ConditionMonitoring and ault Diagnosisin Wind nergy onversion Sstemis: A Reviewi

Y. Amirat, M.E.H. Benbouzid, Senior Member, IEEE, B. Bensaker and R. Wamkeue, Member, IEEE

Abstract-There is a constant need for the reduction of ...A. ......................

operational and maintenance costs of Wind Energy Conversion ..a o......2...FIA o Wu kid Res o Wogd Eut2o 3Systems (WECS). The most efficient way of reducing these costs Muirai ......................

would be to continuously monitor the condition of these systems. 3

This allows for early detection of the degeneration of the 0.....generator health, facilitating a proactive response, minimizing -downtime, and maximizing productivity. Wind generators arealso inaccessible since they are situated on extremely high 0

towers, which are normally 20 m or greater in height. There areReM 5 D 0 5

also plans to increase the number of offshore sites increasing the Fig. 1. Installed wind power.need for a remote means of WECS monitoring that eliminatessome of the difficulties faced due to accessibility problems. Side effect on other components can be reducedTherefore and due to the importance of condition monitoring significantly. Many faults can be detected while the defectiveand fault diagnosis in WECS (blades, drive trains, and cmoeti tl prtoa.Tu,ncsayrpi cingenerators); and keeping in mind the need for future research, compnben ismesi operation al.Ths necessarytrepi tactinsthis paper is intended as a tutorial overview based on a review of ca bepnedi th tme nd edno to etknthe state of the art, describing different type of faults, their immediately. This is important as WECS generators aregenerated signatures, and their diagnostic schemes. inaccessible since they are situated on extremely high towers,

which are normally 20 m or greater in height (Fig. 2). It isIndex Terms-Wind turbine, induction generator, drive also important especially for offshore plants, where bad

train, condition monitoring, fault diagnosis. weather conditions (storm, high tides, etc.) can prevent anyrepair actions for several weeks (Fig.2). Moreover, condition

1. INTRODUCTION ~~~monitoring will also detect extreme external conditions, suchI. INTRODUCTION ~~~asicing or water induced tower oscillations of offshore plants,

Wind energy conversion is the fastest-growing source of and can trigger appropriate control actions to prevent damagenew electric generation in the world and it is expected to of plants components. This way, overall maintenance costsremain so for some time. At the end of 2003 the installed and down times of wind energy converters can bewind capacity stands at over 40000 MW, doubling since significantly reduced [2-5].1999, and it could exceed 95000 MW by the end of 2008 Therefore and due to the importance of condition(Fig. 1). But the higher target is to achieve 120o of the world's monitoring and fault diagnosis in WECS (blades, drive trains,electricity from wind power by 2020. Harnessing wind energy and generators); and keeping in mind the need for futurefor electric power generation is an area of research interest research, this paper is intended as a tutorial overview basedand at present, the emphasis is given to the cost-effective on an exhaustive review of the state of the art, describingutilization of this energy resource for quality and reliable different type of faults, their generated signatures, and theirpower supply. During the last two decades wind turbines have diagnostic schemes. As the Doubly-Fed Induction Generatorbeen developed in size from 20 kW to 2 MW, while even (DFIG) is one of the most used WECS configurations, thelarger wind turbines already are being designed [1]. review will be mainly focused on it (Fig. 3).

Autonomous online condition monitoring systems withintegrated fault detection algorithms allow early warnings ofmechanical and electrical faults to prevent major componentfailures.

Y. Amirat and M.E.H. Benbouzidare with the Laboratoire d'Ing~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~.... .........ier...........e...M&aniqueetElectrique(LIME),UniversityofWesternBrittany,Ruede~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~.........................................Kergoat-BP93169,29231BrestCedex3,France(phone:±332980180~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~HHH....................07; fax: ±33 2 98 01 66 43;e-mail: m.benbouzid@ieee.org). ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~........Y. Amirat is.......... ~also with the Electrical Engineering Department, University of Annaba,~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~......................23000Annaba,Algeria.~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~.............B. Bensaker is with the ElectricalEngineering Department, University of~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~.............Annaba,23000Annaba,Algeria.~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~.....

R. Wamkeue is with the Department of Applied Sciences, University of~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~.....Quebec, Rouyn-Noranda, QC,Canada. Fig. 2. Currently largest onshore wind turbine and offshoreinstallations[6].~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~.

1-4244-0743-51071$20.OO©2007IEEE1434~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~...... ..........

Main Circuit Breaker pennant ± wind meter

Dime gene t grida a 1 1~~~~~~~~~~~~~~~eaboTe........f......nIereariefeenge gererator

I :Medium Voltage SwitLhgear

|Converter 07e[ro ,l li;T!i X .;A!

WCohol-|| indTOnrtrbi"

Vmotbrevlvgmer

Fig. 4. Wind turbine nacelle cross-section.

Fig. 3. Most used WECS configuration (with DFIG arrangement) [6].Entire unit 2,7)

Hub 0,3Structure5 BladnesPitch 13,4

II. FAILURE MODE ANALYSIS Yaw System 6,7 J

Real wind turbine failure data quantitative analyses have H Gdrn13 Generator 5.5shown important features of failure rate values and trends [7- Ha 110]. For illustration, Fig. 4 shows the main wind turbinecomponents that are concerned by the above failure analyses.

M

In the first study concerning Swedish wind power plants MElectric System 17,5[7-8]; it has been shown that most failures were linked to theelectric system followed by sensors, and blades/pitch Gearscomponents. This is clearly illustrated by Fig. 5 that showsfailures number distribution for Swedish wind power plantsthat occurred between 2000 and 2004. Sensors 14,1 129

Another study, concerning Danish and German wind Drive train 1,1power plants [9], shows the same tendency. Indeed, principal Fig. 5. Failures number distribution [0] for Swedish wind power plants [7].contributors to the higher German failure rate are theelectrical control or system subassemblies (grid or electricalsystem, yaw system and mechanical or pitch control system) 0.4rather than mechanical subassemblies such as the gearbox. O 35Figure 6 shows then the failure rate in the two power plantsduring the period 1994-2004.

All the above analyses are consistent with the introductionof variable speed drive technology but are contrary to thereceived wisdom that gearboxes are a major cause of turbine Ufailure [11]. L

For the above reasons, the proposed review will be 005focused on these types of failure. XDOO

III. CONDITION MONITORING AND DIAGNOSIS e X

Itis

well-known that many electrical and mechanicalfaults in induction motors have a direct impact on the motormagnetic field. Indeed, they modulate it [12]. Moreover, ithas been proven that failures in a mechanical drive train Fig. 6. Failures rate for Danish and German wind power plants [9].connected to an induction machine can be detected at theterminal of the machine [13].

In the case of wind turbine condition monitoring, a The basic configuration that is used for WECS conditionnumber of published work are based on the following monitoring and diagnosis is shown by Fig. 7. It should behypothesis: It is possible to detect wind turbine drive train noticed that in modern wind turbines, some signals, such asfaults through the terminals of the associated generator [14- rotational speed, generator temperature, etc., are commonly15]. monitored by the supervisory control system [3].

1435

vibrations ensorl \ ~~~~~~~~~getnerator current

\ X~BO

1>ottrn * Mo11nitoringJ Wt\ECUo sensrSystem Monitor

iWind ~~~~~~Doubly Fed 4 Quadrant WX\ECTurbine o Gearbox + Induction Crtonverter & Por

,Generator Controller O r cuoutput

WfEC

Fig. 7. Basic architecture of WECS condition monitoring and diagnosis [3].

A. Electrical System (DEIG) Condition Monitoring characterization. Indeed, the rotor modulating signals haveThere*aremanytechniquesandte harmonic content that gives evidence pf stator and rotor

use to moio th codto of inuto mahns Someof asymmetries more clearly than the harmonic content of statorthe technology used for monitoring includes sensors, which and rotor currents.

mayd Gearb mo te rtue In [24], authors have raised a key feature of wind turbine

Turbine Geerto .' ovntraollr Ourtosned tputr peoinnl

flux lenstiles, etc. These sensors are together coupled withn c ractoropation. Indeed, th eym ar p ignanlyalgri and architectures, whichalofor rff transient therefore prompting the use of nonstationaryalgorithms and architectures, which allows for efficient trniet thrfr rmtn heueo osainr

monitoring of the machines condition [16]. The most popular techniques for fault detection [25]. In this case, waveletmethods of induction machine condition monitoring utilize analysis has been used to the detection of stator turn faults inthe steady-state spectral components of the stator quantities. a DFIG. The detection algorithm is a combination of theThese stator spectral components can include voltage, cu..ent, Extended Park Vector, wavelet analysis, and statistics. This

1 r 1 1 1^ ^ ~~techniqu a o fetdb hne nDI pe hcand power and are used to detect turn faults, broken rotor .t que was not affected by changes in DFIG speed, whichbars, bearing failures, air gap eccentricities [17-18]. is crucial in WECS applications.

The above techniques that are based on steady-state One of the preferred options at present, for large turbines

analysis are being applied to induction generators. When excess of2-MW rating, is the variable speed DFIG with the

scaning the available literature, it has been found that fault rotor converter connected to the rotor via slip rings [6], [26].detection and diagnosis techniques are mainly arranged for However, in contrast to squirrel cage generators there are

inter-turn stator faults and stator or rotor asymmetries [19- additional wear parts, e.g. the slip ring system (Fig. 8).23].~~~~~Inallthse ors, utor ar jst ppyin wll Therefore, in [27], the authors are suggesting a patentedesalse tecniue to inuto geeaos However,Tin diagnostic technique for the monitoring of the transmission[19-2] th auhr ar usn th roo mouatn sinl roperties and sparking of DFIG with slip rings. In this case,

spectra as a diagnosis index for the stator and rotor faults mdfctosi h rnmsinpoete r igoeusing the monitored rotor currents through FFT analysis.

1436

G inverter A Receptor B C Down D. ^ ~~~~~~~s iS h~~~~~Ncondluctor

Ow,n\const. 3 f =var. XfMetalmesh

Steiel wiren_ G ,\\ '

slip ring systemFig. 8. Slip ring generator system principle.

IEC 1864/02B. Damage Detection ofWECS Blades Fig. 9. Lightning protection for large modern wind turbine blades [31].

Wind turbine blades are a vital component. Due toexternal conditions and internal stress as well as fatigue, the Indeed, in [32] the authors dealt with the demodulation of thecrack and damage may gradually take place as time goes by, current signal of an induction motor driving a multistagethus leading to the performance deterioration of wind gearbox for its fault detection. In this case, amplitudegeneration. In other words, it is crucial to monitor the turbine demodulation and frequency demodulation are applied to theblades such that the operation performance can be better induction motor current for detecting the rotating shaftensured. A condition monitoring package should be capable frequencies. A discrete wavelet transform is applied to theof being retrofitted onto existing wind turbines without demodulated current signal for denoising and removing therequiring additional sensors and wiring on the machine. intervening neighboring features. Spectrum of a particularTherefore, in [28], the authors used the above mentioned level is used for gear fault detection. This technique seems tohypothesis to detect the presence of unbalance and defects in be very interesting to monitor WECS gearbox as it involves athe blades of a small wind turbine by measuring the power nonstationary technique.spectrum density at the generator terminals. In this case, In the case of rolling bearing fault diagnosis and apartbicoherence, a normalized bispectrum, is used. Indeed, it is from the well-established techniques developed for inductionable to monitor small physical changes in the machine using a motor drives [17], a recently published seems to bevery noisy signal. This technique overcomes problems of the convenient for wind turbine bearing fault detection [33].bispectrum that is not convenient for detection purposes [18]. Indeed, it uses the wind turbine generator stator current.The advantage of such an approach is that the generator Moreover, due to the nonstationary nature of this current, theterminal quantities are easily accessible during operation, the wavelet packet transform provides better analysis undercurrent via a current transformer, the voltage via a voltage varying load conditions. The wavelet packet transform alsotransformer and the power by computation. This is a very permits the tailoring of the frequency bands to cover the rangeuseful technique as it requires no additional sensors, of bearing faults induced frequencies resulting from rotorparticularly on the blades, which is the case in [29], where a speed variations (e.g. variable speed DFIG).continuous wavelet transform-based approach is used todetect blades damage.

Unfortunately, wind turbine blades experience faults and D Other Fault Detection and Diagnosis Approachesdamages that could not be monitored using the wind turbine In [34], the authors propose the application of artificialgenerator terminals. They are particularly exposed to a major intelligence techniques for WECS condition monitoring,threat in form of lightning strikes. To prevent damage, blades including tower, nacelle, and power train. This approachare equipped with a lightning protection system, as most requires a learning process for each individual WECS, andmodern WECS [30] (Fig. 9). However, as lightning is random seeks to detect trends, without necessarily linking cause toin nature, a complete protection against its damages is not fault effects. In this case, the proposed approach for WECSachievable. Therefore, in [31] is presented a method for monitoring is shown by Fig. 10, where the layer, calledlightning impact localization and classification using a fiber evaluation or diagnosis layer uses neural networks foroptic current sensor network that helps to detect damages characteristics learning or fuzzy techniques to combinecaused by lightning and to monitor the blades. The system is information from different measurements.connected to the wind turbine control and monitoring system.

E. Data CollectionC. Other Drive Train Components Monitoring and Diagnosis Today, most turbines are fitted with equipment that makes

Othe wid tubin ke comonets ae rlle beaingand it possible to collect condition monitoring data remotely viagear Fo thsetpesof ompnentconitin moitoingand modem or internet. Moreover, since wind turbines are typicallygear Fo ths tye of copnn codto monitoring..... and built in onshore or offshore wind farm configurations; there iS a

diagnosis, the use Of the wind turbine generator terminals has ne o uligu ewrs(i.1)not been explicitly investigated.

1437

vhraio aXra tan tl (adapWulimits.4|ganerator bearing, values iearningB

(BCU, RMS, Cre t tedani)s

Fig.S<d 11.Codto moiorn netor in win farms.1N

th wind far (eg win speed) ill beusdtoipve th vIbrto inue by berigsn gerhesoverallwidfr perorassncile by miiizn operationa

lose de aenrailus [5] [

I i"X1sIV INDSR APICTOShi pIaper ha regview the stt of art ofl widerg

Condition monitor singis a mciemitnnetol cneso ytmscniinmntrngaddanss h

lnlrrow banhd l_l

Fiustratesatesio EScniinmntrn addanss[4.Fg2ntegration of a CMS in a WECS [35-3]]nmtr.Ide,i.emta ti osbet eetwn

thicsete MSmoitrig fncios rebaedonroust tubiearieeran auts hrug ten erinlsofth

accllermterstomesuet bthe nacll oclationin* axialad gneao erias hi a ls h as fWC

transverBsesdirection relatditionthenrtorinaxddansi3is;.gerbox.Howe egr,astiond turbCMSine geeatorC opeatonsar

Pldkh P&|kh4 ......1......438

L~~ ~~~~~~~~~~~~~~~~~~ t'Y- ....................

Dt|ArOWfiusnlRdrnbt;g Adte;6\\ts\......

Fig. I l. Condition monitoring network in wind farms.

With this network approach, fault diagnosis will be Fig. 13. Typical sensor position [35].optimized comparing WECS operation under identicalconditions. Moreover, redundant measurements provide by snos adSaevbainsnost esr hthe wind farm (e.g. wind speed) will be used to improve the virtoinuebyeangadgawhlsOverall wind farm performance by minimizing operationallosses due anemometer failures [5], [34].

V. SUMMARYIV. INDUSTRYAPPLICATIONS This paper has review the state of art of wind energy

Condition monitoring is a machine maintenance tool conversion systems condition monitoring and diagnosis. The(known as Condition Monitoring System - CMS) that is emphasishasbeenputon faults that couldbemonitoredusingbecoming a component of long-term service packages the wind turbine generator (DFIG) terminals in an attempt toprovides by some wind turbine manufacturers. Figure 12 use well-established techniques developed for inductionillustrates the integration of a CMS in a WECS [35-36]. In motors. Indeed, it seems that it is possible to detect windthis, nc,,ase,te M moniorin fucins,ar basedonArobust-1C' fiurbinie drivei train fauilts thiroughl thei tierminals, of thei

predominantly transient, the use of nonstationary techniques [18] M.E.H. Benbouzid, "A review of induction motors signature analysisis required for faults detection. as a medium for faults detection," IEEE Trans. Industrial Electronics,

Finally industry application ofcondivol. 47, n°5, pp. 984-993, October 2000.Finally industry application of condition monitoring has [19] D. Casadei et al., "Diagnostic technique based on rotor modulating

been briefly introduced through CMS. signals signature analysis for doubly fed induction machines in windgenerator systems," in Proceedings ofIEEE IAS'06, vol. 3, pp. 1525-1532, Tampa (USA), October 2006.

[20] D. Casadei et al., "Experimental fault characterization of doubly fedinduction machines for wind power generation," in Proceedings of

REFERENCES IEEE SPEEDAM'06, pp. 1281-1286, Taormina (Italy), May 2006.

[1] BTM Consults Aps, International Wind Energy Development - World [21] I. Albizu et al., "On-line stator winding fault diagnosis in induction

Market Update 2000. ISBN 87-987788-0-3, Ringk6bing, Denmark, 2001. generators for renewable generation," in Proceedings of IEEE[2] P. Caselitz et al., "On-line fault detection and prediction in wind MELECON'04, vol. 3, pp. 1017-1020, Dubrovnik (Croatia), May 2004.

energy converters," in Proceedings of EWEC-94, pp. 623-627 [22] Q.F. Lu et al., "Model of stator inter-turn short circuit fault in doubly-Thessaloniki (Greece), 1994. fed induction generators for wind turbine," in Proceedings of IEEE

[3] P. Caselitz et al., "Development of a fault detection system for wind PESC'04, vol. 2, pp. 932-937, Aachen (Germany), June 2004.energy converters," in Proceedings of EUWEC96, pp. 1004-1007, [23] L. Mihet-Popa et al., "Condition monitoring of wind generators," in

Goteborg (Sweden), 1996. Proceedings of IEEE IAS'03, vol. 3, pp 1839-1846, Salt Lake City[4] P. Caselitz et al., "Application of condition monitoring systems in wind (USA), October 2003.

energy converters," in Proceedings ofEWEC97, pp. 579-582, Dublin [24] H. Douglas et al., "The detection of interturn stator faults in doubly-fed(Ireland), 1997. induction generators," in Proceedings ofIEEEIAS'05, vol. 2, pp. 1079-(Ireland), 1997. ~~~~~~~~1102, Honk Kong, October 2005.

[5] J. Nilsson et al., "Maintenance management of wind power systems [5 K. Kim etna, "ctio n m f d s eusing condition monitoring systems Life cycle cost analysis for two neuropret alndwvtign processing," siE asMeTncase studies," IEEE Trans. Energy Conversion, vol. 22, n, pp. 223- eronics and wavelet s1nal Jrocessine " IEEEASE Trans.

229, March 2007. Mechatronics, vol. 7, n°2, pp. 201-219, June 2002.[6] M.E.H. Benbouzid et al., "The state of the art of generators for wind [26] R. Datta et al., "Variable-speed wind power generation using doubly

energy conversio system" i Proce g of I Chania fed wound rotor induction machine A comparison with alternative(Greecerg Se tember 2006m schemes," IEEE Trans. Energy Conversion, vol. 17, n°3, pp. 414-421,

[7] J. Ribrant et al., "Survey of failures in wind power systems with focus September 2002.onSwedish wind power plant during 1997-2005," IEEE Trans. Energy [27] M. Wurfel et al., "monitoring of the properties of the rotor slip ring

Conversions vol. 22 nl pp. 167-173 March 2007. system of doubly-fed induction generators," in Proceedings of IEEEConerson vol 22 n1, n rIMC0pp. 167-173, Marc 2007.(UA My205[8] J. Ribrant, Reliability performance and Maintenance - A Survey of IEMDC'05, pp. 295-299, San Antonio (USA), May 2005.Failure in Wind Power Systems, MlIaster Thesis, KTH School of [28] W.Q. Jeffries et al., "Experience with bicoherence of electrical powerElectrical Engineering, 2006. for condition monitoring of wind turbine blades," IEE Proc. Vision,

'9 P.J. Tavner.et al."Reliabilityanalysisforwindturbines,Image and Signalprocessing, vol. 145, n°3, pp. 141-148, June 1998.[9]iPnJdTanerg et l.I," ility a alyssAfrwi trie,

[29] C.S. Tsai et al., "Enhancement of damage-detection of wind turbineWi7nd Energy, vol. 10, n° 1 pp. 1-18, March-April 2006..[10] P. J. Tavner et al. "Machine and converter reliabilities in wind blades via CWT-based approaches," IEEE Trans. Energy Conversion,[10] P. J. Tavner et at., "Machine and converter reliabilities in wind vo.2,n3'p 7-81 etme 06

turbines," in Proceedings ofIEE PEMD'06, Dublin (Ireland), 2006. vol. 21, n3, pp. 776-781, September 2006.[11] M. Mohsin Khan et at, "Reliability and condition monitoring of a [30] I. Cotton et al., "Lightning protection for wind turbine blades and

wind turbine," in Proceedings of IEEE CCECE'05, Saskatoon bearings," Wind Energy, pp. 23-37, 2001.(Canada), 2005. [31] S.G.M. Kramer et al., "Fiber optic sensor network for lightning impact

[12] S. Nandi et al. "Condition monitoring and fault diagnosis of electrical localization and classification in wind turbines," in Proceedings of[125.Nani e a. "Cniinmntrn.n al igoi feetia IEEE ICMFIIS'06, vol. 2, pp. 173-178, Heidelberg (Germany),motors - A review," IEEE Trans. Energy Conversion, vol. 20, n°4, pp. SEEE 20'06.719-729, December 2005. September 2006.

[13] R. Yacamini et a., "Monitoring torsional vibrations of [32] A.R. Mohanty et al., "Fault detection in a multistage gearbox byelectromechanica systemsusingstatorurrents,"ASEJodemodulation of motor current waveform," IEEE Trans. Industrialelectromechanical systems using stator currents," ASME Journal of Elcrnc,vl 3 0,pp 2519,Ags 06

Vibration & Acoustics, vol. 120, pp. 72-79, 1998. Electronics, vol. 53, n°4, pp. 1285-1297, August 2006.Vibrat.i. & Acoustics vl.12 pp. 72-79, 198 . [33] L. Eren et al., "Bearing damage detection via wavelet packetr1m iinsonetai "ve tract ig Prcnditio monitrin infora1to decomposition of the stator current," IEEE Trans. Instrumentation &

pp. 591-594a Bristol (UK)d 2004t 0 Measurement, vol. 53, n°2, pp. 431-436, April 2004.pp.591-94 Bristol (U) 20. . .........[34]P. Caselitz et al., "Advanced condition monitoring system for wind[15] M.R. Wilkinson et al., "Condition monitoring of wind turbine drive 3

trains," in Proceedings of*ICE'06, Chania (Greece), September 2006. energy converter," in Proceedings ofEWEC'02, Nice (France), 2002._,,_ MEH BenbOuzid "Bibliography oninductionmoto ldetection [35] European Commission / DG TREN, "Advanced maintenance and[16] M.E.H. Benbouzid, "Bibliography on induction motors faults otcoand diagnosis," IEEE Trans. Energy Conversion, vol. 14, n4, pp. repair for offshore wind farms using fault prediction and condition

1065-1074, December 1999. monitoring techniques," Final Report, NNE5120011710, FP5 Contract,[17] M.E.H. Benbouzid et al., "What stator current processing based 2005.

technique to use for induction motor rotor faults diagnosis?," IEEE [36] P. Caselitz et al., "Rotor condition monitoring for improvedTrans. Energy Conversion,vol. 18, n'2, pp. 238-244, June 2003. operational safety of offshore wind energy converters," Journal ofTrans. Energy Convesion,vol.18,n2p.238244Solar Energy Engineering, vol. 127, n°2, pp. 253-261, May 2005.

1439