had a Lspecimen length of 3000 [mm] and a Larm distance of 1920 [mm].
180
The static test was carried out by contracting the actuator at a constant 7
compression displacement speed of 5 [mm/min] until failure. The fatigue test goal was to load the specimen with a sinusoidal load until fatigue failure to investigate the failure mechanism. For this purpose, the buckling load C2 was initially obtained from a static test until a linearity slope deviation larger than 185
5 % was observed in the resistance curve (see figure 7) and the out-of-plane displacement versus the rotation arm angle (see figure 8). Based on the C2 buckling load moment MC2 , fatigue loading was selected according to the following four-step load spectrum, 1.1 · 106 cycles at 45 % · MC2 , 6.9 · 105 cycles at 62 % · MC2 , 5.1 · 105 cycles at 80 % · MC2 and 1.0 · 105 cycles at 110 % · MC2 until
190
failure during fatigue. The fatigue test was carried out at a test frequency of 2 [Hz]. It is well known that especially the trailing edge of a wind turbine rotor blade is subjected to tension/compression fatigue loads. Tension loads present a number of engineering challenges in relation to the clamp design. Larger Llap widths are required for tension than for compression loads. From both, compres-
195
sion loads are responsible for the trailing edge buckling and relative separation of the pressure and suction panels (also denominated as blade breathing), which is a more demanding loading case. Based on these two constraints, the stress ratio was simplified to R=10 with load control. However, it should be noted that the test setup allows tuning the compression/tension load ratio with an
200
educated clamp design. Static and fatigue tests were instrumented with a 250 [kN] load cell and displacement sensor in the actuator. The rotating arm angle was computed by trigonometry based on the measurement of the actuator displacement sensor. The edgewise moment M was computed based on the measurement of the load
205
cell and the Larm distance. The static and fatigue tests were recorded with cameras from different angles to visualize the trailing edge buckling wave and ultimate failure [27]. In addition, a laser was positioned as referencing system to indicate the original position of the trailing edge. The sub-components were instrumented with strain gauges of 5 [mm] gauge
210
section to measure the longitudinal strains and the transverse strains at several positions. Figure 9 shows the position of the strain gauges on the sub-component 8
surface for the static tests. Suction and pressure side surfaces were instrumented with the same layout of strain gauges. Strain gauges were positioned according to a grid of 4 columns and 4 rows and labeled from position 001F000 to 215
014F000 where the last numbers refer to the orientation (longitudinal or transverse strain). The grid instrumentation layout allowed plotting the longitudinal strains versus the xR axis to visualize the evolution of the edge moment at different zR coordinates. In addition, linear displacement sensors (S10 to S14) were placed perpendicular to the specimen at the trailing edge position to record the
220
amplitude of the buckling wave formed due to the edgewise moment. A similar sub-component instrumentation was carried out in the case of the fatigue test with a finer displacement sensor distribution. The instrumentation layout is shown in figure 10.
3. Sub-component FE model description, development and analysis 225
Test setup development and analysis were based on the construction of a subcomponent FE model. Based on this model, a parametric analysis was carried out. The specimen length (sec. 3.1), the multi-axial load introduction (sec. 3.2) and the clamp overlap [28] parameters were considered as design variables. In addition, the FE model behaviour and results of the static tests were compared.
230
For this purpose, a sub-component FE model was built using 2D quad shell elements (Type 75 Quad4) with MSC Marc. A convergence study was carried out for three different methods to obtain the buckling load at point C2. A linear load bifurcation buckling analysis (LBA) was carried out using the Lanczos and the inverse power sweep method. In addition,
235
the buckling load was computed from the transient simulation at 5% linearity stiffness slope deviation of the force versus displacement curve. Figure 11 shows a good agreement for the three methods and convergence of the buckling load at point C2 above 10000 nodes which correspond to a maximum element length of 0.05 [m] and aspect ratios below 0.6.
240
The FE model was divided into different regions with material properties
9
and thicknesses according to the sub-component structure (see figure 12). The thick laminate CAP regions were modelled as a variable thickness of 24 to 38 [mm] distributed along zR axis elements with a [Biax, UD, Biax] layup. The sandwich regions were modelled with thicknesses from 20 to 30 [mm] with a 245
[Biax, PVC foam, Biax] layup. Since the main purpose of the model was to compare the longitudinal strains and the buckling load with the experiments, dissipative modelling energy methods (i.e. cohesive elements) were avoided. Therefore, it was not possible to model the failure load. The regions located at the trailing edge and leading edge adhesive joints were modelled as a single
250
line of nodes connecting the pressure and suction shells. Using shell elements of a thickness of 5 [mm] and a [Biax, Adhesive, Biax] layup for those locations. The regions of the reinforcements clamps were modelled adding an extra thickness of 200 [mm] of wooden material to the layup to mimic the extra stiffness added by the clamps in those locations. Detailed dimensioning of the clamping
255
size region was performed based on an analytical study of the adhesive shear strength between the clamping blocks and the sub-component. Material properties for Biax, UD (unidirectional), Adhesive and PVC foam were collected from properties reported for the original blade [24] from which the sub-components specimens were obtained.
260
Model boundary conditions were applied according to the scheme from figure 4 and are shown in figure 12. The hinge was modelled at the CAP region central node with fixed displacement in the 3 translation degrees of freedom and allowing rotation about the three axes. Since the hinge boundary condition node was located in the region with a high stiffness due to the wood reinforce-
265
ment, local artificial concentrations of stresses at the hinge position were not considered nor transferred to the specimen gauge (Lgauge ). In a similar manner, a point load was applied at the last node of the trailing edge on both sides of the sub-component to create the edgewise moment. The local stress concentrations caused by the point load were neglected since the node was located in the clamp
270
region. The modelling of the static loading cases was performed with a nonlinear analysis applying a linear incremental force. The buckling load at point 10
C2 was solved with a linear load bifurcation buckling analysis (LBA). C3.1 β and C3.2 ∆α criteria angles were computed by trigonometry. In the case of C3.2 ∆α the displacements of inboard and outboard nodes were used. C3.2 β 275
angle was computed by the derivate of the displacements with respect to the zR coordinates of the last nodes row located in the trailing edge, which join the pressure and suction shells (see Eq. 1)):
β=
d∆ yR d(yR − yR,unload ) = dzR dzR
(1)
3.1. Influence of the specimen length The mentioned FE model was used to evaluate the influence of the spec280
imen length. For this purpose, different models were built with a constant Llap = 200 [mm] modifying the total Lspecimen from 2 to 10 [m]. Figure 13 shows a global edgewise stiffness decrease as the sub-component length increases. However, the buckling load C2 remains stable showing a slight increase in the smallest of sub-component lengths, where for very small Lgauge the buckling
285
load tends to infinity. This suggests that there is a minimum Lgauge length for which the C2 buckling load is not representative of the blade trailing edge buckling in that section. Figure 14 shows the buckling wave amplitudes for sub-component lengths between 2 to 10 [m]. Modal changes can be observed for buckling waves of sub-
290
components 2, 4 and 7 [m] long, where non-symmetrical waves are observed. This corresponds to lengths where several harmonic modes overlap. While the sub-component length increases, the buckling wave amplitudes remain stable, and the number of cycles per buckling wave increases with the length. The increase of the number of cycles implies that the period of the wave remains
295
constant and is not dependent on the sub-component length. Figure 15 shows a comparison of the longitudinal strains for each subcomponent length at the central cross section of the sub-component. Independently of the sub-component length, longitudinal strains do not show a change in the slope plotted along the chord xR . Moreover, longitudinal strain variations 11
300
between sub-component lengths are correlated with variations in the buckling wave amplitudes and buckling loads. Sub-components with lengths of 2 meters show the highest deviations, albeit not larger than 5%. The selection of the sub-component length is a trade-off between short subcomponents which require less test setup effort and lengths that are represen-
305
tative of the full blade performance on edgewise loading. According to this analysis, the data suggest Lspecimen > 3 [m]. However, the selection of the subcomponent length depends on the blade geometry and cross-sectional moment of inertia evolution along the blade span. Therefore it is recommended to perform an FE analysis to select the optimal sub-component length. The analysis
310
of longitudinal strains in the middle section or the buckling load evolution and desired buckling wave mode can be considered as indicators. 3.2. Multi-axial test, specimen inclination angle The longitudinal strains resulting from the application of an edgewise moment can be tailored by varying the test setup parameter related with the
315
xR,hinge hinge position and the hinge rotation axis twist angle (see figure 5). Modifying both parameters allows imposing a given longitudinal strain profile which matches the longitudinal strains obtained e.g. from a full-scale blade model or blade test for a given load case. Figure 16 shows the longitudinal strain profiles in the middle section when
320
the xR,hinge hinge position is translated along the xR axis from -0.74 to 0.51 [m], when the xR axis origin is positioned at the middle of the CAP. This figure shows that the longitudinal strain profiles are dependent on the xR,hinge hinge position, where the highest strains will be obtained when xR,hinge = 0 for identical edgewise moments. For higher or lower xR,hinge values, the strain profile mag-
325
nitude decreases due to the fact that the edgewise stiffness is dependent on the boundary condition, and translating the hinge position to the leading edge or the trailing edge increases the model edgewise stiffness. Figure 17 shows the longitudinal strain profiles in the middle section when the hinge rotation axis twist angle γhinge is varied from 78 to -78 ◦ . Here γhinge = 0 corresponds to 12
330
a position where the profile chord line and rotating arms are aligned. Figure 17 shows that increasing or decreasing the hinge angle opens the longitudinal strain profile, promoting a deviation between the pressure and suction side strains. This difference between pressure and suction side strains is related to a flapwise component contribution. Therefore, according to the parametric anal-
335
ysis, varying the xR,hinge hinge position and hinge angle γhinge allows imposing a longitudinal strain profile which matches a combined flapwise and edgewise load case strain profile distribution.
4. Results and discussion 4.1. Sub-component static test results 340
The static subcomponent specimen was loaded with an edgewise moment at a constant speed until failure. During the static test, a characteristic buckling wave was formed along the trailing edge panels above a certain load. The formation of the buckling wave promoted the full debonding of the trailing edge adhesive joint. The static test showed three main stages (see figure 18): the
345
pre-buckling, the post-buckling and the ultimate failure and debonding stage. The pre-buckling stage: This first stage was characterized by a linear edgewise stiffness until the C2 buckling load was reached. No visible buckling wave was formed by the trailing edge panels, and low amplitudes for the out-of-plane XZ displacement sensors were recorded (see figure 19).
350
The post-buckling stage: Once the C2 buckling load has been reached, a stiffness non-linearity developed during the post-buckling stage until the final failure. The inflection point was determined as the C2 buckling point located at 1◦ of rotating arms angle and at an edgewise moment of 95.47 [kN · m]. During the post-buckling stage, off-plane XZ displacement sensors recorded a rapid
355
increase of the trailing edge buckling wave amplitudes, leading to maximum amplitudes of 30 [mm] before the final failure (see figure 19). The ultimate failure stage: The failure was characterized by a sudden trailing edge adhesive joint fracture. The fracture ocurred all along the trailing edge
13
length and at an edgewise moment of 142 [kN · m]. Once the trailing edge adhe360
sive joint failed, a strong decrease in edgewise stiffness was recorded (see figure 18). The failure occurred at the highest buckling wave amplitudes recorded by the out-of-plane XZ displacement sensors. The longitudinal strains plotted across the blade chord xR were function of the applied edgewise moment. Figure 20 shows that longitudinal strains
365
were distributed from a compression state at the trailing edge, to a tension state at the leading edge. Moreover, the longitudinal strain distribution shows zero values for chord positions crossing the hinge position at the central CAP location or neutral axis. In figure 20, longitudinal strains show a less nonlinearity along the chord direction during the pre-buckling stage than during the post-
370
buckling stage. This change in the strain profiles was related to the buckling wave formation recorded by the out-of-plane XZ displacement sensors which altered the strain field. Figure 21 shows the side view of the trailing edge during the test. The images show how the buckling wave is formed at different levels of edgewise
375
moments. While at 20% of the total edgewise capacity the buckling wave is barely visible, higher edgewise moment show higher buckling wave amplitudes until final adhesive failure. The failure mechanism of the trailing edge adhesive joint is driven by the buckling wave formation (also visible in video [27]). The increment of the buckling wave amplitude is associated with a multiaxial load
380
state in the trailing edge adhesive joint, where the pressure and suction trailing panels tends to separate (this is also denominated as the breathing of the blade), leading to higher peeling loads along the adhesive joint. Similar failure mechanisms have been reported for blade tests [20, 24, 22] and trailing edge blade failures in operation [8], where due to the action of combined flapwise and
385
edgewise moments adhesive joint failures occurred. 4.2. Sub-component FE modelling. Static tests The static test setup was modelled according to the model specifications described in section 3. Figure 22 shows the model deformed shape at different 14
rotation arm loading angles where the maximum stress failure index is plot390
ted. Above the point C2 buckling load (in the post-buckling stage), the model deformed shape shows a buckling wave with a third order profile along the trailing edge. In the pre-buckling stage, the deformed shape shows low buckling wave amplitudes with no visible trailing edge panel buckling. The model deformed shape is comparable to the one recorded during the static test and
395
shown in figure 21. Both the model and static test results (see section 4.1) agree in the buckling wave order which is determined by the Lgauge parameter (see section 3.1). Larger failure index values are located on maximum and minimum buckling wave amplitudes where multiaxial strain fields can be expected. The observed specimen loading and failure suggest that due to large
400
deformation the strains and material damage develop together. However, this is not enough to prevent the specimen loading above the pre-buckling region. Once the post-buckling region is reached without significant structural failure, the further loading increases the nonlinear deformation of the trailing edge adhesive joint and sandwich panel, resulting in large stress mode mixity in the
405
bondline and in the skin laminates. Consequently, the trailing edge adhesive joint fails, resulting in a permanent decrease in edgewise stiffness. The edgewise moments and rotation arm angles were computed from the model based on the reaction forces and nodal displacements. Figure 18 shows the edgewise moment versus the rotation angle from the static test and the
410
model. In this figure, the resistance curve slope relates to the edgewise stiffness. Edgewise stiffnesses from models and experiments show a good agreement, and in both it is possible to observe the characteristic change in stiffness which defines the C2 buckling load. The C2 buckling load could be determined by two different methods. The first one, extracting from the experimental and
415
models resistance curve the inflexion point using the linearity slope deviation. The second one, performing a linear load bifurcation analysis (LBA) with the models. Showing differences lower than a 5 % between the mentioned methods. Since the blade profile is asymmetric and the loading vector and boundary conditions do not coincide with the inertia axis of the section, the structure is 15
420
unstable. Therefore, no further artificial boundary conditions were needed to promote the buckling behaviour in the models. Longitudinal strains (in zR direction) recorded during the test and from the FE model are shown in figure 23. Experimental and FE model strain distribution show that an edgewise moment arises across the blade (in xR axis).
425
Negative strains in the trailing edge area indicate compression loading, while positive strains in the leading edge area indicate tension loading. The strain distribution is zero at the neutral line position located in the central CAP location where the hinge is located. This strain distribution coincides with one of an edgewise moment diagram crossing the neutral line, with positive values in
430
the leading edge and negative values in the trailing edge. The model suggests that the trailing edge opening (or breathing of the blade) caused by an edgewise moment is related to the compression loads formed along the trailing edge. Figure 24 shows the criterion C3.2 ∆α opening angle of the trailing edge plotted versus the rotation arm displacements. The figure shows
435
that the trailing edge opening angle increment exhibits a strong dependency on the pre-buckling and post-buckling stages transition. While during the prebuckling stage the opening angle of the trailing edge shows a weak dependency on the edgewise moment, once the post-buckling stage is reached the trailing edge opening angle varies linearly with the load. Similar behaviour can be
440
observed with the C3.1 β criterion in figure 24. The β and ∆α angles can be evaluated independently by instrumenting sub-component tests through numerical models. Therefore these criteria are measurable and comparable. Criteria from section 1.1 and loading curve are shown in figure 25. This figure was normalized based on the LBA load factor from the C2 criterion. The C1
445
criterion corresponds to the maximum load prior to failure from sub-component tests. Criteria C3.1 and C3.2 were determined at maximum values of 0.01 [rad] and 0.5 [◦ ] respectively based on figure 24 data. Criteria C4.1M ax.stress and C4.2T sai−W u were determined based on the maximum stress failure and TsaiWu [29] stress analysis criteria, respectively, without considering safety factors.
450
These are criteria implemented for in-plane stress states. These two criteria 16
do not account for out-of-plane and peeling stresses. Therefore, it can lead to designs above the trailing edge buckling C2 point and not prevent the failure mechanism described by the tests (see figure 25). To account for out-of-plane and peeling stresses; several modelling alternatives exist such as modelling the 455
trailing edge with a detailed 3D mesh or using cohesive elements. When these techniques are too costly, the use of stability criteria allows predicting design regions for multiple load cases. According to the observations, it is recommended that designs aim to be in the trailing edge pre-buckling region where low β and ∆α angles are expected as well as low mode-mixity values.
460
4.3. Sub-component fatigue tests results A fatigue test was carried out with the same test configuration as the static test (see section 2 and video [30]). Inspections of fatigue tests and the static test found that no observable damage occurred at specimen plywood clamps or the adhesive joint with the specimen. A four-step load spectrum was applied until
465
failure occurred during fatigue loading. These four loading steps are shown in figure 27 and figure 28. In these figures, the evolution of the edgewise moments and the rotation arm angles is described and the loading spectrum can be observed. Since the loads were restricted to the pre-buckling region, the first three fatigue loading steps showed no buckling wave. However, the fourth loading
470
step at 110 % of the C2 buckling criterion showed an out-of-plane displacement and the formation of a buckling wave (see figure 26). This buckling wave was of the same order as the buckling wave observed during the static test. The failure occurred in fatigue after 2.4 million cycles and an edgewise compression loading of 110 % of the C2 criterion buckling load (see figure 26). The adhesive
475
joint failure occurred along the Lgauge in a similar manner as the one observed during the static test. During the fatigue test, a stiffness degradation of 20 % was recorded from the beginning of the test to prior to final failure. It should be pointed out that the failure mechanisms obtained from this test are dependent on specimen geometries, material properties, layup and edge-
480
wise / flapwise load contribution. Therefore, advanced numerical models which 17
can predict the variety of failure modes and allow test design and result evaluation in combination with the experimental observations are recommended.
5. Conclusions A sub-component test setup for full-scale wind turbine blade segments was 485
proposed, designed and constructed. Static and fatigue tests were successfully conducted. The test setup loads wind turbine blade sections with a main edgewise moment, allowing to impose a desired strain field which can be obtained from a blade loading case. Based on a numerical analysis, two main parameters were identified in the test setup design. Since the test is dominated by struc-
490
tural buckling and boundary conditions are influencing the structural stability, the specimen length Lspecimen needs to be considered in the test design. It was proposed that by modifying the specimen inclination or hinge rotation angle γhinge an edgewise and flapwise loading ratio can be imposed and tuned. Two full-scale blade sections of 3×2 meters were evaluated applying an edge-
495
wise moment according to the proposed test setup until failure during static and fatigue tests. The experimental and numerical results were evaluated and described based on the trailing edge strength and stability criteria. Static tests showed the formation of a buckling wave along the trailing edge until the trailing edge adhesive joint failure was reached. Pre-buckling and post-buckling regions
500
were identified in the global loading curve response, showing a nonlinear structural response. The edgewise stiffness slope change is defined by the C2 buckling load and the transition between a linear and nonlinear region. This transition point also coincided with a rapid increment of the out-of-plane displacements, the buckling wave and the trailing edge opening angles C3.2 ∆α and C3.1 β.
505
A fatigue test was carried out until failure was reached after 2.4 million cycles. The failure mechanism was similar to the one observed during the static test. A buckling wave formation was observed during the fatigue test once C2 buckling load maximum amplitude was reached. An edgewise stiffness degradation of around 20 % was observed during the fatigue test.
18
510
Acknowledgments The authors acknowledge the contribution of the EU 7th framework programs FP7-ENERGY-2013 IRPWIND Project (under Grant Agreement 609795) and the Dutch TKI-WoZ research SLOWIND project to motivate and partly fund this research. The authors also acknowledge DTU for providing the blade
515
sections and the feedback of IWES, CRES and CENER as IRPWind consortium members.
References [1] M. Wilkinson, B. Hendriks, F. Spinato, E. Gomez, H. Bulacio, P. Tavner, Y. Feng, H. Long, Methodology and Results of the Reliawind Reliability 520
Field Study, European Wind Energy Conference ( EWEC 2010 ) (2010) 7. [2] Report on Wind Turbine Subsystem Reliability - A Survey of Various Databases, NREL/PR-5000-59111,National Renewable Energy Laboratory. [3] M. Reder, E. Gonzalez, J. J. Melero, Wind Turbine Failure Analysis Targeting current problems in Failure Data Analysis, Journal of Physics:
525
Conference Series 753 (072027). doi:10.1088/1742-6596/753/7/072027. [4] S. Faulstich, B. Hahn, P. J. Tavner, Wind turbine downtime and its importance for offshore deployment, Wind Energy 14 (3) (2011) 327–337. doi:10.1002/we.421. [5] M. Junginger, A. Faaij, W. Turkenburg, Cost reduction prospects for off-
530
shore wind farms, Wind engineering 28 (1) (2004) 97–118. doi:10.1260/ 0309524041210847. [6] Y. Wenxian, Testing and condition monitoring of composite wind turbine blades, in: B. Attaf (Ed.), Recent advances in composite materials for wind turbines blades, The World Academic Publishing Co. Ltd., Hong Kong,
535
2013, Ch. 9, pp. 147–169.
19
[7] Progress and trends in nondestructive testing and evaluation for wind turbine composite blade, Renewable and Sustainable Energy Reviews 60 (2016) 1225–1250. doi:10.1016/j.rser.2016.02.026. [8] S. Ataya, M. M. Z. Ahmed, Damages of wind turbine blade trailing 540
edge: Forms, location, and root causes, Engineering Failure Analysisdoi: 10.1016/j.engfailanal.2013.05.011. [9] P. U. Haselbach, K. Branner, Initiation of trailing edge failure in full-scale wind turbine blade test, Engineering Fracture Mechanics 162 (2016) 136– 154. doi:10.1016/j.engfracmech.2016.04.041.
545
[10] F. Sayer, A. Antoniou, A. Van Wingerde, Investigation of structural bond lines in wind turbine blades by sub-component tests, International Journal of Adhesion and Adhesives 37 (2012) 129–135. doi:10.1016/j.ijadhadh. 2012.01.021. [11] M. Rosemeier, F. Lahuerta, K. Branner, A. Antoniou, Subcomponent test-
550
ing of wind turbine blades, in: IRPWind conference (Amsterdam), 2016. doi:10.13140/RG.2.2.30244.88962. [12] K. Branner, P. Berring, P. Haselbach, X. Chen, A. Antoniou, M. Rosemeier, C. Pueyo, A. Makris, G. Roukis, F. Lahuerta, R. Nijssen, Report on validation and improvement of blade design tools. Work Package 7.1:
555
Efficient blade structure Deliverable number 71.3, Tech. rep., IRPWInd, Grant agreement no 609795 (2018). [13] A. Antoniou, M. Rosemeier, C. Pueyo, A. Makris, G. Roukis, F. Lahuerta, R. Nijssen, Testing report on blade subcomponents. Work Package 7.1: Efficient blade structure Deliverable number 7.1.2, Tech. rep., IRPWind,
560
Grant agreement no 609795 (2017). [14] K. Branner, P. Berring, P. U. Haselbach, Subcomponent testing of trailing edge panels in wind turbine blades, in: ECCM17 - 17th European Conference on Composite Materials, no. June, Munich, Germany, 2017. 20
[15] L. Mishnaevsky, K. Branner, H. N. Petersen, J. Beauson, M. McGugan, 565
B. F. Sørensen, Materials for wind turbine blades: An overview, Materials 10 (11) (2017) 1–24. doi:10.3390/ma10111285. [16] F. Lahuerta, M. J. D. Ruiter, L. Espinosa, N. Koorn, D. Smissaert, Assessment of wind turbine blade trailing edge failure with sub-component tests, in: ICCM21 (Ed.), 21st International Conference on Composite Materials
570
(ICCM21, Xi’an), no. August, Xi’an, 2017, pp. 20–25. [17] M. Rosemeier, G. Basters, A. Antoniou, Benefits of sub-component over full-scale blade testing elaborated on a trailing edge bond line design validation, Wind Energy Science 2017 (August) (2017) 1–14. doi:10.5194/ wes-2017-35.
575
[18] M. Rosemeier, M. B¨ atge, A. Antoniou, A novel single actuator test setup for combined loading of wind turbine rotor blade sub-components, in: 2nd International Symposium on Multiscale Experimental Mechanics: Multiscale Fatigue, Lyngby, 2017, pp. 1–10. [19] M. A. Eder, R. D. Bitsche, F. Belloni, Effects of geometric non-linearity on
580
energy release rates in a realistic wind turbine blade cross section, Composite Structures 132 (2015) 1075–1084. doi:10.1016/j.compstruct.2015. 06.050. [20] M. A. Eder, R. D. Bitsche, Fracture analysis of adhesive joints in wind turbine blades, Wind Energy 18 (6) (2015) 1007–1022. arXiv:arXiv:1006.
585
4405v1, doi:10.1002/we.1744. [21] M. Eder, R. Bitsche, M. Nielsen, K. Branner, A practical approach to fracture analysis at the trailing edge of wind turbine rotor blades, Wind Energy 17 (3) (2014) 483–497. doi:10.1002/we.1591. [22] L. T. T. Data, M. Nielsen, F. M. Jensen, P. H. Nielsen, K. Martyniuk,
590
A. Roczek, T. Sieradzan, V. Roudnitski, P. Kucio, R. Bitsche, P. Andresen, T. Lukassen, Z. Andrlov´a, K. Branner, C. Bak, M. Mcgugan, H. Knudsen, 21
A. B. Rasmussen, J. J. Rasmussen, V. J. Wedel-heinen, A. M. Nielsen, H. Per, B. Kallesøe, T. Lindby, F. Sørensen, C. Jensen, U. Uldahl, B. Rasmussen, Full scale test of SSP 34m blade, edgewise loading LTT. Data 595
report 1, Tech. Rep. January, Risø National Laboratory for Sustainable Energy, Technical University of Denmark (2010). [23] Find M. Jensen, A. S. Puri, J. P. Dear, K. Branner, A. Morris, Investigating the impact of non-linear geometrical effects on wind turbine bladesPart 1: Current status of design and test methods and future challenges in design
600
optimization, Wind Energy 14 (2) (2010) 239–254. doi:10.1002/we.415. [24] P. U. Haselbach, Ultimate Strength of Wind Turbine Blades under Multiaxial Loading, Phd, Technical University of Denmark (2015). [25] CEN, Eurocode 3: Design of steel structures - Part 1-6: Strength and stability of shell structures; EN 1993-1-6:2007 + AC:2009, European Committee
605
for Standardization, 2009. [26] GL, Guideline for the Certification of Wind Turbines, Germanischer Lloyd Industrial Services. Hamburg, Germany, 2010. [27] G. de Winkel, YouTube video: Wind turbine rotor blade trailing edge test to failure (2016).
610
URL https://www.youtube.com/watch?v=6pM6SlzNooQ&t=1s [28] F. Lahuerta, Buckling tests on trailing edges. Preliminary FE. Hinge at the middle of the CAP. WMC-2016-007-cc, Tech. rep., Knowledge Centre WMC, Knowledge Centre WMC, Wieringerwerf (2016). [29] S. W. Tsai, E. M. Wu, A General Theory of Strength for Anisotropic
615
Materials, Journal of Composite Materials 5 (1) (1971) 58–80.
doi:
10.1177/002199837100500106. [30] F. Lahuerta, G. de Winkel, YouTube video: Wind turbine rotor blade trailing edge fatigue test to failure (sub-component) (2018). URL https://www.youtube.com/watch?v=U4G9h1h3A04 22
Load factor
C1 C4
Post-buckling stage
C2 C3.1 C3.2
Pre-buckling stage ∆zR
Figure 1: Edgewise stiffness resistance curve. Trailing edge strength and stability criteria.
yR xR
βload
yR
αload = α0 + ∆α
zR C3.1. Load factor at largest tolerable rotation angle C3.2. Load factor at largest tolerable opening & of a buckling wave along the trailing edge closing increment angle between unloaded and loaded βload (zR ) < βmax case along the trailing edge ∆α(zR ) < ∆αmax
Figure 2: Trailing edge strength and stability criteria C3.1 & C3.2.
Trailing edge detail
Trailing edge Adhesive joint
Sandwich panel
Trailing edge panels
Thick laminate CAP Spar CAP Load carrier.
Leading edge Adhesive joint Leading edge panels
Figure 3: Blade segment tested as sub-component specimen. Main parts.
23
Force
Force
xR zR
Llap
Lgauge
Llap Larm
Trailing edge
CAP (Thick laminate)
Hinge
Rotating arm
Wood clamp
Leading edge
Hinge
Wood clamp
Rotating arm
Figure 4: Scheme of the test method. Main parts.
hinge rotation axis γhinge hinge rotation axis twist angle
principal axis chord coordinate xR hinge rotation axis xR chord coordinate
Figure 5: Test setup hinge position parameters.
Figure 6: Test rig setup. Detail of strain gauges and displacement sensors.
24
( G J H Z L V H P R P H Q W 0 > N 1 P @
3 U H E X F N O L Q J 3 R V W E X F N O L Q J % X F N O L Q J O R D G P R P H Q W 0 & D W > N 1 P @
5 R W D W L R Q D U P D Q J O H > @
2 X W S O D Q H G L V S O D F H P H Q W > P P @
Figure 7: Edgewise moment versus rotating arm angle. Fatigue test determination of the buckling load moment MC2 .
3 U H E X F N O L Q J 6 3 R V W E X F N O L Q J 6 % X F N O L Q J O R D G P R P H Q W 0 &
5 R W D W L R Q D U P D Q J O H > @
Figure 8: Out-plane displacement sensor S10 versus rotating arm angle. Fatigue test determination of the buckling load moment MC2 .
25
90 direcction
5,00
004F000
001F000, 001F090
80,00
5.00
S13
007F000 010F000, 010F090 008F000 011F000, 011F090 012F000, 012F090
005F000 002F000, 002F090 006F000
009F000
91 ,2 1
003F000, 003F090
15,00
S14
40,00
S12
S11
S10
40,00
71,80
35.9
80,00
71,80
15.00
*1.21 with spar cap axis
91 ,2 1
0 direction*
(spar cap axis) 013F000
10,00
10,00
014F000
32,00
245,50
32,00
Figure 9: Instrumentation plan for static test. Strain gauges oriented according to axis 0◦ ,90◦ .
90 direction 31.5
S11
S12
5.00
S10
31.5
31.5
S13
S14
005F000, 005F090
58.3
29.1
001F000
31.5
15,00
S16
5.00
31.5
S15
29.1
31.5
009F000
58.3
15.00
0 direction* *1.21 with spar cap axis
006F000, 006F090 002F000
147.5
010F000
91
,2 1
007F000, 007F090 011F000
(spar cap axis)
91 ,2 1
122.5
003F000
012F000 004F000
10
008F000, 008F090
45
219
45
Figure 10: Instrumentation plan for fatigue test. Strain gauges oriented according to axis 0◦ ,90◦ .
26
( G J H Z L V H P R P H Q W 0 0 P D [
/ D Q F ] R V P H W K R G , Q Y H U V H S R Z H U V Z H H S P H W K R G O L Q H D U L W \ V O R S H G H Y L D W L R Q
) ( P R G H O Q X P E H U R I Q R G H V
Figure 11: FE model convergence analysis.
Figure 12: FE model description. Regions division and boundary conditions.
27
( G J H Z L V H P R P H Q W 0 0 P D [
6 X E F R P S O H Q J W K > P @ 6 X E F R P S O H Q J W K > P @ 6 X E F R P S O H Q J W K > P @ 6 X E F R P S O H Q J W K > P @ 6 X E F R P S O H Q J W K > P @ 6 X E F R P S O H Q J W K > P @ % X F N O L Q J O R D G
( G J H Z L V H G L V S O D F H P H Q W
P D [
Figure 13: Edgewise moment versus edgewise displacement (both normalize to a maximum value) for sub-components lengths from 2 to 10 [m].
/ V S H F L P H Q / V S H F L P H Q / V S H F L P H Q / V S H F L P H Q / V S H F L P H Q / V S H F L P H Q / V S H F L P H Q / V S H F L P H Q / V S H F L P H Q
> P P @ V F D O H % X F N O L Q J Z D Y H V D W R I W K H E X F N O L Q J O R D G 6 X E F R P S O H Q J W K ] 5 > P @
Figure 14: Buckling wave amplitudes (displacement in [mm]) along zR trailing edge for specimen lengths from 2 to 10 [m]. Buckling waves shown at 150% of the C2 buckling load.
28
/ R Q J L W X G L Q D O V W U D L Q > V W U D L Q @
/ H Q J W K > P @ / H Q J W K > P @ / H Q J W K > P @ / H Q J W K > P @ / H Q J W K > P @ / H Q J W K > P @ / H Q J W K > P @ / H Q J W K > P @ / H Q J W K > P @
[ 5 F K R U G S R V L W L R Q
Figure 15: Longitudinal strains along the chord xR (at the middle section) for subcomponents lengths from 2 to 10 [m]. Strains shown at 30% of the C2 buckling load.
/ R Q J L W X G L Q D O V W U D L Q > V W U D L Q @
[ 5 K L Q J H [ 5 K L Q J H [ 5 K L Q J H [ 5 K L Q J H [ 5 K L Q J H [ 5 K L Q J H [ 5 K L Q J H
[ 5 D [ L V > P @ R U L J L Q D W P L G G O H & $ 3
Figure 16: Longitudinal strains (in zR direction) along the chord xR (at the middle section) of a 3 [m] sub-component at different xR,hinge hinge position. Strains shown at 30% of the C2 buckling load.
29
/ R Q J L W X G L Q D O V W U D L Q > V W U D L Q @
K L Q J H
K L Q J H K L Q J H K L Q J H K L Q J H K L Q J H K L Q J H
[ 5 D [ L V > P @ R U L J L Q D W P L G G O H & $ 3
Figure 17: Longitudinal strains (in zR direction) along the chord xR (at the middle section) of a 3 [m] sub-component at different hinge rotation axis twist angle γhinge . Hinge xR coordinate at middle of the CAP. Strains shown at 30% of the C2 buckling load.
( G J H Z L V H P R P H Q W 0 > N 1 P @
3 U H E X F N O L Q J 3 R V W E X F N O L Q J ) D L O X U H 8 Q O R D G % X F N O L Q J P R P H Q W > N 1 P @ ) ( P R G H O
5 R W D W L R Q D U P D Q J O H > @
Figure 18: Static test edgewise moment versus rotating arm angle (compression loading). Static test results and FE model curve.
30
' L V S O D F H P H Q W R I I S O D Q H > P P @
6 6 6 6 6
5 R W D W L R Q D U P D Q J O H > @
Figure 19: Static test off-plane XZ displacements versus rotating arm angle. Measurements of liners displacement sensors S10 to S14.
/ R Q J L W X G L Q D O V W U D L Q V > V W U D L Q @
0 H G J H Z L V H > N 1 P @ 0 H G J H Z L V H > N 1 P @ 0 H G J H Z L V H > N 1 P @ 0 H G J H Z L V H > N 1 P @ 0 H G J H Z L V H > N 1 P @
[ 5 D [ L V > P @ R U L J L Q D W P L G G O H & $ 3
Figure 20: Static test longitudinal strain (in zR direction) versus chord position (in xR axis) at different edgewise moments. Strain gauges 004F000,005F000,006F000,013F000.
31
Figure 21: Side view of the trailing edge during the test. Buckling wave formation and adhesive joint failure.
1.13 0.72 0.30 0.01 0.52 0.93
Rotation arm angle 0.22°
1.34
Rotation arm angle 0.56°
1.76 2.17
Max imum stress f ailure index
2.58 3.00 Rotation arm angle 1.0°
Figure 22: FE model at different rotation arm loading angles. Color gradient represent the maximum stress failure index.
32
/ R Q J L W X G L Q D O V W U D L Q V > V W U D L Q @
0 H G J H Z L V H > N 1 P @ 0 H G J H Z L V H > N 1 P @ 0 H G J H Z L V H > N 1 P @ ) ( P R G H O D W > N 1 P @ ) ( P R G H O D W > N 1 P @ ) ( P R G H O D W > N 1 P @
[ 5 D [ L V > P @ R U L J L Q D W P L G G O H & $ 3
* H R P H W U L F D O D Q J O H V D Q G H Y R O X W L R Q > @
Figure 23: Static test longitudinal strain (in zR direction) versus chord position (in xR axis) at different edgewise moments. Comparison between FE model and strain gauges measurements from tests. Strain gauges 004F000,005F000,006F000,013F000.
/ 7 7 O R D G 6 & 7 O H Q J W K > P @ & D Q J O H / 7 7 O R D G 6 & 7 O H Q J W K > P @ & D Q J O H
' L V S O D F H P H Q W
&
Figure 24: C3.1 β and C3.2 ∆α versus rotation arm displacement. Axes normalized based on C2 criterion.
33
/ R D G 3 3 &
/ 7 7 O R D G 6 & 7 O H Q J W K > P @ & P D [ O R D G & H D U O L H U E L I X U F D W L R Q S R L Q W / % $ & > U D G @ & > @ & 0 D [ V W U H V V V W U H V V D Q D O \ V L V & 7 V D L : X V W U H V V D Q D O \ V L V
' L V S O D F H P H Q W &
Figure 25: Buckling resistance criteria applied to specimen loading curve. Axes normalized based on C2 criterion.
34
Figure 26: Side view of the trailing edge during fatigue tests. Buckling wave formation and adhesive joint failure. Four steps load spectrum, 1.1 · 106 cycles at 45 % · MC2 , 6.9 · 105 cycles at 62 % · MC2 , 5.1 · 105 cycles at 80 % · MC2 and 1.0 · 105 cycles at 110 % · MC2 until failure during fatigue.
35
( G J H Z L V H P R P H Q W 0 > N 1 P @
0 D Y J W U D I D W B V 1 P D [ > F \ F O H V @ 0 P D [ W U D I D W B V 0 P L Q W U D I D W B V
& \ F O H V W R I D L O X U H > N F \ F O H V @
Figure 27: Maximum, minimum and average edgewise moment versus cycles to failure. Fatigue tests. Four steps load spectrum, 1.1 · 106 cycles at 45 % · MC2 , 6.9 · 105 cycles at 62 % · MC2 , 5.1 · 105 cycles at 80 % · MC2 and 1.0 · 105 cycles at 110 % · MC2 until failure during fatigue.
5 R W D W L R Q D U P D Q J O H > @
D Y J W U D I D W B V 1 P D [ > F \ F O H V @
P D [ W U D I D W B V P L Q W U D I D W B V
& \ F O H V W R I D L O X U H > N F \ F O H V @
Figure 28: Maximum, minimum and average rotation arm angle versus cycles to failure. Fatigue tests. Four steps load spectrum, 1.1 · 106 cycles at 45 % · MC2 , 6.9 · 105 cycles at 62 % · MC2 , 5.1 · 105 cycles at 80 % · MC2 and 1.0 · 105 cycles at 110 % · MC2 until failure during fatigue.
36