Spectral Entropy and Frequency Cepstral Coefficients of Stray Flux Signals for Sparking Detection in DC Motors

In this work, the use of stray flux signals is proposed to detect sparking faults in the commutation system of DC motors. The proposed method uses an algorithm based on the spectral entropy of stray-flux signals. This is combined with a quantitative indicator of the sparking activity that is based on the Mel frequency cespstral coefficients. The proposed algorithm is applied to flux signals captured both under the starting and at steady state to comparatively evaluate the results and analyze their feasibility to be used as the basis of the sparking detection method. The results prove the potential of the proposed approach for sparking fault detection and assessment in DC motors as well as its suitability for possible future incorporation in stand-alone diagnosis systems.