Research Of The Detection Method For Motor Fault Features Based On Cross Higher-Order Cumulants For Two Phase Currents

The advantages of the high reliability, low cost, simple configuration, and better practicability, make induction motor widely used in industry and project. But once the motor breaks out faults, it could lead to termination of the product lines, cause economic loss. So if at the early fault can effectively extract fault features, determine the type of faults and ensure the plan of maintenance pertinencely, it is possible to avoid economic loss. For the medium and small motors of low-voltage, bearing fault is the main fault type, which can account for 40%-60% of all fault types, for small motors even up to 90%. This thesis mainly studies on the diagnostic method of induction motor bearing fault.This thesis uses the method of Motor Current Signature Analysis(MCSA) to describe the bearing fault characteristic. When the bearing has different faults, the corresponding fault characteristic frequency will generate in the stator current. But the amplitude of fault characteristic frequency is small, and there are fundamental waves, harmonic waves and even the background colored noise of power grid in the stator current. Therefore, the early warning of the induction motor faults can be summarized to the problem of extracting weak signal feature from nonlinearity, non-stationary, non-Gaussian signals against the background of colored noises.The thesis uses the phenomenon of phase coupling between fault characteristic frequency and other characteristic frequencies to extract weak fault feature. The traditional spectrum estimation based on second-order statistics ignores the phase information of the signal. However higher-order cumulants can effectively deal with nonlinearity, non-stationary, non-Gaussian signals, which can not only restrain colored noise but also analyze the phenomenon of quadratic phase coupling. So this thesis uses higher-order cumulants, while it does not depend on the faint characteristics'amplitude of slice spectrum too much, but use the phenomenon of phase coupling between the characteristics and other characteristic frequencies to detect bearing fault.By detecting two-phase current of the motor, using the ability of the cross higher-order cumulants to suppress Gaussian colored noise, combining with the ability of Multiple Signal Classification(MUSIC) to extract weak signal characteristics, this thesis puts forward the detection method of motor fault feature based on the cross higher-order cumulants. However, at the same time of improving the resolution, it will result in false peaks inevitably, so that fault characteristics are not well determined. This thesis uses the method of Singular Value Decomposition-Total Least Squares (SVD-TLS) to resolve the problem of false peaks.The validity of above methods have been verified by simulation. At last, this thesis designed a system of detecting bearing fault online, used above methods to recognize the status of the bearing fault effectively.