Robust actuator fault reconstruction for LPV systems using sliding mode observers

This paper presents a fault reconstruction method for LPV systems which is robust against uncertainty and corrupted measurements. The design is analyzed based on a virtual system resulting from factorizing the input distribution matrix associated with the monitored actuators. The key observer parameter (which is used to define the observer gains) is designed using LMIs to minimize the effect of the uncertainty and measurement corruption on the fault reconstruction. The output error injection signals are used to reconstruct the virtual faults, and the input distribution matrix factorization is used to map back to the actual faults. Simulations using an LPV model of a large transport aircraft are presented.