A New Entropy Bi-Cepstrum Based-Method for DC Motor Brush Abnormality Recognition

Abnormal arcs in dc motors are often associated with various potential failures or operational defaults. Although they may not directly led to motor breakdown, they can be causes to faults, further damages, and fire hazard. There can be arcs between brushes and rotors when motor running under normal condition, known as normal arcs. However, abnormal arcs, which are difficult to be visually distinguished from normal arcs, occur when there is loosen or contamination of brushes. Therefore, detecting the existence and identifying the type of unusual arcs can be applied as an effective method for brush condition monitoring. This paper presents a detection strategy for abnormality in brush based on the online electromagnetic field (EMF) analysis with advanced feature extraction techniques. The techniques aim at finding the unusual changes in EMF to identify abnormal arc among normal ones. Entropy bi-cepstrum applied as feature extraction method is an inverse spectrum of cumulant. Bi-cepstrum is insensitive to noise, and entropy reflects the complexity of the target signal. In the experiment, three typical types of unusual arcs occurring in brush area are successfully identified, and the result shows the accuracy as high as 91.4%. The new strategy with algorithms can serve as a very useful tool for abnormality recognition of the motor brush.