Sensor Fault Detection in a System with Random Disturbances

Sensor faults are detected in an operating automatic system by a simplified version of the dedicated observer scheme. Control inputs are augmented by a random disturbance of moderate intensity. The dedicated observer in this case is a Kalman filter, driven by a single sensor. This filter provides estimates of the outputs from the other, nonredundant, sensors. A logical combination of these functionally redundant signals with the actual sensor signals provides prompt detection of incipient faults on all instruments without false alarms. The scheme is applied to a simulation of the lateral axis control system of a hydrofoil boat in which four sensors are to be covered by the fault detection scheme. Tests indicate that the scheme is robust with respect to variations in the intensity of the random disturbance.