Review of failures and condition monitoring in wind turbine(3)

B. Wavelet analysis

examples are given for condition monitoring parameters of

The wavelet analysis is a function that divides a signal different manufacturer in [29], [37]. Among all these

alternatives, electrical sensors installed around the generator into different scale components and that assigns a frequency are highly recommended in recent works [20]-[26] as they to each component. The continuous wavelet transform of the are non-invasive and easy to implement compared to the output power is proposed to monitor bearing failures in the

gearbox and the generator [23]. This technique is also used mechanical ones.

As a wind turbine is an electromechanical system, to detect the rotor mass unbalance in a synchronous electrical failures have almost the same importance as generator and rotor electrical unbalance in an induction mechanical failures. In [38] fault detection of power converter for a variable speed wind turbine has been

proposed. Conventional condition monitoring systems of wind turbines use acoustic and vibration analysis of mechanical parts to detect main bearing [12], blade [13] and drive-train failures [19]. Nevertheless, a potential approach for detecting drive-train mechanical faults using generator electrical signals has been introduced [14], [39]. Since the generator provides electromechanical coupling, the monitoring of its output power could lead to electrical and mechanical fault detection [23]-[26]. The use of the

generator output power for condition monitoring compared to conventional vibration, temperature and oil lubrication monitoring has the following advantages [25]: Reduced number of sensors. Electrical measurements are cheaper than mechanical measurements. Generator power output is already available and can be easily accessed. Both mechanical and electrical signatures are contained in generator output power. The effectiveness of diagnosis based on the measurement of generator currents in double-fed induction generators has been also validated to characterize both stator and rotor electrical faults [21], [22], [40], [41]. Because of the non-stationary nature of signals, the selection of a proper signal

processing method is important to have an accurate

condition monitoring system. Inaccurate signal analysis

leads to a condition monitoring system to several with false alarms which makes the fault detection unreliable. Several signal processing have been proposed for fault detection from various measurements and the most important are mentioned in this section. A. Spectral methods

One of the most known methods in condition monitoring and fault diagnosis is the fast Fourier transform (FFT) which is used in fixed speed wind turbines [17]. In fact, it is not useful in variable speed wind turbine diagnosis because of non-stationary nature of signals. Spectrogram which shows the spectral density of a signal varying with time can be computed from time signal using the short time Fourier transform (STFT) also known as windowed Fourier transform. This type of signal processing has been proposed to analyze short-circuits in stator coils

and rotor unbalance in wind turbine under strong load transient conditions [8].

generator [24].

The conventional continuous wavelet transform involves more intensive computation than the discrete wavelet transform. However, in order to preserve its superior status, a wavelet-based adaptive filter is designed to track the energy in the output power in prescribed fault related frequencies rather than at all frequencies. By using this last technique, the output power and the rotational speed are proposed to detect mechanical and electrical perturbations in wind turbines [25].

Other works based on wavelet transforms are presented to detect both electrical and mechanical faults in wind turbine systems [14]. C. Empirical Mode Decomposition

The empirical mode decomposition (EMD) is a practical method used to decompose a non-stationary and non-linear signal into a finite number of intrinsic modes without any previous knowledge about the signal. It is shown that this technique is potentially an interesting tool to detect modes associated to twice the slip frequency in the output power of induction generator. These frequencies

would appear in a drive train mechanical or electrical fault [26].

V. CONDITION MONITORING SYSTEMS

In wind turbine systems, high power output requires high

levels of torque and consequently high gear-mesh forces.

Because of the low speed of the turbine, the various gearbox

components are usually supported by rolling element bearings. These bearings are subject to significant radial loads and need to be carefully monitored to detect any degradation. Recently, some condition monitoring systems (CMS) have been developed for wind turbines. With these

CMS, wind turbines are instrumented with many accelerometers externally mounted around the generator, the

gearbox and the most critical rolling element bearings (Fig. 5) [42]-[44]. The gearbox has generally a planetary first stage and one or two additional parallel shaft stages and it produces many gear mesh frequencies and their harmonics which are modulated between them. Each bearing produces other characteristic frequencies related to outer race ball pass, inner race ball pass, cage and element spin. All these frequency components are modified by mechanical faults affecting each of them.

Generally, condition monitoring techniques are defined around of vibration analysis, oil analysis, thermography,

strain measurements, acoustic monitoring, electrical effects, process parameters, visual inspection, performance

monitoring and self diagnostic sensors [29], [35], [45]. The current monitoring systems are usually applied to gearbox, generator and main bearing which cause more down time to the whole …… 此处隐藏：5954字，全部文档内容请下载后查看。喜欢就下载吧 ……