Abstract
Electrical induction machines are widely used in the modern wind power production. As their repair cost is important and since their down-time leads to significant income loss, increasing their reliability and optimizing their proactive maintenance process are critical tasks. Many diagnosis systems have been proposed to resolve this issue. However, these systems are failing to recognize accurately the type and the severity level of detected faults in real time. In this chapter, a remote automated control approach applied for electrical induction machines has been suggested as an appropriate solution. It combines developed Fast-ESPRIT method, fault classification algorithm, and fuzzy inference system interconnected with vibration sensors, which are located on various wind turbine components. Furthermore, a new fault severity indicator has been formulated and evaluated to avoid false alarms. Study findings with computer simulation in Matlab prove the satisfactory robustness and performance of the proposed technique in fault classification and diagnosis.
TopBackground
In the literature, numerous recent works are available treating fault detection and recognition in induction machines. These researches can be classified into four kinds: time-domain, frequency-domain, time-frequency-domain, and artificial-intelligence-based methods (Peter Tavner et al., 2008). Each method of the previously mentioned categories usually have advantages of simple implementation, nevertheless they generally suffer from some difficulties in terms of fault recognition accuracy. Besides, the traditional method using vibration measurement may have several disadvantages, for example: technical difficulties of access to the machine, influence of the transmission path, environment noise and sensitivity to the sensor position. Consequently, the most logical implementing of the diagnosis procedure is to compare constantly the outputs of the system with its inputs. The most popular technique used by frequency-domain methods is MCSA which is based on signal spectral analysis (S. Chakkor et al., 2014a). It has been widely applied in the fault discrimination of induction machines (René Husson, 2009), (Vedreño Santos, 2013). Indeed, it has provided good results in many industrial applications. However, this technique can lead sometimes to erroneous diagnostic conclusion because it presents diverse practical limitations due mainly to the following reasons (Bonaldi, et al., 2012), (S. Chakkor, 2014a, 2014b): the spectral leakage, the need for a high frequency resolution, the variation in load conditions during the sampling period, the confusion between the electromechanical frequencies which are similar to others caused by real faults, long measurement period. For these reasons, several signal-processing techniques have been recently applied to facilitate the identification of faults and to enhance MCSA (F. Giri, 2013). All these drawbacks have prompted our motivation to develop and to integrate new diagnosis tool issued of advanced artificial intelligence. In fact, it seems necessary to implement robust detection and classification technique in the objective to ensure an operative monitoring and control of electrical induction machines. Actually, many researches studies have been realized to employ ANN (Z. Chen et al., 2008), fuzzy logic, soft computing algorithms, classification methods and machine learning to avert false alarms (R. E. Bourguet et al., 1994), (Tom M. Mitchell, 1997), (Peter Vas, 1999), (B. K. Bose, 2007), (B. M. Wilamowski et al., 2011). In this framework is sited the contribution of this paper. It consists to develop an approach allowing four tasks: increasing fault detectability, improving fault classification, fault localization and making decision taking into consideration the true quantification of its severity degree.
Key Terms in this Chapter
MDL: Minimum description length. Is a principle in which the best hypothesis (a model and its parameters) for a given set of data is the one that leads to the best compression of the data. MDL was introduced by Jorma Rissanen in 1978. It is an important concept in information theory and computational learning theory.
FIS: Fuzzy inference system. Is defined as a system that uses fuzzy membership functions to make a decision. It uses the “IF…THEN” rules along with connectors “OR” or “AND” for drawing essential decision rules.
MCSA: Motor current signature analysis. A condition monitoring technique used to diagnose problems in induction motors. It was first proposed for use in nuclear power parks for inaccessible motors and motors placed in hazardous areas. It is rapidly applied in industry today. Tests are performed online without interrupting production with motor running under the load at normal operating conditions. It can be used as predictive maintenance tool for detecting common motor faults at early stage and as such prevent expensive catastrophic failures, production outages and extend motor lifetime
CAFH: Classification algorithm of fault harmonics. A classification method of Fault Harmonics for induction machines based on advanced signal-processing technique proposed by S. Chakkor et al.
TLS: Total least squares. In applied statistics, it is a type of errors-in-variables regression, a least squares data modeling technique in which observational errors on both dependent and independent variables are taken into account. It is a generalization of Deming regression and also of orthogonal regression, and can be applied to both linear and non-linear models.
IESRCM: Intelligent embedded system for control and remote monitoring. It is an intelligent electronic embedded device for remote monitoring of wind turbine components in wind parks. It was been proposed by S. Chakkor et al. in 2014.
ESPRIT: Estimation of signal parameters via rotation invariance techniques. A technique to determine parameters of a mixture of sinusoids in a background noise. This technique is first proposed for frequency estimation. However, with the introduction of phased-array systems in daily use technology, it is also used for Angle of arrival estimations as well.
GPS: Global positioning system. A space-based radio-positioning and time-transfer system. GPS satellites transmit signals to proper equipment on the ground. These signals provide accurate position, velocity, and time (PVT) information to an unlimited number of users on ground, sea, air, and space.