Robust FDI for fault-tolerant thrust allocation with application to spacecraft rendezvous

This paper deals with the design and validation of an active fault-tolerant control system to detect, isolate and accommodate a single thruster fault affecting the thruster-based propulsion system of an autonomous spacecraft. The proposed method consists of a fault detector for robust and quick fault detection, a two-stage hierarchical isolation strategy for fault isolation, and an online control allocation unit scheduled by the isolation scheme for fault tolerance. A new factorization approach for the uncertain inertia matrix inverse is proposed. Thanks to this factorization, a novel robust Nonlinear Unknown Input Observers (NUIO) approach is proposed based on LMIs which ensure maximization of the admissible Lipschitz constant while at the same time satisfying an L2 gain bound and some constraints on the observer dynamics. At the first stage of the isolation scheme, a bank of NUIOs is used to identify a subset of possible faulty thrusters. Then, at the second stage, an EKF is introduced to estimate the torque bias directions. Using these directions, jointly with the detector's residual and the information obtained from the first stage, a set of explicit rules is derived to unambiguously isolate the faulty thruster. A Monte Carlo campaign, based on a simulator developed by Thales Alenia Space industries, is conducted in the context of a terminal rendezvous phase of the Mars Sample Return mission. Mission oriented criteria demonstrate that the proposed strategy is able to cope with a large class of realistic thruster faults and to achieve mission success.