This paper focuses on segmented vibration signal analysis for precise condition monitoring of electrical machines. A statistical classification based signal decomposition algorithm has been proposed for identification of denser vibrating regions dynamically under various machine operating conditions and thereby to enumerate adaptive thresholds for quick and accurate prediction of abnormalities. The proposed signal decomposition algorithm segments the vibration signal amplitude into classes of equal width over the range of maximum and minimum values and determines the oscillations at multiple levels of the signal amplitude using the transition matrix obtained through statistical classification. The analysis has been carried out over 3,96,000 samples of real-time vibration signal acquired from the shaft of DC motor coupled to AC generator at different operating conditions. The observed variations in the range of oscillations within the scope of segmented classes, with respect to the operating conditions of starting to no load speed and loading along with environmental disturbances emphasize the significance of computation of thresholds dynamically. The technique further traces the changes in non-stationary vibration signal oscillations at every class level accurately. Thus, the deceptive threshold that hides the incipient changes in the behavioral pattern is clearly outlined, thereby resulting to effective condition monitoring.
Author(s) by submitting the manuscript to the International Journal of COMADEM agree to transfer the rights to COMADEM International UK with some exceptions as described in COPYRIGHT TRANSFER AGREEMENT available here.