Kalman filter based sensor fault detection and identification in an electro-pump system

A successful fault detection (FD) procedure depends on correct sensory measurements which may suffer from different sensory soft faults in the form of bias, drift, scaling factor and hard faults which cannot be identified and detected in a standalone use but in combination with other sensors. Thus in this paper the problem of sensory fault detection is considered fusing sensory information. The sensory soft faults are modeled and augmented to electro-pump state space model. Nonlinear model of induction motor is linearized and a state space model for pump subsystem is developed using electrical analogy approach. Both system states and augmented sensory soft faults are then estimated employing a Kalman filter. The efficiency of the method in finally evaluated through simulation.