The rapid expansion of wind power has been accompanied by reported reliability problems and the aim is to provide a means of increasing wind turbine reliability, prevent break downs, increase availability and reduce maintenance costs and power outages. This research work reports the development of condition monitoring (CM) for early fault detection in wind turbine blades based on vibration measurements. The research started with a background and a survey of methods used for monitoring wind turbines. Then, finite element modelling (FEM) of three bladed horizontal axis wind turbine (HAWT) was developed to understand the nature and mechanism of the induced vibration. A HAWT test rig was constructed and equipped with computerised vibration measuring system for model verification. Statistical and spectral processing parameters then were used to analyse vibration signals that collected in healthy and faulty cases. Results obtained using time and frequency based techniques are not suitable for extracting blades condition related information. Consequently, empirical mode decomposition method (EMD), principal component analysis method (PCA) and continuous wavelet transform (CWT) are applied for extraction blade condition related features from the measured vibration. The result showed that although these methods generally proved their success in other fields, they have failed to detect small faults or changes in blade structure. Therefore, new techniques were developed using the above mentioned methods combined with feature intensity level (FIL) and crest factor. Namely, those are EDFIL, RMPCA and wavelet based FIL. The new techniques are found to be reliable, robust and sensitive to the severity of faults. Those analysis techniques are suitable to be the detection tool for an integrated wind turbine condition monitoring system. Directions for future work are also given at the end of the thesis.
展开▼
