Fault reconstruction for stochastic hybrid systems with adaptive discontinuous observer and non-homogeneous differentiator

This paper investigates the state estimation and fault reconstruction problems for continuous-time Markovian jump systems, where unknown additive sensor and actuator faults, and actuator degradation are considered simultaneously. First, an augmented descriptor system is proposed where the extended vector is composed of state vector, additive sensor fault and actuator fault vectors. Then, an adaptive sliding mode observer is presented where a switching term is injected to eliminate the effect of actuator degradation. The developed robust observer can achieve estimation of state, additive sensor and actuator fault vectors simultaneously. Based on the observer technique, two methods, namely equivalent output error injection method and non-homogeneous differentiator method, are employed to reconstruct the actuator degradation. Finally, a practical example with an F-404 aircraft engine system is exploited to illustrate the effectiveness of the proposed observer approaches, and make comparisons on these two fault reconstruction schemes.