Fault-Tolerant State Estimation for Large-Scale Systems Based on Structural Analysis*

This paper presents a distributed fault-tolerant state estimation approach for a large-scale system modeled by nonlinear differential algebraic equations. The approach uses structural analysis to decompose the global system into minimal structurally overdetermined (MSO) sets. Then, state observers for the MSOs are calculated. The state observer will only be sensitive to a predefined subset of faults directly affecting the MSO. The proposed approach is implemented in a multi-engine propulsion cluster model, and an unscented Kalman filter (UKF) is computed from four selected MSOs. Simulation results with measurement noise are presented to analyze the efficiency of the proposed approach.