Most wind turbines have a rotor made up of three blades and may drive a generator through either a gearbox or be direct drive. The main shaft is supported by large bearings. All of this is housed in a tower-mounted nacelle, (interior pictured above), which is driven into the wind.
Common failures include cracking and roughness on blades, electrical short circuits within the generator and overheating of the gearbox. Misalignments can cause bearing failures and are particularly easy to detect using vibration analysis. Build-up of ice on rotor blades can cause imbalance leading to rapid bearing wear, and is also easily detected by vibration analysis.
Technical consultant David Clark of CMS Wind says: “We see lead times to failures on par with 24 months on the main bearing, 12+ months in the planetary, 7-9 months on the HSS and generator – obviously all speed-associated. We see failure distribution regardless of manufacturer as 1-2% main bearing, 50% gearbox and almost 50% generator. We see lead times from 7-24 months to these failures using properly configured and -accessed vibration condition monitoring.”
The most common method of condition monitoring (CM) for wind turbines is vibration analysis. It is used extensively to monitor the bearings and gearbox, but also has applications on the blades, rotor and tower. Other important methods are acoustic emission, ultrasound testing, oil analysis, strain measurement, electrical effects, shock pulse method (SPM), analysis of process parameters and performance, radiographic inspection, and thermography. Each has benefits.
Oil analysis is used both to determine the quality of the oil itself and to indicate the condition of moving parts. Wear in mechanical parts can often be detected through oil contamination before resulting vibration levels become noticeable. Most oil analysis is still performed manually by taking samples, although sensors can continuously monitor temperature, moisture level and even contamination.
Acoustic emissions are generated by structures when there is a rapid release of strain energy. This is particularly noticeable during crack initiation and propagation in metal parts. The detection of acoustic emissions is similar to vibration analysis, but while vibration sensors are mounted directly on machinery being monitored, acoustic emission sensors detect sounds propagated through the air. In some cases, this can identify faults in bearings, gearboxes and blades more effectively than vibration monitoring.
Ultrasound is used primarily to detect cracks and delamination in structural elements of a wind turbines, most notably the blades. Real-time strain measurement can be measured, using both conventional resistance-based strain gauges and optical fibre sensors. This information is used predominantly as an input to CM models, for example in predicting the rate of crack propagation and therefore the size of cracks which require repair. Radiographic inspection using X-rays may also occasionally be used for similar purposes.
Measurements of key process parameters such as wind speed, rotor speed, blade angle and power output can also provide insight into the condition of a wind turbine. Current and voltage analysis are used to monitor the generator, motors and accumulators. For example, spectral analysis of the generator’s stator current can provide early detection of isolation faults in cabling.
Thermography, although less commonly used for online CM, can identify faults in electronic components, and variations in blade surface temperatures which indicate cracks.
VIBRATION MONITORING
Compared to many other machines, wind turbines have very low shaft rotation speeds, and the speed is also quite variable, dependent on wind speed. The type of sensor best used to detect vibration depends on the frequency. At very low frequencies it is best to use a position transducer: between around 10Hz and 2kHz velocity is ideal, and over 2kHz acceleration gives the most information. Although the blade tip speed may reach 200mph, the rotational speed is relatively slow: perhaps in the range of 10-20rpm, producing vibrations at less than 1Hz.
Other vibrations caused by aerodynamic and sea swell disturbances also cause very low frequency vibrations. However, gearboxes and generators have higher frequencies, of up to 1kHz. Low frequencies and inconsistent shaft speed require longer sample durations as well as appropriate sensors and signal analysis.
The most common method used for vibration analysis in general is spectral analysis. This using the mathematical method of Fast Fourier Transforms (FFT) to decompose a complex signal into a number of underlying simple waveforms. Individual sources of vibration, such as a shaft imbalance or a worn bearing, each produce a waveform with a clearly defined frequency. Considering how this frequency relates to the shaft speed can often provide an insight into what is causing it. Spectral analysis is often used to produce a spectral plot, with the amplitude of vibration at each frequency shown. Spikes at a particular frequency indicate a particular source of vibration and an increase in that spike indicates a fault in this area.
Standard methods for condition monitoring, such as covered by ISO 10816-3, are often poorly-suited to wind turbines. Wind flow disturbances around the tower and nacelle excite resonances within these structures resulting in complex vibrations. These may also combine with sea swell disturbances. This results in vibration signals that look very different to a typical machine in a factory. The ISO 10816 series has therefore been expanded with specific guidance for horizontal axis wind turbines; part 21 covers those with a gearbox and part 22 direct drive turbines.
The wind turbine specific vibration standards highlight the much greater time intervals required to obtain valid vibration signals for the very low frequency vibrations seen in these machines. The lowest frequencies of between 0.1Hz and 10Hz require sampling durations of over 10 minutes. These measurements are typically taken within the nacelle. ISO 10816-21 specifies measurements should record all three cardinal directions – both radial directions and the axial direction, relative to the main shaft. However, in practice only a single direction is usually sufficient.
Somewhat higher frequencies are seen in gearboxes and geared generators, with characteristic frequencies of between 10 Hz and more than 1 kHz. These are suited to shorter sampling durations of around one minute. Again, all three cardinal directions should be measured (both radial directions and the axial direction) according to ISO 10816-21.
Gearbox measurements should be made on the housing close to the input shaft bearing and output shaft bearing. Vibration is either measured on the housing for integrated gearbox-generator designs or on the bearing housings for elastically-coupled designs.