The Research On Detection Method For Motor Bearing Faults Based On Bicoherence Spectrum

As we all know, induction motors are widely used in numerous fields,such as industrial production,national defence and military. In order to ensure the security of producting,avoid failures and higher-priced maintenance, fault detection will be significative in nascent stage. Under the circumstances of the highest incidence rate of the bearing fault,we study on the detection method of the bearing fault in induction motors in this paper.Motor Current Signature Analysis sets a functional relation between the common faults in the bearing and characteristic frequencies in stator current.While the motors are running with bearing faults, some specific fault characteristic frequencies will be generated accordingly in stator current. However, there are nonlinear phase coupling phenomena between those fault characteristic frequencies and fundamental wave,harmonic wave,which we can consequently confirm the fault category from.While we extract fault characteristic frequencies from the signal, there is a problem of restraining the background of colored nosies from the power network. In addition,weak fault characteristic information has been submerged in noises for incipient faults.As a result,it will be difficult to extract characteristics.We analyse the sampled data by means of the bicoherence spectrum from higher order statistics in this paper. The bicoherence spectrum not only makes use of frequency information in stator current, but also preserves the phase information of the signal,which can reflect well on the phenomenon of phase coupling between each frequency. Additionally, higher order statistics,which can completely restrain gaussian noises in theory, have exceptional ability to eliminate noises.Making use of the higher order statistics in restraining gaussian colored noises,and the bicoherence spectrum in describing the extent of the quadratic phase coupling quantitatively,we put forward the detection method of motor fault characteristics based on the bicoherence spectrum. The general algorithm of the bicoherence spectrum has also been improved in this paper. Firstly,we add windows to the signal in time domain with a set of mutually orthogonal discrete prolate spheroidal sequence rather than the hanning- window.Secondly, we take an average after the sampled data has been sectioned to process in the general algorithm.The frequency spectrum leakage will occur as a result of less data in each section.False peaks will also been found in the result of spectrum estimation.However the modified algorithm process the sampled data in entire section,which can avoid the phenomenon frequency spectrum leakage mentioned above.Last but not the least, the set of mutually orthogonal discrete prolate spheroidal sequence is better to reduce the sidelobe leakage in frequencies of the spectrum estimation,and make the fourier transform of the signal much more energy concentrated.It has been proved that the method mentioned above is effective through the simulation research in this paper.We can recognize the fault state of the bearing significantly via the system of the bearing faults detection online, which is based on the above method.