Fault detection and isolation for deep space satellites

Abstract

A method for detecting faults in the navigation and control system of deep space satellites is presented. A new method for computing the probability of a fault given multiple different types of residuals processors is presented. The method uses the Shiryayev sequential probability ratio test to estimate the probability of the presence of a fault signal given the residuals generated from either parity relationships or fault detection filters, a fault map of the impact of each fault signal on the residuals, and an adaptive fault estimation scheme that enables processing with fewer residuals. This new methodology is applied to the detection of the fault signals in the attitude control system and navigation system of deep space satellites. First a sensor fusion process is presented for blending star tracker data, gyro data, accelerometer data, and information from the vehicle control system to form the best estimate of the navigation state. Then a set of fault detection filters are developed that detect and uniquely identify faults in each of the sensors or actuators. Decision-making is handled through the sequential processing. Simulation results for a single-satellite system are presented.