针对旋转机械振动信号中的工频干扰问题，提出了基于奇异向量频谱的消除方法． 振动信号经奇异值分解后，信号中的有用成分和工频分量被分解到不同的正交子空
间． 因奇异值谱表征信号中各个成分的能量分布，故先由奇异值谱估计与工频相应的奇异向量所在的子空间范围，然后在该范围的子空间内根据奇异向量的频谱特征确定与工频成分对应的奇异向量，进而确定工频成分对应的奇异值，将这些奇异值置0，用其余的奇异值和奇异向量重构为时域信号，从而达到消除工频干扰的目的． 仿真和试验结果表明，所提出的工频干扰消除方法是有效的，在消除信号中工频干扰成分的同时不会影响信号的其他成分．

Proposed is a method on the basis of singular vector spectrum for the removal of power interferences in vibration signals of rotating machinery.In this method,firstly,vibration signals are processed through singular value decomposition,and the useful components as well as the power frequency components are divided into different orthogonal subspaces.Then,as the singular vector spectrum reflects the energy distribution of each signal component,the subspace range of the singular vector corresponding to power interference is estimated according to the singular vector spectrum,in which singular vectors corresponding to power interferences are obtained by analyzing the spectrum features of singular vectors.Finally,the singular values corresponding to power interferences are determined and set to zero,and other singular values as well as singular vectors are used to reconstruct desired time-domain signals without power interferences.Simulated and experimental results demonstrate that the proposed method can effectively remove power interferences without any effect on other components of the vibration signal.