Induction motor fault detection by spectral principal components analysis of the supply currents

A new method of obtaining diagnostic data from induction motors, derived from the three supply currents using principal components analysis, is presented in this paper. The techniques presented here focus on extracting relevant information from spectral matrices. These techniques are qualified as parsimonious tools for exploring the behaviour of current vector valued signals in the frequency domain with minimal loss of information. In fact, the new diagnostic method obtains data from the three stator currents by exploring special fault characteristic frequencies in the power spectral density of the first principal component. The main advantage of this new diagnostic tool is its ability to extract automatically the characteristic frequencies relative to the different machine operating modes. This is accomplished using the proportion of the power attributed to the first principal component and/or using the sensor contribution to the power at specific frequencies. Thus, the new diagnostic method gives a good basis for an automatic non intrusive condition monitoring for rotating machinery.