Computer Engineering Commons^™

Mathematical Justification Of Spectral/Covariance Techniques: On The Example Of Arc Detection, Jan Beck, David Nemir, Vladik Kreinovich

Departmental Technical Reports (CS)

Detecting arcing faults is an important but difficult-to-solve practical problem. Many existing methods of arc detection are based upon acquiring a signal that is proportional to current and then making an analysis of the signal's power spectrum (or, equivalently, its covariance function). Since the power spectrum, i.e., the absolute values of the Fourier transform, carries only partial information about the signal, a natural question is: why should we restrict ourselves to the use of this partial information? A related question is caused by the fact that even the most efficient methods still miss some arcing faults and/or lead to false …