On the applications of SVD in fault diagnosis

Practically measured time series signals are reconstructed in phase space. By using singular value decomposition (SVD) theory, the method of singular spectrum characteristic based on the reconstructed attractor track matrix is proposed to increase the signal noise ratio (SNR), then the detection of abrupt information and early rub-impact detection of rotating machinery based on SVD is further proposed. Based on the concept of singular entropy, the problem of reasonable selection for the order of noise reduction in singular spectrum is solved. This method has been applied to the experiment of fault diagnosis. The experiment suggests that it is a simple and effective way to increase the SNR and emphasizes the fault characteristics of the original vibrant signals. It makes the signals in time domain after noise reduction more clear, and it increases the precision of fault diagnosis for equipment and provide for machinery fault diagnosis with an effective tool.