A signal-based approach for detection and isolation of current sensor faults in induction motors

Incipient fault detection and isolation (FDI) in the current sensors of a controlled induction motor is addressed by an original diagnosis system based exclusively on the three-phase signals model. In this way we avoid a reduction in the performance of the FDI system due to the uncertainty in the machine model. Assuming that the three current signals are measured, the proposed FDI system is made of a combination of Kalman filters and statistical change detection and isolation algorithms (CUSUM or Cumulative Sum). The approach is validated by a closed-loop simulation using a squirrel cage induction motor. The latter is controlled by a linear quadratic regulator combined with an extended Kalman filter.