Amplitude filtering characteristics of singular value decomposition and its application to fault diagnosis of rotating machinery

In this paper, two important properties of singular value decomposition (SVD) are deduced theoretically: (1) number law of singular values: one frequency corresponds to two singular values; (2) order rule of singular values: the larger the amplitude of signal is, the greater the corresponding two singular values are. The above two properties are collectively referred as amplitude filtering characteristics of SVD, and a signal separation algorithm (SVD-AF) based on this characteristic is proposed. Research shows that the algorithm shows excellent characteristics in both extracting multiple and single frequency components. What’s more, the purified signal does not contain redundant components, nor does phase deviation occur. Finally, the proposed algorithm is used to purify axis orbits of the rotor of large sliding bearing test bed, the obtained axis trajectories are clear and concentrated, and the misalignment as well as rub impact fault of the rotor is identified successfully.