A subspace-based rejection method for detecting bearing fault in asynchronous motor

Fault detection and diagnosis of asynchronous machine is became a central problem for industrial domain since the past decade. A solution to tackle this problem is to use stator current for a great condition monitoring, as referred to the ldquoMCSArdquo (Motor Current Signal Analysis). Indeed, it is known that due to the motor structure, mechanical spectral signature fault, as for example bearing fault, is present in the stator current spectrum. However, a basic spectral analysis is not sufficient to detect a bearing fault, because of the high dynamic of 50 Hz. So, we go further into the existing methods, to introduce a new subspace algorithm. This new method, takes into account the a priori knowledge of the electrical signal frequency (50 Hz) by totally canceling its influence on the desired bearing fault signature. This algorithm is based on the MinNorm algorithm and therefore we name it the Weighted Prior-MinNorm algorithm (WP-MinNorm). Accordingly, we improve the detection and we facilitate the diagnosis of an outer raceway fault. Finally, we illustrate for different load conditions the ability of the method proposed to effectively detect with few collected data, the bearing fault.