Stator current analysis by subspace methods for fault detection in induction machines

This paper aims to develop a condition monitoring architecture for induction machines, with focus on bearing faults. The main objective of this paper is to identify fault signatures at an early stage by using high-resolution frequency estimation techniques. In particular, we present two subspace methods, which are Root-MUSIC and ESPRIT. Once the frequencies are determined, the amplitude estimation is obtained by using the Least Squares Estimator (LSE). Finally, the amplitude estimation is used to derive a fault severity criterion. The experimental results show that the proposed architecture has the ability to measure the faults severity.