The Theory And Application Research On Fault Detection And Diagnosis In Electromechanical System

This doctorate thesis investigates into the techniques of Fault Detection and Diagnosis (FDD) in Electromechanical System. Some novel methods of fault signature extraction are presented, with which many kinds of fault are diagnosed in simulations and experiments.In chapter 1, Firstly, fault feature extraction methods and their characteristics in FDD are introduced briefly, research progress, existing problems and application prospects on the fault signal processing with wavelet are stated in detail. Secondly, the research status of fault diagnosis in roller bearing and asynchronous motor is summarized. Finally, significances of the project, structure of dissertation and fruit of research are presented.A novel method for multi-scale denoising based on wavelet transform is given in chapter 2. The new method improves the threshold-processing algorithm and present the form of white noise standard deviation at each scale. Simulations show that while getting rid of noise mostly, this algorithm keeps the main edges of signal well.In chapter 3, a new method for real-time harmonics cancellation based on frequency shifting principle is deduced. The method can be applied to cancellation of a given frequency in fault signal, so the fault feature signal can be detected easily. The function of phase-locked loop (PLL) in the algorithm is analyzed theoretically in detail. The method solved the problem of real-time harmonics and reactive detecting successfully and it can be realized easily.In chapter 4, the Newland's harmonic wavelet is improved and a new type of wavelet with better time-frequency characteristic is constructed. Frequency band of signal can be divided arbitrary with the improved harmonic wavelet so that the phenomena of aliasing are reduced. Several typical fault signals in Electromechanical System are analyzed with it. In addition to the improved harmonic wavelet in chapter 4, a new half-complex wavelet is presented to improve the complex wavelet, which is given by Chaari.O. A unilateral recursive algorithm is derived to improve the real-time characteristic. A parameter n is imported to make the time-frequency and filtering characteristics of it better and its ability of extracting fault feature is enhanced.In chapter 5, the faults diagnoses of stator and rotor in asynchronous motor are mainly investigated. Fault feature in stator current is extracted by singularity detection of signals with half-complex wavelet transform and it can be helpful to the real-time detection of motor fault. Harmonics caused by the fault of motor is detected accurately with real-time harmonics cancellation method. Steady state current and starting current in stator are analyzed with real-time harmonics cancellation method and the improved harmonic wavelet separately, after that the feature component of rotor fault can be detected successfully.In chapter 6, the new half-complex wavelet is adopted to diagnosis defects of rolling bearings. Inherent resonance frequency can be found out by time-frequency analysis with half -complex wavelet, and high frequency resonance analysis is implemented with half-complex wavelet to locate the failure. Experiment result shows that the high frequency resonance analysis based on half-complex wavelet can improved the sensitivity of rolling bearing fault diagnosis...