Holistic Condition Monitoring of Wind Turbines

Felix Diller, winter term 2019/20

Wind turbine Condition Monitoring is one of the most often used applications of Condition Monitoring (CM). Condition Monitoring is the permanent monitoring of parameters representing the wind turbine´s condition to identify their changes, as these can indicate a developing fault.

Introduction to Condition Monitoring of Wind Turbines 

Wind energy is one of the most promising renewable energy technologies. Wind turbines (WTs) are used to convert the power of wind into electrical energy. Condition Monitoring techniques are used to guarantee reliable operation and economic maintenance. CM is often referred to as Structural Health Monitoring (SHM), which is the most important subcategory of CM, because it prevents catastrophic structural failure. Such failure can lead to human damage and high financial loss.

Condition Monitoring consists of two levels: First is the sensor level, whose task is to measure quantities which represent the system´s condition in order to detect a condition change. Second is the information processing level, which must automatically and efficiently evaluate the large amount of data collected. ^[1] Whenever the Condition Monitoring system detects irregularities, it should inform the wind turbine supervisor of a potential fault. Modern algorithms and software solutions such as machine learning are used for data processing.

Reasons Condition Monitoring is predestined for wind turbines ^[2] :

• Wind turbine operation is usually non-supervised, implying that without Condition Monitoring, damage is detected only after total breakdown of the system or in regular maintenance work. Condition Monitoring can be used to provide a warning of possible failure before the breakdown.
• The exchange of parts is very costly and complex because of large and expensive structures, and is connected to significant logistic problems including transport and required infrastructure such as cranes.
• Since wind turbines use wind to produce energy, they must be located in windy areas, which goes along with harsh environments. Other rough climatic factors are storms, including lightning, moisture, ice and a wide operating temperature range. These climatological conditions promote damage and, further, hamper maintenance work and pose risks for the upkeep workers.
• The facts stated are especially applicable to offshore WTs: They are located far away from the coast and are hardly accessible. Additionally, they are subjected to the very rough maritime climate including sea breeze and salt. The location in the ocean and its hydrodynamic environment goes along with a periodic excitation by the waves, which can lead to material fatigue damage and erosion holes in the tower below sea level (scour) due to sediments ^[3] .
• In difference to monitoring other civil engineering structures like bridges, there is a high number of wind turbines of the same type. Monitoring systems can be reused which leads to decreasing prices and the high amount of collected data can be used to improve algorithms to make the information processing more efficient and accurate.
• Maintenance work can be better planned: Unnecessary work can be avoided, and the maintenance work can be placed in periods of suitable weather periods.

Application of Condition Monitoring:

• Restorative CM: Failure is detected.
• Preventive CM: Failure is predicted and prevented.
• Plan maintenance: CM can be used for efficient scheduling of maintenance work to reduce service expenses and can be placed in periods of suitable weather conditions.
• Real-time data, e.g. eigenvalues, can be used to optimally adjust the system to avoid certain running states, e.g. resonance ^[4] . Inappropriate operating states are often a reason for damage ^[3] .
• Lifetime extension decision making ^[5] .
• Results obtained by CM can be used for the development of new wind turbines.

Condition Monitoring methods face different challenges. The system runs in diverse, non-stationary conditions due to varying wind speed. This leads to varying rotational speed of the shaft, frequency and load on the wind turbines. Different operating frequencies complicate the detection of irregularities during frequency analysis. An additional challenge is to ensure that CM saves cost. Therefore, low cost sensors and information processing hardware such as computers are required. Finally, stable systems with a low error rate are needed. Both false alarms and missed alarms should be avoided. Specific considerations are needed to select the monitoring devices, their precision and the influence of detected damage, also called Effect of Defect.

CM methods can be divided into global and local monitoring. Local techniques focus on the monitoring of smaller parts, while global techniques focus on big structures and systems. However, defects monitored globally usually have their origin in local phenomena. ^[6]

To understand the different applications of Condition Monitoring, it is necessary to know the structure of a wind turbine, showed in figure 1. For convenience only, the most used and most efficient wind turbine topology with a horizontal axis and a three-bladed rotor is described. On top of a large tower is the nacelle, which houses the generator, gearbox and electronics. The generator is connected through the shaft, with or without a gearbox (“Direct Drive”) to the blades (usually three). ^[2]

Figure 2 shows the parts which are vulnerable to defects and to down time.

Modal Analysis

see Vibration analysis (Overview) and Schwingungsanalyse: Tragstruktur Windenergieanlage

Modal analysis can be used on every wind turbine component and is the most widely used monitoring method for WTs ^[1] . Therefore, it is discussed in a separate section. 

The modal parameters eigenvalues, eigenmodes and damping ratios are functions of the mechanical properties of the structure. Changes within these parameters can indicate structural changes like cracks. To detect such changes, a continuous monitoring of the dynamical structural response is needed. ^[6] The structural excitation can be active or passive. Active means an artificial excitation by exciters like impulse hammers, but usually passive methods are used (Operational modal analysis OMA). Here, ambient excitation like wind loads, the rotating shaft and other loads causes the structure to oscillate. The big advantages of OMA are the possibility to apply it during operation, the excitation of the whole structure and no need of artificial excitation.

Sensors for vibration measurement are accelerometers, strain gauges or piezoelectric elements. Accelerometers should be placed in the antinodes of the eigenmodes. Other methods to monitor displacements are Digital Image Correlation for Structural Health Monitoring, Continuous Monitoring with Real Time Kinematics GPS and Laser Vibrometry. The location of defects can be detected by using curvature mode shapes und wavelet maps ^[1] .

Tower

The wind turbine tower is the structure which connects the nacelle and the blades with the ground. Usually welded steel tube towers are used. Concrete towers are rare because of high cost. For taller wind turbines a hybrid tower design with reinforced concrete and steel tubes is implemented. ^[2] The tower is the most difficult interchangeable and expensive part of the wind turbine, but the failure rate is usually low.

Electromechanical impedance method

This method uses piezoelectric materials and an electrical alternating field. The piezoelectric material is mechanically coupled to the investigated structure, the host structure. Is the piezoelectric sensor actuated by the electrical field, it deforms. This deformation is impaired by the stiffness of the host structure, which resists the primary electrical field. Therefore, the impedance can be measured. Changes the stiffness of the host structure due to defects, a change in the impedance of the piezoelectric material can be observed. Hence, it is possible to detect local defects within the structure. ^{[1] [8]}

Blades

The blades convert the flow energy of the wind into rotational energy. They are manufactured from fiber-reinforced plastic. Usually glass fiber is used, but for larger structures also carbon fiber is needed. Two construction methods are common, one with a load carrying box, the main spar, with weak outer shells, or with load carrying outer shells. The upwind and downwind shells are manufactured separately and bonded together at the trailing and leading edge. ^[2] As a result of their enormous dimensions and composite material, their maintenance and replacement are complex and expensive ^[1] . Due to their rotation, they are exposed a bending moment each revolution, with an order of magnitude of  load cycles during their lifetime ^[2] . Additionally, aerodynamic turbulences, collisions with flying obstacles such as birds, ice and other bad weather conditions can lead to damage, which can be detected by the following Condition Monitoring methods.

Acoustic emission

Defects in structures can be the source for elastic waves and acoustic emission, which can be detected by piezoelectric sensors. Sources of acoustic emission can be divided into primary and secondary events. A primary event is the event of a crack. During all mechanical cracks, elastic energy is released and an elastic wave is emitted. Brittle cracks are more characteristic because of their sudden occurring and the higher released energy compared to brittle cracks. The elastic waves of brittle cracks have a higher altitude and a smaller frequency range and are therefore easier to detect than ductile failure. Before a total failure of the structure occurs, smaller cracks and debonding and their corresponding waves can be observed. Secondary events are based on mechanism after the defect has occurred. Examples are the rubbing of fracture surfaces and the brittle cracking of corrosion products in the gap in corrosive environments, in which (especially offshore) wind turbines are operated. Figure 3 shows different sources of acoustic emission. ^[9]

Using an array of sensors, the damage can be localized, using the amplitude and the energy of the wave, the type of failure can be evaluated. The technique is also applied to drivetrain components. ^{[1] [10]}

see Acoustic emission testing of fractures in pre-stressed concrete wire , Acoustic emission testing and Composite testing with Lamb waves

Fiber optics method

Optical fibers can be used to monitor strain in fiber-reinforced composites. One method is the use of plastic optical fibers, which are transparent as long as they are not exposed to strain. The optical fibers can be laminated into the laminate of the wind turbine blades. Light is transmitted from a light emitting diode to a photo detector through the optical fiber. If strain is applied to the optical fiber, the fiber gets more and more opaque and the light cannot be transmitted by the optical fiber. This can be detected by the photodetector and it is possible to make assumptions about the strain of the composite.

A more advanced method are Fiber Bragg gratings. Periodic changes of the refractive index within the optical fiber are applied by a laser. The resulting grating reflects waves with the wavelength $//<![CDATA[ \begin{array}{l}\lambda_B=2n_{eff}\Lambda\end{array} //]]>$ . $//<![CDATA[ \begin{array}{l}n_{eff}\end{array} //]]>$ describes the effective refractive index and $//<![CDATA[ \begin{array}{l}\Lambda\end{array} //]]>$ the period of the grating. Waves with other wavelengths are transmitted. If the grating is exposed to strain (mechanical, thermal, …), the grating is stretched, the period is changed and other wavelengths are reflected. Through the detection of the reflected wavelength, the applied strain can be evaluated. The advantages of this method are the feasibility of different fiber Bragg gratings in one single fiber, called multiplexing, the easy integration into the glass fiber reinforced composite and the transmission of information using the wave length making the method robust.

The third type used for fiber optical methods are optical fuses. They are laminated into the composite. If the composite is exposed to an impact, the optical fuses break and transmit light different, which can be detected during optical investigation. ^{[1] [10]}

Electrical resistance-based damage detection

Electrical resistance measurement can be used to detect failure within a composite laminate of electrical conducting fibers, e.g. carbon fibers. Electricity is conducted by the fibers, whereas the matrix (usually polymer such as epoxy) acts as an isolator. The electrical conductivity in fiber direction is much better than perpendicular to the fibers, because of the matrix-fiber interfaces. Less contact of adjacent fibers, e.g. because of a delamination or lower fiber volume fraction, leads to an increasing electrical resistance. Though, a reduced stiffness results in a reduced electrical conductivity. ^[1]

Strain memory alloy method

Ceramics, which change their grid structure irreversible when they are exposed to strain, can be used to investigate the maximum occurring strain during operation. The changes in the grid structure result in a change of the magnetic properties, which can be evaluated by the measurement of the magnetic susceptibility. They do not need an electricity supply and act completely passive. ^[1]

Radar

The use of Radar (Radio Detection and Ranging) is a new method in the CM of wind turbine blades. Usually this technique is used for the detection of failure and their position.

The drawback of this method is, that it is only applicable on electrically conductive materials such as carbon fiber composite. A sensor array is mounted on the tower, where every blade passes by once each revolution. Through differential signal processing, where the signal of the previous measurement is subtracted from the actual measurement, certain changes in the blade structure can be detected.

Another application of radar in the blades is the through-transmission from the root to the tip. Since the blades are hollow, water intrusion can be a problem, which can be detected using radar. ^[11]

Gearbox and Drivetrain

CM solutions for rotating machinery, gears, bearings, and electrical components are taken from other fields of industry and became state of the art for wind turbines ^[6] .

Analysis of general wind turbine parameters and electrical analysis

An economic method to detect faults in wind turbine drivetrain is the use of electrical quantities measured during the operation, such as shaft speed and generated power. Already installed sensors can be used and no additional sensors are required, but complex information processing. Due to varying wind speed and loads, the running parameters are not constant, which complicates the development of needed algorithms. On the other hand, different modes can be investigated. Examples are the power monitoring to detect generator rotor misalignment using fast Fourier and wavelet transformation, and the shaft speed monitoring to detect shorted windings.  With additional and expensive torque measurement sensors, rotor faults can be detected, which cause torsional oscillation or a shift in the torque-speed ratio. ^[12]   Another application is the spectral analysis of the stator current to detect stator winding faults (machine current signature analysis MCSA) ^[13] .

Lubrication analysis

All moving parts in the wind turbine power train are lubricated by lubricants like oil. It is possible to investigate the oil with respect to viscosity, water content, particle count and temperature. One objective is the evaluation of the oil quality to identify a point to change the lubricant. The other objective is the detection of mechanical faults of the bearing or the gears, e.g. broken gear teeth can cause metal swarf contamination. Normally, this method is used more in offline monitoring, but the trend is towards online monitoring. ^{[10] [14]}

Shock pulse method

The shock pulse method (SPM) is used to monitor the bearing of the wind turbine. If the bearing surface has a defect, a perfect rolling of the bearing elements is impossible, and a shock will occur when surfaces are colliding. This impact causes a transient compression wave in the bearing, which can be detected with a shock pulse transducer. The shock pulse is a string of pulses with varying magnitudes. Their frequencies are a much higher than the operating frequency of the bearing and can be distinguished. ^[15]

Thermography

Infrared Thermography can be divided into active and passive thermography. Passive thermography monitors the temperature of a component. Fire or heat due to unintended high friction can be detected. Active methods use an artificial thermal excitation. Irregularities of heat transfer in the structure can be an indicator for damage within the component. The disadvantage of this method is the high price of infrared cameras, which prevents an extensive monitoring application. ^[9]

Literature

1. Ciang, C. C., Lee, J.-R., Bang, H.-J.: Structural health monitoring for a wind turbine system: A review of damage detection methods in Meas. Sci. Technol. (2008), 19:12.
2. Hau, E.: Windkraftanlagen: Grundlagen, Technik, Einsatz, Wirtschaftlichkeit. 5th ed. Springer Vieweg., Berlin (2014).
3. Boon, J. H. den, Sutherland, R.; Whitehouse, R.; Soulsby, R.; Stam, C. J. M.; Verhoeven, K.; Høgedal.: Scour Behaviour and Scour Protection for Monopile Foundations of Offshore Wind Turbines in European Wind Energy Conference & Exhibiton 2004, London, UK.
4. Tewolde, S., Krieger, J., Tesfay, Y., Larson, K.: Improved Structure Safety and Reduced Operational Cost with Structural Health Monitoring for Offshore WTGs in Awea Offshore Windpower Conference 2019, Boston, USA.
5. Rubert, T., Zorzi, G., Fusiek, G., Niewczas, P., McMillan, D., McAlorum, J., Perry, M.: Wind turbine lifetime extension decision-making based on structural health monitoring in Renewable Energy (2019) 143 p. 611–621.
6. Häckell, M. W.: A Holistic Evaluation Framework for Long-Term Structural Health Monitoring. Dissertation, Hannover, 23.09.2015.
7. Hahn, B., Durstewitz, M., Rohrig, K.: Reliability of Wind Turbines in: Wind Energy. Springer, Berlin, Heidelberg (2007), p. 329–332.
8. Rosiek, M., Martowicz, A., Uhl, T.: An Overview of Electromechanical Impedance Method for Damage Detection in Mechanical Structures in 6th European Workshop on Structural Health Monitoring Dresden (EWSHM 2012), July 3-6, 2012, Dresden, Germany.
9. Scruby, C. B.: An introduction to acoustic emission in J. Phys. E: Sci. Instrum. (1987), 20:8, p. 946–953.
10. Coronando, D., Fischer, K.: Condition Monitoring of Wind Turbines: State of the Art, User Experience and Recommendations. Project Report of Fraunhofer Institute for Wind Energy System Technology IWES Northwest, Bremerhaven (2015).
11. Arnold, P., Moll, J., Mälzer, M., Krozer, V., Pozdniakov, D., Salman, R., Rediske, S., Scholz, M., Friedmann, H., Nuber, A.: Radar-based structural health monitoring of wind turbine blades: The case of damage localization in Wind Energy (2018), 21:8, p. 676–680.
12. Lu, B., Li, Y., Wu, X., Yang, Z.: A review of recent advances in wind turbine Condition Monitoring and fault diagnosis in 2009 IEEE Power Electronics and Machines in Wind Applications, IEEE (2009), p. 1–7
13. Huang, S., Wu, X., Liu, X., Gao, J., He, Y.: Overview of Condition Monitoring and operation control of electric power conversion systems in direct-drive wind turbines under faults in Front. Mech. Eng. (2017) 12:3, p. 281–302.
14. Kokila, M., Isakki, P.: A survey of wind turbine control monitoring and fault detection on wind energy in 2016 International Conference on Computing Technologies and Intelligent Data Engineering (ICCTIDE'16), IEEE (2016), p. 1–4.
15. Morando, L. E.: Technology Overview: Shock Pulse Method in Technology Showcase; Integrated Monitoring, Diagnostics and Failure-Prevention in Proceedings of a Joint Conference, April 22-26, 1996, Mobile Alabama.