The ability to give a prognosis for failure of a system is an invaluable tool. In this work, four wavelet-based methods have been developed for use with DC motors used in automotive applications that achieve this goal. Wavelet and filter bank theory is reviewed, as well as the nearest neighbor rule, the Minkowski p metrics and linear discriminant functions. The framework for the development of a fault detection and classification algorithm is described. Additionally, an experimental setup based on RT-Linux, and results from testing are presented, verifying the analysis. |