Renewable Energy 46 (2012) 169e178

Contents lists available at SciVerse ScienceDirect

Renewable Energy
journal homepage: www.elsevier.com/locate/renene

Condition monitoring of wind turbines: Techniques and methods

Fausto Pedro Garca Mrquez a, *, Andrew Mark Tobias b, Jess Mara Pinar Prez a, Mayorkinos Papaelias b
a b

Ingenium Research Group, Universidad Castilla-La Mancha, 13071 Ciudad Real, Spain University of Birmingham, Birmingham B15 2TT, United Kingdom

a r t i c l e i n f o
Article history: Received 15 September 2011 Accepted 3 March 2012 Available online 5 April 2012 Keywords: Fault detection and diagnosis Condition monitoring Maintenance management Wind turbines

a b s t r a c t
Wind Turbines (WT) are one of the fastest growing sources of power production in the world today and there is a constant need to reduce the costs of operating and maintaining them. Condition monitoring (CM) is a tool commonly employed for the early detection of faults/failures so as to minimise downtime and maximize productivity. This paper provides a review of the state-of-the-art in the CM of wind turbines, describing the different maintenance strategies, CM techniques and methods, and highlighting in a table the various combinations of these that have been reported in the literature. Future research opportunities in fault diagnostics are identied using a qualitative fault tree analysis. Crown Copyright 2012 Published by Elsevier Ltd. All rights reserved.

1. Introduction Only time will tell whether the forecasts by WWEA [1] for continuing growth in global wind power capacity [2] will be realised as shown in Fig. 1. But to make wind power competitive with other sources of energy, availability, reliability and the life of turbines will all need to be improved. As the wind energy sector grows, business economics will demand increasingly careful management of costs. For a 20-year life, the operations and maintenance (O&M) costs of 750 kW turbines might account for about 25%e30% of the overall energy generation cost [3] or 75%e90% of the investment costs [4]. Furthermore, one projection in 2002 was that the O&M costs for 2 MW turbines (which together with 2.5 and 3 MW turbines have since become the workhorses of the wind power industry) might be 12% less than an equivalent project of 750 kW machines [5]. But new wind farms typically have higher capacity and comprise more machines. The turbine data in Fig. 2 [6] suggest that larger turbines fail more frequently and thus require more maintenance. Reducing inspection and maintenance costs has thus become increasingly important as wind turbine size and numbers have continued to rise. Of course, some WT components fail earlier than expected and, because unscheduled downtime can be so costly [7], condition monitoring systems (CMS) are employed to improve WT

availability and reduce the O&M costs [8]. However there is a degree of uncertainty about the appropriateness of applying specic maintenance policies to WT components. This paper discusses the applicability of various maintenance strategies to WT condition monitoring, reviews the available techniques and methods in the literature, and presents a fault tree analysis (FTA) summarising the ways in which WTs can fail. The discussion focuses on the three blade up-wind variable speed turbine with double feed asynchronous generator which is the dominant type of WT currently in use worldwide [9]. 2. Wind turbines Most WT machines are three-blade units comprising the major components illustrated in Fig. 3 [10]. Driven by the wind, the blades and rotor transmit energy via the main shaft through the gearbox to the generator, the main shaft being supported by the bearings, and the gearbox being such that the generator speed is as near as possible to optimal for the generation of electricity. Alignment with the direction of the wind is controlled by a yaw system and the housing (or nacelle) is mounted at the top of a tower. Some defects such as leaking and corrosion can be detected by visual inspection; discolouration of component surfaces may indicate slight temperature variations or deteriorating condition, and the sound coming from the bearings can also indicate physical condition [11,12]. However, many of the most typical failures like cracking and roughness on the surfaces of the blades, electric short circuits in the generator, and overheating of the gearbox all demand a more sophisticated approach to maintenance.

* Corresponding author. E-mail addresses: Faustopedro.garcia@uclm.es (F.P. Garca Mrquez), a.m.tobias@bham.ac.uk (A.M. Tobias), jesusmaria.pinar@uclm.es (J.M. Pinar Prez), m.papaelias@bham.ac.uk (M. Papaelias).

0960-1481/$ e see front matter Crown Copyright 2012 Published by Elsevier Ltd. All rights reserved. doi:10.1016/j.renene.2012.03.003

170

F.P. Garca Mrquez et al. / Renewable Energy 46 (2012) 169e178

Fig. 1. Wind Energy: Global Capacity (green) and Forecast (red) [1]. (For interpretation of the references to colour in this gure legend, the reader is referred to the web version of this article.)

Fig. 3. Main parts of a turbine (Source: [11]) showing (1) blades, (2) rotor, (3) gearbox, (4) generator, (5) bearings, (6) yaw system and (7) tower.

3. Maintenance theory Maintenance is required to make sure that the components continue to perform the functions for which they were designed. The basic objectives of the maintenance activity are to deploy the minimum resources required to ensure that components perform their intended functions properly, to ensure system reliability and to recover from breakdowns [13]. 3.1. Corrective, scheduled and condition based maintenance Classical theory sees maintenance as either corrective or preventive. The former (also known as unscheduled or failure based maintenance) is carried out when turbines break down and when faults are detected or failures occur in any of the components. Immediate refurbishment or replacement of parts may be necessary [14] and unscheduled downtime will result. Corrective maintenance is therefore the most expensive of strategies and wind farm operators will hope to resort to it as little as possible. The various stages are shown in Fig. 4. By contrast, the objective behind preventive maintenance (PM) is to either repair or replace components before they fail as shown in Fig. 5 [14]. This has most straightforwardly been achieved by scheduled maintenance, also known as time based (or planned)

maintenance and involving repair or replacement at regular time intervals as recommended by the supplier and regardless of condition. Scheduled maintenance activities in WT include the changing of oil and lters, and the tightening and torqueing of bolts. But reducing failures in this way comes at the cost of completing maintenance tasks more frequently than absolutely necessary and not exhausting the full life of the various components already inservice. An alternative is to mitigate against major component failure and system breakdown with condition based maintenance (CBM) in which continuous monitoring and inspection techniques are employed to detect incipient faults early, and to determine any necessary maintenance tasks ahead of failure [15]. This involves acquisition, processing, analysis and interpretation of data and selection of optimal maintenance actions [16] and is achieved using condition monitoring systems [17e19]. CBM has been shown to minimise the costs of maintenance, improve operational safety, and reduce the quantity and severity of in-service system failures. Byon and Ding [20] have demonstrated its applicability to WTs, and McMillan and Ault [21] have used Monte Carlo simulation to evaluate its cost effectiveness when applied to WTs. CBM is now the most widely employed strategy in the WT industry. 3.2. Reliability centred maintenance The state-of-the-art way of deciding upon maintenance strategy in the WT industry is reliability centred maintenance (RCM), which has been formally dened as a process used to determine what must be done to ensure that any physical asset continues to do whatever its users want it to do in its present operating context [22]. It involves maintaining system functions, identifying failure modes, prioritizing functions, identifying PM requirements and selecting the most appropriate maintenance tasks [23] with the objective of managing system failure risk e((ffectively [24]; operational and maintenance policies are optimised such that the overall maintenance task is reduced [14]. RCM has been recognized and accepted in many industrial elds, such as steel plants [25], railway networks [26] and [27], ship maintenance and other industries [28]. RCM in the wind turbine industry is addressed by Andrawus et al. [29].

Fig. 2. Distribution of failure frequencies between different turbine types, sorted by turbine size [6].

Fig. 4. Stages of a corrective maintenance task.

F.P. Garca Mrquez et al. / Renewable Energy 46 (2012) 169e178

171

Fig. 5. Stages of a preventive maintenance task.

4. Condition monitoring of wind turbines On the basis that a signicant change is indicative of a developing failure [30], condition monitoring systems (CMS) [13] comprise combinations of sensors and signal processing equipment that provide continuous indications of component (and hence wind turbine) condition based on techniques including vibration analysis, acoustics, oil analysis, strain measurement and thermography. On WTs they are used to monitor the status of critical operating major components such as the blades, gearbox, generator, main bearings and tower. Monitoring may be on-line (and hence provide instantaneous feedback of condition) or off-line (data being collected at regular time intervals using measurement systems that are not integrated with the equipment) [31]. With good data acquisition and appropriate signal processing, faults can thus be detected while components are operational and appropriate actions can be planned in time to prevent damage or failure of components. Maintenance tasks can be planned and scheduled more efciently, resulting in increased reliability, availability, maintainability and safety (RAMS) whilst downtime, maintenance and operational costs are reduced [8]. CM techniques are thus used throughout the industry [32,33] and benets are especially shown for offshore wind farm(s) [34] because of not only the high costs of O&M at sea but also the typically larger turbines. Several techniques are available. 4.1. Vibration analysis Vibration analysis continues to be the most popular technology employed in WT, especially for rotating equipment [35]. Different sensors are required for different frequencies: position transducers are used for the low-frequency range, velocity sensors in the middle frequency area, accelerometers in the high frequency range and spectral emitted energy sensors for very high frequencies [36]. As for applications, it is appropriate for monitoring the gearbox [37e43], the bearings [11,12,44e46] and other selected WT elements as listed in Table 1; the sensor conguration in a nacelle is illustrated in [37]. Furthermore Tandon and Nakra [47] presented a detailed review of the different vibration and acoustic methods, such as vibration measurements in the time and frequency domains, sound measurements, the shock pulse method and the acoustic emission technique, for CM of rolling bearings [48]. 4.2. Acoustic emission Rapid release of strain energy takes place and elastic waves are generated when the structure of a metal is altered, and this can be analysed by acoustic emissions (AE). The primary sources of AE in WTs are the generation and propagation of cracks, and the technique has been found [49e51] to detect some faults earlier than others such as vibration analysis [52]. The measurement and interpretation of AE parameters for fault detection in radially loaded ball bearings has been demonstrated at different speed ranges in [53]. In addition, the application of AE for the detection of bearing failures has been presented in [54]. Acoustic monitoring has some similarities with vibration monitoring but whereas vibration sensors are mounted on the component involved [36] so as to detect movement, acoustic sensors are attached with

exible glue with low attenuation and record sound directly. AE sensors have been used successfully not only in the monitoring of bearings and gearboxes but also for damage detection in blades of a WT as discussed in [55]. Its application is also possible to an inservice WT for a real-time rotating blade [56,57]. Non-destructive testing techniques using acoustic waves to improve the safety of wind turbine blades are presented in [58], and to enable the assessment of the damage criticality for blades of small WT based on AE data in [59,60]. The use of AE is gradually growing for both CM of rotating WT components as well as blades. 4.3. Ultrasonic testing techniques Ultrasonic testing (UT) techniques are used extensively by the wind energy industry for the structural evaluation of wind turbine towers and blades. UT is generally employed for the detection and qualitative assessment of surface and subsurface structural defects [13,25,61]. Ultrasonic wave propagation characteristics allow estimation of the location and type of defect detected, thus providing a reliable method of determining the material properties of the principal turbine components. Signal-processing algorithms including time-frequency techniques and wavelet transforms can be used to extract more information [62e65]. An ultrasound technique to visualize the inner structure of wind turbine blades is presented in [66], the deployment of UT techniques for inspection of the whole multi-layered structure of the WT blade and to nd defects like delaminations, lack of glue, etc is illustrated in [67]; and the ultrasonic air-coupled technique has been used to research internal defects in wind turbine blades [68]. Ultrasonicallyobtained images make it possible to recognise the geometry of defects and to estimate their approximate dimensions. Jasiuniene et al. [69] adapted air-coupled ultrasonic technique for better identication of the shape and size of defects in a WT blade. A review of other methods can be found in [70]. 4.4. Oil analysis Whether for the ultimate purpose of guaranteeing oil quality or the condition of the various moving parts, oil analysis is mostly executed off-line by taking samples [35] despite on-line sensors having (for years) been available at an acceptable price level [35,36] for monitoring oil temperature, contamination and moisture [71]. Little or no vibration may be evident while faults are developing, but analysis of the oil can provide early warnings; a case study of a WT gearbox is described in [72]. In the case of excessive lter pollution, oil contamination or change in component properties, characterization of the particulates can give an indication of excessive wear [36]. Such approaches are particularly effective and cost-efcient in avoiding catastrophic failures [73,74]. On-line oil analysis is gradually becoming more important with several on-going pilot projects. 4.5. Strain measurement Strain measurement using strain gauges can be very useful for lifetime forecasting and protecting against high stress levels, especially in the blades. An assessment of strain gauge signal interpretation from strain gauge sensors installed on the blade has been performed in order to adjust calibration practices and sensor selection [75]. Optical bre sensors are still very expensive [35] but cost-effective systems based on bre optics are being developed. References [56,76,77] illustrate how load monitoring can be performed using strain sensors in the rotor blades. Strain measurement can be expected to grow in importance as an input to CM.

172

F.P. Garca Mrquez et al. / Renewable Energy 46 (2012) 169e178

Table 1 Condition monitoring techniques in WT, where. a: Statistical methods; b: Time domain analysis; c: Cepstrum analysis; d: Fast Fourier transformation(FFT); e: Amplitude demodulation; f: Wavelet transformation; g: Hidden Markov models; h: Novel techniques. Blades Vibration [56], a [134], f[135] Rotor Gearbox Generator Bearings Tower

Acoustic emission Ultrasonic techniques Oil analysis Strain Electrical effects Shock Pulse methods Process parameters Performance monitoring Radiographic inspections Thermography Other

c[37], c[108], a[113] d[10], d[32], a[37], c[38], b[39], f[40], b[41], e[42] bef[43], eb[104], c[107], f[119], f[125], fg[126,136], c[106], c[137] [56], a[59,88,140], a[141], f[143] a[95], f[125] a[55,57,58,60,142] f[63,140], a[67e70] b[72], a[95] [146] [56,75,88,147], c[77] [148] [79,81] a[94] a[87] [88] [69,91] [93e95] [95,150,151]

d[10], a[37], c[38], a[44], d[11], d[10] d[12] a[45], df[46], c[107], f[120], f[123], g[130], f[135], f[138], f[139] a[53], de[84], b[105,144], a[145]

a[74] [80] de[84], a[82], f[19,83]

c[149] h[87]

a[86] a[86], b[89] cd[149] h[87] [11,12,151]

4.6. Electrical effects CM of electrical equipment such as motors, generators and accumulators is typically performed using voltage and current analysis. Discharge measurements are used for medium and high voltage grids. A spectral analysis of the stator current [78] in the generator can be used for detecting isolation faults in the cabling without inuencing WT operation. Electrical resistance can also be used for the structural evaluation of certain WT components. Electrical resistance varies with stiffness and abrupt changes can be used to detect cracks, delaminations, and fatigue. Hence, the technique can be applied to in-service WTs. References [79e81] demonstrate how the resistance principle is useful for detecting fatigue damage in particular. These techniques are at the moment conned to research-related activities but there is signicant potential for applying them successfully in the eld. 4.7. Shock pulse method (SPM) The SPM has been used as a quantitative method for condition monitoring of bearings, and works by detecting the mechanical shocks that are generated when a ball or roller in a bearing comes in contact with a damaged area of raceway or with debris [82]. Signals are picked up by transducers, and analysis - e.g. using a normalized shock value [83]- gives an indication of system condition [84]. Low frequency signals of vibration collected in the nacelle and caused by other sources can easily be ltered electronically. A case study of SPM with a piezoelectric transducer is described in [85]. The method is occasionally used by the industry to support vibration measurements. 4.8. Process parameters Maintenance based on process parameters and the detection of signals exceeding predened control limits is common practice in WTs, control systems becoming increasingly sophisticated and diagnostic capabilities ever better. Transient and oscillatory stability were analysed with different wind scenarios for electricity generation process in [86]. For explanation of the use of signals and trending for fault detection based on parameter estimation, see [87]. 4.9. Performance monitoring The relationship between parameters such as power, wind speed, blade angle and rotor speed can also be used for an

assessment of WT condition and for the early detection of faults [88]. Previous work includes power and voltage icker analysis with variable wind speed and turbulence [86]. Similar to estimation of process parameter(s), more sophisticated methods, including trending, are not often used [36]. The torque and power generated with wind time series taken in the eld from an anemometer is considered in [89]. 4.10. Radiographic inspection Radiographic imaging of critical structural turbine components using X-rays is only rarely used although it does provide useful information regarding the structural condition of the component being inspected. Radiographic imaging depends on the different level of absorption of X-ray photons as they pass through a material [70]. So as to detect tight delaminations or cracks, having gaps less than 50 mm, the backscatter X-ray imaging technique [70] is employed. X-Ray imaging is useful to locate the internal defects of the WT, and the main advantage of X-ray inspection is the accuracy of this technique [90]. A transportable radiographic system for WT blades has been recently demonstrated as a solution to nd defects and reduce the cost of inspection [91]. 4.11. Thermography Thermography is often used for monitoring electronic and electric components and identifying failure [92]. The technique is only applied off-line, and often involves visual interpretation of hot spots that arise due to bad contact or a system fault. At present the technique is not particularly well-established for online CM, but cameras and diagnostic software that are suitable for on-line process monitoring are starting to become available. Infrared cameras have been used to visualize variations in blade surface temperature [93e95] and can effectively indicate cracks as well as places threatened by damage [96]. In the longer term, this might be applicable to WT generators and power electronics too. Pulsed thermography can be employed for the structural evaluation of blades but due to the bulky equipment involved this is not a standard methodology amongst wind turbine operators. Early investigations of carrying out thermographic measurements of inservice blades using helicopters to deploy the IR cameras has not yet been proven satisfactorily and faces serious difculties of implementation.

F.P. Garca Mrquez et al. / Renewable Energy 46 (2012) 169e178

173

5. Sensory signals and signal processing methods Regardless of the technique, the capability of a CMS relies upon two basic elements: the number and type of sensors, and the associated signal processing and simplication methods utilized to extract important information from the various signals. An electronic measurement system will acquire the data, and then process and distribute them to an observer or other technical control system. Data acquisition will involve measuring the required variables (e.g. current, voltage, temperature, speed) and turning them into electronic signals. To do so effectively will involve judicious choice and placement of the right type and number of sensors; conditioning (performing basic operations including amplication, ltering, linearization, and nally modulation/demodulation) may be necessary to reduce the susceptibility of the signals to interference. Optimization techniques may then be employed [97] in the processing of the signals by a digital signal processor (DSP), involving not only the processing itself but also sorting and manipulation as necessary. Subsequent distribution will be to either a screen, computer, storage device or other system. There are several options, including e.g. Ethernet networks with TCP/IP protocol together with WLAN for a WT for communicating with either a Farm Server or a supervisory command and data acquisition (SCADA) system. The latter is a particular computer-based system that allows local and remote control of the functions of a WT, gathering data from the wind farm and analysing them in order to report operational performance, and hence ensure efcient operation. SCADA uses various signal processing methods, the most relevant to WT being those covered below and in Table 1. 5.1. Statistical methods One common application of statistical algorithms for the purposes of CM is to analyse the data signals from the various sensors in WTs. Common statistical measures such as root mean square (RMS) and peak amplitude are widely used for the diagnosis of failures but more advanced features are also being developed [37,74,82,86,98,99]. Other important statistical parameters are the maximum value, minimum value, mean, peak to peak, standard deviation, shape factor, crest factor, impulse factor, denite integral, energy ratio and kurtosis. 5.2. Trend analysis When applied to WT, trend analysis refers to the concept of collecting data from the various sensors and looking for trends. It requires particular algorithms [37,100,101], and applications include the monitoring of pitch mechanisms although most common is its use on power output patterns from WT generators. Note that trend estimation is a different technique specically for forecasting. 5.3. Filtering methods Sometimes there are redundant data that contain information that is not useful and must be eliminated so as to avoid compromising the computations. For example, the vibrations from a nacelle will have to be ltered whilst measuring the vibration of the gearbox inside [84]. A method using least median squares that lters in parallel with estimating the power curve is presented in [102], and ltering with and without a classical statistical method (based on standard deviation) is described and compared in [103]. The drawback of ltering whilst monitoring trends is that the

parameters have to be adjusted to take account of varying operating conditions. 5.4. Time-domain analysis Analysis in the time domain is a way of monitoring WT faults like resistive and inductive imbalances between the rotor and the stator phases, and turn-to-turn faults in the rotor windings of the generator. Variations in current signals and trends are typically used for vibration analysis [39,41,104], oil analysis [72] and AE [105]. 5.5. Cepstrum analysis The power cepstrum is a time based approach dened as the inverse Fourier transform of the logarithmic power spectrum [106]. The cepstrum is well suited for applications to equipment diagnostics in a WT. Gear vibration spectra commonly show sidebands of meshing frequency and its harmonics arising from the modulation of tooth meshing waveform [38,107]. For gearboxes in good condition, the sideband level generally remains constant with time. Therefore, changes in the number and amplitude of the sidebands normally indicate deterioration. The presence of several families of sidebands and other components can complicate the distinction and evaluation of the sideband spacing [108]. 5.6. Time synchronous averaging (TSA) TSA, also called time domain averaging, is a signal processing technique that serves as the basis for many gear fault detection algorithms. The method is for identifying a rotating bearing defect by measuring the vibration of a rotating bearing and obtaining a waveform signal. It may also help in identifying the source of vibration in WT gearboxes [109]. For instance, a cracked gear tooth in a WT gearbox that meshes once per revolution produces a highly periodic vibration signature which can be very weak. TSA can be used to highlight the vibration signal features taking place over a given period, and can work with non-periodic signal components, ltering the noise of the time-domain signal. Once the average time signal is achieved, it is possible to compute the FFT (as below). A review of other TSA algorithms is presented in [110]. 5.7. Fast-Fourier transform (FFT) The FFT [111] algorithm is used for the conversion of a digital signal from the time domain into one in the frequency domain. Particular frequency ranges correspond to particular states (e.g. fault-free, defective bearing), the ranges reecting the rotational speed of the main shaft and the shape and size of the element concerned. All bearing elements generate vibration at specic frequencies (referred to as fault frequencies) and hence FFT nds most usage in gearbox monitoring [10,32]. FFT is also used for bearings [10,11] wherein, if damage starts to develop, the shape of the vibration distribution deviates from the nominal Gauss-shaped curve [11,12]. The advantage of frequency domain analysis over time domain analysis is its ability to identify and isolate certain frequency components of interest [112]. 5.8. Amplitude demodulation This approach can extract very low-amplitude and low-frequency periodic signals that might be masked by other higher energy vibrations as in WT gearboxes [43,104]. While the raw spectrum can be useful for monitoring gear mesh frequencies, the envelope spectrum provides superior sensitivity to bearing defect frequencies in WT applications [84]. With its high sensitivity, demodulation has

174

F.P. Garca Mrquez et al. / Renewable Energy 46 (2012) 169e178

been proven to be good for evaluating defects that produce impacting, e.g. rolling contacts in bearings and tooth-to-tooth contacts in the gear meshes [42]. It also helps to reduce the complexity of the analysis, the main advantage being that it provides excellent visibility of bearing defects frequencies without the interference of gear mesh frequencies in the same spectrum. Despite the advantages for WT gearboxes, some barriers to implementation remain. The selection of transducer location is crucial to guarantee the results and, if the demodulation process is based on low-pass ltering, the original amplitude of the original defect signals is not preserved. 5.9. Order analysis FFT is useful for studying oscillations in constant speed wind energy converters (WEC) when applied to time series signals. However, this algorithm is not suitable for variable speed WEC where a different algorithm based on rotational angles is required, i.e. so-called order analysis, particularly well-suited for rotor imbalances and aerodynamic asymmetries [38]. Torsional oscillation of the nacelle generates a signal which is out of phase with any transverse oscillations, so they can be separated and analysed individually [113]. Interpolation and production of the order spectrum then leads to the computation of the FFT. Such analysis may be used for monitoring the overall rotor condition including surface roughness, mass imbalance and aerodynamic asymmetry. 5.10. Wavelet transforms Wavelet transformation is a time-frequency technique similar to short time Fourier transform (STFT) but more appropriate for

non-stationary signals. It provides a time-frequency 3D map of the signal being analysed in [114,115] and involves decomposing it into a set of sub-signals or levels with different frequencies [116,117]. It is applied to WTs in order to monitor the vibration level caused by misalignment, bearing and other problems, and can be used as a general sign or indication of a faulty WT. Wavelet transforms have been applied to waveform data analysis in fault detection and diagnostics of various WT parts including gears [118,119], bearings [120,121], and other mechanical systems [122,123]. Assessment of the effectiveness and reliability of wavelet transforms and comparison with other vibration analysis techniques has been completed by Dalpiaz and Rivola [124]. Baydar and Ball [125] successfully applied wavelet transforms to vibration signals and acoustic signals [40,43,126]. FFT (and indeed RMS) have also been used to estimate the maximum amplitude of the wavelet coefcients which helps in discriminating between normal and damaged components [127]. 5.11. Hidden Markov models Hidden Markov models (HMM) have successfully been applied to the classication of patterns in trend analysis [128] and CM [126]. Atlas et al. developed a method to predict wear accurately [129], when applying a method for monitoring of milling processes with HMM. Ocak and Loparo [130] presented the application of HMM in bearing fault detection [126], and the dynamic statistical characteristics that exist in the current observations of vibration signals in the machine have been modelled using HMM utilizing a Markov chain.

Fig. 6. Fault tree for WT.

F.P. Garca Mrquez et al. / Renewable Energy 46 (2012) 169e178

175

5.12. Novel techniques Fault detection and diagnosis (FDD) is a sophisticated adaptation of CMS that incorporates intelligent algorithms suitable for early detection of incipient faults providing an insight into the corresponding level of criticality [19]. FDD methods can be modelbased or non-model based depending upon the way process knowledge is incorporated within the signal processing unit; a more specic classication is shown in [38]. Articial intelligence (AI) is essentially employed to reproduce human reasoning as accurately as possible, the reasoning process being based on the behaviour of the system, and written down in terms of rules. The dynamic nature of the environments in which WTs operate has led to the emergence of predictive maintenance plans that take qualitative account of the environment as well as its actual effect on the condition of the components. Expert systems or so-called rule-based diagnostic systems detect and identify incipient faults in accordance with the rules representing the relation of each possible fault with the actual monitored equipment condition [131]. They have to full two capabilities so as to be effective: the acquisition and integration of new knowledge, and also the explanation of their reasoning [132]. 6. Fault tree analysis (FTA) For detailed CM study, identication of potentially hazardous events and an assessment of their consequences and frequency of occurrence are necessary. One of the most popular approaches for this purpose is FTA [133], a so-called fault tree (FT) being a diagrammatic way of describing the complete set of possible causes that can lead to failure. FTA thus allows identifying and then quantifying the initiating failure causes that will help set the stage for developing a PM program t to maintain system reliability at the required level with particular attention given to aggressive environmental factors. The process involves two stages: selection of the various components that are to be considered for analysis, and then assessment of how the component states affect the condition variables that will be measured by the CM system. Both are critical and have a signicant impact on the accuracy of the model. Here, the FT reproduced in Fig. 6 was constructed starting with the top event: wind turbine failure. For each of the same components as highlighted in Fig. 3 in turn, this was then successively broken down via the various possible intermediate failures using AND and OR logic gates so as nally to yield the basic failure events at the lowest levels e.g. corrosion of pins, pitting, deformation of the outer bearing race and other rolling elements, or indeed abrasive wear [35]. 7. Conclusions The primary focus of this review of CM and the various mathematical methods for signal processing is upon WT gearboxes and bearings, rotors and blades, generators and power electronics, rather than system-wide turbine diagnosis. An inventory of the available CM techniques along with signal processing algorithms has been provided and selection of a set of techniques which is feasible and better suited for WTs has been made possible. The review is summarised in Table 1, which may be read from the viewpoint of either techniques (row-wise) or components (column-wise). The references of the different methods used by the researchers in the components of WT and techniques for CM are collected in the next table. The methods reported in the literature are indicated by the letters. For each element of the WT there are different techniques that can be employed, and for all of these techniques there are

mathematical methods available and referenced in the literature. The main obstacles facing the designers of condition monitoring systems for wind turbines clearly continue to be: i. selection of the number and type of sensors; ii. selection of effective signal processing methods associated with the selected sensors; and iii. design of an effective fusion model (i.e., the combination of sensors and signal processing methods which give an improved performance); Acknowledgement This work has been supported by the European Project Development and Demonstration of a Novel Integrated Condition Monitoring System for Wind Turbines (NIMO, Ref.:FP7-ENERGY2008-TREN-1: 239462, www.nimoproject.eu). Acronyms

AE AI CBM CM CMS FDD FFT FT FTA HMM O&M PM RMS RCM SCADA SPM TSA UT WEC WT

acoustic emission articial intelligence condition based maintenance condition monitoring condition monitoring system fault detection and diagnosis fast Fourier transform fault tree fault tree analysis hidden Markov model operation and maintenance preventive maintenance root mean square reliability centred maintenance supervisory command and data acquisition shock pulse method time synchronous averaging ultrasonic testing wind energy converter wind turbine

References
[1] WWEA. World wind energy report 2009, http://www.wwindea.org; 2009. [2] McGowan JG, Connors SR. Windpower: a turn of the century review. Annual Review of Energy and the Environment 2000;25:147e97. [3] Milborrow D. Operation and maintenance costs compared and revealed. Wind Stats 2006;19(3):3. [4] Vachon W. Long-term O&M costs of wind turbines based on failure rates and repair costs. In: Proceedings WINDPOWER, American wind energy Association Annual conference. Portland: Oregon; 2002. p. 2e5. June. [5] Walford C. Wind turbine reliability: understanding and minimizing wind turbine operation and maintenance costs. Sandia report, SAND2006-1100. Albuquerque, New Mexico 87185 and Livermore, California 94550: Sandia National Laboratories; 2006. [6] Tavner PJ, Spinato F, van Bussel GJW, Koutoulakos E. Reliability of different wind turbine concepts with relevance to offshore application. Brussels: European Wind Energy Conference; April 2008. [7] Anon. Managing the wind: reducing Kilowatt-Hour costs with condition monitoring. Refocus 2005;6(3):48e51. [8] Yang WX, Tavner PJ, Crabtree CJ, Wilkinson M. Cost effective condition monitoring for wind turbines. IEE Transactions on Industrial Electronics 2010;57(1). [9] Wind Power in the United States: Technology, Economic, and Policy Issues http://www.fas.org/sgp/crs/misc/RL34546.pdf (02-09-09). [10] de Novaes Pires G, Alencar E, Kraj A. Remote conditioning monitoring system for a hybrid wind diesel system-application at Fernando de Naronha Island, Brasil. http://www.ontario-sea.org (19-07-10).

176

F.P. Garca Mrquez et al. / Renewable Energy 46 (2012) 169e178 [43] Miller AJ, 1999. A New wavelet basis for the decomposition of gear motion error signals and its application to gearbox diagnostics, Master of Science Thesis, The Pennsylvania State University. [44] Dyer D, Stewart RM. Detection of rolling element bearing damage by statistical vibration analysis. ASME Journal of Mechanical Design 1978;100: 229e35. [45] Sun Q, Tang Y. Singularity analysis using continuous wavelet transform for bearing fault diagnosis. Mechanical Systems and Signal Processing 2002; 16(6):1025e41. [46] Tse PW, Peng YH, Yam R. Wavelet analysis and envelope detection for rolling element bearing fault diagnosis-their effectiveness and Flexibilities. ASME Journal of Vibration and Acoustics 2001;123:303e10. [47] Tandon N, Nakra BC. Vibration and acoustic monitoring techniques for the detection of defects in rolling element bearings e a review. Shock and Vibration Digest 1992;24(3):3e11. [48] Tandon N, Choudhury A. A review of the vibration and acoustic measurement methods for detection of defects in rolling element bearings. Tribology International 1999;32(8):469e80. [49] Yoshioka T. Detection of rolling contact subsurface fatigue cracks using acoustic emission technique. Lubrication Engineering 1992;49(4):303e8. [50] Yoshioka T, Fuhjiwara T. Application of acoustic emission technique to detection of rolling element-bearing failure. In: Dorneld DA, editor. Acoustic Emission Monitoring and Analysis in Manufacturing. New York: ASME; 1984. p. 55e75. [51] Yoshioka T, Takeda M. Classication of rolling contact fatigue initiation using acoustic emission technique. Lubrication Engineering 1994;51(1):41e4. [52] Tandon N, Yadava GS, Ramakrishna KM. A comparison of some condition monitoring techniques for the detection of defect in induction motor ball bearings. Mechanical Systems and Signal Processing 2007;21(1):244e56. [53] Tandon N, Nakra BC. Defect detection in rolling element bearings by acoustic emission method. Journal of Acoustic Emission 1990;9(1):25e8. [54] Tan CC. Application of acoustic emission to the detection of bearing failures. In: Proceedings Tribology Conference, Brisbane; 1990. p. 110e4. [55] Wei J, McCarty J. Acoustic emission evaluation of composite wind turbine blades during fatigue testing. Wind Engineering 1993;17(No. 6):266e74. [56] Giebel G, Oliver G, Malcolm M, Kaj B. Common Access to wind turbines data for condition monitoring. Riso National Laboratory. In: Proceedings of the 27th Riso International Symposium on Material Science, Denmark; 2006. p. 157e64. [57] Blanch MJ, Dutton AG. Acoustic emission monitoring of eld Tests of an operating wind turbine. Key Engineering Materials 2003;245-246:475e82. [58] Jungert A. Damage detection in wind turbine blades using two different acoustic techniques. The e-Journal of Nondestructive Testing; 2008:25. [59] Anastassopoulos AA, Kouroussis DA, Nikolaidis VN, Proust A, Dutton AG, Blanch MJ, et al. Structural integrity evaluation of wind turbines blades using pattern recognition analysis on acoustic emission data. Journal of Acoustic Emission 2002;20:229e37. [60] Joosse PA, Blanch MJ, Dutton AG, Kouroussis DA, Philippidis TP, Vionis PS. Acoustic emission monitoring of small wind turbine blades. Journal of Solar energy Engineering 2002;124(4):446e55. [61] Endrenyi J, McCauley J, Singh C. The present status of maintenance strategies and the impact of maintenance on reliability. IEEE Transaction Power System 2001;16(4):638e46. [62] Castaings M, Cawley P. The generation, propagation and detection of Lamb waves in plates using air-coupled ultrasonic transducers. Journal of Acoustic Society of America 1996;100(5):3070e7. [63] Castaings M, Cawley P, Farlow R, Hayward G. Single Sided inspection of composite materials using air coupled ultrasound. Journal of Nondestructive Evaluation 1998;17(1):37e45. [64] Raghavan A, Carlos E, Cesnik S. Review of guided-wave structural health monitoring. The Shock and Vibration Digest 2007;39:91e114. [65] Tsai CS, Hsieh CT, Huang SJ. Enhancement of damage detection of wind turbines. IEEE Transactions on Energy Conversion 2006;21(3):776e81. [66] Juengert A, Grosse CU. Inspection techniques for wind turbine bladesusing ultrasound and sound waves. In: Proceedings of non-destructive testing in Civil Engineering, Nantes, France, 2009; 2009.   niene _ E, Zukauskas [67] Rai sutis R, Jasiu E. Ultrasonic NDT of wind turbines blades using guided waves. Ultrasound, vol. 63. Kaunas: Technologija; 2008. 7e11.   niene _ E, Rai [68] Jasiu sutis R, Sliteris R, Volei sis A, Jakas M. Ultrasonic NDT of wind turbine blades using contact pulse-echo immersion testing with moving water container. Ultragarsas (Ultrasound) 2008;63(3):28e32. [69] Jasiuniene E, Raisutis R, Sliteris R, Voleisis A, Vladisauskas A, Mitchard D, et al. NDT of wind turbine blades using adapted ultrasonic and radiographic techniques. Insight- Non-Destructive Testing and Condition Monitoring 2009;51(9):477e83.   niene _ E, Sliteris [70] Rai sutis R, Jasiu R, Vladi sauskas A. The review of nondestructive testing techniques suitable for inspection of the wind turbines blades. Ultragarsas (Ultrasound) 2008;63(1):26e30. [71] Yin YH, Wang WH, Yan XP, Xiao HL, Wang CT. An integrated online oil analysis method for condition monitoring. Measurement Science and Technology 2003;14(11):1973e7. [72] Leske S, Kitaljevich D. Managing gearbox failure, Dewek. Dewi Magazine; 2006. N. 29. [73] Barron T. Engineering condition monitoring. Addison Wesley Longman; 1996. [74] Toms LA. Machinery oil analysis: methods, automation and benets. Coastal; 1998.

[11] Igarashi T, Hamada H. Studies on the vibration and sound of defective roller bearings (First report: vibration of ball bearing with one defect), Bull. JSME June 1982;25(204):994e1001. [12] Igarashi T, Yabe S. Studies on the vibration and sound of defective roller bearings (First report: sound of ball bearing with one defect), Bull. JSME 1983;26(220):1791e8. [13] Knezevic J. Reliability, maintainability and supportability engineering: a probabilistic approach. McGraw Hill; 1993. [14] Ben-Daya MS, Duffuaa AR. Handbook of maintenance management and engineering. Springer Verlag London Limited; 2009. [15] Pedregal DJ, Garcia Marquez FP, Roberts C. An algorithmic approach for maintenance management. Annals of Operations Research 2009;166: 109e24. [16] Campbell JD, Jardine AKS. Maintenance excellence: optimizing equipment life-cycle decisions. New York: Marcel Dekker; 2001. [17] Garcia Marquez FP. An approach to remote condition monitoring systems management. The IET International Conference on Railway Condition Monitoring; 2006:156e60. [18] Garcia Marquez FP, Pedregal DJ, Roberts C. Time series methods applied to failure prediction and detection. Reliability Engineering & System Safety 2010;95(6):698e703. [19] Garcia Marquez FP, Roberts C, Tobias A. Railway point mechanisms: condition monitoring and fault detection. In: Proceedings of the Institution of mechanical Engineers, Part F, Journal of Rail and Rapid Transit, vol. 224(1). Professional Engineering Publishing; 2010. p. 35e44. [20] Byon E, Ding Y. Season-dependent condition-based maintenance for a wind turbine using a partially observed markov decision process. IEEE Transactions on Power Systems 2010;25(4):1823e34. [21] McMillan D, Ault GW. Condition monitoring benet for onshore wind turbines: sensitivity to operational parameters. IET Renewable Power Generation 2008;2(1):60e72. [22] Moubray J. Reliability-centered maintenance. New York: Industrial Press; 1997. [23] Smith AM. Reliability-centred maintenance. New York: McGraw-Hill, Inc; 1993. [24] Garcia Marquez FP, Pedregal DJ. Applied RCM2 algorithms based on statistical methods. International Journal of Automation and Computing; 4:109e116. [25] Deshpande VS, Modak JP. Application of RCM for safety considerations in a steel plant. Reliability Engineering and System Safety 2002;3(78):325e34. [26] Garcia Marquez FP, Schmid F, Collado JC. A reliability centered approach to remote condition monitoring. A railway points case study. Reliability Engineering & System Safety 2003;80(1):33e40. [27] Garcia Marquez FP, Schmid F, Collado JC. Wear assessment employing remote condition monitoring: a case study. Wear 2003;2(255):1209e20. [28] Deshpande VS, Modak JP. Application of RCM to a medium scale industry. Reliability Engineering and System Safety 2002;1(77):31e43. [29] Andrawus JA, Watson J, Kishk M, Adam A. The selection of suitable maintenance strategy for wind turbines. Wind Engineering 2009;30(6):471e86. [30] Wiggelinkhuizen E, Verbruggen T, Xiang J, Watson SJ, Giebel G, Norton E, et al. CONMOW: condition monitoring for offshore wind farms. In: Proceedings of the 2007 EWEA European Wind Energy Conference (EWEC2007), Milan, Italy; 7e10 May 2007. [31] Scarf PA. A framework for condition monitoring and condition based maintenance. Quality Technology &. Quantitative Management 2007;4(2):301e12. [32] Cozens NJ, Watson SJ. State of the art condition monitoring techniques suitable for wind turbines and wind farm applications, report for CONMOW project; 2003. [33] Watson SJ, Ineld DG, Xiang J. Condition monitoring of wind turbines e Measurements and methods. IET Renewable Power Generation; 2008. [34] Nilsson J, Bertling L. Maintenance management of wind power systems using condition monitoring systems e Life Cycle costs analysis for two case Studies. IEE Transactions on Energy Conversion 2007;22(1). [35] Hameed Z, Hong YS, Choa YM, Ahn SH, Song CK. Condition monitoring and fault detection of wind turbines and related algorithms: a review. Renewable and Sustainable Energy Reviews 2009;13:1e39. [36] Verbruggen TW. Wind turbine operation & maintenance based on condition monitoring, WT-O. Final report, ECN-Ce03e047; April 2003. [37] Caselitz P, Giebhardt J. Fault prediction techniques for offshore wind farm maintenance and repair strategies. In: Proceedings of the EWEC2003; 2003. [38] Caselitz P, Giebhardt J, Mevenkamp M. On-line fault detection and prediction in wind energy converters. In: Proceedings of EWEC94, Thessaloniki, Greece; 1994. p. 623e7. [39] Futter DN. Vibration monitoring of industrial gearboxes using time domain averaging. In: IMechE 2nd International conference on gearbox noise, vibration and diagnostics. Mechanical Engineering Publications; 1995. [40] Lin ST, McFadden PD. Vibration analysis of gearboxes by the linear wavelet transform. In: IMechE 2nd International conference on gearbox noise, vibration and diagnostics. Mechanical Engineering Publications; 1995. [41] McFadden PD. Examination of a technique for the early detection of failure in gears by signal processing of the time domain average of the meshing vibration. Mechanical Systems and Signal Processing 1987;1(2):173e83. [42] McFadden PD. Detecting fatigue cracks in gears by amplitude and phase demodulation of the meshing vibration, Journal of vibration, acoustics. Stress and Reliability in Design 1996;108:165e70.

F.P. Garca Mrquez et al. / Renewable Energy 46 (2012) 169e178 [75] Moradakis E, Papadopoulos K, Philippidis TP. Assessment of the strain gauge technique for measurement of wind turbine blade loads. Wind Energy 2000;3(1):35e65. [76] Schroeder K, Ecke W, Apitz J, Lembke E, Lenschow G. A bre Bragg grating sensor system monitors operational load in a wind turbine rotor blade. Measurement Science and Technology 2006;17(5):1167e72. [77] Wernicke J, Shadden J, Kuhnt S, Byars R, Rhead P, Damaschke. Field experience of bre optical strain sensors for providing real time load information from wind turbine blades during operation. In: Paper presented at the European wind energy conference, 22e25 November, London, UK; 2004. [78] Schoen RR, Lin BK, Habetler TG, Schlag JH, Farag S. An unsupervised, on-line system for induction motor fault detection using stator current monitoring. IEEE Transactions on Industry Applications 1995;31(6):1280e6. [79] Seo DC, Lee JJ. Damage detection of CFRP laminates using electrical resistance measurement and neural network. Composite Structures 1999;47:525e30. [80] Todoroki A, Tanaka Y. Delamination identication of cross-ply graphiteeepoxy composite beams using electric resistance change method. Composites Science and Technology 2002;62:629e39. [81] Matsuzaki R, Todoroki A. Wireless detection of internal delamination cracks in CFRP laminates using oscillating frequency changes. Composites Science and Technology 2006;66:407e16. [82] Butler DE. The shock pulse method for the detection of damaged rolling bearings. NDT International 1973;6(2):92e5. [83] Zhen L, Zhengjia H, Yanyang Z, Xuefeng C. Bearing condition monitoring based on shock pulse method and improved redundant lifting scheme. Mathematics and Compuers in Simulation ;79(3):318e38. [84] Tandon N, Nakra BC. Comparison of vibration and acoustic measurement techniques for the condition monitoring of rolling element bearings. Tribology International 1992;25(3):205e12. [85] Morando IE. Measuring shock pulses is Ideal for bearing condition monitoring. Pulp and Paper 1988;62(12):96e8. [86] Mller H, Pller M, Basteck A, Tilscher M, Pster J. Grid compatibility of variable speed wind turbines with directly coupled synchronous generator and hydro-dynamically controlled gearbox. In: Sixth International Workshop on Large-Scale integration of wind power and transmission networks for offshore wind farms, Delft, NL; 2006. p. 307e15. [87] Zaher AS, McArthur SDJ. A multi-agent fault detection system for wind turbines defect recognition and diagnosis. In: Proceedings IEEE Lausanne POWERTECH 2007; 2007. p. 22e7. [88] Sorensen BF, Lading L, Sendrup P, McGugan M, Debel CP, Kristensen OJD, et al. Fundamentals for remote structural health monitoring of wind turbines blades - A Preproject. Riso-R-1336(EN); 2002. [89] Dale SL, Dolan, Lehn PW. Real-time wind turbine emulator suitable for power quality and dynamic control Studies. In: Proceedings of international conference on power systems transients, Montreal, Canada; 2005. p. 1e6. [90] Peters ST. Handbook of composites. London: Chapman & Hall; 1998. 839e855. [91] Fantidis JG, Potolias C, Bandekas DV. Wind turbine blade nondestructive testing with a transportable Radiography system. Science and Technology of Nuclear Installations; 2011. ID 347320. [92] Smith BM. Condition monitoring by thermography. NDT International 1978; 11(3):121e2. [93] Rumsey MA, Musial W. Application of infrared thermography nondestructive testing during wind turbine blade Tests. Journal of Solar Energy Engineering 2001;123(4):271. [94] Smith GM, Clayton BR, Dutton AG, Irving AD. Infra-red thermography for condition monitoring of composite wind turbine blades feasibility studies using cyclic loading tests. In: 15th British Wind Energy Association Conference (BWEA15), York, UK; 1993. [95] Beattie AG, Rumsey M. Non-destructive evaluation of wind turbine blades using an Infrared camera. In: ASME wind energy Symposium, 18th, Aerospace Sciences Meeting and Exhibit, 37th, Reno, NV; 1999.  ski L, Krawczuk M. Damage detection in turbine wind blades by [96] Dolin vibration based methods. In: 7th International conference on Modern practice in stress and vibration analysis; 2009. Journal of Physics: Conference series 181 012086. [97] Levitin G. Genetic algorithms in reliability engineering. Reliability Engineering and System Safety 2006;91(9):975e6. [98] Thiringer T. Power quality measurements performed on a low-voltage grid equipped with two wind turbines. IEEE Transactions on Energy Conversion 1996;11:601e6. [99] Tana CK, Irvinga P, Mba D. A comparative experimental study on the diagnostic and prognostic capabilities of acoustics emission, vibration and spectrometric oil analysis for spur gears. Mechanical Systems and Signal Processing 2007;21:208e33. [100] Garcia Marquez FP, Pedregal DJ. Failure analysis and diagnostics for railway trackside equipment. Engineering Failure Analysis 2007;14(8):1411e26. [101] Garcia Marquez FP, Pedregal DJ, Schmid F. Unobserved component models applied to the assessment of wear in railway points: a case study. European Journal of Operational Research 2007;176:1702e3. [102] Llombart-Estopia A. Improving the operation and maintenance of wind farms: determination of wind turbines performance. In: Proceedings of ICREPQ; 2008. [103] Giebhardt J. Condition monitoring for wind turbines; 2007. State of the Art Report As Deliverable 7.1.1, WP7 Condition monitoring. UPWIND.

177

[104] Cheng J, Yang Y, Yu D. The envelope order spectrum based on generalized demodulation timeefrequency analysis and its application to gear fault diagnosis. Mechanical Systems and Signal Processing 2010;24:508e21. [105] Li CJ, Li SY. Acoustic emission analysis for bearing condition monitoring. Wear 1995;185:64e74. [106] Wismer NJ. Gearbox analysis using cepstrum analysis and comb liftering; 1994. Application Note Brel & Kjaer, Denmark. [107] Caselitz P, Giebhardt J, Mevenkamp M, Reichardt M. Application of condition monitoring systems in wind energy converters. In: Proceedings of EWEC97, Dublin; 1997. p. 579e82. [108] Zheng GT, Wang WJ. A new cepstral analysis procedure of recovering excitations for transient components of vibration signals and applications to rotating machinery condition monitoring. ASME Journal of Vibration and Acoustics 2001;123:222e9. [109] Combet L, Gelman L. An automated methodology for performing time synchronous averaging of a gearbox signal without speed sensor. Mechanical Systems and Signal Processing 2007;21(6):2590e606. [110] Bechhoefer E, Kingsley M. A review of time synchronous average algorithms. In: Annual conference of the prognostics and Health management Society, San Diego, CA; 2009. [111] Burrus CS, McClellan JH, Oppenheim AV, Parks TW, Schafer RW, Schuessler HW. Computer-based exercises for signal processing using MATLAB. Englewood Cliffs, NJ: Prentice-Hall; 1994. 43e59. [112] Jardine AKS, Lin D, Banjevic D. A review on machinery diagnostics and prognostics implementing condition-based maintenance. Mechanical Systems and Signal Processing 2006;20(7):1483e510. [113] Caselitz P, Giebhardt J. Rotor condition monitoring for improved operational safety of offshore wind energy converters. Transactions of the ASME. Journal of Solar Energy Engineering 2005;127(2):253e61. [114] Yang WX, Tavner PJ, Wilkinson M. Condition monitoring and fault diagnosis of a WT with a synchronous generator using wavelet transforms. In: The 4th International conference on power electronics, machines and Drives 2008 (PEMD2008), York, UK; 2008. p. 6e10. [115] Abbate A, DeCusatis CM, Das P. Wavelets and Subbands: fundamentals and applications. Boston: Birkhauser; . 101e503. [116] Newland DE. An introduction to random vibrations, spectral and wavelet analysis. New York: Longman; 1993. [117] Strang G, Nguyen T. Wavelets and lter banks. Wellesley-Cambridge Press; 1997. [118] Staszewski WJ, Tomlinson GR. Application of the wavelet transform to fault detection in a Spur gear. Mechanical Systems and Signal Processing 1994;8: 289e307. [119] Wang WJ, McFadden PD. Application of wavelets to gearbox vibration signals for fault detection. Journal of Sound and Vibration 1996;192:927e39. [120] Luo GY, Osypiw D, Irle M. On-Line vibration analysis with fast continuous wavelet algorithm for condition monitoring of bearing. Journal of Vibration and Control 2003;9:931e47. [121] Rubini R, Meneghetti U. Application of the envelope and wavelet transform analyses for the diagnosis of incipient faults in ball bearings. Mechanical Systems and Signal Processing 2001;15:287e302. [122] Aretakis N, Mathioudakis K. Wavelet analysis for Gas turbine fault diagnostics. Journal of Engineering for Gas Turbines and Power 1997;119: 870e6. [123] Peng ZK, Chu FL, Peter, Tse W. Singularity analysis of the vibration signals by means of wavelet Modulus Maximal method. Mechanical Systems and Signal Processing 2007;Vol. 21(2):780e94. [124] Dalpiaz G, Rivola A. Condition monitoring and diagnostics in automatic machines: comparison of vibration analysis techniques. Mechanical Systems and Signal Processing 1997;11:53e73. [125] Baydar N, Ball A. Detection of gear failures via vibration and acoustic signals using wavelet transform. Mechanical Systems and Signal Processing 2003; 17:787e804. [126] Miaoa Q, Makis V. Condition monitoring and classication of rotating machinery using wavelets and Hidden Markov models. Mechanical Systems and Signal Processing 2007;21:840e55. [127] Xiang J, Watson SJ. Practical condition monitoring techniques for offshore wind turbines. In: EWEC 2008 conference proceedings, Brussels, Belgium, PO, vol. 195; 2008. p. 1e8. [128] Kwon KC, Kim JH. Accident identication in nuclear power plants using Hidden Markov models. Engineering Applications of Articial Intelligence 1999;12:491e501. [129] Atlas L, Ostendorf M, Bernard GD. Hidden Markov models for monitoring machining tool-Wear. IEEE. ICASSP 2000;6:3887e90. [130] Ocak H, Loparo KA. A New bearing fault detection and diagnosis Scheme based on Hidden Markov modeling of vibration signals. IEEE ICASSP 2001;5:3141e4. [131] Yam RCM, Tse PW, Li L, Tu P. Intelligent predictive decision support system for condition-based maintenance. The International Journal of Advanced Manufacturing Technology 2001;17(5):383e91. [132] Garcia MC, Sanz-Bobi MA, del Pico J. SIMAP: intelligent system for predictive maintenance application to the health condition monitoring of a windturbine gearbox. Computers in Industry 2006;57(6):552e68. [133] Andrews JD, Moss TR. Reliability and risk assessment. Longmans; 1993. [134] Ghoshal A, Sundaresan MJ, Schulz MJ, Pai PF. Structural health monitoring techniques for wind turbines blades. Journal of Wind Engineering and Industrial Aerodynamics 2000;85:309e24.

178

F.P. Garca Mrquez et al. / Renewable Energy 46 (2012) 169e178 [143] Menon S, Schoess JN, Hamza R, Busch D. Wavelet-Based acoustic emission detection method with adaptive thresholding. In: Proceedings of the SPIE 3986; 2000. p. 71e7. [144] Heng RB, Mohd MJ. Condition monitoring using acoustic signals. In: Proceedings of the Fifth International Conference on Protable Condition Monitoring Fluids and Machinery Performance Monitoring. UK: Mechanical Engineering Publications Limited; 1996. p. 145e58. [145] Kim JS, Kang MC, Yyu BJ, Ji YK. Development of an online tool-life monitoring system using acoustic emission signals in gear shaping. International Journal of Machine Tools and Manufacture 1999;39(11):1761e77. [146] Prabhakaran A, Jagga CR. Condition monitoring of Steam turbine-generator through contamination analysis of used lubricating oil. Tribology International 1999;32:145e52. [147] Van Paepegem W, Degrieck J. Simulating damage and permanent strain in composites under in-plane fatigue loading. Computers and Structures 2005; 83:1930e42. [148] Mba DU, Bannister RH. Protable condition monitoring of low speed rotating machinery using stress waves. In: Proceedings of the Fifth International Conference on Protable Condition Monitoring Fluids and Machinery Performance Monitoring. UK: Mechanical Engineering Publications Limited; 1996. p. 159e68. [149] Hatch C. Improved wind turbines condition monitoring using acceleration enveloping. Orbit; 2004:58e61. [150] Ness S, Sherlock CN. Nondestructive Testing Handbook. In: Nondestructive Testing Overview, vol. 10. American Society for Nondestructive Testing; 1996. [151] Drewry MA, Georgiou GA. A review of NDT techniques for wind turbines. Insight 2007;49(3):137e41.

[135] Hong J-C, Kim YY, Lee HC, Lee YW. Damage detection using the Lipschitz exponent estimated by the wavelet transform: application to vibration modes of a Beam. International Journal of Solids and Structures 2002;Vol. 39: 1803e16. [136] Robinson JC, Canada RG, Piety KR. Vibration monitoring on slow speed machinery: new methodologies covering machinery from 0.5 to 600 RPM. In: Proceedings of the Fifth International Conference on Protable Condition Monitoring Fluids and Machinery Performance Monitoring. UK: Mechanical Engineering Publications Limited; 1996. p. 169e81. [137] Becker E, Poste P. Keeping the blades turning: condition monitoring of wind turbine gears. Refocus 2006;7(2):26e32. [138] Sun Q, Tang Y, Yang W, Ji Y. Feature extraction with discrete wavelet transform for drill wear monitoring. Journal of Vibration and Control 2005;11:1375e96. [139] Nikolaou NG, Antoniadis IA. Rolling element bearing fault diagnosis using wavelet packets. NDT & E International 2002;35(3):197e205. [140] Gieske JH, Rumsey MA. Nondestructive evaluation (NDE) of composite/metal bond interface of a wind turbines blade using an acousto-ultrasonic technique. In: Musial W, Berg DE, editors. ASME Wind Energy Symposium. AIAA/ ASME; 1997. p. 249e54. [141] Beattie AG. Acoustic emission monitoring of a wind turbines blade during a fatigue test. In: . Musial W, Berg DE, editors. ASME Wind Energy Symposium. AIAA/ASME; 1997. p. 239e48. [142] Dutton AG, Blanch MJ, Vionis P, Lekou D, van Delft DRV, Joosse PA, et al. Acoustic emission condition monitoring of wing turbine rotor blades: laboratory certication testing to large scale in-service deployment. In: Proceedings of the 2001 European wind energy conference, Copenhagen, Denmark; 2001.