UKF-based fault detection for nonlinear stochastic systems with strong noise and communication protocols.

This paper investigates the problem of unscented Kalman filter-based fault detection for nonlinear stochastic systems with strong noise and communication protocols. Firstly, the state estimation for nonlinear systems with communication protocols (Round-Robin protocol and weighted Try-Once-Discard protocol) is achieved using the unscented Kalman filter. Then, the new residual generator and residual evaluation function are designed using the exponentially weighted moving average method to counteract the impact of strong noise. Based on the stochastic analysis theory, a quantitative inequality is established to analyze the fault detectability in the sense of probability, thereby deriving the relationships among fault amplitude, weight coefficient, false alarm rate, and missed detection rate. Finally, the effectiveness of the proposed fault detection method is verified for the rotary steerable drilling tool system through numerical simulations and experiments.