Content uploaded by Kim Branner
Author content
All content in this area was uploaded by Kim Branner on Nov 05, 2022
Content may be subject to copyright.
General rights
Copyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright
owners and it is a condition of accessing publications that users recognise and abide by the legal requirements associated with these rights.
Users may download and print one copy of any publication from the public portal for the purpose of private study or research.
You may not further distribute the material or use it for any profit-making activity or commercial gain
You may freely distribute the URL identifying the publication in the public portal
If you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediately
and investigate your claim.
Downloaded from orbit.dtu.dk on: Mar 29, 2019
Integrated dynamic testing and analysis approach for model validation of an innovative
wind turbine blade design
Luczak, Marcin; Peeters, B.; Manzato, S.; di Lorenzo, E. ; Csurcsia, P. Z.; Reck-Nielsen, Kasper; Berring,
Peter; Haselbach, Philipp Ulrich; Branner, Kim; Ruffini, Valentina
Publication date:
2018
Document Version
Publisher's PDF, also known as Version of record
Link back to DTU Orbit
Citation (APA):
Luczak, M., Peeters, B., Manzato, S., di Lorenzo, E., Csurcsia, . P. Z., Reck-Nielsen, K., ... Ruffini, V. (2018).
Integrated dynamic testing and analysis approach for model validation of an innovative wind turbine blade
design. Paper presented at 28th edition of the Biennial ISMA conference on Noise and Vibration Engineering,
ISMA 2018, Leuven, Belgium.
brought to you by COREView metadata, citation and similar papers at core.ac.uk
provided by Online Research Database In Technology
Integrated dynamic testing and analysis approach for
model validation of an innovative wind turbine blade
design
M.M. Luczak1, B. Peeters2, S. Manzato2, E. Di Lorenzo2, P. Z. Csurcsia2,5, K. Reck-Nielsen3, P.
Berring1, P. U. Haselbach1, K. Branner1,V. Ruffini4
1 Technical University of Denmark, Department of Wind Energy
Frederiksborgvej 399, 4000, Roskilde, Denmark
e-mail: mluz@dtu.dk
2 Siemens Industry Software NV, RTD Test Division
Interleuvenlaan 68, 3001, Leuven, Belgium
3 CEKO Sensors ApS
Diplomvej 381, DK-2800 Kgs. Lyngby, Denmark
4 University of Bristol, Faculty of Engineering
Queen's Building, University Walk, Clifton BS8 1TR, UK
5 Vrije Univeriteit Brussel, Department of Engineering Technology
Pleinlaan 2, 1050 Elsene, Belgium
Abstract
DTU Wind Energy continues the experimental investigation of the wind turbine blades to assess innovative
designs of long and slender blades. This paper presents an experimental structural dynamics identification
and structural model validation of the 14.3m long research blade. Unique feature of the blades is that its
internal layup design has been highly optimized w.r.t. stretching the rotor and substantial mass reduction at
the same time. As the result, the blade is more flexible than the traditional one. The results of the modal
tests following analyses were performed: (i) Uncertainty Quantification of the experimental modal
parameters for the blades, (ii) non-linearity assessment, (iii) numerical model correlation – frequencies and
mode shapes of the experimental model comparison with those from Finite Element (FE). Finally, the
outlook for the future experimental blade research activity is outlined.
1 Introduction
Modern wind turbine blades are expected to withstand the in-service and extreme loads for a lifetime of 25
years. The IEC 61400-23 standard [1] requires full-scale structural testing of wind turbine blades. The
standard defines the static load, fatigue load and natural frequencies testing and interpretation of the results.
Requirements for the structural dynamic identification are limited to measurements of the first and second
flapwise and first edgewise frequencies. Remaining blade modal properties such as damping and mode
shapes are listed under the optional tests.
The length of recently developed offshore wind turbine blades exceeds 100 meter [2]. In-depth
understanding of the dynamic behavior of such slender and flexible structures requires detailed experimental
investigation exceeding the current scope required by the standard since unstable vibrations of wind turbine
blade may lead to its failure. Adequate identification of the blade‘s mode shapes is critical for the analysis
of stability problems. Experimental modal testing is a well-established tool [3-5] for the estimation of the
modal model of a mechanical system. At DTU Wind Energy (former Risø National Laboratory), modal
testing has been applied to full-scale wind turbine blades. The work of Larsen et al [6] and Pedersen et al
[7] revealed some discrepancies between experimentally and numerically obtained mode shapes. Large
differences were observed especially for torsional modes. Extensive and well documented activities on
modal testing of research-scale blades was conducted at Sandia National Laboratory. Different support
configurations for the investigated blades have been studied [8] followed by the uncertainty analysis of the
obtained results [9]. Uncertainty quantification has been a subject of investigations for the modal test of
helicopter rotor blades [10]. The test setup has been found to have a large effect on the performed tests, as
the mass of the sensors or damping added from cables may significantly influence the values of the observed
frequencies and damping. Furthermore, the influence of the support structure has also been taken into
account in the parameter estimation and numerical model validation [11].
Progress in the development of measurement equipment has enabled application of different sensors for
modal testing of wind turbine blades. Optical methods are commonly used for full-field blade response
measurement [12-14]. One of the main advantages of the application of camera-based full field methods is
the collection of strain and acceleration data combined in one measurement system. Strain modal analysis
[15] is an attractive alternative for vibration acceleration measurements, as the blade is often instrumented
with a dense network of strain gauges for static and fatigue testing. Alternative excitation methods are being
investigated with the application of Piezoelectric Transducer (PZT) patches as actuators [16].
Characterization of blade dynamics is one area of the modal testing results application. The modal models
are also used for the numerical model validation for the model improvement [17, 18] and damage detection
[19].
This paper presents a study on the integrated dynamic testing and analysis for model validation of a research-
scale wind turbine blade. Section 2 describes the experimental setup and measurement results. Section 3
provides information on the validation of the FE model of the blade. The final section concludes on the
research done so far and sets the future outlook.
2 Numerical modal analysis
The blade was modelled using the commercial finite element package pre/post-processing software MSC
Patran (version 2014). The model was generated by using Risø-DTU’s in-house software Blade Modeling
Tool (BMT). MSC Marc was applied as the solver in all analyses.
The blade structure was discretized using 20-noded layered continuum solid elements, which requires a
volume representation of the geometry. The entire blade geometry was modelled based on input data of 99
blade cross sections generated by DTU Wind Energy’s in-house Beam Cross section Analysis Software
(BECAS) optimizing scheme [20]. The blade airfoil defines the outer cross-section geometry. An offset
according to the layup and layer thickness of the composite material determines the element thickness.
Finally, the individual cross-sections were connected by spline curves and interpolation surfaces to obtain a
volume representation of the blade. This process was handled automatically by BMT, which uses 60
regions/solids to assign the different cross-sectional properties computed by the BECAS optimizing scheme.
Figure 1 shows segments of the blade model.
Figure 1: Left: Cross-section segment at radial position of around 5m, right: mid segment.
Depending on the region, the composite layup consists of 6 to 32 plies through the thickness. The properties
of the composite, with its material, layup and ply orientations were assigned to 20-noded layered continuum
elements defining the element stiffness matrix. The model was discretized with approximately 130,000
layered 20-noded continuum elements.
The blade manufacturer has not stated the exact properties for the fibers and matrix material, so properties
from comparable materials described in the literature have been used to some extent. For this purpose, the
elastic properties were calculated using Autodesk Simulation Composite tool, which uses a calculation
scheme based on the Classical Laminate Theory. The estimated characteristic mechanical material
properties and the estimated design material strengths applied in these studies are presented in Table 1.
Mechanical
properties
UNIAX
BIAX
TRIAX
Chop
Core
Glue
Gelcoat
E1 [MPa]
37800
9550
18700
13600
48.5
3009
2000
E2 [MPa]
11100
9550
10900
13600
G12 [MPa]
3270
10100
7720
5130
ν12
0.24
0.62
0.55
0.32
0.4
0.3
0.3
ϱ [kg/m3]
1850
1780
1780
1684
80
1540
1500
Thickness [mm]
0.95
0.5
0.75
0.3
5-10
-
0.6
Design strength
(PSF 2.205)
XT [MPa]
360.0
69.3
186.0
56.2
x
x
x
XC [MPa]
257.0
64.9
152.0
56.2
x
x
x
YT [MPa]
24.8
69.3
30.5
56.2
x
x
x
YC [MPa]
63.5
64.9
51.5
56.2
x
x
x
S [MPa]
16.6
55.9
42.3
23.9
x
x
x
Table 1: Material properties
The natural frequencies and mode shapes of the blade are shown in Table 2:
1st Flapwise FE mode 4.12 [Hz]
1st edgewise FE mode 9.87 [Hz ]
2nd Flapwise FE mode 11.36 [Hz]
3rd Flapwise FE mode 20.81 [Hz]
Table 2 Mode shapes examples from FEM analysis
3 Experimental modal analysis
DTU Wind Energy department has recently commissioned a Large Scale Test Facility built for the
experimental research of full-scale wind turbine blades and their subcomponents. The facility has three test
rigs capable of accommodating blades of different lengths, ranging from 12 up to 45 meters, for static,
fatigue and dynamic testing as shown in Figure 2.
Figure 2 Large Scale Test Facility layout with three test rigs for a full scale blade tests.
The blade tested for this study is a research-scale blade, with a length of 14.3 m and a mass of approximately
530 kg. It has been entirely designed at the DTU Wind Energy department, and manufactured by Olsen
Wings. Modal tests have been performed before destructive static and fatigue testing. An estimated modal
model of the intact blade will serve as a reference data for the post-fatigue damage detection. After the
fatigue testing, the modal tests will be repeated and the results will be compared to the initial modal analysis
data. In addition, blade modal models will be analyzed in order to investigate the detection possibilities of
damage in the blade.
3.1 Test setup
In the presented measurement campaign, the test blade was supported with two elastic cords in the edgewise
direction to provide free-free boundary conditions as presented in Figure 3. This configuration is particularly
convenient for the updating of the numerical model of the blade, as it does not require modelling of the
boundary conditions.
The investigated blade has been simultaneously excited in the flapwise and edgewise directions with two
electrodynamic shakers. As measurement grid, it was decided to split the blade into 15 equidistant “stations”,
one every 1m. Additionally, each station was measured at 8 locations (Figure 4) , including: the trailing and
leading edge, the point of maximum stiffness on the pressure and suction sides of the airfoil, plus 4
intermediate points between these. The high-fidelity finite element model presented in the previous section
was used to simulate the test results and optimize sensor placement, in terms both of coordinates and
orientation. Using the proposed locations, it was verified that the current measurement setup could guarantee
a good description of the first 30 modes.
To ensure a consistent and repeatable excitation of the blade, and avoid damaging it, two relatively stiff
points on the flap and edgewise directions were selected. As shown in Figure 3, the shaker along the
edgewise direction was placed approximately in the middle of the blade, close to the expected nodal line of
the first edgewise mode, to keep the maximum resonant response of the blade to safe levels. For the
horizontal shaker, a point close to the center of gravity was selected. However, ensuring a good connection
of the shaker to the blade in this direction has been more difficult, as in this direction the blade responds
like a pendulum, causing the stinger to detach often from the blade. In addition, as the stingers were
disconnected from the blade after each run before moving the accelerometers, a consistency check was
performed to ensure the system remained invariant over the different test runs.
Figure 3 Instrumented blade on the test stand in the free-free support configuration (left) and geometry of
the measurement and excitation points (right).
Figure 4: Sensor instrumentation along the airfoil.
The excitation forces and the corresponding driving-point Frequency Response Functions (FRFs) were
measured by two impedance heads, which encapsulate force and acceleration sensors in one housing. The
dynamic response of the blade was measured with 15 tri-axial piezoelectric accelerometers. Considering the
total number of degrees-of-freedom on the measured grid (15 stations x 8 locations x 3 directions = 360
DOFs) it was decided to perform the measurement in a roving accelerometer configurations, so that the 15
sensors were moved over the 8 locations for each measurement station. As the driving-point FRFs at the
excitation locations were measured for all runs, they are used as reference to recombine the different datasets
and scale the mode shapes.
3.2 Uncertainty analysis
In this section, the main sources of uncertainties that were analyzed correspond to the measurement setup.
As discussed in the previous section, the whole measurement grid was covered in 8 runs, where the
excitation was applied always at the same location while the accelerometers were moved around the profile
of the airfoil. Moreover, as the two stingers connecting the shakers to the blade were disconnected before
moving the accelerometers, inconsistencies in the actual direction of application of the force could arise.
Variations in modal parameters between the different runs can therefore be attributed to:
• Misalignment of the stinger axis,
• Loose connection between shakers and structure,
• Time of measurements (campaign was performed over a period of 3 days, with significantly
different meteorological and temperature conditions),
• Mass loading effects due to accelerometers and cables (this is considered to be negligible considered
the total mass of the blade),
• Change in the stiffness of the supporting elastic cords over time.
In Figure 5 the variability of the excitation over the 8 runs is shown. For the excitation in the edgewise
direction, the Driving Point FRFs are highly repeatable and consistent. On the other hand, the corresponding
flapwise Driving Points show much higher variability between the different runs, because of the difficulties
in ensuring a good and reliable connection between the shaker and the blade.
Figure 5: Driving Point FRF for the edgewise and flapwise excitation over the 8 different runs
Starting from these observations, modal analysis was performed on the 8 datasets and the individual
estimates are compared. The following procedure was applied:
1. The Polymax modal parameter estimation was applied to the FRFs collected during each run using
the same settings: (i) a maximum model order of 100 was set to estimate modes in the 2.5 to 150
Hz frequency band; (ii) upper and lower residuals were included; (iii) poles were selected manually;
and (iv) modes were assumed to be complex.
2. Only modes consistently appearing in all modesets were included in the statistical calculation.
3. For each mode, the average and the standard deviation of the frequency and damping were
computed.
4. Individual estimates were normalized to the corresponding average value to simplify visualization.
5. To correlate the results with the actual test, the average coherence on the edgewise driving point is
also added to the plot.
The results of the analysis are displayed in Figure 6 both for the frequencies and damping. In general, the
scatter of the estimates is very limited between the different runs. For the frequencies, we observed some
higher variability in correspondence of the first mode (the 1st flapwise bending), the torsional mode around
43 Hz, and finally on a cluster of modes between 100 and 110 Hz. For the damping, a larger variability can
also be observed at higher frequency, around 130 Hz. A higher uncertainty on the damping can generally be
expected as damping estimates are more sensitive to inconsistencies in the data and on the selection of one
specific pole in the stabilization diagram. In general, however, despite the difficulties in connecting the
flapwise shaker to the structure, the actual duration of the test over few days and the roving of the
accelerometers, the results are quite consistent and as a consequence, the FRFs from the different runs can
be processed all together into a single database.
Figure 6: Modal Frequency and Damping uncertainty analysis results.
150.002.40 Hz
0.10
1.00e-6
Log
g/N
Driving Point Edgewise
150.002.40 Hz
0.10
0.01e-3
Log
g/N
Driving Point Flapwise
3.3 Nonlinearity assessment
In modern system identification special excitation signals are available to assess the underlying systems in
a user-friendly, time efficient way [21]. In order to avoid any spectrum leakage, to reach full nonparametric
characterization of the noise, and to be able to detect nonlinearities, a periodic signal is needed. Many users
prefer noise excitations, because they are simple to implement, but in this case nonlinearities are not
identifiable, and there is a possible leakage error.
3.3.1 Multisine excitation technique
The best signal that satisfies the properties listed above is the multi-sine signal (see Figure 7), which looks
and behaves like Gaussian white noise, but is deterministic. It is important to highlight that multi-sine
excitation is not equivalent to stepped sine excitation [22,23].
Figure 7 Different excitation signals in time and in frequency domain
3.3.2 Best Linear Approximation approach
The Best Linear Approximation (BLA) method has been widely used in the last decades to efficiently
estimate FRFs [21, 24, 25]. The BLA of a nonlinear system minimizes the mean square error between the
true output of a nonlinear system and the output of the linear model, see Figure 8. With this technique,
instead of using the classical H1 estimate (cross-power spectral density estimate [21]) and its coherence
function, a BLA model is estimated, and the coherence function is split into a) noise level and b) nonlinear
contribution estimates.
Figure 8 The baseline model and the cost function (S) of the best linear approximation framework
The proposed BLA technique makes use of the knowledge that the excitation signal has both stochastic and
deterministic properties. In this work, random phase multi-sines are used and generated in the frequency
domain such that the magnitude is set by the user, and the phases of the cosines are chosen randomly from
a uniform distribution [26].
A further difference w.r.t. H1 estimation is that instead of directly using the averaged input and output data,
a partial BLA estimate is calculated for each period of the excitation, and several different realizations of
the excitation signal are repeated multiple times. A BLA FRF estimate, for a given signal, is then calculated
via the average of different BLAs. In this case it is possible to easily estimate the noise levels and standard
deviation of the estimates at each frequency line. The difference between the total variance of BLA and the
noise variance is an estimate of the variance of the stochastic nonlinear contributions. In [22] it has been
shown that the above statements are only true when some additional assumptions are satisfied.
3.3.3 Measurement analysis
Multiple multisine experiments have been done at relatively low and high power levels as it is recommended
for the BLA measurement procedure [24].
Figure 9 shows the magnitude spectrum of the flapwise measured excitation signal at low (left figure, grey
line) and high (right figure, black line) levels (with an approximate overall signal level difference of 10 dB,
i.e. factor of 3). The reference signal – a computer-generated ideal waveform – has a completely flat
magnitude. The differences between the ideal flat and measured magnitude characteristic, and the level of
nonlinearities w.r.t the noise level illustrate the dynamic behaviour of the shaker and its interaction with the
underlying structure. Black dots show the corresponding noise level estimates, which gives a rough estimate
of the SNR (around 46 dB at low level excitation at mid frequencies, and around 50 dB at high level
excitation). Orange and blue dots refer to the odd and even nonlinearities on the non-excited detection lines
(the frequency bins where the type of nonlinearities can be captured). The even nonlinearities – compared
to the noise level – become dominant at the high-level excitation profile, especially above 60 Hz. A further
interesting thing to mention is that the influence of electrical noise (harmonics of the mains frequency 50
Hz) is less significant at the higher level excitation. Apart from this, the nonlinearities have limited effect,
as they are of the same order as noise.
Figure 9 The measured excitation spectrum on the Y (flap) direction is shown for low level (gray line, on
the left side) and high levels (black line, on the right side) together with the estimated noise level (black
dots), even (blue dots) and odd (orange dots) nonlinearities.
Figure 10 shows the frequency response matrix estimate of the underlying system in the X and Y directions
at the output channel S13, close to the wing tip. Black and grey lines show the high and low level FRFs with
the corresponding noise (black and grey dots) and nonlinearity (red and orange dots) estimates. The low-
level FRFs below -60 dB are quite noisy, and the nonlinearities have a limited effect (their level is
approximately 5 dB higher than the noise). At the high excitation level, the SNR is significantly improved,
but the level of the nonlinearities increases as well. From these results, it is also possible to see that the
torsional mode at 42 Hz exhibits a particularly large difference in level between noise and nonlinearity, at
both low and high excitation levels. In general, the nice overlay between the BLA and the measured FRFs
is an indication that the system can be reasonably assumed to behave linearly. These results warrant further
investigation of the behavior of this mode.
Figure 10 The best linear aproximation of the frequency response matrix is shown for low level (gray line)
and high levels (black line) together with the estimated noise (grey and black dots) and nonlinearity
(orange and red dots) levels at S13 channel.
3.4 Measurement results
Based on the preliminary observation and analysis performed in Sections 3.2 and 3.3, here a summary of
the identified modes is given. Because of the repeatability between the different test runs, it was decided to
fit a model including all FRFs at the same time, so that the fitted FRF and mode shapes can be more
consistent to the actual blade behavior.
Figure 11: Validation of the identified modal model. Left: Comparison of measured (red) and synthesized
(green) FRFs for a specific DOF (S11:7:-Y). Right: AutoMAC for the selected modeset.
150.002.00 Hz
0.10
1.00e-6
Log
g/N
F FRF S11:7:-Y/Drive:Flap:+Y 0 #
F Synthesized FRF S11:7:-Y/Drive:Flap:+Y 1 #
150.002.00 Hz
0.10
1.00e-6
Log
g/N
F FRF S11:7:-Y/Drive:Edge:-X 0 #
F Synthesized FRF S11:7:-Y/Drive:Edge:-X 1 #
In total, 33 modes have been reliably identified in the frequency range between 2 and 150 Hz. Figure 11
shows the validation of the modeset identified from the measured experimental data. On the left, the
synthesized FRFs between one measured Degree of Freedom and the two excitations are displayed. In both
cases, the correlation between the measured and reconstructed FRFs is very high, confirming that the
identified model is able to correctly represent the structure in the frequency range of interest. On the right,
the AutoMAC (autocorrelation between all mode shapes) is displayed: as the matrix has a dominant
diagonal, the identified set contains independent vector and can thus be reliably used to validate (and later
update) the numerical model from Section 2.
3.5 Optical accelerometer measurements
Figure 12 The optical accelerometer (CEKO OA1) was mounted flap-wise on the test blade. The
accelerometer measures 27 mm in diameter. The first 7 resonance frequencies measured using the optical
accelerometer and a reference piezoelectric accelerometer are shown in the table.
Two uniaxial all-optical MEMS accelerometers (CEKO Sensors, model OA1) were used to measure the
transfer function of the test blade. The optical accelerometers are completely metal-free, thus immune to
electromagnetic interference from e.g. lightning strikes, and therefore safe to operate on wind turbines under
all conditions. They are based on refractive index modulation technology, which is a high sensitivity
frequency modulated sensing principle. The accelerometers were mounted 12.5 m and 9 m from the root,
respectively. Data was recorded using a CEKO Sensors S-DAS monitoring system operating at a 3000 Hz
sampling rate. A piezoelectric accelerometer was located next to the optical accelerometer as reference. The
measured transfer functions for the optical and piezoelectric accelerometer during pseudorandom excitation
are seen in Figure 13. Figure 13 Comparison of measured transfer functions. Numerically calculated
resonance frequencies are shown with dashed lines.Both the reference and the optical accelerometer detects
practically all resonances below 100 Hz, though some have very low amplitudes due to the measurement
locations. Differences in amplitudes are primarily due to the two accelerometers measuring along slightly
different angles. At the 9 m position the amplitude of the fundamental mode is reduced compared to the
resonances around 30 Hz. This is expected as this location is close to a node of the fundamental mode. The
absolute error between resonance frequencies measured using the optical accelerometer and those of the
piezoelectric reference, as well as the numerically calculated frequencies, is of the order of 1% or less.
Mode
Frequency,
Optical [Hz]
Frequency,
Reference [Hz]
Deviation
[%]
1 4.100 4.048 1.280
2 9.300 9.430 1.380
3 11.000 10.966 0.310
4 11.800 11.807 0.060
5 21.699 21.655 0.200
6 30.149 29.982 0.560
7 34.299 34.290 0.030
Figure 13 Comparison of measured transfer functions. Numerically calculated resonance frequencies are
shown with dashed lines.
4 Test-simulation correlation
The numerical model described in Section 2 was validated with the experimental model identified in Section
3.4. As the experimental grid was derived from the numerical model, no geometric correlation is required
as the measurement points already coincide with a node in the model. Consequently, the Modal Assurance
Criterion between the numerical and experimental model can be calculated, and the results are displayed in
Figure 14. Generally good correlation can be seen for the first 13 modes, while at higher frequencies the
model and the experiments start to diverge. This is to be expected as the modal response at higher frequency
is more localized and subject to uncertainties.
Figure 14 Modal Assurance Criterion between
the numerical and experimental mode shapes
Test
FEM
MAC
Freq. %
Error
4.05
4.13
0.98
2.14
10.96
9.87
0.98
-9.91
11.80
11.36
0.97
-3.71
21.65
20.81
0.98
-3.86
29.98
26.89
0.96
-10.31
34.28
32.86
0.97
-4.16
43.30
37.25
0.92
-13.98
48.93
45.76
0.91
-6.49
58.00
49.85
0.92
-14.06
62.67
53.14
0.88
-15.20
66.84
60.82
0.86
-9.00
80.00
71.75
0.87
-10.31
86.41
83.19
0.69
-3.73
100.10
96.78
0.66
-3.32
111.40
109.80
0.53
-1.44
133.59
110.73
0.59
-17.11
Table 3: Results of correlation analysis between test
and simulation results
Based on the correlation between the modes, Table 3 compares the numerical and experimental natural
frequency of matching modes. Here, only modes that show a correlation above 50% are compared and 16
pairs are reported. In general, very good agreement is found between all modes with a dominant flapwise
direction, with errors between the natural frequencies ranging between 1 and 5%. Bigger discrepancies (with
errors around 10%), on the other hand, are observed for the modes in the edgewise and torsional direction.
Before drawing further conclusions, it is necessary to include the bungee cords used to suspend the blade in
the model, as they add stiffness in the edgewise direction. Indeed, the natural frequencies of the numerical
model in this direction are consistently softer than the experimental results. In general, the good correlation
between the global modes at low frequencies gives confidence in the general validity of the model, and an
update of the material properties is not deemed necessary at this stage. In Figure 15, a comparison of some
of the mode pairs is given.
1st Flapwise Bending (4.05-4.12 Hz)
1st Edgewise Bending (10.96 – 9.87 Hz)
2 Edgewise Bending (58 – 49.85 Hz)
2nd Torsion (80 – 71.75 Hz)
Figure 15 Examples of experimental and numerical identified mode shapes.
5 Conclusions and future outlook
This paper presented dynamic testing and analysis for model validation of an innovative wind turbine blade
design. Motivation for the work was to investigate the dynamic properties of the blade to a larger extent
than required by a certification standard and apply the results for the blade model validation. An
experimental modal analysis was implemented and provided an estimation of the full modal model
parameters in the wide frequency band. Uncertainty quantification has confirmed high consistency in the
experimental data collection. Based on this analysis result, the estimated experimental modal model was a
credible reference for the validation of the numerical model. Performed tests did not show the presence of
nonlinearity. Both uncertainty and nonlinearity analysis revealed problem with the precision of the
identification of the torsional mode. Within the test campaign use of the innovative optical accelerometer
has been successfully demonstrated. Future work will encompass use of the operational modal analysis for
the blade bolted to the concrete block of the test rig. A certification standard requires static and fatigue
testing with the use of the strain gauges. Therefore it is planned to use a test setup and the blade
instrumentation for the vibration test.
Acknowledgements
The experimental work described herein has been conducted using mechanical testing and measurement
equipment from Villum Center for Advanced Structural and Material Testing (CASMaT). The support from
Villum Fonden (Award ref. 00007293) is gratefully acknowledged.
This work was supported by the Danish Centre for Composite Structures and Materials for Wind Turbines
(DCCSM), Grant no. 09-067212 from the Danish Strategic Research Council.
The work is supported by the Danish Energy Agency through the Energy Technology Development and
Demonstration Program (EUDP), Grant No. 64016-0023. The supported project is named ‘‘BLATIGUE:
Fast and efficient fatigue test of large wind turbine blades”, and the financial support is greatly appreciated.
This work was funded by the VLAIO Innovation Mandate project number HBC.2016.0235.
References
[1] International Electrotechnical Commission. International standard iec 61400-23 Wind turbine
generator systems – part 23: full-scale structural testing of rotor blades. Wind Turbines-Part 1: Design
Requirements. (2014).
[2] GE Renewable Energy, GE announces Haliade-X, the world’s most powerful offshore wind turbine,
(2018).
[3] P Avitabile, Modal testing : a practitioner's guide, John Wiley & Sons Ltd 2018.
[4] DJ Ewins, Modal testing: theory and practice, Research studies press Letchworth 1984.
[5] W Heylen, S Lammens, P Sas, Modal Analysis Theory and Testing, 2nd ed., Katholieke Universiteit
Leuven, Departement Werktuigkunde, Leuven, 1998.
[6] GC Larsen, MH Hansen, A Baumgart, I Carlén, Modal analysis of wind turbine blades, Risø–R–
1181(EN) (2002).
[7] HB Pedersen, OJD Kristensen. Applied modal analysis of wind turbine blades. (2003).
[8] DT Griffith, G Smith, M Casias, S Reese, TW Simmermacher. Modal Testing of the TX-100 Wind
Turbine Blade, Sandia National Laboratories Technical Report, Report# SAND2005-6454. (2006).
[9] D Griffith, TG Carne. Experimental Uncertainty Quantification of Modal Test Data, Sandia National
Laboratories (SNL-NM), Albuquerque, NM (United States). (No. SAND2006-6507C) (2006).
[10] MM Luczak, Uncertainty Quantification of the Main Rotor Blades Measurements, Recent Progress in
Flow Control for Practical Flows, Springer, 2017, pp. 429-482.
[11] DT Griffith, TG Carne, Experimental modal analysis of 9-meter research-sized wind turbine blades,
Structural Dynamics and Renewable Energy, Volume 1, Springer, 2011, pp. 1-14.
[12] US Paulsen, O Erne, M Klein, Modal analysis on a 500 kW wind turbine with stereo camera
technique, International Operational Modal Analysis Conference. IOMAC. (2009).
[13] P Poozesh, J Baqersad, C Niezrecki, P Avitabile, E Harvey, R Yarala. Large-area photogrammetry
based testing of wind turbine blades, Mechanical Systems and Signal Processing. 86 (2017) 98-115.
[14] P Poozesh, A Sarrafi, Z Mao, C Niezrecki. Modal parameter estimation from optically-measured data
using a hybrid output-only system identification method, Measurement. 110 (2017) 134-145.
[15] FL dos Santos, B Peeters, Strain Modal Analysis, Recent Progress in Flow Control for Practical
Flows, Springer, 2017, pp. 405-428.
[16] V Ruffini, T Nauman, C Schwingshackl, Impulse excitation of piezoelectric patch actuators for
modal analysis, Topics in Modal Analysis & Testing, Volume 10, Springer, 2017, pp. 97-106.
[17] JR White, DE Adams, MA Rumsey, Modal analysis of CX-100 rotor blade and micon 65/13 wind
turbine, Structural Dynamics and Renewable Energy, Volume 1, Springer, 2011, pp. 15-27.
[18] M Luczak, S Manzato, B Peeters, K Branner, P Berring, M Kahsin. Updating finite element model of
a wind turbine blade section using experimental modal analysis results, Shock Vibrat. 2014 (2014).
[19] GC Larsen, P Berring, D Tcherniak, PH Nielsen, K Branner, Effect of a damage to modal parameters
of a wind turbine blade, (2014).
[20] J Blasques, R Bitsche, V Fedorov, MA Eder, Applications of the beam cross section analysis software
(becas), (2013) 46-49.
[21] J Schoukens, R Pintelon, Y Rolain, Mastering system identification in 100 exercises, John Wiley &
Sons 2012.
[22] R Pintelon, J Schoukens, System identification: a frequency domain approach, John Wiley & Sons
2012.
[23] R Priemer, Introductory signal processing, World Scientific Publishing Company 1990.
[24] TP Dobrowiecki, J Schoukens. Linear approximation of weakly nonlinear MIMO systems, IEEE
Transactions on Instrumentation and Measurement. 56 (2007) 887-894.
[25] P Guillaume, R Pintelon, J Schoukens. Accurate estimation of multivariable frequency response
functions, IFAC Proceedings Volumes. 29 (1996) 4351-4356.
[26] P Csurcsia. Static nonlinearity handling using best linear approximation: An introduction, Pollack
Periodica. 8 (2013) 153-165.
