Multifractal 1-D Wavelet Leader based on Spectral Kurtosis of Armature Currents for Sparking Detection in DC Motors

Sparking is a common phenomenon in brushed DC motors. The early detection of the sparking may prevent its aggravation and adopt proper maintenance actions in due time. We propose to apply a Multifractal 1-D wavelet leader analysis to the spectral kurtosis of armature currents to detect the presence of sparks in the commutation system of DC motors. We apply this method to signals captured both under starting and under steady-state regimes to compare and analyze their viability to be used as a basis of the sparking detection method. Likewise, we also propose a quantitative indicator to determine the severity of the failure based on the kurtosis of the signals in the temporal domain. The results prove the potential of the approach for sparking detection and sparking level assessment in DC motors and its suitability for its possible future incorporation in autonomous systems.