A non-intrusive method for sparking assessment in brush dc-motors based on wavelet analysis

Despite the use of brush dc motors in industry has been decremental during recent decades, they are still employed in many industrial sites, even in high output power applications. However, the limited number of research works addressed to them yields a lack of predictive maintenance techniques to diagnose relevant faults in these machines, which may have catastrophic repercussions. Sparking in the commutator/brushes system is a symptom of incorrect operation linked to different possible failures (brush wear, deficient contacts, commutator defects...) and its detection can be crucial to prevent a further development of these anomalies. This work proposes a method for sparking assessment in dc motors based on the computation of the energy of specific signals resulting from the DWT analysis of the armature current. The experimental results confirm that the increment of the sparking activity yields low frequency components that provoke an increment in the energies of high order wavelet signals. Based on this fact, a new indicator of the sparking activity is proposed that shows a high effectivity according to the obtained results. The developed indicator can be a valuable tool for field engineers, satisfying their need of an online, simple and reliable method to assess the condition of these critical parts of dc motors.