Identifiability of Additive Actuator and Sensor Faults by State Augmentation

Actuator and sensor faults can cause poor performance or even instability in dynamic systems. In flight control systems for aircraft and spacecraft, such faults can lead to loss of control and serious incidents. Therefore, rapid detection and identification of actuator and sensor faults is important for enhancing flight safety. One approach to fault detection and identification (FDI) in actuators and sensors is based on multiple-model methods [1,2]. These methods have been extended to detect faults, identify the fault pattern, and estimate the fault values [3,4]. Such methods typically use banks of Kalman–Bucy filters (or extended Kalman filters) in conjunction with multiple hypothesis testing and have been reported to be effective for bias-type faults, such as aircraft control surfaces getting stuck at unknown values, or sensors (e.g., rate gyros) that develop unknown constant or slowly varying biases. A basic requirement for these methods is that the faults should be identifiable. Identification of biases in the inputs and sensors was initially considered in [5]. Identifiability of bias-type faults was considered in [4] and preliminary identifiability conditions were presented. A more detailed analysis of identifiability was presented in [6]. This Note provides a complete characterization of the conditions for identifiability of constant bias-type actuator faults, sensor faults, and simultaneous actuator and sensor faults. Section II considers actuator faults and presents necessary and sufficient conditions (NASC) for their identifiability. Section III presents NASC for identifiability of sensor faults occurring in some of the sensors, and Sec. IV presents NASC for identifiability of simultaneous faults in actuators and sensors. Numerical examples are included to illustrate the results. II. Actuator Faults