http://shm.sagepub.com/Structural Health Monitoring

http://shm.sagepub.com/content/13/6/660The online version of this article can be found at:

 DOI: 10.1177/1475921714553732

2014 13: 660Structural Health MonitoringAnanth Grama and Janette Jaques Meyer

Sungmin Kim, Douglas E Adams, Hoon Sohn, Gustavo Rodriguez-Rivera, Noah Myrent, Ray Bond, Jan Vitek, Scott Carr,Crack detection technique for operating wind turbine blades using Vibro-Acoustic Modulation

  

Published by:

http://www.sagepublications.com

can be found at:Structural Health MonitoringAdditional services and information for    

  http://shm.sagepub.com/cgi/alertsEmail Alerts:

 

http://shm.sagepub.com/subscriptionsSubscriptions:  

http://www.sagepub.com/journalsReprints.navReprints:  

http://www.sagepub.com/journalsPermissions.navPermissions:  

http://shm.sagepub.com/content/13/6/660.refs.htmlCitations:  

What is This? 

- Nov 14, 2014Version of Record >>

at KOREA ADV INST OF SCI & TECH on December 7, 2014shm.sagepub.comDownloaded from at KOREA ADV INST OF SCI & TECH on December 7, 2014shm.sagepub.comDownloaded from

Original Article

Structural Health Monitoring

2014, Vol. 13(6) 660–670

� The Author(s) 2014

Reprints and permissions:

sagepub.co.uk/journalsPermissions.nav

DOI: 10.1177/1475921714553732

shm.sagepub.com

Crack detection technique foroperating wind turbine blades usingVibro-Acoustic Modulation

Sungmin Kim1, Douglas E Adams2, Hoon Sohn3,Gustavo Rodriguez-Rivera1, Noah Myrent2, Ray Bond1, Jan Vitek1,Scott Carr1, Ananth Grama1 and Janette Jaques Meyer2

AbstractThis article presents a new technique for identifying cracks in wind turbine blades undergoing operational loads usingthe Vibro-Acoustic Modulation technique. Vibro-Acoustic Modulation utilizes a low-frequency pumping excitation signalin conjunction with a high-frequency probing excitation signal to create the modulation that is used to identify cracks.Wind turbines provide the ideal conditions in which Vibro-Acoustic Modulation can be utilized because wind turbinesexperience large low-frequency structural vibrations during operation which can serve as the low-frequency pumpingexcitation signal. In this article, the theory for the vibro-acoustic technique is described, and the proposed crack detec-tion technique is demonstrated with Vibro-Acoustic Modulation experiments performed on a small Whisper 100 windturbine in operation. The experimental results are also compared with two other conventional vibro-acoustic techniquesin order to validate the new technique. Finally, a computational study is demonstrated for choosing a proper probing sig-nal with a finite element model of the cracked blade to maximize the sensitivity of the technique for detecting cracks.

KeywordsVibro-Acoustic Modulation, crack detection, micro fiber composite, nonlinear method, frequency response function,wind turbine blades

Introduction

Wind power is considered by many to be the most feasi-ble renewable power source due to its competitiveenergy cost compared to other options, including solarpower. For this reason, the wind power industry hasgrown rapidly in recent decades.1,2 As wind turbineshave been installed and operated, economic data haveshown that maintenance costs consume a significantportion of the total energy costs, which are borne bywind farm owners/operators.3,4 In particular, wind tur-bine blades require a robust structural health monitor-ing strategy because the blades are key elements of windturbines, cost up to 20% of the total turbine cost,2 andare exposed to many unexpected events such as lighting,storms, and extreme winds. Moreover, a failure of thewind turbine blades can lead to failures of other subsys-tems in the turbines such as the tower, drive train, aswell as neighboring infrastructures around the turbine.1

Therefore, detecting damage before catastrophic failurecan help to minimize unscheduled maintenance of wind

turbines, which is typically more costly than scheduledmaintenance.

The maintenance of wind farms involves severalchallenges due to trends in wind turbine technology.First, the size of utility-scale wind turbines hasincreased significantly. The wind turbine industry con-tinues to develop larger wind turbines to increase theefficiency of power harvesting. This trend has contin-ued for several decades, and there are plans to installan 8-MW wind turbine which has 164-m rotor diameterin 2014.5 Considering that the span of the wing in theAirbus 380, among the largest aircraft, is 80 m, wind

1Purdue University, West Lafayette, IN, USA2Vanderbilt University, Nashville, TN, USA3Korea Advanced Institute of Science and Technology, Daejeon, South

Korea

Corresponding author:

Douglas E Adams, Vanderbilt University, Nashville, TN 37240, USA.

Email: douglas.adams@vanderbilt.edu

at KOREA ADV INST OF SCI & TECH on December 7, 2014shm.sagepub.comDownloaded from

turbines are the largest aerodynamic structures in theworld. This large size causes difficulties for maintainingthe wind turbines since repair technicians must climbthe towers atop where blades and nacelles are installed.In addition, the speed of a large wind turbine blade atthe tip can be up to 370 km/h.6 It can be dangerous forwind turbine repair technicians to safely approach thewind turbine, and several accidents have been reportedin recent years. Second, many wind turbines are beingbuilt in geographic locations where the wind resource isabundant, and these locations are often difficult formaintenance crews to reach. Many wind turbines havebeen built on mountains and offshore, and these loca-tions dictate higher costs as it becomes more difficultfor construction and maintenance crews to access theturbine. To address these challenges, the developmentof an efficient maintenance strategy is needed.

This article focuses on the development of an auton-omous crack detection technique. Cracks in wind tur-bine blades are a structural concern if they initiate infatigue critical areas due to the cyclic gravitationalbending, shear, and axial loading together with the cen-trifugal loading, and cracks also offer a pathwayfor moisture intrusion, which can lead to other issues.The proposed technique is based on Vibro-AcousticModulation, which measures the response of the bladeto two excitation signals with different frequencies. Ifthe blade is cracked, the response will contain nonlinearcharacteristics that can be quantified to identify thepresence of damage. There are several benefits of usingVibro-Acoustic Modulation for crack detection onwind turbine blades. First, Vibro-Acoustic Modulationis a nonlinear crack detection technique; therefore, thesensitivity of this technique to cracks is much higherthan linear vibration-based techniques.7 Next, paststudies have reported that Vibro-Acoustic Modulationcan effectively detect cracks in structures that are fabri-cated using various materials including compositematerials, which are used to manufacture wind turbineblades.8,9 Finally, this technique is less affected by vary-ing environmental and loading conditions,10 which ishelpful for structural health monitoring of wind tur-bines because these turbines operate continuouslyunder a wide range of conditions.

The new technique proposed in this article utilizesthe structural vibration of the turbine blades duringoperation as the low-frequency pumping signal forVibro-Acoustic Modulation. Using the vibration thatis present during normal operation, the size and weightof the health monitoring hardware are reduced.Furthermore, the use of structural vibrations of thewind turbine blades as the pumping signal enablescrack detection when wind turbines are in operation,whereas many current wind turbine structural health

monitoring techniques must be performed when theturbine is offline.

In this article, a theoretical review of Vibro-AcousticModulation is described, and the proposed crack detec-tion method is experimentally demonstrated. Two con-ventional vibro-acoustic crack detection tests were alsoconducted to validate the proposed method. Finally, acomputational study with a finite element model ofwind turbine blades is conducted, and the results of thisstudy can be used to determine the most sensitive prob-ing excitation frequency for crack detection.

Theoretical review

Vibro-Acoustic Modulation is one type of crack detec-tion technique. It has been demonstrated in the litera-ture that certain kinds of defects in materials such ascracks increase the nonlinear behavior of the material.11

When the blade material is damaged, the integrity andstiffness of that material change thereby causing a non-linear vibrational response. Therefore, by measuringthe increase in nonlinearity, damage in the specimencan be ascertained.

The Vibro-Acoustic Modulation test measures thevibrational response of the specimen when it is excitedby two sinusoidal signals at different frequencies. Thesesignals are called the pumping signal and the probingsignal. The pumping signal is produced at a low fre-quency called the pumping frequency, fpu, and the prob-ing signal is produced at a high frequency defined asthe probing frequency, fpr.

Figure 1 shows the response when the two excitationsignals are theoretically applied to a system. If the sys-tem is linear and damped, the response in the steadystate is the linear superposition of the responses of eachsignal, and only the linear components of Figure 1 willappear in the frequency spectrum of the response.When cracks cause nonlinear behavior within the sys-tem, the response contains both the probing frequencyand the pumping frequency in addition to other fre-quency components such as harmonics of each signaland sidebands around the probing signal. When thesystem is nonlinear, the stress–strain relationship canbe expressed as follows

s = E(e + b1e2 + b2e

3 + � � � ) ’ E(e + b1e2) ð1Þ

where s is the stress, e is the strain, E is Young’s modu-lus, and b1 and b2 are nonlinear coefficients. When thestrain contains two signals at two different frequencies,the strain can be expressed as follows

e = epu sin vput + epr sin vprt ð2Þ

where vpu = 2pfpu and vpr = 2pfpr.

Kim et al. 661

at KOREA ADV INST OF SCI & TECH on December 7, 2014shm.sagepub.comDownloaded from

The stress is then expressed as

s = E epu sin vput + epr sin vprt� �

+ Eb1

e2pu + e2

pr

2�

e2pu

2cos 2vput �

e2pr

2cos 2vprt

!

� Eb1epuepr cos vpr + vpu

� �t � cos vpr � vpu

� �t

� �ð3Þ

The first term contains the linear response compo-nents, and these are the only components that linear

systems exhibit. The components in the second term

contain frequencies of 2vpu and 2vpr which are forced

harmonics. The third term contains frequencies of

vpu6vpr which are called modulation sidebands. The

vpu6vpr components can be further expanded into

vpu6nvpr (where n = 1, 2, 3, ...) components when

higher-order nonlinear terms are considered. Vibro-

Acoustic Modulation elicits these modulation sideband

components at fpr6nfpu. Since nonlinearities due to

defects in the system are not limited to quadratic types

of nonlinearities, the second- and third-order sidebands

can also be observed.The mechanism of how a crack increases the non-

linear behavior is not clearly understood. However,

there are several explanations which suggest why a

crack introduces nonlinear behavior. An intuitive

explanation for understanding the relationship between

the crack and nonlinearity in the system is the opening

and closing action of the crack or equivalently the

change in contact stress along the interface of the crack

as a function of load. When the specimen is excited by

the pumping signal, the crack in the specimen opens

and closes according to the period of excitation. The

stiffness changes due to the change in contact stress

when the crack opens, which introduces nonlinearity

into the specimen response because the effective contact

area changes (Figure 2(a)). Figure 2(b) shows a non-

linear function in the stress–strain relationship based

on the opening and closing effect of the crack. This

function is linear when the strain is either positive or

negative. However, this function cannot be expressed

as a linear function because its slope is discontinuous at

the zero strain point.

Experimental setup

A Whisper 100 wind turbine manufactured bySouthwest Wind Power was used for the experiments inthis article. The Whisper 100 generates 900 W at12.5 m/s of approaching wind speed,12 and this turbinemodel was installed on a tower with a height of 1.626 m.

To produce operating conditions of the wind tur-bine, a wind tunnel was built as shown in Figure 3(a).The wind tunnel consists of six industrial fans at theoutlet, a polycarbonate enclosure, and polycarbonatehoneycomb at the inlet. The six industrial fans are con-trolled by an industrial motor controller, and the airflow created by the six fans produces wind in the windtunnel. The dimensions of the wind tunnel are4.089 3 3.175 3 12.954 m (width 3 height 3 length).These dimensions were selected to reduce the effects ofwalls of the tunnel on the aerodynamic phenomena.The wind turbine was positioned at 2.4 m from the endof the wind tunnel downstream where the fans arelocated.

The rotor of the wind turbine carries three carbon-reinforced fiberglass blades, and its rotor diameter is2.1 m. In the experiments presented in this article, onedamaged blade and two healthy blades were installed.

Two micro fiber composite (MFC) transducers13

were installed on each blade—one for the probingexcitation and the other for sensing as shown inFigure 3(b). An MFC consists of rectangular piezoceramic rods sandwiched between layers of adhesive,electrodes, and polyimide film. The electrodes areattached to the film in an interdigitated pattern whichapplies voltage directly to and from the ceramic rods.It is a low-profile transducer that can actuate as well assense vibration as the transducer expands, bends, ortwists. Using these low-profile and lightweight MFCs,the effect of the experimental setup on the aerodynamicperformance of the blade was minimized.

The blades for the Whisper 100 are made ofcarbon-reinforced fiberglass. However, none of thematerial properties of the blades were provided bythe manufacturer. The material properties such asYoung’s modulus (4.847 GPa) were measured fromtensile tests performed on the specimens taken from

Linear ResponseNonlinear Response

Frequency

Amplitude

2 puf pr puf f+prfpuf pr puf f−3 puf

Figure 1. Response to two excitation signals.

662 Structural Health Monitoring 13(6)

at KOREA ADV INST OF SCI & TECH on December 7, 2014shm.sagepub.comDownloaded from

the blades. The measured material density for theblades was 1064 kg/m3.

To maximize the nonlinear behavior of the blade,the location of the crack was carefully chosen such thatthe dynamic stress on the blade in operation is the max-imum at the crack. In previous studies,14–16 it wasshown that alternating bending stress due to gravity isthe major source of vibration of wind turbine blades inoperation. Therefore, only the vibration due to thegravitational force of the blades was considered as thesource of dynamic stress in operation.

Figure 4(a) shows the dynamic stress distributionalong the span direction of the blade while in opera-tion. A theoretical centrifugal force acting at the fre-quency of 3 Hz was plotted for comparison. The stresscaused by bending is much larger than tensile or shearstress as commonly shown in other beam structures.The bending stress is even larger than the centrifugalforce over most of the range.

Figure 4(b) shows the stress distribution in the crosssection of the blade in which the highest stress is causedby bending. Due to the twisting angle, the local maxi-mum points may not exactly be the leading and trailingedges. However, because the twisting angle is smallenough in this cross section, the leading and trailingedges are local maxima in the cross section in this case.

This result also indicates that the Vibro-AcousticModulation method should be sensitive to cracks alongthe trailing edge of the airfoil near the mid-span area ofthe blade. This characteristic sensitivity correspondswell with reports that blades often fail in the mid-spanarea of the blade.2 Another noticeable feature in thisresult is that a crack located at points where thedynamic strain is large will grow more rapidly underfatigue loading. In other words, cracks that are not aseasily detected using Vibro-Acoustic Modulationshould be less likely to grow under fatigue (periodic)loading than the cracks that can be detected. This stress

(b)(a)

Crack

MFC transducers

Figure 3. Experimental setup for Vibro-Acoustic Modulation tests: (a) Whisper 100 wind turbine in the wind tunnel and (b)sensors and actuator installed on the blade.MFC: micro fiber composite.

(a) (b)

strain

Crack closed

Crack opened

stressCrack opened

Effective contact

Crack closed

Effective contact

Figure 2. (a) Change of effective contact area during crack opening and closing and (b) nonlinear stress–strain function.

Kim et al. 663

at KOREA ADV INST OF SCI & TECH on December 7, 2014shm.sagepub.comDownloaded from

distribution can be useful for pinpointing the locationof the crack with modal analysis after the existence ofdamage is verified.

Based on this result, the trailing edge of the bladeat 0.3048 m from the tip of the blade was chosen forthe crack location where the maximum stress occurs asshown in Figure 4(a). The crack was induced by apply-ing a sudden load to the tip while the blade wasclamped at the intended crack location.

As shown in Figure 5(a), wireless data acquisitionsystem was installed on the rotor to eliminate the need

to use a slip ring to transmit the MFC signals becauseslip rings can be susceptible to electrical noise. The dataacquisition system includes a small netbook with soft-ware that generates the probing signal and acquires theresponse data, as well as an amplifier, which is poweredby a 9-V battery. This system provided signals with goodsignal-to-noise ratios, despite the 60 Hz noise from theelectric power system in the laboratory where the experi-ments were conducted. The rotor-installed laptop, whichtransferred the acquired data, was controlled remotelyby another computer off the turbine.

Figure 4. Stress distribution on the blade: (a) stress distribution along span direction and (b) stress distribution on a cross-sectional airfoil surface.

Figure 5. Data acquisition system: (a) netbook installation on the hub and (b) amplifier installation on the hub.

664 Structural Health Monitoring 13(6)

at KOREA ADV INST OF SCI & TECH on December 7, 2014shm.sagepub.comDownloaded from

Vibro-Acoustic Modulation tests

Three different types of Vibro-Acoustic Modulationtests were conducted on the wind turbine blades to vali-date the idea of utilizing the structural vibration ofwind turbine blades as a pumping signal. First, Vibro-Acoustic Modulation tests were performed on the oper-ating wind turbine blades, the method proposed in thisarticle. Next, Impact Modulation tests and conven-tional Vibro-Acoustic Modulation tests using twoactuators were conducted for comparison.

Vibro-Acoustic Modulation tests in operation

Vibro-Acoustic Modulation tests were conducted onthe wind turbine operating in the wind tunnel. Whilethe wind turbine was rotating, a MFC actuatorattached on the blade was excited, and another MFCsensor on the blade was used to measure the responseof the blade. The sampling rate was 96 kHz, and themeasurement period was 5 s. The rotational speed ofthe wind turbine was 3.0–3.2 Hz. Sinusoidal probingsignal excitations for the MFC actuator at frequenciesof 5–10 kHz were used. The acquired data were trans-formed using a discrete Fourier transform with aHanning window, and the sideband levels at fpr6fpu

were observed in the frequency spectrum.Figure 6 shows one of the response spectra acquired

in the tests. The peak at 3.1 Hz is the rotational speedof the rotor and was observed in the low frequencyrange with its second and third harmonics. This signalprovided the pumping signal for the Vibro-AcousticModulation tests. Around the peak at the probing fre-quency, the sideband peaks due to the modulation withthis pumping signal are observed at fpr6fpu.

In addition to the sidebands at the specified fre-quency of fpr6fpu, most of the low-frequency compo-nents were also present on both sides of the peak at theprobing frequency. Therefore, it can be concluded thatmost of the low-frequency components modulated theprobing signal. This conclusion implies that any vibra-tional components at low frequencies can be utilized asa pumping signal. For the tests conducted in the windtunnel, the excitation at 9.4 Hz was particularly effec-tive as a pumping signal because it was not only thethird harmonic of the rotational speed of the blade, butit was also near the first natural frequency of the blade;thus, the sidebands at fpr69:4Hz are relatively strong.

Some researchers have reported that the sidebandlevel acquired from Vibro-Acoustic Modulation testsvaries by probing frequency.17,18 Therefore, Vibro-Acoustic Modulation tests on the operating wind turbinewere conducted using several different probing frequen-cies. Figure 6(a) shows the measured amplitudes of thesidebands and the amplitude of response at the probingfrequency. The sideband levels plotted on this figure werecalculated from the sum of the sideband amplitudes atfpr6fpu. Note that only blade 3 is damaged and the othertwo blades do not contain cracks. It can be seen thatstronger sidebands were observed for the damaged bladeat some of the probing frequencies. However, the healthyblades also have relatively strong sidebands. There couldbe several possible reasons for this difference. First, theblade material itself can exhibit nonlinearity as shown inthe tensile tests of the specimens taken from the blades.16

Another possible source of nonlinearity is the boundaryconditions of the blades. The root of the blade is clampedon the nacelle by bolts, and it has been reported thatbolts often cause nonlinearities.19 Other joints betweenthe parts of the wind turbine structure such as the tower,foundation, and rotor can cause nonlinear responses aswell. While all these other factors can cause nonlineari-ties, they should be fairly constant across the blades.Therefore, if one measures a much larger modulation inthe damaged blade than in the healthy blades, one canassume that an increase in the sideband amplitudes iscorrelated with the presence of the crack.

It is also important to note that the curves of theresponse at the probing frequency and the curves of thesideband level are very similar. It is thought that thesideband levels are affected by similar values of the fre-quency response function of the specimen because thefrequency difference between the probing frequencyand sideband frequencies is very small.

Impact Modulation tests

The Impact Modulation test is a nonlinear defect detec-tion technique that uses free vibrations at the naturalfrequencies (excited by an impact) as the pumping

Figure 6. Vibro-Acoustic Modulation test result on thedamaged blade.

Kim et al. 665

at KOREA ADV INST OF SCI & TECH on December 7, 2014shm.sagepub.comDownloaded from

signal.20 One of the major benefits of impact modula-tion testing is that the impact can produce a largeamplitude pumping signal. In this test, the tips of theblades were impacted to excite the first modes of vibra-tion of the blades in flapping motion rather than byoperating the turbine. The probing signal was excitedby the MFC transducer. The tests were performedusing the same probing frequencies used in the Vibro-Acoustic Modulation test in operation.

Figure 7 shows the frequency spectrum of theresponse. The feature shown in the low frequency rangeis the frequency response function of the blades betweenthe impact point and the sensor. The impact excited abroadband of response frequencies in the low frequencyrange. It is also interesting to compare the low fre-quency response in Figure 6 with Figure 7. The curve at0–20 Hz in Figure 7 is similar to that in Figure 6. Sincethe third harmonic in Figure 6(a) is located at the natu-ral frequency of the blade, it has a larger amplitudethan the second harmonic and almost the same ampli-tude as the pumping signal. The first natural frequencyof the blade was approximately 9.4 Hz.

The free vibration at this frequency was used as thepumping signal in this Impact Modulation test. The mea-sured probing signal response and sideband responsesare plotted in Figure 8(b). Even though the pumping sig-nal was different from that used in the Vibro-AcousticModulation tests in operating conditions, the sidebandlevel changes are similar, suggesting that because thesideband level changes are similar, it proves that usingthe vibrations from operation is effective.

Conventional Vibro-Acoustic Modulation tests

Conventional Vibro-Acoustic Modulation tests usingtwo actuators were also performed on the same set of

blades in order to validate the proposed technique. Inthese tests, the wind turbine blades were taken off theturbine, and the roots of the blades were clamped ontoa fixture to provide similar boundary conditions to thetest conditions of on-turbine tests. To provide thepumping signal, a PCB Piezotronics disk actuator wasinstalled. Due to its response characteristic, the actua-tor was operated at 400 Hz for the pumping signalexcitation rather than using the pumping frequencyused in the previous tests. The probing signal was pro-vided by an MFC attached to the blades, and the signalwas measured by the MFC used in the previous tests.In order to determine the dependence of the Vibro-Acoustic Modulation results on the probing frequency,Vibro-Acoustic Modulation tests were also conductedwith a signal swept between 5 and 10 kHz as the prob-ing signal. The responses at the probing frequenciesand sideband levels over the different probing frequen-cies are plotted in Figure 8(c). Figure 8(c) also showssimilar curves to those in Figure 8(a) and (b).

These results indicate that the proposed technique,consisting of the Vibro-Acoustic Modulation test onthe operating wind turbine blades and utilizing thestructural vibration of the blades as the pumping signal,can measure the nonlinear characteristics of the bladesas well as conventional types of pumping signals.

Computational study

For a better understanding of the mechanism thatcauses a cracked blade to exhibit an increased non-linear response during Vibro-Acoustic Modulation test-ing, the wind turbine blade used in this article wasmodeled using finite elements. The blade was modeledas 32 Timoshenko beam elements, which include sheardisplacement as well as displacement due to the bend-ing moment, connected in series along the span direc-tion. To design this model, chord lengths and twistingangles were measured at 33 cross sections along thespan of the blade. The airfoil shape was assumed asWortmann FX63-137.

The crack was modeled by decreasing the cross-sectional area at the trailing edge of the airfoil only whencrack opened. Therefore, the stiffness matrix for the simu-lation was recalculated with the reduced cross-sectionalarea and second moment area only when crack was open.The mass decrease due to the crack was not considered.In this simulation, a crack was propagated from the trail-ing edge at 0.3048 m from the blade tip, which is the loca-tion of the crack in the experiments. The size of the crackwas assumed as r = ldamaged=c = 0:1 where ldamaged is thelength of the crack and c is the chord length.

Figure 9 shows the calculated natural frequencies ofthe blade. The first calculated natural frequency wasFigure 7. Impact Modulation test result.

666 Structural Health Monitoring 13(6)

at KOREA ADV INST OF SCI & TECH on December 7, 2014shm.sagepub.comDownloaded from

9.09 Hz, a good match for the first natural frequencyconsidering that it only differed from the experimen-tally determined natural frequency by 3%.

Setups for the simulated Vibro-Acoustic Modulationtests were modeled as follows. The force from the MFCactuator was modeled as a moment input because theforce from the MFC is in-plane and expected to cause abending moment rather a transverse force. The displace-ment, measured by the MFC sensor, was modeled as atransverse displacement because the MFCs measurestrain on the surface that is proportional to the curva-ture. In theory, the curvature is not always proportionalto the transverse displacement, but there is a strong rela-tionship between the two, so transverse displacement isassumed to be a reasonable indicator of the curvature.

Figure 10 shows the frequency response functionbetween the MFC actuator and MFC sensor of thehealthy blade and damaged blade models. A frequencyresolution df = 0:1Hz was chosen because the minimum

frequency difference between the natural frequencieswas 2 Hz (Figure 10).

Figure 10(a) shows that the frequency response func-tion in the range of f = 2000–6000 Hz is relatively flatat most frequencies. This is partially attributed to fewernatural frequencies in this range as shown in Figure 9.The sideband estimates from the model, based on thecrack opening and closing motion theory, suggest thatthe modulation sidebands emerge due to a shift of thepeaks in the frequency response function and changesin the frequency response function value at the pump-ing frequency.7

Figure 10(b), the frequency response function in thelow frequency range, has several peaks and has a highlyfrequency-dependent curve. However, it is difficult tofind the difference between the curve of the healthyblade and that of the damaged blade. Therefore, thechoice of probing frequency in this range would be aneffective choice.

Figure 8. Result of Vibro-Acoustic Modulation tests with different probing frequencies: (a) Vibro-Acoustic Modulation test resultin operating conditions, (b) Impact Modulation test result, and (c) conventional Vibro-Acoustic Modulation test result.

Kim et al. 667

at KOREA ADV INST OF SCI & TECH on December 7, 2014shm.sagepub.comDownloaded from

Figure 10(c), the frequency response function in lowfrequency range, has fewer peaks initially. Moreover,the curves throughout most of this range are flat.Therefore, it would be difficult to find modulation side-bands with the probing frequency in this range as well.

Figure 10(d) shows the frequency response functionin an ideal frequency range for selecting a probing fre-quency. The frequency response functions are highlyfrequency dependent, and the difference between thecurve of the healthy blade and that of the damagedblade is also large.

This result suggests that the probing frequencyshould be chosen carefully to effectively produce non-linear modulation. This result also corresponds wellwith the experimental results. Figure 8(a) to (c) showsthat there are larger differences in Vibro-AcousticModulation test results when the probing frequency

Figure 10. Frequency response function of the blade: (a) overall range, (b) low frequency range, (c) mid-frequency range, and (d)high frequency range.

Figure 9. Calculated natural frequencies.

668 Structural Health Monitoring 13(6)

at KOREA ADV INST OF SCI & TECH on December 7, 2014shm.sagepub.comDownloaded from

was chosen at 9000–10,000Hz compared to when theprobing frequency was chosen at 5000–6000 Hz.

Conclusion

Wind turbines require a robust structural health moni-toring strategy for wind turbine blades, and thisresearch developed a new crack detection technique forwind turbine blades in operation using Vibro-AcousticModulation. The technique utilizes the structural vibra-tion of the wind turbine blades as a pumping signalwhen the rotor rotates and measures the sideband levelswhich are the result of the modulation between theprobing signal generated from the piezoelectric actua-tor on the blade and the pumping signal.

When considering the application of this damagedetection technique in the field with a utility-scale hori-zontal axis wind turbine, modifications may be neces-sary to account for the increased size of the structureand the varying rotational speed of the turbine due tothe stochastic input provided by the wind resource.For example, the rotational speed of a utility-scale windturbine will be much lower than that of the Whisper100 turbine. Therefore, to detect modulation sidebandsof the probing frequency, increased frequency resolu-tion may be required. Another area of concern risesfrom the variable nature of the wind speed and environ-ment outside of the laboratory. The blade rotationalspeed of the turbine will vary as the wind changes speedand direction. However, the modulation sidebands ofthe probing signal can be located as long as the locationof the pumping frequency is known. Therefore, a mea-surement of the low-frequency vibrational response ofthe blade would identify the location of the pumpingfrequency and resultant locations of the modulationsidebands. If such a measurement is not possible, thenthe rotor encoder signal could be used to determine therotational speed of the turbine. Another common windturbine environmental factor is lightning. The dataacquisition system used to acquire the vibration datawill be susceptible to lightning strikes like the other elec-tronics and controls on the turbine, but this danger canbe mitigated through the use of surge suppression andgrounding with a lightning protection system. In addi-tion, the MFC transducer is coated in a polyimide film,and these materials are known to be good insulators.

This technique has many benefits for structuralhealth monitoring of wind turbine blades. First, thetechnique does not require the wind turbines to bestopped in order to inspect the blades. Therefore, notonly is the maintenance cost reduced, but there is alsono loss of generated power from the turbines when theyare taken offline for inspection. The technique is alsobeneficial because it is based on nonlinear property

changes which are sensitive to small defects. Therefore,the technique can detect small cracks in blades earlierthan other techniques normally would. Furthermore,the technique is less affected by various environmentaland loading conditions such as temperatures, humidity,wind profile, and so on than the presence of the cracksin the blades. Finally, the technique can be performedwith cost-effective excitation system because the non-linear characteristics of the excitation system do notaffect sideband levels which present nonlinear charac-teristics of the material.

In this article, the Vibro-Acoustic Modulation tech-nique on an operating wind turbine blade was demon-strated with a Whisper 100, 900 W small-scale windturbine. The test results showed that the proposed tech-nique can measure the nonlinearity in the response ofthe blades using the structural vibration as the pumpingsignal in operation as effectively as using other conven-tional Vibro-Acoustic methods that are implementedusing different types of pumping signals. In addition, acomputational frequency response function comparisonbetween a healthy blade and a damaged blade showedthat it is important to choose a proper probing fre-quency in order to maximize the sensitivity of the pro-posed technique to cracks.

Funding

Funding for this research was provided by the Division ofComputer and Network Systems (CNS) for the NationalScience Foundation (NSF) (Award Number 1136045).

References

1. Rumsey MMA and Paquette JA. Structural health moni-

toring of wind turbine blades. In: Smart sensor phenom-

ena, technology, networks, and systems, San Diego, CA, 9

March 2008.2. Ciang CC, Lee J and Bang H. Structural health monitor-

ing for a wind turbine system: a review of damage detec-

tion methods. Meas Sci Technol 2008; 19: 122001.3. European Wind Energy Association. Wind energy—the

facts: costs & prices, vol. 2. Oxford, UK: Earthscan, 2012.4. European Wind Energy Association. Wind energy—the

facts: a guide to the technology, economics and future of

wind power. New York, NY: Routledge, 2009.5. http://www.mhivestasoffshore.com/Products-and-services/

The-Turbines/V1646. http://www.aweo.org/windmodels.html7. Yoder NC. The robust detection of cracks in complex

aerospace structures using nonlinear Vibro-Acoustic Modu-

lation. PhD Dissertation, Purdue University, West Lafay-

ette, IN, 2010.8. Van Den Abeele K, Johnson PA and Sutin A. Nonlinear

Elastic Wave Spectroscopy (NEWS) techniques to discern

material damage, part I: Nonlinear WaveModulation Spec-

troscopy (NWMS). Res Nondestruct Eval 2000; 12: 17–30.

Kim et al. 669

at KOREA ADV INST OF SCI & TECH on December 7, 2014shm.sagepub.comDownloaded from

9. Brush ER, Adams DE, Zwink BR, et al. Passive healthmonitoring coupled with full field active inspection ofmultilayered composite armor structures. In: AIP confer-

ence proceedings, vol. 1096 (Review of progress in quanti-tative nondestructive evaluation), Chicago, IL, 22–25July 2008, pp. 1083–1090. College Park, MD: AIP.

10. Yoder N, Haroon M, Adams DE, et al. Multi-dimen-sional sensing for impact load and damage evaluation ina carbon filament wound canister (invited paper). Mater

Eval 2009; 66(7): 756–763.11. Sutin AM and Nazarav VE. Nonlinear acoustic methods

of crack diagnostics. Radiophys Quantum El 1995; 38(3–4):109–120.

12. Southwest Windpower. Whisper 100/200 datasheet, http://pdf.wholesalesolar.com/wind%20pdf%20folder/Whisper_100_200_Spec_Sheet.pdf

13. http://www.smart-material.com/MFC-product-main.html

14. Dana SR. Detection of blade damage and ice accretion for

health monitoring of wind turbines using integrated blade

sensors. Master’s Thesis, Purdue University, West Lafay-ette, IN, 2011.

15. Adams DE, White J, Rumsey M, et al. Structural healthmonitoring of wind turbines: method and application toa HAWT. Wind Energy 2011; 14(4): 603–623.

16. Kim S, Adams DE and Sohn H. Crack detection on windturbine blades in an operating environment using Vibro-Acoustic Modulation technique. In: AIP conference pro-

ceedings, vol. 1511, Denver, CO, 15–20 July 2012.17. Yoder NC and Adams DE. Vibro-Acoustic Modulation

utilizing a swept probing signal for robust crack detec-tion. Struct Health Monit 2010; 9: 257–267.

18. Sohn H, Lim HJ, DeSimio MP, et al. Fatigue crack detec-tion using guided waves nonlinear modulation. In: SPIEsmart structures and materials+ nondestructive evaluation

and health monitoring, international society for optics and

photonics, San Diego, CA, 10–14 March 2013.19. Jaques JL. Using impact modulation to identify loose bolts

in a satellite structure. PhD Dissertation, Purdue Univer-

sity, West Lafayette, IN, 2010.20. Duffour P, Morbidini M and Cawley P. A study of the

Vibro-Acoustic Modulation technique for the detectionof cracks in metals. J Acoust Soc Am 2006; 119: 1463.

670 Structural Health Monitoring 13(6)

at KOREA ADV INST OF SCI & TECH on December 7, 2014shm.sagepub.comDownloaded from