Multiple fault diagnose and identification based on multi-scale principal component analysis

When using conventional principal component analysis to detect multiple fault, it can lead to the problem of precision decreased. At the stage of fault identification, it canpsilat accurately identify various of faults. For this problem, this paper introduces multi-scale principal component analysis method based on fully analysis that different types of faults have different frequently characteristics. In term of the characteristic that wavelet space in the fine scale shows the big objectpsilas burst and reflects the small objectpsilas slow change in the coarse scale, it can identify efficiently specific types of faults and reduce alarm rates. Simulation results show the efficiency of this method.