Zoom-MUSIC frequency estimation method for three-phase induction machine fault detection

Abstract

Fault diagnosis of electrical machines based on frequency analysis of stator current has been the interest of many researchers for the past twenty years. Several frequency estimation techniques have been developed and are used to help the induction machine fault detection and diagnosis and it is still necessary to test new methods. This paper presents a technique to improve the diagnosis by using the well known MUSIC (MUltiple SIgnal Classification) method. This method is a powerful tool extracting meaningful frequencies from the signal and it has been largely used in different areas included electrical machines. In this application, fault sensitive frequencies have to be found in both the stator current and the stray flux signatures. They are often numerous in a given frequency range and they are affected by the signal-to-noise ratio. Then, the MUSIC method takes a long computation time to find more frequencies by increasing the order of the frequency signal dimension. To solve this problem, an algorithm based on zooming in a specific frequency range is proposed with MUSIC in order to improve the diagnosis performances to extract frequencies. The proposed method has been applied to detect incipient broken bar rotor fault in a three-phase squirrel-cage induction machine.

Stator Short Circuits Detection in PMSM by means of Higher Order Spectral Analysis (HOSA)

“…These methods are classified in non-parametric, parametric and subspace techniques [16,19,20]. The non-parametric methods have the same problems compared to the classical FFT concerning the frequency resolution.…”

“…Then, the parametric methods do not validate the traditional assumption that the data outside the window are equal to zero and their performances are affected by the signal-to-noise ratio. The subspace methods are known as high resolution methods as MUSIC, bispectrum or trispectrum which detect frequencies with low signal-to-noise ratio [19].…”

“…Multiple Signal Classification (MUSIC) method associates the noise to the minimum eigenvalue of the noise variance allowing low signal-to-noise frequency components estimation and the frequency estimation is based on the orthogonality of the signal vectors to the noise eigenvectors [18,19].…”

“…To solve this problem, an algorithm based on zooming in a specific frequency range is proposed with MUSIC in order to improve the diagnosis performances to extract frequencies [18,19]. The Zoom-MUSIC algorithm is a technique to make the MUSIC faster and more accurate in extracting frequencies in a specified bandwidth [18][19][20].…”

“…The disadvantages of HOSA is that has a high computational overhead, and their complex interpretation [16,17]. However, the model order and the processing time can be reduced using filtering and frequency decimating techniques as have described in [18][19][20]. This paper suggest that is possible to study and identify short circuits in the windings of the PMSM determined by means of HOSA as PSD, MUSIC and bispectrum in the whole operation range.…”