Available online at www.sciencedirect.

com

ScienceDirect
Procedia CIRP 11 (2013) 287 290

2nd International Through-life Engineering Services Conference

Review of condition monitoring and fault diagnosis technologies for wind

turbine gearbox
Mengyan Nie*, Ling Wang
National Centre for Advanced Tribology at Southampton (nCATS), Engineering Sciences, Faculty of Engineering and the Environme nt, University of
Southampton, Southampton SO17 1BJ, UK
* Corresponding author. Tel.: +44-23-8059-3761; fax: +44-23-8059-3016. E-mail address:m.nie@soton.ac.uk

Abstract
In order to develop robust condition monitoring and prognosis technologies and systems for wind turbine gearboxes, a comprehensive review
of the state-of-art of condition monitoring and fault diagnosis techniques has been carried out. The challenges and opportunities are identified
to guide future research in improving the accuracy and ability of condition monitoring and prognosis systems for wind turbine gearboxes. This
review also focuses on the fault diagnosis technologies and application of novel sensors in wind turbine gearbox condition monitoring.
2013
B.V.
2013 The
TheAuthors.
Authors.Published
PublishedbybyElsevier
Elsevier
B.V. Open access under CC BY-NC-ND license.
nd
International
Through-life
Engineering
Selection
ofof
thethe
International
Scientific
Committee
of the
Selection and
andpeer-review
peer-reviewunder
underresponsibility
responsibility
International
Scientifi
c Committee
of "2
the 2nd
International
Through-life
Services
Conference"
the Programme
Chair
Ashutosh
Tiwari.
Engineering
Servicesand
Conference
and the
Programme
Chair
Ashutosh Tiwari
Keywords: Condition monitoring; wind turbine; gearbox; oil analysis; signal processing; operation and maintenance

1. Introduction
Since the early 1980s, wind power technology has
undergone an immense growth with respect to both the
turbine size and the capacity installed worldwide. Global
Wind Energy Council (GW EC) reported that during the
period of 1996 to 2011, the average cu mulat ive growth rate of
wind power was over 20%, and the co mmercial wind power
capacity installed in about 80 countries totalled about 240 GW
at the end of 2011, as shown in Fig. 1 [1]. As a mainstream
electricity generation source, wind power plays a central role
in an increasing number of countries immed iate and longer
term energy plans. For example, the US targets 20% wind
based electricity generation, i.e. over 300 GW, by 2030 [ 2]. In
2009, EU Co mmun ication has also set a target share from
wind energy of 20% for the EU electricity by 2020 and 33%
by 2030 [3]. China aims for 15% renewable power generation
by 2020 [4].
However, wind turbines (WT) are co mplex electro mechanical systems wh ich ext ract kinetic power of the wind
and convert it into electrical power. The wind turbines often

work under very harsh environment, with rapidly changing

environment temperature, air p ressure and alternating load
operation conditions. Thus wind turbines are subjected to
different sort of failures, and the wind energy industry has
been affected by failu res of wind turbine co mponents, such as
main bearings, gearbo xes, and generators. Moreover, wind
turbines are hard-to-access structures, and often located in
rural areas, such as a mountainous area or at sea with t raffic
inconvenience. In addition to downtime loss of power
generation, replacing failed components usually involves
cranes and lift ing equip ment as well as service crews, which
put huge impacts on the cost of wind energy. The relatively
high operation and maintenance cost is a critical issue for
making wind energy competitive to conventional sources and
achieving the desirable renewable targets. Therefore, research
in condition monitoring and fault diagnosis of wind turbines
has gained dramatically increasing attention, as illustrated in
Fig. 2. It clearly shows that the research effort in the areas of
condition monitoring (CM) for wind turbine has dramatically
increased since 2005. In part icular, in the area of condition
monitoring for wind turbine gearbox (CM_ GB), there were

2212-8271 2013 The Authors. Published by Elsevier B.V. Open access under CC BY-NC-ND license.

Selection and peer-review under responsibility of the International Scientific Committee of the 2nd International Through-life
Engineering Services Conference and the Programme Chair Ashutosh Tiwari
doi:10.1016/j.procir.2013.07.018

288

Mengyan Nie and Ling Wang / Procedia CIRP 11 (2013) 287 290

only two papers reported pre -2004, and about 3 papers per
year in the period of 2004 to 2009. But the nu mber of the
publications in this area dramatically increased to 15 in 2010,
and then rapidly grew in the follo wing years. This reflects on
the fact that the failure of gearbo x contributes 21% to the total
downtime of wind turbines, with the highest rate of
downtimes per co mponent among all the WT co mponents,
although it only accounts for 4% of the faults [5,6,7].
Several rev iews on wind turbine gearbox condition
monitoring have already been published in the last few years.
Becker et al reviewed a nu mber of commercial available
condition monitoring solutions for wind turbine gearbo x pre2006 [8]. Also in 2006, Hyers et al reviewed the state of the
art condition monitoring and prognosis technologies used in
aerospace applications and evaluated their applicability to
wind turbine components (e.g. gearbox) [9]. In 2009, Lu et al
reviewed the advancement made in wind turbine condition
monitoring and fault diagnosis between 2006 and 2009
focusing on the monitoring of gearbo x and bearing, rotor and
blades, generator and power electronics [ 10]. They also
discussed that it is more appropriate to use time -frequency
analysis techniques such as wavelet as a key signal processing
tool for vibrat ion signals for wind turbine condition
monitoring and fault diagnosis due to the transient nature.

2. Failure modes of wind turbine gearbox

250
Annual Installed Wind Capacity
Cumulative Installed Wind Capacity

200
Wind Capacity / GW

The review also revealed there was an increasing trend in

using acoustic emission (AE) for condition monitoring and in
developing model based reasoning algorithms for fault
detection. Recently, Hamilton and Quail published a detailed
review on the oil monitoring techniques for wind turbine
gearboxes and recommended that online ferrography,
selective fluorescence spectroscopy, scattering measurements,
Fourier Transform Infrared, photoacoustic spectroscopy and
solid state viscometers are all suitable for gearbo x CM based
on their cost, size, accuracy and develop ment. They also
recommended that a combination of a number of sensors that
analyse different characteristics of the oil would provide a
greater accuracy than an individual one [ 11]. Zhong et al
briefly discussed the common failures occurring in wind
turbine gearboxes and their root causes. They also reviewed a
number of mon itoring techniques, including vibrat ion
analysis, oil analysis and AE techniques for gearbox condition
monitoring, and outlined the challenges and future research
directions for wind turbine gearbox condition monitoring
[12].
Co mpared to the previous reviews, this paper aims to
review the most recent advances in condition monitoring and
fault diagnosis techniques for wind turbine gearbo x, with the
focus on current and novel sensing techniques as well as new
signal processing methods.

150

100

50

0
1996199719981999200020012002200320042005200620072008200920102011

Year

Fig. 1. Global annual/cumulative installed wind capacities up to end of 2011

140
120

CM_WT
CM_GB

Wind turbines generally operate under extremely harsh

environment conditions including dustiness, humidity,
temperature, air pressure and unpredictable loads due to gust
wind. Wind turbine drive trains are thus sus ceptible to severe
tribological conditions such as wear, fatigue and corrosion,
leading to increased component damages and machine
malfunctions. According to a number of surveys of wind
farms cross Europe and A merica, gearbo x is by far the most
liab le subsystem that is responsible for wind turbine
downtime and maintenance cost. Wind turbine gearbo x
components are found to be subjected to a number of wear
mechanis ms, such as abrasive wear, pitting, scuffing, erosion,
crack, breakage, and ch ipping etc. The co mmon faults
observed in wind turbine gearboxes are gear damage, bearing
damage, b roken shaft, leaking oil and high o il temperature,
where bearing failures due to micro-p itting, scuffing and
white structure flaking (WSF) are found to in itiate majority of
the gearbox failures [10,12-15].

Papers in year

100

3. Recent advances in wind turbine gearbox condition

monitoring and fault diagnosis

80
60
40
20
0
1990

1995

2000

2005

2010

Publication year

Fig. 2. Publications in research fields of condition monitoring for wind

turbine (CM_WT) and wind turbine gearbox (CM_GB) in the period of 1991
to 2012.

SCA DA (Supervisory Control and Data Acquisition) is the

most common practice in wind turbine condition monitoring.
It monitors the conditions of wind turbines based on their
historical data, including wind speed, temperature and power
outputs. Some SCADA systems also include vibration and oil
debris monitoring data. SCADA is used to relate wind turbine
operating conditions (speed, temperature and power) with the
condition of the system, e.g. the rate at which damage
accumulated [15]. SCA DA systems have been imp lemented

Mengyan Nie and Ling Wang / Procedia CIRP 11 (2013) 287 290

in many wind farms as requested by insurance policy.

Although a good and useful practice, SCADA systems are not
effective especially in detecting incipient turbine fau lts to
provide sufficient warn ings for operators to take maintenance
actions. As more and larger wind farms are being developed,
the amount data collected by SCADA also becomes difficult
to manage using conventional data processing methods.
Therefore advanced data fusion/min ing strategies and reliable
theoretical models are desirable to improve the ability of
SCA DA systems in detecting wind turbine failures early and
accurately. This is in fact required by all other sensing
systems too.
Apart fro m SCADA systems, vibration-based condition
monitoring systems (CMS) are also widely used in wind
turbine condition monitoring, especially for wind turbine
gearboxes. Majority of the existing commercially available
vibration-based CMS use time -do main or frequency-domain
analysis based on Fourier t ransform (FT) for fault detection
and diagnosis. However, they are not part icularly suitable due
to the high transient nature of wind turbine vibration signals
as Fast Fourier transform (FFT) provides only the global
frequency averaged over a relatively large time span, which
means the time information is lost during the transformation.
Many efforts have been made to develop advanced signal
processing techniques for vibrat ion signals to extract useful
informat ion in recent years. Wavelet analysis, as one of the
time-frequency analysis methods, has been developed to
provide the change of frequencies in the signal over a period
of time. A number of wavelets have also been adopted to
remove non-stationary noise fro m the recorded vibrat ion
signals. Among them, continuous wavelet transform, discrete
wavelet transform as well as harmonic wavelet transform
have been accepted as key signal processing methods for wind
turbine gearbox mon itoring [16-19]. Emp irical mode
decomposition (EM D), intrinsic mode function (IMF) as well
as local mean deco mposition (LM D), which are widely used
for non-stationary and non-linear signal processing, have also
been applied in wind turbine gearbo x failure detection and
diagnosis, especially for slow speed WT components [20,21].
Other statistical analysis techniques, such as higher-order
statistical analysis, cepstrum analysis and outlier analysis,
have also been used to reduce the Gaussian noise in vibration
signals and to reveal the fault-related non-Gaussian
informat ion, thus identifying fault-related signal features [2224]. Apart fro m the above mentioned conventional statistics
based data analysis, novel data mining approaches are also
being developed for gearbox condition monitoring to
effectively deal with the rapid ly growing data size and
diversity of measurement datasets [25].
A method based on the monitoring of motor and stator
current has recently developed for WT gearbo x mon itoring.
The advantage of this method is that the motor and stator
current are already available in machine control processes and
the data is easy to be obtained [26-30]. Their results showed
that changes in the current can be related to the wind turbine
operation conditions and mach ine fau lts. Renaudin et al
developed optical or magnetic encoders that measure angular
speed fluctuations due to localised faults in bearings and thus
were used for bearing fault detection [31].

289

Apart from developing robust individual sensing

techniques, it is becoming more and more recognised that
mu ltip le-sensor integrated systems with mu lti-parameter
approaches can achieve much better condition monitoring
results. Each sensing technique has its own merits as well as
limitat ions. For examp le, acoustic emission is found to be able
to detect incipient failures in low-speed wind turbine
components while vib ration monitoring is better in early fault
diagnostics for the high-speed components. Crabtree et al [32]
developed a mult i-parameter approach based on comparison
of independent signals (conventional vibrat ion condition
monitoring signals with operational signals, such as load or
energy) to increase the confidence in fault signal
interpretation and thresholds generated. Also, vibration and
AE based multi-sensing technology together with model
based reasoning algorithms were developed to enhance fault
diagnosis for wind turbine gearbox [19,33,34].
4. Conclusions
Recent advances in sensing and signal processing
technologies have significantly imp roved the condition
monitoring and prognosis of wind turbine gearbo xes, and
reduced the wind turbine operation and maintenance costs.
However, accurate fau lt detection and reliable prognosis in
wind turbine gearbo x remain challenging due to the inherited
complexit ies in their mechanical systems and operation
conditions. In order to achieve robust condition monitoring
for WT gearbo xes, the following research areas need to be
further strengthened:
x Sensor Technologies: Most current condition monitoring
systems are based on vibration and oil analysis.
Accelerometers are often mounted on gearbox housings,
which often leads to low signal-to-noise ratio that affecting
the detection. This issue can be dealt with by using
miniaturizing and/or embedding the sensors close to the
fault sites. Apart from the existing sensing techniques,
novel sensors such as Barkhausen noise inspection, in-situ
hydrogen sensors, oil quality monitoring devices, as well
as multi-sensor combination, should be further exploited
for wind turbine gearbox monitoring.
x Signal Processing and Data Management: As huge amount
of data are being collected from wind farms, the fast
growing databases require advanced signal processing
techniques and data mining strategies to extract the most
useful information for timely wind turbine condition
monitoring and maintenance.
Acknowledgements
The authors would like to thank the EPSRC and TES
Centre fo r their financial support to this feasibility study
under grant number EP/1033246/ 1.
References
[1] Global wind report Annual market update 2011. Global Wind Energy
Council, March 2012.

290

Mengyan Nie and Ling Wang / Procedia CIRP 11 (2013) 287 290

[2] 20% Wind energy by 2030: increasing wind energys contribution to US

electricity supply. United States Department of Energy, Report DOE/GO102008-2567, July 2008.
[3] Investing in the development of low carbon technologies (SET -Plan).
European Commission Communication, Oct. 7, 2009.
[4] Mid and long range plan for renewable energy development. Chinese
Committee for National Development and Reform, Aug. 2007.
[5] Hahn B, Durstewitz M, Rohrig K. Reliability of wind turbine experiences
of 15 years with 1,500W. Wind Energy. Berlin: Springer; 2007, p. 329332.
[6] MacMillan D, Ault GW. Quantification of condition monitoring benefit
for offshore wind turbines. Wind Energy 2007; 31(4): 267-285.
[7] Wilkinson MR, Spinato F, T avner PJ. Condition monitoring of generators
and other subassemblies in wind turbine drive trains. 2007 IEEE
International Symposium on Diagnostics for Electric Machines, Power
Electronics and Drives, Sep. 2007, p. 388-392.
[8] Becker E, Poste P. Keeping the blades turning: condition monitoring of
wind turbine gears. Refocus 2006; 7(2): 26-32.
[9] Hyers RW, McGowan JG, Sullivan KL, Manwell JF, Syrett BC.
Condition monitoring and prognosis of utility scale wind turbines. Energy
Materials 2006; 1(3): 187-203.
[10] Lu B, Li Y, Wu X, Yang Z. A review of recent advances in wind turbine
condition monitoring and fault diagnosis. 2009 IEEE Power Electronics
and Machines in Wind Applications, Jun. 24-26, 2009, Lincoln, NE, USA.
[11] Hamilton A, Quail F. Detailed state of the art review for the different on line / in-line oil analysis techniques in context of wind turbine gearboxes.
Journal of Tribology-Transaction of ASME 2011; 133(4): 971-978.
[12] Zhong X, Zeng L, Zhao C, Zhang J, Wan S. Research of condition
monitoring and fault diagnosis techniques for wind turbine gearbox.
Applied Mechanics and Materials 2012; 197: 206-210.
[13] Bartelmus W. Gearbox damage process. 9 th International Conference on
Damage Assessment of Structures, Jul 11-13, 2011, University of Oxford,
Oxford, England.
[14] Cibulka J, Ebbesen MK, Hovland G, Robbersmyr KG, Hansen MR. A
review on approaches for condition based maintenance in application with
induction machines located offshore. Modelling Identification and
Control 2012; 33(2): 69-86.
[15] Feng Y, Qiu Y, Crabtree CJ, Long H, Tavner PJ. Monitoring wind
turbine gearboxes. Wind Energy 2012. DOI: 10.1002/we.1521.
[16] Yang S, Li W, Wang C. The intelligent fault diagnosis of wind turbine
gearbox based on artificial neural network. Proceedings of 2008
International Conference on Condition Monitoring and Diagnosis (CMD
2008), Apr 21-24, 2008, Beijing, China, p1327-1330.
[17] Kou HJ, Yuan HQ, Zhao XY. Application of wavelet de-noising in fault
diagnosis of wind turbine gearboxes. Advanced Materials Research 2012;
562-564: 1091-1094.
[18] Zhou FC. Research on online monitoring of gear fault of wind power
gearbox based on undecimated wavelet transformation. Manufacturing
Science and Technology 2012; 383-390: 4928-4931.
[19] Lu Y, Tang J, Luo H. Wind turbine gearbox fault detection using
multiple sensors with features level data fusion. Journal of Engineering
for Gas Turbines and Power 2012; 134(4): 042501.
[20] Antoniadou I, Manson G, Staszewski WJ, Worden K. Damage detection

in gearboxes considering intermittent faults and time-varying loads. Key

Engineering Materials 2012; 518: 76-86.
[21] Liu WY, Zhang WH, Han JG, Wang GF. A new wind turbine fault
diagnosis method based on the local mean decomposition. Renewable
Energy 2012; 48: 411-415.
[22]Liu Y, Zhou Y, Xin W, He Y, Fan P. Bispectrum analysis for feature
extraction of pitting fault in wind turbine gearbox. 2010 IEEE
International Conference on Mechatronics and Automation(ICMA 2010),
Aug. 4-7, 2010, Xi'an, China, p. 1298-1301.
[23] Teng W, Wu X, Gao Q, Liu Y. Cepstrum Analysis of Vibrat ion Signals
of Wind T urbine Gearbox. Proceedings of the 2011 IEEE International
Conference on Multimedia and Expo (ICME 2011) 2011, p. 4237-4240.
[24] Antoniadou I, Manson G, Dervilis N, Staszewski W, Worden K. On
damage detection in wind turbine gearboxes using outlier analysis.
Proceedings of SPIE - The International Society for Optical Engineering,
2012, 8343.
[25] Baqqar M, Ahmed M, Gu F. Data mining for gearbox condition
monitoring. 17th International Conference on Automation and
Computing, Sept. 10, 2011, Huddersfield, UK, p138-142.
[26] Mohanty AR, Kar C. Monitoring gear vibrations through motor current
signature analysis and wavelet transform. Mechanical Systems and Signal
Processing 2006; 20(1): 158-187.
[27] Kar C, Mohanty AR. Vibration and current transient monitoring for
gearbox fault detection using multiresolution Fourier transform. Journal
of Sound and Vibration 2008; 311(1-2): 109-132.
[28] Amirat Y, Choqueuse V, Benbouzid M, Turi S. Hilbert transform-based
bearing failure detection in DFIG-based wind turbines. International
Review of Electrical Engineering 2011; 6(3): 1249-1256.
[29] Baqqar M, Wang T, Ahmed M, Gu F, Lu J, Ball A. A general regression
neural network model for gearbox fault detection using motor operating
parameters. 2012 UKACC International Conference on Control
(CONTROL 2012), Sept. 3-5, 2012, Cardiff, UK, p584-588.
[30] Cibulka J, Ebbesen MK, Robbersmyr KG. Bearing Fault Detection in
Induction Motor-Gearbox Drivetrain. 25th International Congress on
Condition Monitoring and Diagnostic Engineering (COMADEM), Jun.
18-20, 2012, Huddersfield, UK.
[31] Renaudin L, Bonnardot F, Musy O, Doray JB, Remond D. Natural roller
bearing fault detection by angular measurement of true instantaneous
angular speed. Mechanical Systems and Signal Processing 2010; 24(7):
1998-2011.
[32] Crabtree CJ, Feng Y, T avner PJ. Detecting incipient wind turbine
gearbox failure: a signal analysis method for on-line condition
monitoring, European Wind Energy Conference and Exhibition 2010
(EWEC 2010), Apr. 20-23, 2010, Warsaw, Poland.
[33] Lekou DJ, Mouzakis F, Anastasopoulos AA. Fused acoustic emission
and vibration techniques for health monitoring of wind turbine gearboxes
and bearings. European Wind Energy Conference and Exhibition 2009
(EWEC 2009), Mar. 16-19, 2009, Marseille, France.
[34] Soua S, Lieshout PV, Perera A, Gan TH, Bridge B. Determination of the
combined vibrational and acoustic emission signature of a wind turbine
gearbox and generator shaft in service as a pre-requisite for effective
condition monitoring. Renewable Energy 2013; 51: 175-181.