Energy Procedia 24 ( 2012 ) 328 – 339

1876-6102 © 2012 Published by Elsevier Ltd. Selection and/or peer-review under responsibility of SINTEF Energi AS.doi: 10.1016/j.egypro.2012.06.116

DeepWind, 19-20 January 2012, Trondheim, Norway

On the Development of Condition Based Maintenance Strategy for Offshore Wind Farm: Requirement Elicitation Process

Idriss El-Thalji and Erkki Jantunen Industrial Systems, VTT Technical Research Centre of Finland, Espoo 02044, FINLAND

Abstract

Motivation: Condition monitoring and diagnostic practices have become significantly important part of offshore wind farms in order to cut down operation and maintenance costs. The challenge is to enable Condition Based Maintenance (CBM) strategy to be implemented to provide maintenance decisions and services at the right time i.e. maintenance is performed when it is needed and not too early and in vain and not too late i.e. causing breakdown and downtime. CBM implementation is challenging due to the complexity and scalability of remote sensing, data acquisition, data manipulation, state detection, health and prognostic assessment and advisory generation. Objective: The paper describes the condition monitoring and prognostic challenge from both academic and industrial perspectives and presents their requirements in order to enable them to cope with multiple and complex fault and failure mode types and provide cost-effective monitoring and prognostic techniques. Method: The paper utilises needs gathering techniques (i.e. interviews, reports, etc.) and a developed requirements analysis framework in order to provide traceable and high quality research and development requirements. Result: The state of the art of Predictive Health Monitoring (PHM), industrial stakeholders’ requirements for CBM strategy and requirement analysis framework for better traceability and high quality elicitation process.Implication: The requirement elicitation process is importantly required for current offshore wind farms due to advanced technological development, rapid scaling up of systems, and harsher installation sites in order to achieve a successful CBM strategy and higher level of cost effectiveness than exists today. © 2012 Published by Elsevier Ltd. Selection and/or peer-review under responsibility of [name organizer] Keywords: Condition Based Maintenance; Industrial Stakeholders; Wind Energy, Requirement Elicitation; Requirements management

1. Introduction

Condition monitoring and diagnostic practices have become significantly important part of offshore wind farms in order to cut down operation and maintenance costs. The statistics and reports of failures and downtime events are significantly highlighting the need to take into use CBM strategy. Following CBM strategy provides optimal maintenance decisions and services at the right time i.e. maintenance is performed when it is needed and not too early and in vain and not too late i.e. causing breakdown and downtime. In total that impacts the overall system availability and operation and maintenance costs.

Available online at www.sciencedirect.com

© 2012 Published by Elsevier Ltd. Selection and/or peer-review under responsibility of SINTEF Energi AS.Open access under CC BY-NC-ND license.

Open access under CC BY-NC-ND license.

Idriss El-Thalji and Erkki Jantunen / Energy Procedia 24 ( 2012 ) 328 – 339 329

Technically this is quite a challenge for condition monitoring and especially PHM i.e. prognosis. Therefore, this application type requires better understanding of multiple fault and failure modes, symptoms related to loads, environment conditions related to site and season, and operating, and maintenance use-case scenarios. For instance, the detection of wear requires sophisticated measurement and signal analysis techniques in order to be able to give prognosis of the development of wear.

Recently, a number of review papers have been published with focus on condition monitoring techniques please see e.g. Hameed, Ahn, et al. [1], Bin, Yaoyu, et al. [2], Hameed, Hong, et al. [3], Swiszcz, Cruden, et al. [4], Ciang Chia, Lee, et al. [5] ,Amirat, Benbouzid, et al. [6] , Khan, Iqbal, et al. [7]. For blade and rotor system; Jeffries, Chambers, et al. [8] explored the application of the normalized bi-spectrum or bi-coherence for condition monitoring of wind turbine blades. Chin-Shun, Shyh-Jier, et al. [9] proposed a signal-processing approach for the inspection of wind turbine blades. Daneshi-Far, Capolino, et al. [10] reviewed the recent works done on condition monitoring and fault diagnosis for wind turbine generators. Zhang and Li [11] investigated a development platform of offshore wind turbine foundation monitoring system software.

There are a number of commercial condition monitoring suppliers who have been developing and continuously enhancing their systems for wind energy applications. Due to the rapid development within wind energy sector, the PHM systems as supportive systems for wind farms are expected to continually support and monitor the installed wind turbines of different design and configurations, in different installation contexts, with different use-case scenarios.

However, both academic and CM suppliers need to understand what the whole life cycle stakeholders expect of their research and development activities. The needs are provided by customers and industrial stakeholders like wind farm owners, operators, maintenance service providers, etc.

Basically, research and development needs require being collected and engineered into requirements in a form of how engineers and designers can understand and consider them in their system development. The extracted requirements will be later defined as functions and allocated by physical components and/or specifications of new innovative systems. However, there are a number of challenges within CBM requirement engineering in general and within wind energy in particular; first, a PHM system is an integrated system, it deals with multiple physical components, multi-faults, multi-detection, diagnostic and prognostic techniques and has multi physical and organisational interfaces; second, the stakeholder needs by itself are challenging, since they have different types of pitfalls (i.e. incomplete, conflicts, unspecific, context-less, etc.) that impact their traceability and quality.

Therefore, the paper, first reviews the academic contributions within PHM issues, and then collects and analyses the industrial stakeholders needs using developed requirements analysis framework.

It is worth to provide a reference line for both research contributions and industrial development in order to enhance their research interests and directions. The paper starts in the following section with providing the state-of-the-art. In third section, it describes how the needs were collected together with their pitfalls. In the fourth section, it presents the developed framework and analysis, where the empirical findings in terms of high-level and detailed-level requirement are presented in section six. Finally, section seven provides the main conclusions.

2. State of the art of academic research

The papers have been collected and classified into the eight main sub sections of standard PHM. Table 1 illustrates the main academic contributions within PHM, in particular, in wind energy sector. No contributions have been detected within the research database within external systems and technical displays categories. The reason might be that those categories seems as a supportive functions to main categories, thus they are not separately discussed.

330 Idriss El-Thalji and Erkki Jantunen / Energy Procedia 24 ( 2012 ) 328 – 339

Table 1, Academic contributions within wind energy PHM systems No. PHM sub-systems Contribution issues

1 Data acquisition Design , Integration, Simulation, Integration with SCADA

2 Data manipulation Operational parameter classification, Electrical signal analysis

3 State detection Incipient fault detection, Power signal analysis using EMD

4 Health assessment/Diagnostic ANN , BPNN, Markov modelling

5 Prognostic assessment Residual lifetime estimation using sojourn time distribution

6 Advisory generation Maintenance Optimization

7 External systems and data archiving X

8 Technical displays and information presentation X

2.1. Data acquisition

In the European project CONMOW (Condition Monitoring for Offshore Wind Farms) Edwin, Theo, et al. [12] investigated whether a cost-effective integral condition monitoring system could be realized in practice. Verbruggen [13] presented results of a part of the project which had been carried out and entitled WTOMEGA (Wind Turbine O&M based on Condition Monitoring). Zaher, Cruden, et al. [14] proposed a design of a data acquisition platform which could be mounted on a wind turbine. Papadopoulos and Cipcigan [15] presented an ontology model developed with knowledge based editor, Protége 3.3.1, for condition monitoring of wind turbines. Joon-Young, Jae-Kyung, et al. [16] proposed a simulator for a 3MW wind turbine and its condition monitoring system. Kusiak and Li [17] explored the fault data provided by the supervisory control and data acquisition system for fault prediction purposes.

2.2. Data manipulation

Wengang, Fei, et al. [18] proposed multiple operation parameters to classify the complex and non-stationary wind turbine operation conditions related to vibration. Crabtree, Djurovic, et al. [19] examined the possibilities for analyses of electrical signals from the stator of wound rotor induction generators, commonly used in wind turbines.

2.3. State detection

The main scientific contributions regarding experimental methods, simulation methods and field testing have been summarized as follows: Popa, Jensen, et al. [20] performed an experimental investigation for incipient fault detection and adapted existing fault detection methods in the literature suitable for use in wind generator systems using doubly fed induction generators (DFIGs). Wenxian, Jiesheng, et al. [21] proposed a new technique through analysing the total power signals measured from the terminals of the WT generator by using the approach of empirical mode decomposition (EMD). Wenxian, Tavner, et al. [22] proposed a technique that has a versatile function, able to detect both the mechanical and electrical faults in the wind turbine. Wenxian, Tavner, et al. [23] proposed a WT condition monitoring technique that used the generator output power and rotational speed to derive a fault detection signal. Huang, Luo, et al. [24] demonstrated the principle of interrogation system and the selection of the operational method of fiber Bragg grating (FBG) sensors. Amirat, Choqueuse, et al. [25] proposed a new fault detector based on the amplitude demodulation of the three-phase stator current.

Idriss El-Thalji and Erkki Jantunen / Energy Procedia 24 ( 2012 ) 328 – 339 331

2.4. Health Assessment

Garcia, Sanz-Bobi, et al. [26] developed the SIMAP which is an intelligent system for predictive maintenance. It is software addressed to the real-time diagnosis of industrial processes. The incipient detection of anomalies allows for an early diagnosis and the possibility to plan effective maintenance actions. Modelling dynamic non-linear industrial processes by means of artificial neural networks (ANN); characterizing both quantitative knowledge coming from historical data by means of artificial neural networks as well as qualitative knowledge coming from maintenance and operation experts by means of expert systems; performing a dynamical multi-objective non-linear optimization with constraints by means of generic algorithms; representing the uncertainty by means of fuzzy logic. Shulian, Wenhai, et al. [27] indicated that Back Propagation neural network had the excellent abilities of parallel distributed processing, self-study, self-adaptation, self-organization, associative memory , and simultaneously its highly non-linear pattern recognition technology was an efficient and feasible tool to solve complicated state identification problems in the gearbox fault diagnosis. Abeygunawardane and Jirutitijaroen [28] proposed a suitable Markov model for a wind turbine with condition monitoring (CM).

2.5. Prognostic assessment & advisory generation

Eggen [29] defined residual lifetime estimation as a tool for extended planning, where the technical condition of degraded component might be followed until failure occurs. Maximum likelihood or least square methods are useful methods to estimate expected values for the sojourn times. A Bayesian approach for estimating of sojourn time distribution parameters was used, in order to utilize both data and expert judgment based on life time prediction models, operational statistics, and knowledge of specified technical condition. Tian, Jin, et al. [30] proposed an optimal CBM solution. Fonseca, Farinha, et al. [31] presented a system to manage wind turbine maintenance through a predictive model. Table 1 summarized the main research contributions with relation to PHM categories.

Beside the academic researcher, the industrial developers are carrying out significant development efforts to provide robust technology for real applications. Therefore it is important to review their contributions or industrial state-of-the-art in their development process. Crabtree [32] has listed 20 commercial condition monitoring systems. The survey shows that vibration based condition monitoring is the main technique implemented, with trends to be integrated with the oil debris monitoring and fiber optic strain measurement. The survey highlighted three main issues: (1) that major innovation will be in terms of developing signal processing techniques; (2) the industry noticed the needs and importance of monitoring operational parameters such as load, speed, etc.; (3) automation of condition monitoring and diagnostic tasks is acquired by WT operators, specially, for large wind farms with large number of WT units, where manual inspection and data collection are not practical.

3. Stakeholders Needs Collection

Within the process of gathering the stakeholders’ needs a number of problem appeared and that complicated the requirement engineering process with wind energy applications; the PHM system has the following issues:

1. A wind turbine has a number of life cycle processes (i.e. development, design, productions, installation, operation, maintenance, and disposal/update) and within each process there are a number of stakeholders. This leads to increased number of gathered needs and requirement elicitation processes.

2. Stakeholders have either similar or conflicting criteria, different trade-off criteria or different level of importance. Furthermore, there are different stakeholder types, for instance; customers, active stakeholders and passive stakeholders.

332 Idriss El-Thalji and Erkki Jantunen / Energy Procedia 24 ( 2012 ) 328 – 339

3. Stakeholders provide their votes of current systems and image(s) of enhanced systems. These two categories should be processed in different ways. On the other hand, a typical stakeholder doesn’t usually describe the contexts and constraints which are elementary to define the needed constraints and their ranges. That leads to what is called requirement leakage.

4. The needs (i.e. verbal or image) will be transformed into written form; however different stakeholders use different writing styles or procedures to describe their needs. This means that the need or requirement descriptors are not unified or standardized. The main problem is that the transform technique is not providing reversible relationship between needs and requirements and that also leads to requirement losses.

5. Stakeholders describe what the main system needs (i.e. in this case the main system is wind turbine or wind farm) and where the problem ends. That is misleading and shifts the criticality within requirements from cause-root systems to effect-end systems.

6. The stakeholders’ needs are not always enabling the classifying of their needs into main PHM subsystems; remote sensing, data acquisition, data manipulation, state detection, health and prognostic assessment and advisory generation. Furthermore, some of the needs are related to the whole system, the interfaces of two subsystems or of more subsystems, or even to system context and its external system.

7. There is a risk and it often happens that stakeholder needs could be biased, or shifted to handle such a rapid development process. For instance, requirements creep takes place when some functions or specification within written statement of requirements could be modified. Requirement change can occur when stakeholder decides to change his needs or brings in new images to be considered.

8. Each stakeholder is usually prioritizing his own and other stakeholder needs based on his subjective experience and within the limits of his solution. Basically, it is a must to have three level of requirements; (1) scope requirements, (2) high-level or system-level requirements, (3) detailed level requirements.

i. Scope requirements: they describe or contain information of business targets, proposed solution, and confirm an understanding of client’s business idea.

ii. High-level or system-level requirements: they describe how business targets will be facilitated into clear detailed level requirements.

iii. Detailed level requirements: they describe complete and verifiable requirements including functions, specifications, etc. in order to realize a specific system or a specific process.

9. Whenever a stakeholder wants to update his requirements list, he needs to go through the same procedures again which mean repetitive efforts with poor verification level.

These challenges make the requirement elicitation process difficult, time consuming and could impact the validity of outcomes, thus it need to be solved using a practical and effective technique. Therefore, the developed framework will be discussed in following section.

4. Developed Requirement Analysis framework

The developed requirement analysis utilises a rational database and requirement elicitation technique in order to provide system requirement reviews with relation to specific wind turbine issues; scales, installation sites, operating conditions, faults, failures, symptoms, and operating, and service use-cases scenarios.

The need to develop a framework and systematic procedures is important in order to provide systematic transform technique that enables both stakeholders and developers to communicate. It enables a stakeholder to tell what he wants, needs, in the form which the developers are able to understand. The PHM system is one of problematic applications, in which it is really difficult to extract requirements from

Idriss El-Thalji and Erkki Jantunen / Energy Procedia 24 ( 2012 ) 328 – 339 333

the market and the stakeholder needs. The developed framework (illustrated in figure 1) consists of the following modules:

1. User interface for entering needs: the form shown in figure 2 provides stakeholders a user interface to enter the needs using their language.

2. Need Requirement Transform Matrix (NRT): this matrix transforms needs to requirements in a systematic way based on identifiers that are given in spoken or written format. The NRT matrix illustrated in table 2, consists of need column (i.e. right side), identifiers relationships matrix (i.e. top side of matrix), requirement classification and stakeholder importance (i.e. left side), and need identifiers in the middle of matrix. Sequentially, four steps need to be performed; (1) extract identifier(s) as illustrated comprehensively in figure 3, (2) construct related requirements; (3) classify requirements based on three levels (i.e. scope, high level and detailed level) and (4) assign the priority or importance of each requirement from different stakeholders’ point of view.

3. Requirement statement generator: this module generates systematic well defined written requirements based on extracted identifiers. The generic statement is as follows:

The <OBJECT> that support <SYSTEM-TO-SUPPORT> under <USE CASE/SCENARIO> and using <CONCEPT> should/shall be able to <FUNCTION> with <SPECIFICATION> allocated by <COMPONENT> interfaced by <INTERFACE> with <INTERFACED COMPONENT> that makes <OUTPUTS> based on <INPUTS> triggered by <TRIGGERS> constraint by <CONSTRAINT> and possible <RISK>

4. Requirement analysis and Query form: grouping similar requirements is much easier based on identifiers and for different level of system (i.e. wind turbine, gearbox, bearings). Identifier approach is also helpful when analysing requirement conflicts. Based on different identifiers the requirements can be clustered.

Thus, these four modules consist of a number of requirements for engineering procedures and steps as follows:

1. Identify relevant stakeholders in order to cover the whole life cycle processes and their needs. 2. Gather the relevant stakeholder needs (in their language) and insert their needs into user-

interface form. 3. Use NRT matrix and analyse the needs based on their identifiers. 4. Classify needs into three main categories; scope, high level and detailed level requirements. 5. Assign importance or priority based on stakeholder perspective. 6. Formulate the extracted requirements into well-defined written format.

Figure 1, Developed Requirement Analysis Framework

334 Idriss El-Thalji and Erkki Jantunen / Energy Procedia 24 ( 2012 ) 328 – 339

Figure 2, Need entry user interface

Figure 3, Extracted Identifiers

Idriss El-Thalji and Erkki Jantunen / Energy Procedia 24 ( 2012 ) 328 – 339 335

Table 2, illustrative example of Need-Requirement Transform Matrix

we would like to have vibration baseddetect system for plantary gear stage

is it possible to monitor our electricalpanels by thermal measurement

quick repsonse for spare part inventoriesand optimization services

the system should be able to measure theoil quality and ageing

one of the challenge is to detectmislaginment of low speed shaft

enhance the web based featrues andminimize the wind farm operator learningcurve

the system should be able to detect andsort the repetitive and abnormal shocks

the reporting time for alrge wind farm (i.eexisting one takes 2 hr/year, for 100 units ittakes 200 hr/year)

the displacement sensors should be noncontact type

the vibration measurement are notsensitive due to their allocation method

diagnose correctly the measurements meanless service inspections

it is revealed that O&M cost become lowerper installed kW

336 Idriss El-Thalji and Erkki Jantunen / Energy Procedia 24 ( 2012 ) 328 – 339

5. Empirical findings

The extracted requirements have been formulated in well written format based on the need identifiers. Extracted requirements classified based on PHM sub-system are illustrated in a summary shown in Table 3.

Table 3, Extracted Stakeholder Requirements

No. PHM sub-systems # of related requirements 1 Data acquisition 9 2 Data manipulation 8 3 State detection 27 4 Health assessment/Diagnostic 0 5 Prognostic assessment 0 6 Advisory generation 3 7 External systems and data archiving 10 8 Technical displays and information presentation 5 Total 62

5.1. Scope level requirements (PHM business objective)

Operation and maintenance cost was clearly one of the most important business targets that required to be reduced or cut down. Furthermore, most of stakeholder needs highlight the acceptance of the idea toward the shift from being scheduled and corrective maintenance to CBM strategy. Therefore, the need for PHM as enabling technology for CBM strategy is stated in their needs.

5.2. High-level requirements (PHM Concept)

According to stakeholders, cutting down the operation & maintenance cost is directly related to the frequency and scale of duty of planned maintenance, inspections, corrective maintenance and damage severity. Therefore, these issues are related to the capabilities and characteristics of implemented PHM system, which defines functional and technical requirements. This level of requirements mainly is interesting for researchers and developers when they are still talking about different concepts, sub concepts and how to assess, evaluate and integrate them. For example, blade, gearbox, and generator monitoring are three different systems to be monitored with different technologies. However, the operator might emphasize that the whole wind turbine needs to be monitored and as an integrated system instead of an existing system which covers some systems and in a fragmented way. The following points are the main high level requirements:

Integrated PHM for whole wind turbine, drive train, gearbox stages Site and seasonal disturbances and their impact on monitoring scenarios Failure, symptoms and monitoring techniques analysis for different failure modes

5.3. Detailed requirements

Due to the industrial nature the stakeholder requirements are highly focused in this level. Stakeholders can specifically describe their needs in relation to the system, functions, technical specification, etc. Consequently the technical requirements mainly focus on the following issues:

Physical system to support: rotor bearings, planetary gear, electrical panels, ring gear, LSS and HSS shafts and their bearings.

Idriss El-Thalji and Erkki Jantunen / Energy Procedia 24 ( 2012 ) 328 – 339 337

Measurement type: temperature, lubricant temperature, rpm, load, oil age, vibrations. Detection object: wear, abnormal shocks, low rotational vibration, early damage states, imbalance, misalignment, ice. Sorting object: fault alarms, false alarms, damage alarms Data type: sampled signals, information, symptoms, and technicians’ descriptors. Data processing: acquire, transfer, analyse, integrate, and visualize data Monitoring constraints: marginal conditions, pitching and rapid changes Data transfer: within turbine, within wind farm, with server, with service providers. Monitoring considerations: blade angle error, uneven rotor blade profile, blade damage, pitch error, external exciters, drive train misalignment, component scale, operational risks. Remote monitoring and acting: field balancing, alignment, lubrication interpretations Use cases and monitoring scenarios: stand still, idling, normal operating, extreme operating, braking, cut-off, etc. Specifications: scalability, time to response, integrity, user friendly, adaptive, remoteness, analysis performance, reporting performance, sampling intervals, allocation sensitivity, Mean time between failure, mean time to repair/replace, accuracy of detection, diagnosis, prognosis, data security.

6. Conclusions

The outcome of the state-of-the-art: On basis of the state of the art, academic contributions and research efforts are focusing on implementation of different techniques that have been successfully implemented in other industrial sectors as artificial intelligence. In fact, wind turbine has a number of sub systems such as blade, gears, bearings, and generator that require to be monitored. Therefore, still the research contribution is toward enhancing how to monitor the unified system of such aerodynamic, mechanical and electrical sub systems. Therefore, the expectation is also toward on an integrated monitoring system. Remoteness and scale of both wind farm and turbine are two points making wind turbines complicated to monitor and highlight the need of innovative technology enabling PHM systems.

Stakeholders’ needs problems: Requirement engineering and management process for CBM in wind energy sector (i.e. as a relatively new industrial application) is more complicated than other well understood industrial applications. In wind energy sector, the complexity arise due to a number of issues related to the scale of wind turbines, involvement of transfer technologies, number of relevant stakeholders, conflicts and trade-off criteria, responsibilities, and lack of standards. These issues are impacting the whole process and its outcomes. Thus, the extracted requirements could be ambiguous, incomplete, unverifiable, inconsistent, untraceable, and ill-prioritized, that tend to lead the further product development processes to have serious pitfalls.

Developed requirements analysis framework: the developed framework provides a helpful tool and features. that enables the analysis process to be systemized, verified and validated with outcomes from different perspectives taking into consideration the business conflicts, overlapping, trade-off criteria and other subjective issues due to the involvement of human organizational factors.

Industrial stakeholders’ requirements: The extracted requirements highlight a number of development issues such as concepts, functions, specifications, components, and integration. It is also clear that the stakeholders have short term needs, which means that they much more focus on monitoring enhancement and less on diagnostic and prognostic development. Furthermore, their needs are more focused on technical and detailed issues.

Co-R&D between academia and industry: it is clear that long-time horizon and the high order of key acceptance criteria are main characteristics of academia practices, compared to industrial practices which are characterised by short-time horizon and by more balanced key acceptance criteria. Therefore, collaboration should begin in early stage of research and development life cycle, in order to prioritise and

338 Idriss El-Thalji and Erkki Jantunen / Energy Procedia 24 ( 2012 ) 328 – 339

create more relevant academic research for industrial partners and at the same time, to share the associated expenses and responsibilities over whole research and validation processes.

Acknowledgements

The authors wish to acknowledge the research co-operation within AERTOs OMO project. This work has been partially financed by the VTT Graduate School. The authors would also like to thank numerous Wind energy industrial stakeholders that have been directly or indirectly involved in the work.

References

1. Hameed, Z., Ahn, S. H. and Cho, Y. M., 2010, Practical aspects of a condition monitoring system for a wind turbine with emphasis on its design, system architecture, testing and installation, 35, 879-894,

2. Bin, L., Yaoyu, L., Xin, W. and Zhongzhou, Y., 2009, A review of recent advances in wind turbine condition monitoring and fault diagnosis, 1-7,

3. Hameed, Z., Hong, Y. S., Cho, Y. M., Ahn, S. H. and Song, C. K., 2009, Condition monitoring and fault detection of wind turbines and related algorithms: A review, 13, 1-39,

4. Swiszcz, G., Cruden, A., Booth, C. and Leithead, W., 2008, A data acquisition platform for the development of a wind turbine condition monitoring system, 1358-1361,

5. Ciang Chia, C., Lee, J.-R. and Bang, H.-J., 2008, TOPICAL REVIEW Structural health monitoring for a wind turbine system: a review of damage detection methods, 19, 122001,

6. Amirat, Y., Benbouzid, M. E. H., Bensaker, B. and Wamkeue, R., 2007, Condition Monitoring and ault Diagnosis in Wind Energy Conversion Systems: A Review, 2, 1434-1439,

7. Khan, M. M., Iqbal, M. T. and Khan, F., 2005, Reliability and condition monitoring of a wind turbine, 1978-1981,

8. Jeffries, W. Q., Chambers, J. A. and Infield, D. G., 1998, Experience with bicoherence of electrical power for condition monitoring of wind turbine blades, 145, 141-148,

9. Chin-Shun, T., Shyh-Jier, H. and Cheng-Tao, H., 2009, Inspection enhancement of wind turbine blades using novel signal-processing approaches, 1-7,

10. Daneshi-Far, Z., Capolino, G. A. and Henao, H., 2010, Review of failures and condition monitoring in wind turbine generators, 1-6,

11. Zhang, Y.-l. and Li, J., 2010, Design of Offshore Wind Turbine Foundation Monitoring System Based on Excel, 1-4,

12. Edwin, W., Theo, V., Henk, B., Luc, R., Miguel Catalin, T., Andrew, M., Axel Juhl, C., Edwin, B., Pr ftechnik Cm, G., Dirk, S. and Nordex Energy, G., 2008, CONMOW: Condition Monitoring for Offshore Wind Farms,

13. Verbruggen, T. W., 2008, Wind Turbine Operation &amp;amp; Maintenance based on Condition Monitoring,

14. Zaher, A., Cruden, A., Booth, C. and Leithhead, B., 2009, Database management for high resolution condition monitoring of wind turbines, 1-5,

15. Papadopoulos, P. and Cipcigan, L., 2009, Wind turbines' condition monitoring: an ontology model, 1-4,

16. Joon-Young, P., Jae-Kyung, L., Ki-Yong, O., Jun-Shin, L. and Beom-Joo, K., 2010, Design of Simulator for 3MW Wind Turbine and Its Condition Monitoring System, 2181, 930-933,

17. Kusiak, A. and Li, W., 2011, The prediction and diagnosis of wind turbine faults, 36, 16-23, 18. Wengang, S., Fei, W., Yue, Z. and Yongqian, L., 2010, Research on Operation Condition

Classification Method for Vibration Monitoring of Wind Turbine, 1-6,

Idriss El-Thalji and Erkki Jantunen / Energy Procedia 24 ( 2012 ) 328 – 339 339

19. Crabtree, C. J., Djurovic, amp, x, S., Tavner, P. J. and Smith, A. C., 2010, Fault frequency tracking during transient operation of wind turbine generators, 1-5,

20. Popa, L. M., Jensen, B. B., Ritchie, E. and Boldea, I., 2003, Condition monitoring of wind generators, 3, 1839-1846 vol.3,

21. Wenxian, Y., Jiesheng, J., Tavner, P. J. and Crabtree, C. J., 2008, Monitoring wind turbine condition by the approach of Empirical Mode Decomposition, 736-740,

22. Wenxian, Y., Tavner, P. J. and Wilkinson, M., 2008, Wind turbine condition monitoring and fault diagnosis using both mechanical and electrical signatures, 1296-1301,

23. Wenxian, Y., Tavner, P. J., Crabtree, C. J. and Wilkinson, M., 2010, Cost-Effective Condition Monitoring for Wind Turbines, 57, 263-271,

24. Huang, X.-f., Luo, D., Wang, G.-q., Ding, N., Xu, J.-r., Liu, Y. and Fan, H., 2010, State of the Art in Condition Monitoring in Wind Turbines Based on Embedded FBG Sensors, 1-4,

25. Amirat, Y., Choqueuse, V. and Benbouzid, M., 2010, Condition monitoring of wind turbines based on amplitude demodulation, 2417-2421,

26. Garcia, M. C., Sanz-Bobi, M. A. and del Pico, J., 2006, SIMAP: Intelligent System for Predictive Maintenance, 57, 552-568,

27. Shulian, Y., Wenhai, L. and Canlin, W., 2008, The intelligent fault diagnosis of wind turbine gearbox based on artificial neural network, 1327-1330,

28. Abeygunawardane, S. K. and Jirutitijaroen, P., 2009, A Markov model for a wind energy conversion system with condition monitoring, 1-6,

29. Eggen, A., 2009, Handbook on Condition Monitoring of Wind Turbines, Marseille, France 30. Tian, Z., Jin, T., Wu, B. and Ding, F., 2011, Condition based maintenance optimization for wind

power generation systems under continuous monitoring, 36, 1502-1509, 31. Fonseca, I., Farinha, T. and Barbosa, F. M., 2009, On-condition maintenance for wind turbines,

1-6, 32. Crabtree, C., Survey of Commercially Available Condition Monitoring Systems for Wind

Turbines, Supergen Wind Energy Technologies Consortium and Durham University School of Engineering and Computing Sciences, November 2010.