Particle Filter Fault Diagnosis of Highly Automated Aircraft

This paper addresses the challenge of Fault Diagnosis (FD) in automated aircraft in the presence of actuator faults. We introduce a particle filter-based FD method explicitly crafted to detect and isolate prescribed fault modes within this category of systems. Particle filtering, a Sequential Monte Carlo (SMC) simulation methodology, emerges as a highly effective model-based state estimation and change detection method in nonlinear stochastic dynamical systems. Our approach utilizes a hybrid dynamic system formulation, featuring binary and continuous-value states, that streamlines the identification of multiple, potentially concurrent faults. The resulting filter eliminates the computational overhead associated with the realization of the bank of estimators exhibited in conventional Multiple Model FD (MMFD) approaches. In addition, our method boasts a straightforward algorithmic implementation that is well-suited for real-world integration in embedded flight control systems. We exemplify the adaptability of the method by applying it to a classical quadrotor setup, where the targeted fault type involves power loss to any of the four rotors. The applicability and efficacy of the approach are supported by numerical simulations.