Fault Classification and Detection by Wavelet-Based Magnetic Signature Recognition

A noninvasive methodology to evaluate and classify electrical system failures is presented in this work. It is based on the electrical system magnetic signature recognition by using the wavelet signal decomposition and the resulting variance spectrum evaluation, respectively. The proposed methodology was validated by comparing theoretical and experimental results. The finite-element method was used in the numerical simulations of the magnetic flux density, and a postprocessing approach was adopted in the signal decomposition and analyses. An experimental setup was built to obtain the magnetic signature regarding some preselected fault configurations.