Perspectives on condition monitoring techniques of wind turbines

Fixed-speed WTs

Squirrel cage generators (SCGs) are used for fixed-speed WTs and are directly connected to the grid (Chompoo-Inwai et al., 2005; Hau, 2005) as shown in Figure 3. The drive train consists of a gearbox (in most WTs), a mechanical brake and the electrical generator. The generators are very cost-effective, especially for small and medium WTs. Due to the narrow range of operations and the requirements for reactive power, these generators are not preferred for large WTs. The fixed-speed model is mainly used for small and medium WTs. The generator rotor speed has a constant value for all wind speeds. This means that the tip ratio can only be set by a variation in wind speed. However, the maximum efficiency of a WT is determined at a certain value of tip ratio. Therefore, fixed WTs cannot operate at maximum efficiency based on wind speeds (Variable speed wind turbine, 2017).

OPEN IN VIEWER

Figure 3.

Diagram of a fixed-speed WT (squirrel cage generator).

Variable-speed WTs with DFIG

Variable-speed WTs can generate the maximum efficiency at most wind speeds. The electrical system of a WT with a DFIG is shown in Figure 4. This type of system is used in large WTs. A DFIG has a stator winding, which is directly connected to the grid, while the rotor is supplied by a power converter (Tafticht et al., 2006). For this reason, this type of generator is called doubly fed, because both the stator and rotor windings are energised. A back-to-back power converter is used to rectify the grid voltage at grid frequency and then convert it into the desired three-phase AC voltage at slip frequency for rotor excitation. A DC-link capacitor is connected between the DC and AC sides of the converter as an energy buffer (Yingvivatanapong and Methaprayoon, 2004). A controller is used to regulate separately the real and reactive powers of the stator.

OPEN IN VIEWER

Figure 4.

Diagram of a doubly fed induction generator.

Variable-speed WTs with synchronous generator type

The electrical system of a WT with a permanent magnet synchronous generator (PMSG) is shown in Figure 5. The power converter is connected to the stator winding, and the machine is excited by the rotor magnets. The converter has two stages with an intermediate DC-link to convert the variable frequency of the generator into the constant frequency of the grid (Zafarani et al., 2015). The efficiency and the power density of permanent magnet generators are significantly higher than those of SCGs and DFIGs, because of the lack of excitation current (Polinder et al., 2016). However, they are more expensive because rare-earth magnets (e.g. NdFeB) are necessary for the rotor. In the case of permanent magnet generators, stator faults, bearings and converters must be monitored. These are the only types of generators that can be used without a gearbox, thereby improving the reliability of the WT, although at the premium of a higher cost of the machine (Hansen and Michalke, 2008).

OPEN IN VIEWER

Figure 5.

Diagram of a permanent magnet synchronous generator.

CM of generators

The main parts of the electrical drive system found in fixed-speed WTs are the SCG and gearbox, as shown in Figure 3. A common failure of the SCG is cracking of the rotor bars. The health status of the rotor bars can be monitored by analysing the spectrum of the stator current, since any geometric and magnetic imbalance produces a harmonic having sideband frequencies with respect to the grid frequency (Chompoo-Inwai et al., 2005). Many harmonic detection techniques have been proposed to identify the presence of undesired harmonics in the spectrum of the stator current, using discrete Fourier transform (DFT) (Penman et al., 1994; Siddique et al., 2005). DFT method is broader so it covers all those faults. Each fault can then be detected by their different signatures in the spectrum.

The most common failure mechanism of the generator stator windings is degradation of windings insulation due to high temperature, electrical and mechanical stresses. Many CM methods have been proposed based on thermal monitoring, fault detection of winding turns and partial discharge (Amirat et al., 2009; Nandi and Toliyat, 1999). Electrical or mechanical faults have an impact on the magnetic field in SCGs. This can be identified by measuring the stator current. By the analysis of the variation of current signals, most mechanical and electrical generator faults can be detected. Generator current signature analysis (GCSA) is a non-invasive monitoring approach that is commonly used in fault detection of SCGs (Mehala and Dahiya, 2007). One of the key advantages of this approach is that physical access to the machinery is not required. Current transducers, usually in the form of a clamp, are used to avoid any interference with the system’s original wiring. This method is capable of identifying faults, and it can be used to improve the maintenance regime and to prevent further damage and avoid long downtimes (Mehala and Dahiya, 2007).

The robustness of SCGs is higher than that of DFIGs and PMSGs which are used in variable-speed WTs. This is because of the differences in structure of these generators. There are no brushes or slip-rings in SCGs, and there is also no limitation on the temperature of the magnet in comparison with synchronous generators. Hence, SCGs have lower maintenance costs. However, the output efficiency of DFIGs and PMSGs is much higher than that of SCGs (Blaabjerg and Ma, 2013). Another main advantage of DFIGs and PMSGs is that they capture the maximum power at different wind speeds.

Faults in the rotor winding can be detected by measuring the leakage flux and the shaft voltage, in the case of DFIGs (Jinkyu et al., 2007). In the leakage flux test, a detection coil is mounted near to the rotor, so that it can measure an induced electromotive force due to leakage flux. A decrease in the leakage flux due to a fault in one of the rotor coils can be detected by a decrease in the induced electromotive force. In the case of a fault of the insulation of the rotor winding, a leakage current flows through the shaft and causes a variation in voltage that can be applied as a failure detector.

Thermal monitoring is carried out using temperature sensors locally or by thermographic cameras (Márquez et al., 2012). This can be used to identify faults that cause overheating, such as overloading, stall situations and an unbalanced rotor. For example, it has been shown that fluctuations in stator winding insulation can thoroughly reduce the lifetime; an increase of about 5°C can reduce the life of the machine by around 8 years (Hafezi and Jalilian, 2006). The installation of temperature sensors in between rotor windings and stator windings is crucial. This makes the technique an invasive approach; to achieve the best performance, the sensors must be fitted during the manufacture of the motors. However, the use of thermal cameras to measure the temperature gradients can be advantageous in unsafe conditions (Picazo-Rodenas et al., 2011). The drawbacks of a perfect thermal camera are usually cost and difficulty of the analysis. Another downside is that using the rise in temperature to detect the fault is not fast enough to achieve an optimum fault detection procedure, as the fault has already developed (Neelam, 2010). The rotor winding’s faults can be detected by measuring both the leakage flux and the shaft voltage (Han and Song, 2003). In the leakage flux test, a detection coil is mounted near to the rotor, so that it can measure induced voltage. Any reduction of the leakage flux in one of the rotor coils can be detected via the decrease in the induced voltage. In the shaft voltage test, the shaft-to-ground voltage is monitored, so that any leakage current that flows through the shaft generates a voltage that can be detected by the CMS.

CM of power electronic converters

For variable-speed WTs, a power converter has to be added to the electrical system. The power converter has the function to harmonise the voltage output of the generator with the voltage of the grid, either in terms of magnitude or frequency or both. Statistically, IGBTs are the most fragile parts in converters accounting for 30% of failures, followed by DC-link capacitors and IGBT drivers at nearly 20% and 15%, respectively (Wechsler et al., 2012). Both IGBT and DC-link capacitors should be accurately monitored during the WT operation. IGBTs suffer from thermal stresses due to load variations (wind speed variations). Various types of failure mechanisms occur in semiconductor modules, such as bond wire fatigue (Miao and Makis, 2007), bond wire lift-off (BWLO; Yam et al., 2001), bond wire heel cracking (Schilling et al., 2012), aluminium metallization (Lehmann et al., 2003) and solder fatigue (SF; Ramminger et al., 2000). Almost all of them originate from thermal swings during converter operations and the different thermal expansion coefficient between different layers of the modules (Black, 1969). In fact, thermal stresses originate from both power and thermal cycling of power switching components. Power cycling losses are generated by load variations, while thermal cycling losses are generated by thermal variations (Pang et al., 2004).

The main challenge for the next generation of CMSs for power electronics is to track the temperature variation of the hottest point (junction terminal) as an early failure indicator. The problem is that the temperature of the junction terminal is not accessible for direct measurement. In other words, direct measurement of temperature requires amendment of IGBT module, which is destructive method. This method is not really applicable in non-research-based CM methods.

Estimation of temperature is an alternative method to track variation of temperature and also detect early failure mechanisms of IGBTs. This can be conducted through thermal sensitive electrical parameters (TSEPs) measurements (Ji et al., 2013; Smet et al., 2011). TSEP is a promising approach to detect failure mechanisms of IGBTs, mainly due to the direct accessibility of the required terminals. Another advantage is that different TSEPs have shown altered sensitivity to different failure mechanisms. Therefore, the type of failure mechanism can be distinguished and correlated with other failure mechanisms (Jadhav and Roy, 2013; Selig and Coverstone-Carroll, 1996). Some TSEPs are recognised as offline methods and others as online methods for early failure detection.

In addition, another advantage of TSEPs measurement instead of direct measurement of junction temperature is that the temperature profile is affected in the presence of all mentioned failure mechanisms, as well as load variations (Held et al., 1997; Selig and Coverstone-Carroll, 1995).

Threshold voltage is ensured for CM even at low switching frequency. However, it is not distinguished as an online CM method (Wang et al., 2014), since tracking of this electrical parameter is difficult due to the required high sampling rate and it is inaccurate while the system operates in different load conditions. In addition, gate-emitter voltage, V_GE is also considered as an offline method, since it requires injection of external signals to the gate in order to make this electrical switching parameter measurable (Galdi et al., 2008). Measuring some parameters such as junction temperature T_j, or thermal impedance, or resistance thermal path, t_on, t_off and V_CE,on can be applied for an online CM-based method (Patil et al., 2009). In order to provide valid online CM methods for WTs, it is essential to verify the variation of most appropriate electrical parameters based on the type of failure mechanism. Multi-fault indicators should also be cross-related to obtain accurate CM methods. To cross-relate multi-failure mechanisms, TSEPs can be chosen based on their sensitivity to the presence of each failure mechanism. Ghimire et al. (2014) and Busca et al. (2011) revealed that V_CE,on shows a significant sensitivity to the BWLO compared to the other TSEPs. D Xiang et al. (2012) revealed that T_c changes noticeably in the presence of SF. Zhou et al. (2016) studied odd harmonics of converters and revealed that in the presence of BWLO, the odd harmonic current output of the converter increases. However, the odd harmonic does not show a noticeable sensitivity to the SF. Considering V_CE,on, T_c and the odd harmonic of the converter current as failure detectors can help cancel the adverse effect of multi-failure mechanisms. There is a lack of study in understanding the effects of wind speed variation on failure detectors of the electrical drive systems, that is, DC-link, IGBTs and generators. In the application of WTs, the effect of wind condition on the operation of the electrical system should be studied in order to correlate wind speed variations to failure detectors.

Variable wind conditions cause variations of the electrical parameters and temperature swings. Due to the non-linear relations between the parameters and the ageing of components, a significant computational effort is expected and, hence, advanced monitoring algorithms based on artificial intelligence (AI) are necessary. Thanks to these algorithms, the detection of faults and the consequent decision-making can be reliably carried out even when there are multiple factors. Starting from an initial healthy condition, wind speed, electrical parameters and temperature swings are measured online and any variation is analysed by an AI to provide real-time information of the operation of the components. In other words, each component faces different fault mechanisms and root causes. For instance, for generators, electrical faults can originate from short circuits or electrical asymmetry or inductive imbalance. An early detection of faults based on the elaboration of information related to the fault mechanism is beneficial to significantly enhance the accuracy of the CMS. The AI monitoring system can be based on a fuzzy controller. This type of controller is particularly suitable when the mathematical model of the system is unknown or too complex to be managed by deterministic controllers such as a proportional–integral–derivative (PID) controller. In fact, a fuzzy controller is based on empirical rules and can be readily upgraded for a more complex system by simply increasing the number of fuzzy rules. For example, to monitor the health status of an IGBT, a fuzzy description can prepare the precise capability of analysis of the IGBT’s health status by considering the variation in electrical parameters (V_CE, V_GE and I_GE) and T_j. In addition, many other algorithms such as genetic algorithms (GAs), bee colonies (BC) and neural networks (NN) will be investigated. The amount of variation of the electrical parameters (T_j and V_CE,on) plays a key role in understanding the health status of the switching devices. For instance, a 3% increment of V_CE,on is the symptom of wear-out, whereas an increment between 5% and 10% can be considered an early fault, while a 20% increment can be considered as a critical failure. Similarly, a 5% increment of T_j indicates that wear-out has started, a 10% increment indicates an early fault and a 20% increment indicates a critical failure (Smet et al., 2012). These variations can be introduced as fuzzy rules to estimate the health status of the IGBTs. The CM of power devices is complicated by the fact that some failure mechanisms are positively correlated with an electrical parameter, while other failure mechanisms are negatively correlated to the same electrical parameter. For instance, BWLO produces an increased trend of V_CE,on, while die degradation causes a decreased trend of the same parameter. Therefore, it is essential to understand the correlation between all the parameters related to each fault mechanism to avoid error in the fault detection.

CM of DC-link

DC-link capacitors are used in power electronic convertors in order to balance output power and input source of converter. Statistically, DC-link capacitors are also one of the most vulnerable parts in power electronic converters. Open circuit and short circuit are the most common failure mechanisms of DC-link capacitors (Wechsler et al., 2012). Open circuit occurs due to increase in the ambient temperature, current ripple, voltage stresses of capacitors and humidity, which is led to moisture corrosion and electrical vaporisation. Short circuit causes severe vibration, leading to failure of oxide layer, capacitor body and dielectric film of capacitors. Both failure mechanisms, short circuit and open circuit, cause variations in capacitor electrical parameters, such as leakage current, equivalent series resistance (ESR) and equivalent series inductance (ESI). Tracking variation of those electrical parameters can help detect failure mechanisms (Sankaran et al., 1997). Proper failure detection of DC-link capacitors highly depends on the size of capacitors (Kim et al., 2007). Electrolyte vaporisation is dominant for small capacitors due to limited available surface to temperature dissipation and higher ESR. Breakdown of oxide layer is the most dominant failure mechanisms of DC-link, causes increase in leakage current. A typical method used to monitor capacitors is based on the measurement of the ESR, which is an index of their residual lifetime (Lahyani et al., 1998). An alternative method is based on the measurement of the voltage ripple across the capacitor, which is related to current ripple (Pang and Bryan, 2010; Wechsler et al., 2012). However, the effect of load variations on voltage ripple variations is not considered in this method. Furthermore, acoustic emissions (AEs) are also utilised to distinguish the discharge of capacitor due to the ageing process (Smulko et al., 2011).

CM of bearings

The main reason for bearing faults is torque variations due to variable power, which lead to surface and bending fatigue. Misalignment and improper lubrication are other important causes of bearing faults (Qingbo et al., 2007; Shaeffler, 2007). Degraded bearings can lead to inefficient operation and even failures and associated downtime (Frosini and Bassi, 2010).

The performance of rolling element bearings is critical for the functioning of machinery. Thus, it is important to monitor the condition of bearings continuously to avoid system failures. Early fault detection would also be advantageous for service and maintenance. A number of techniques are currently being used for the detection of defects in rolling element bearings. These include vibration, AE measurements and oil monitoring, using methods such as overall level of kurtosis, spectral analysis, high-frequency resonance and spectrographic oil analysis. Techniques such as particle counting, ferrography and chip detection are also used in certain circumstances (Tandon and Nakra, 1992).

When there is damage, the strike between the damaged area and the components such as races and rollers within the rolling element generates additional noise and vibrations. Rolling element bearing faults are categorised into four groups: roller damage, inner race defect, outer race defect and cage damage (Zhang et al., 2015).

The vibration frequency varies depending on the type of fault. The following equations define the characteristic frequencies (Scheffer and Girdhar, 2004):

FTF (fundamental train frequency) fault on the cage or mechanical looseness

f = ω r 2 ( 1 − R D P D cos β )

(1)

BPFO (ball passing frequency outer race) local fault on outer race

f = ω r 2 ( 1 + R D P D cos β )

(2)

BPFI (ball passing frequency inner race) local fault on inner race

f = n 2 ω r ( 1 + R D P D cos β )

(3)

BFF (ball fault frequency) = 2 × RSF (roller spin frequency) local fault on rolling element

f = P D 2 R D ω r ( 1 − ( R D P D cos β ) 2 )

(4)

where f is the frequency of the item (Hz), ω r is the shaft rotation rate (Hz), R D is the roller diameter (m), P D is the mean roller race diameter (m), β is the contact angle (rad) and n the number of rolling elements (balls) (Scheffer and Girdhar, 2004).

High-frequency resonance technique (HFRT), also known as envelope analysis, extracts fault-related high-frequency components of a signal, which are compared to known bearing fault characteristic frequencies (Darlow et al., 1974).

However, one of the drawbacks of the vibration and AE techniques is that they require direct access to the bearing within the machinery. In some conditions, due to health and safety or for practical reasons, the system cannot be accessed, and direct mounting of the sensors is therefore problematic. Faults in a bearing within a mechanical system that operates by an induction motor cause changes in flux and therefore the current spectrum can be used to identify the changes (Benbouzid, 2000; Singh et al., 2014; Tandon et al., 2007).

Tandon et al. have performed a number of experiments to evaluate the performance of vibration, AE and Motor Current Signature Analysis (MCSA) techniques to detect bearing faults. It was concluded that all the techniques can successfully detect simulated defects in the test bearings. AE was shown to be the most advanced technique, capable of detecting micro-size defects (Graney and Starry, 2011). However, stator current monitoring is the most cost-effective technique (Tandon et al., 2007).

CM of gearbox

In WTs, a gearbox increases the speed of the low-speed shaft to a suitable value that is required in the electricity generators. It is stated that gearboxes cause the longest downtime due to their failures (Spinto et al., 2009).

Most gearbox failures in WTs start in the bearings, even though they usually have high-quality bearings (Musial and Butterfield, 2007). Gear tooth damage, high-speed and low-speed shaft faults are also failures in the WT gearbox. Vibration measurement and spectrum analysis are usually chosen for gearbox monitoring. Vibration methods might not be able to detect faults at early stages due to slow rotation of the main shaft; therefore, AE is often used, as this method can detect stress on the surface that has been caused by a rubbing action in faulty components (Lu et al., 2009).

Time synchronous averaging (TSA) is a well-known vibration monitoring method to reduce signal components that are not synchronous with the shaft rate of rotation, which include electrical and environmental noise, bearing vibrations and vibrations related to other machinery (Bechhoefer and Kingsley, 2009; Zhang, 2011).

TSA extracts periodic waveforms from noise. It is appropriate to use TSA in analysis of gearbox data, as it separates the vibration signature from other sources of noise in the gearbox, which are not synchronous with main shaft speed (Bechhoefer and Kingsley, 2009). Vibration signals related to the gears repeat every rotation. Averaging every single one of these individuals enhances the rotational components of the gear and reduces the amplitude of the other components. This will amplify the gear mesh frequency and its harmonics. This technique requires a constant speed, therefore variable-speed case re-sampling and interpolation are necessary (Bechhoefer and Kingsley, 2009).

The main tooth mesh frequency created by the gears can be used for spectrum analysis and monitoring the gear condition. In the same load and speed, if the peaks rise, this could be an indication of a fault. The gear mesh frequency can be calculated using

f m = N ⋅ f r

(5)

where f m is the mesh frequency (Hz), N is the number of teeth and f r is the shaft rotational speed (Hz).

Frequency analysis is shown to be able to detect faults in a gearbox; however, it has some weaknesses. One drawback is that frequency-domain analysis is effective to identify faults, but it is less efficient for online CMSs due to the overheads caused by the complexity of processing and required parameters. Moreover, frequency-domain analysis, for example, the Fourier transformation, assumes that the system is linear and stationary. Thus, the time factor is not used in the frequency-domain. Therefore, a combination of a time-domain technique and a frequency-domain technique can improve the efficiency of the detection algorithms (Zhang et al., 2012).

Oil debris monitoring using particle count devices has also been shown to be effective. The existence of particles larger than 200 µm in the gearbox oil can indicate bearing spall or gear pitting at an early stage (Dupuis, 2010). The main disadvantages of this method are cost and installation of the sensor, which is intrusive. To install such a sensor, the WT must shut down so that the sensor can be placed within the oil pipe between the gearbox and the cooling system.

CM of brake system

It is vital that WTs stop their operation when a failure of a critical component occurs or at wind speeds greater than their rated limits (Entezami, 2013; Entezami et al., 2012). There are three main types of braking system: aerodynamic, mechanical and electrical. Aerodynamic braking, feathering or blade tips, and electrical braking (using the generator) if available, are mainly applied to reduce the speed, and then, the mechanical brake is used to stop the turbine (Friedrich and Andersen, 2010). Entezami (2013) introduced a new technique to monitor the turbine blade tips using speed measurements of the high-speed shaft. Entezami et al. also state that the wear of the discs can be monitored by measuring the speed of the shaft when the brakes are applied. In WTs, a hydraulic power unit provides an active hydraulic pressure to keep the WT brakes disengaged (Entezami et al., 2012).

The active hydraulic pressure is maintained by a hydraulic system that is driven by a three-phase induction motor. Induction motors can be monitored using different techniques such as MCSA and vibration. In addition to this, the hydraulic pressure can also be monitored. However, this is an intrusive technique, and similar results can be achieved using the other methods.

Mechanical braking is performed using friction between the pads and the disc. This is a large force which is applied on both sides and the temperature of this friction can rise up to 700°C (Entezami et al., 2012; Stiesdal, 1999). Temperature sensors can be used to prevent overheating of the pads and disc by disengaging and engaging the callipers when the temperatures exceed the threshold (Jepsen et al., 2010; Wang et al., 2012).

CM of yaw system

A yaw system turns the nacelle so that the rotors face the wind. This allows the WT to achieve the highest amount of energy by facing the direction of the wind. The yaw system consists of yaw drives, yaw brakes and a yaw bearing. The yaw drive has an electric motor, gears and a shaft pinion engaged with a ring gear on the yaw bearing. The yaw brakes are mounted between the nacelle flange and the yaw bearing (Kim and Dalhoff, 2014). According to the authors’ knowledge, there is no existing CMS for the yaw.

Hotspots and damage scenarios

Damage of different structural components, rotor blades, tower or foundations can occur, depending on the circumstances, but a report that summarises the WT accident data up to 30 September 2017 (Wind Turbine Accident Data, 2017) shows that the most common type of failure is blade-related failure, followed by fire and structural failure. Structural failure does not only include the complete collapse of the structure or the structural components but also consists of all the changes in the system, such as cracks, bolts loosening and scouring. These changes are more likely to lead to variations in the boundary conditions, resulting in changes in the expected behaviour of the structure.

The locations on the structure where the damage is most likely to occur are referred to as hotspots. These locations are the most suitable for optimum sensor placement in order to catch the change and damage at an early phase. For the rotor blades, these locations are considered to be at 30%–35% and 70% along the length of the blade from the root section (Ciang et al., 2008).

Tower structures, made out of steel or concrete, are transported in smaller sections to ease the installation demands and are assembled from parts, which are connected over welded flanges with pre-tensioned bolts. Hotspots on the tower include the sections with stress accumulation such as openings and most importantly sections near the connections between tower substructures, where loosening of the pre-tensioned bolted connections along the flanges can occur, along with failure of the welds.

In the case of offshore WTs, hotspots also include the transition piece, such as grouted connections between tower and foundation structure. Another failure scenario is change in the foundation of the structure due to scouring of the sea bed, which leads to changes in boundary conditions and can influence the natural frequencies and overall behaviour of the structure (Tewolde et al., 2017).

Optimum sensor placement

Strategic placement of sensors, along with the application of a global SHM strategy based on an RBI approach, can significantly reduce maintenance costs and the number of unnecessary site inspections. As previously mentioned, the best positions for sensor placement are the hotspots where damage is most likely to occur, but for the future developments in SHM, it is desirable to position sensors in such way that accurate extrapolation of modal parameters, such as natural frequencies and mode shapes, is possible. Many new strategies rely on tracking the changes in dynamic structural parameters, based on which different damage scenarios are developed using numerical simulation software and later recognised in case of their occurrence. The most commonly used sensors for SHM are accelerometers, displacement transducers, strain gauges, temperature sensors and, in the case of offshore WTs, inclinometers.

Accelerometers measure the vibrations of the structure and are used for modal investigation in output-only or ambient techniques, such as operational modal analysis. They are used to obtain natural frequencies, mode shapes and damping. Tracking the changes in modal parameters, such as lower natural frequency or higher damping, indicates the existing damage in the structure. Accelerometers must be combined with temperature sensors for the temperature effect compensation.

Displacement transducers and strain gauges measure relative displacements of the tower based on the local referent system and are used to monitor the strain history. This information can be used for stress and fatigue analysis and residual lifetime estimation. Figure 8 shows displacement time histories from displacement transducers (W1–W6) placed at two sections at a WT in Dortmund, Germany. They are accompanied by temperature sensors (T1–T6) for temperature effect compensation.

OPEN IN VIEWER

Figure 8.

Time histories of displacements.

In order to develop more advanced structural fatigue and lifetime assessments of wind energy conversion systems (WECS), authors would like to propose additional load identification, since the wind force is generally not being measured. A new wind load identification model is currently being developed by the authors, which is based on the response analysis for the inverse wind load reconstruction.

Nowadays, due to the costs and risk factors of corrective maintenance, SHM systems are regularly being implemented. A detailed overview of CM and SHM strategies can be found in Coronado and Fischer (2015). The trend in recent years has seen a transition from scheduled maintenance to risk-based global monitoring of the structure, which is based on monitoring of dynamic structural characteristics and real-time evaluation of measured data. The objectives and benefits of risk-based SHM include the following:

The disadvantages of risk-based SHM include the risk of sudden unexpected events and the existence of uncertainties in parametric structural investigation, which can lead to unpredicted damage. Nevertheless, a well-studied structural model with good knowledge on possible damage scenarios and structural dynamics can benefit the global monitoring systems and lower the unwanted uncertainties.