Nonintrusive Condition Monitoring and Fault Diagnosis for Wind Turbine Gearboxes Using Generator Current Signal

The goal of this dissertation research is to develop nonintrusive condition monitoring and fault diagnosis methods for wind turbine gearboxes. The proposed methods use only the current signals measured from the terminals or used in the control system of wind turbine generators. Current-based gearbox fault diagnosis has advantages over traditional vibration-based methods in terms of cost, hardware complexity, implementation, and reliability.;This dissertation first provided a comprehensive survey on the state-of-the-art condition monitoring and fault diagnostic technologies for wind turbines. The survey briefly reviewed the gearboxes that are commonly used in wind turbines and their reliability, discussed the common failure mechanisms in wind turbine gearboxes, provided a summary on the condition monitoring and fault diagnostic techniques for wind turbine gearboxes, and focused on the review of the signals and signal processing methods used for wind turbine condition monitoring and fault diagnosis.;This dissertation then analyzed the principle of using nonstationary stator current signals of a generator for the fault detection of a multistage gearbox connected to the generator operating in varying-speed conditions. Based on the analysis, the characteristic frequencies of various gearbox faults in the frequency spectra of the generator stator current signals were identified. A method was then proposed for the fault detection of the gearbox using the current signals. The method consisted of an adaptive signal resampling algorithm to convert the nonstationary characteristic frequencies of gearbox faults in the current signals to constant values, a statistical analysis algorithm to extract the fault features from the frequency spectra of the resampled stator current signals, and a fault detector based on the extracted fault features.;Then, this dissertation proposed a multiclass support vector machine (SVM) classifier for fault type identification of wind turbine gearboxes operating in varying-speed conditions using the fault features extracted. The parameters of the SVMs were optimized by machine learning techniques to achieve the best classification accuracy. The proposed current-based fault detection and multiclass-SVM-classifier-based fault type identification methods were validated by experimental results on a wind turbine drivetrain test rig consisting of a gearbox connected with a permanent-magnet synchronous generator with different faults.