基于改进多信号分类法的异步电机转子故障特征分量的提取

在基于定子电流信号进行异步电机故障诊断时，转子断条故障特征频率分量常常被电流的基频分量淹没。针对这一情况，该文提出一种新的改进的MUSIC方法来提取这一故障特征频率。MUSIC方法通过特征值分解把自相关矩阵中包含的信息空间分成信号子空间和噪声子空间两个正交的子空间，该文提出的改进方法是将信号子空间中对应最大主分量的特征向量移到噪声子空间，这样构成两个新的正交子空间I和II。子空间I由信号中的最大主分量和噪声所对应的特征向量张成，子空间II由其他分量的特征向量张成。把不同频率的信号投影到子空间I，基频信号在该空间的投影将远大于其他的频率分量，因此在投影的倒数谱中，基频分量被抑制，凸显出了故障频率分量。仿真和实验表明，该方法用于提取转子断条故障特征是可行并且是有效的。

Rotor Fault Feature Extraction of Motor Faults of Induction Motor Based on a Modified Music Method

It is difficult to detect the rotor broken-bar fault feature component as it always hides behind the strong supply frequency component in the spectrum of the stator current. A modified MUSIC(multiple signal classification) method is proposed in this paper to extract the rotor broken-bar fault feature component in the spectrum of the stator current. In MUSIC, the information space of the current i is divided into two orthogonal subspaces---the signal subspace and the noise subspace, by the eigenvalue decomposition of the autocorrelation matrix Ri of i. In the modified MUSIC，the eigenvectors related to the maximum eigenvalue are moved from the signal subspace to the noise subspace. Then the information space of the current i is divided into another two orthogonal subspaces---the subspace I and the subspace II. The subspace I is spanned by the eigenvectors related to the maximum principal component and the noise. It is clear that the maximum principal component is mainly made up of the supply frequency component. The subspace II is spanned by the other eigenvectors related to the rest components. By projecting signals with different frequencies into the subspace I, the signal with the supply frequency has the maximum value. Consequently, in the spectrum of the reciprocal of the projecting value, the supply frequency component is cut off and the fault feature component emerges in the spectrum. The results of simulation and experiment show that the proposed method is feasible and effective.