Comparison of wavelet-based methods for the prognosis of failures in electric motors

The ability to give a prognosis for failure of a system is an invaluable tool and can be applied to electric motors. In this paper, three wavelet based methods have been developed that achieve this goal. Wavelet and filter bank theory, the nearest neighbor rule, and linear discriminant functions are reviewed. A framework for the development of a fault detection and classification algorithm based on the coefficients calculated from the discrete wavelet transform and using clustering is described. An experimental setup based on RT-Linux is described and results from testing are presented, verifying the analysis.